How the double-ring ClpAP protease motor grips the substrate to unfold and degrade stable proteins

Abstract

Loops in the axial channels of ClpAP and other AAA+ proteases bind a short peptide degron connected by a linker to the N- or C-terminal residue of a native protein to initiate degradation. ATP hydrolysis then powers pore-loop movements that translocate these segments through the channel until a native domain is pulled against the narrow channel entrance, creating an unfolding force. Substrate unfolding is thought to depend on strong contacts between pore loops and a subset of amino acids in the unstructured sequence directly preceding the folded domain. Here, we identify such contact sequences that promote grip for ClpAP and use ClpA structures to place these sequences within ClpA's two AAA+ rings. The positions and chemical nature of certain residues within an unstructured segment that are positioned to interact with the D2 ring have major positive effects on substrate unfolding, whereas segments located within the D1 ring have little consequence. Within the D2-bound segment, two short elements are critical for accelerating degradation; one is at the “top” of D2 and consists of at least two properly positioned nonslippery residues. In contrast, the second D2 element, which can be as short as one residue, is positioned to contact pore loops near the “bottom” of this ring. Comparison with similar studies for ClpXP reveals that positioning a well-gripped substrate sequence within the major unfoldase motor is more important than its proximity to the folded domain and that charged, polar, and hydrophobic residues all contribute favorable contacts to substrate grip.

The health of cells depends upon rapid alterations of their proteomes in response to developmental and environmental changes. A key component of proteome remodeling is targeted protein degradation by proteases of the ATPases associated with diverse cellular activities (AAA+ superfamily), enzymes important in all domains of life. These proteases degrade damaged and misfolded proteins, regulatory proteins that are no longer required, and proteins marked for recycling by exposure of a destabilizing N-end-rule residue, cotranslational tagging with a ssrA degron, or posttranslational ubiquitylation or pupylation (1, 2, 3, 4, 5).

The bacterium Escherichia coli (E. coli) contains five AAA+ proteases, each consisting of an ATP-powered AAA+ unfoldase/translocase motor and a barrel-shaped self-compartmentalized peptidase (1). In the ClpAP, ClpXP, and HslUV proteases, separate AAA+ motor and peptidase subunits assemble to form the active enzymes. By contrast, in the Lon and FtsH proteases, the AAA+ unfoldase and peptidase are encoded in the same polypeptide. In each of these proteases, the AAA+ portions of the enzyme form a hexameric ring with a central axial channel. The ClpA ring is unique among these enzymes in that it contains two distinct AAA+ modules called D1 and D2 (Fig. 1), although similar double-ring AAA+ unfoldases and remodeling enzymes are common in prokaryotes, archaea, and eukaryotes (1, 3, 6).

Figure 1.

Model of protein degradation by the AAA+ protease ClpAP. Model of protein degradation by the AAA+ protease ClpAP. Steps include degron binding and entry of the substrate's unstructured tail into the ClpA channel (steps 1 and 2). Key to the enzyme's success in substrate unfolding is its ability to establish a good grip on this unstructured region of the substrate (step 3). If ClpA can establish/maintain grip, the reaction proceeds to substrate/domain unfolding (step 4) and processive degradation (step 5). Failure to establish and maintain grip can result in substrate back-slipping and/or substrate escape from ClpAP degradation (red arrows, step 3). AAA, ATPases associated with diverse cellular activities.

In ClpAP, one or two double-ring ClpA hexamers bind to heptameric rings of the partner ClpP₁₄ peptidase, which encloses a degradation chamber with 14 internal Ser-His-Asp catalytic triads for peptide-bond cleavage (7, 8, 9). The entrance portals of ClpP₁₄ are too narrow to allow native proteins to enter, and thus active unfolding of substrates and their translocation as denatured polypeptides by ClpA₆ into ClpP₁₄ is required for ClpAP degradation.

Figure 1 outlines five steps involved in ClpAP degradation. For substrate recognition, ClpA binds to a short degron sequence, typically near the substrate’s N or C terminus (step 1) (1, 10). During the early stages of substrate recognition, the degron and, in some cases, additional unstructured linker regions of the substrate are bound by pore loops (and likely other surfaces) within the ClpA axial channel (steps 1 and 2). Formation of these initial ClpA–substrate complexes is reversible (steps 1–3). At a certain point, however, the enzyme becomes committed to unfolding, translocating, and complete degradation of the substrate (steps 4 and 5) (11, 12).

A key feature determining whether degradation ensues (becomes “committed”) is the enzyme's success in unfolding the stable portion of the protein substrate (step 3). Single-molecule analyses indicate that it is the time required for the enzyme to make the first power stroke that succeeds in substantially disrupting the substrate protein’s fold that is rate-limiting for degradation of numerous proteins/domains (13, 14, 15). Factors determining unfolding success depend on features of both the enzyme and substrate. For example, it is becoming increasingly clear that characteristics of the polypeptide sequence that lie within the axial channel when a AAA+ enzyme encounters the folded structure of the substrate can play a vital role (16). These observations support models where certain peptide sequences provide high-quality interactions with the enzyme’s pore loops; these interactions are called enzyme-substrate grip. Sequences providing residues that allow strong grip with this substrate tail increase the probability that a power stroke will successfully unfold the native portion of the substrate (Fig. 1, green arrows). Results of single-molecule optical trapping reveal that ClpAP, ClpXP, and Lon all use, on average, many more futile (nonproductive) ATP-fueled power strokes prior to successful unfolding of stably folded substrates than for nearly identical proteins carrying destabilizing mutations (15, 17, 18). Likewise, recent work with the 26S proteasome demonstrates that better grip is most important for the highest stability substrates (19). If an unfolding power stroke fails because of poor grip on the domain-adjacent substrate linker, then the probability of the substrate back-slipping and/or dissociating from the AAA+ motor increases (Fig. 1, red arrows, steps 3 and 4).

Previous studies on slippery substrate sequences, the opposite of well-gripped sequences, reveal examples in which long stretches of glycines, glycine-alanine repeats, or glutamines cause specific proteins to escape complete degradation and, in some cases, lead to remodeling or partial processing instead (20, 21, 22, 23, 24, 25, 26). In another example, Tian et al. (27) identified a signal consisting of two features that inhibit the complete degradation of a pair of transcription factors by the proteasome: first, a sequence of low complexity (i.e., rich in only one or a few amino acid types) and second, that this sequence directly precedes a tightly folded domain. To date, however, there has only been one systematic examination, a study with ClpXP, of the amino acid determinants in a substrate linker/tail that provide superior grip (16). Here, we perform a similar analysis for ClpAP, which has the added complexity of having two distinct AAA+ rings and an axial channel that is substantially longer than that of ClpX.

To analyze grip by ClpA, we characterize a series of model GFP-linker-ssrA substrates to determine which amino acids are preferentially gripped by the ClpA channel. A 25 poly-Gly sequence between an ssrA degron and a folded GFP domain inhibits ClpAP degradation. By inserting different combinations of residues in multiple locations within this poly-Gly sequence and determining rates of substrate degradation by ClpAP, we establish an initial map of how substrate side chains within the linker and their predicted locations within the two rings of ClpA’s axial channel contribute to grip. We report that substrate–side chain contacts mediated by pore loops in the D1 ring have little role in efficient unfolding and degradation. By contrast, the linker segment within the D2 ring contains two regions of sequence important for grip. Based on these data, we infer that the ClpP-distal (upper, as ClpA is drawn in (Fig. 1)) pore loops of the D2 ring contribute to strong grip by interacting with two or three non-Gly side chains, with hydrophobic and polar groups preferred over negatively charged sequences. Importantly, these contacts must work synergistically with a second grip element composed of nearly any non-Gly residues, provided they are near the ClpP-proximal (bottom, as ClpA is drawn in (Fig. 1)) region of ClpA. Our results further clarify the division of labor between the D1 and D2 rings of ClpA and demonstrate clear differences between determinants of grip for the AAA+ unfolding motors of ClpA and ClpX.

Results

Substrate design for identifying poorly versus well-gripped sequences

Aequorea victoria GFP with a C-terminal ssrA tag (GFP-ssrA) is a convenient model substrate for ClpAP degradation because GFP fluorescence is lost upon unfolding/degradation (28, 29, 30). The ssrA tag is a natural degron that is added to polypeptides during translation in response to the absence of a stop codon on the mRNA and other disturbances in translation (6). The tagging is done by a proteinquality-control system, conserved throughout bacteria, consisting of the ssrA RNA, called transfer-messenger ribonucleic acid (tmRNA), and its required protein factor, SmpB. tmRNA and SmpB are recruited to the stalled ribosome and dislodge defective transcripts to release this ribosome. As part of this process, translation switches from the mRNA to a short ORF on tmRNA, resulting in the addition of the ssrA peptide tag to the C terminus of the nascent polypeptide. This tag sequence is a good degron for multiple AAA+ proteases (i.e., ClpXP, ClpAP, and FtsH in E. coli), thereby ensuring that the mis-translation product is efficiently degraded. In E. coli, the ssrA degron has the sequence AANDENYALAA-COO^- and mutagenesis indicates that the underlined residues are especially important for recognition by ClpA (6, 28, 29, 31, 32).

Using GFP-ssrA at a concentration (5 μM) near the K_M for degradation, and ClpP₁₄ in excess of ClpA₆ (0.4 μM and 0.2 μM, respectively) to minimize free ClpA and favor protease formation, ClpAP degraded GFP-ssrA at a rate of 3.9 ± 0.13 min^-1 ClpA₆^-1 (Fig. 2A). The E. coli ssrA tag is 11 residues long and GFP contains nine unstructured C-terminal amino acids (THGMDELYK) (32, 33, 34). Thus, GFP-ssrA has a total of 20 unstructured C-terminal residues. For our studies, we wanted to test the effects of different amino acids directly abutting the last structured residue of GFP (Ile²²⁹) and thus used the GFP^min variant, which lacks the disordered C-terminal residues of wild-type GFP (16). However, GFP^min-ssrA is not degraded by ClpAP (35), presumably because the 11-residue unstructured segment of this substrate could not reach important binding sites in the axial channel of ClpA. Thus, we constructed substrates with linkers between GFP^min and the full length ssrA tag (Fig. 2B). A cartoon illustrating how an unstructured sequence attached to GFP^min is predicted to be positioned in the axial channel of the double-ring ClpA enzyme is shown in Fig. 2C. This figure assumes that ∼25 residues of the unstructured substrate fill the entire channel as is observed in recent ClpAP cryo-EM structures (7, 9).

Figure 2.

Design of poorly gripped model substrates with baseline degradation rates.A, representative trace showing the loss of GFP-ssrA fluorescence as a function of time during ClpAP degradation. The inset shows the degradation rate of 5 μM GFP-ssrA; each dot is one replicate and the bar is the value of the mean. B, cartoon representation of the model substrate. GFP^min consists of the first 229 residues of GFP, which are those that are stably folded based on structural analysis. The pink region is the 11 amino acid sequence of the ssrA degron. The slip/grip test region (yellow) is inserted in between the last folded residue (Ile229) of GFP and the ssrA degron. C, model of how the 25-residue unstructured linker (slip/grip region, in yellow) may likely be positioned within the ClpA D1 and D2 rings if ClpAP directly abuts GFP^min and the first residue (from the N-terminal side) of the linker can enter the channel. This model assumes each residue takes up about the same amount of space inside the pore. ClpP and the ssrA tail are not drawn to scale. D, sequences of the different length poly-Gly linkers substrates (top) and degradation traces and rates (bottom) of 5 μM GFP^min containing these different linkers. The gray dotted line marks the mean degradation rate of GFP-ssrA (not GFP^min). Note that the Gly₁₂, Gly₁₅, and Gly₁₈ linker substrates are predicted to contain entire or partial ssrA degron sequences (pink underline) within D2, as the length of the linker does not take up the entire ClpA axial channel. The rest of the ssrA degron is predicted to be within the ClpP peptidase chamber or already cleaved.

We inserted poly-Gly linkers of 12, 15, 18, and 25 residues between GFP^min and the full-length ssrA tag and assayed ClpAP degradation of 5 μM of each substrate (Fig. 2, D and E). Among these substrates, the protein with the Gly₁₂ linker was degraded fastest, with the Gly₁₅, Gly₁₈, and Gly₂₅ linkers resulting in successively slower degradation. In the following sections, we operationally consider substrates degraded at a rate of less than 0.5 min^-1 ClpA₆^-1 to be poorly gripped. (See below for consideration of the separate contributions of the steps of substrate recognition and unfolding rate on our degradation values). To search for linker sequences that improved grip by ClpA, we first introduced a Leu-Tyr-Val (LYV) tripeptide at six different positions, each offset by three residues, along the Gly₁₈ linker (Fig. 3A). These substrates are named in the format: ⁷LYV₁₈, where the superscript denotes the first position of the three-residue block within the linker (numbered as in Fig. 2C) and the subscript signifies the poly-Gly background in which the substitutions were introduced (e.g., Gly₁₈ in these first experiments).

Figure 3.

Degradation of GFP^minvariants with initial Leu-Tyr-Val mutations.A, sequences and degradation rates of 5 μM GFP^min containing scanning LYV cassettes positioned as shown. Gray dotted line marks the mean degradation rate of GFP-ssrA. B, degradation rates of 5 μM GFP^min containing single hydrophobic residue substitutions in position 13, 14, or 15 in Gly₁₈ background. The green dotted line marks the average degradation rate of the current “best” GFP^min-linker-ssrA substrate, ¹³LYV₁₈ from panel a, and the gray dotted line marks the mean degradation rate of GFP-ssrA. C, degradation rates of 5 μM GFP^min containing scanning LYV cassettes in Gly₂₄- and Gly₂₅-linkers. The gray and green dotted lines are as in (B). D, degradation rates of 5 μM GFP^min containing scanning LYV cassettes in a 24-residue linker with an invariant ¹³LYV. The gray and green dotted lines are as in (B). (All) Sequences of each variant are shown in the chart in their location relative to D1 and D2 rings (a dash represents a glycine residue). Large hydrophobic residues are colored in green. LYV, Leu-Tyr-Val.

LYV scanning mutagenesis identifies a key grip region

For this initial scan, we used relatively large, hydrophobic residues, as previous studies indicated that bulkier residues are easier for AAA+ proteins to hold onto, perhaps in an analogous way to big knots on a rope improving a climber’s ability to grip compared to a smooth rope or one with small knots (16). LYV tripeptide segments starting at linker positions one, four, seven, and ten had little effect on improving the degradation rate, whereas those starting at positions 13 and 16 increased the degradation rate to ∼1.8 and ∼1.3 min^-1 ClpA₆^-1, respectively, compared to the background rate of ∼0.4 min^-1 ClpA₆^-1 observed with Gly₁₈ (Fig. 3A). As ¹³LYV₁₈ was degraded most rapidly, we focused subsequent analyses on this block of residues. To test if having just one of the three ¹³LYV residues was sufficient to stimulate degradation by ClpAP, we made and tested Gly₁₈ linker substrates containing ¹³L, ¹⁴Y, or ¹⁵V. ClpAP degraded these single-substitution variants at rates similar to the Gly₁₈ parent (Fig. 3B), indicating that the faster degradation of ¹³LYV₁₈ involves synergistic grip effects mediated by at least two of the side chain positions. We return to investigating the minimum number of non-Gly residues required for ClpA to strongly grip the test sequence below.

Cryo-EM structures of ClpAP (7, 9) show ∼24 to 25 residues filling the ClpA axial channel, with ∼10 residues in the D1 ring, ∼12 amino acids within the D2 ring, and ∼2 residues in the small gap between the rings (Fig. 2C). If residues in the slip/grip test linker are bound by ClpA as in Fig. 2C, then positions 13-14-15 would contact the ClpP-distal portion of the D2 ring, suggesting that grip between the ClpP-distal D2 pore loops and side chains in the substrate’s test linker contribute substantially to grip in the context of the Gly₁₈ substrates. However, ssrA-tag residues occupy positions 19 through 25 in the variants constructed in the Gly₁₈ background (Figure 2, Figure 3D, 3, A and B) and could contribute to better grip. Indeed, when we extended the original LYV-scanning study to include positions 19-20-21, 22-23-24, and 23-24-25 using Gly₂₄ or Gly₂₅ linker backgrounds, we observed that these substrates were degraded at very slow rates, similar to that of the parental all-Gly linker substrates (Fig. 3C). Hence, these experiments suggest that ¹³LYV residues interacting with the ClpP-distal region of the ClpA D2 ring contribute to substrate grip, but importantly good grip also requires interactions with non-Gly sequences farther into the D2 ring (compare Gly₂₅ in Fig. 2D to ¹³LYV₂₅ in Fig. 3C).

To probe for this second grip element, in an otherwise Gly₂₅ background, we held the ¹³LYV motif constant and assayed the effects of an additional LYV tripeptide at different linker positions (Fig. 3D). Although addition of ⁴LYV, ¹⁹LYV, and ²²LYV modestly stimulated degradation (to ∼0.7 min^-1 ClpA₆^-1) compared to ¹³LYV₂₅, none of the double LYV test sequences supported GFP^min degradation to near the rate of ∼2.0 min^-1 ClpA₆^-1 observed with ¹³LYV₁₈ (Fig. 3D). These results indicate that grip is not markedly improved simply by the presence of more aromatic/hydrophobic residues along the unstructured test sequence. We reasoned, therefore, that the partial ssrA-tag sequence present in ¹³LYV₁₈ but not in ¹³LYV₂₅ likely contributes to ClpA grip.

Two distinct types of peptide contacts within ClpA D2 mediate grip

The ¹³LYV₁₈ substrate has the sequence ¹³LYVGGGAANDENY₁₈ (ssrA-tag residues underlined) in the region predicted to interact with the ClpP-proximal half of D2 (Fig. 3A). Hence, we cloned, purified, and tested substrates that each carried part of the ssrA sequence: ¹³LYVGGG¹⁹AANGGGG₂₅, ¹³LYVGGGGGG²²DENG₂₅, and ¹³LYVGGGGGGGGG²⁵Y₂₅ (Fig. 4A). The results revealed that the ¹³LYV-²²DEN₂₅ substrate was degraded the fastest, ¹³LYV-¹⁹AAN₂₅ supported an intermediate rate, and ¹³LYV-²⁵Y₂₅ was degraded at the threshold rate, similar to the all-Gly linker substrate. Thus, the ²²DEN and ¹⁹AAN segments of the ssrA tag contribute to grip when they are positioned to interact with the middle or ClpP-proximal portions of the D2 ring. Note however that ²²DEN is also not an autonomous grip element but must be present with the well-gripped ¹³LYV sequence; this conclusion is clearly illustrated by the observation that the Gly₁₈ substrate has the sequence ¹³GGGGGGAANDENY₂₅ within the D2 ring but is degraded very slowly (Fig. 2D). Therefore, we propose that an appropriately positioned DEN sequence acts as a ClpP-proximal-grip motif that functions together with a hydrophobic sequence located in the ClpP-distal region of the D2 ring. We also note that at these ClpP-proximal-grip positions (22-23-24), the polar and charged DEN residues were preferred over the hydrophobic residues provided by the LYV sequence block (compare Figure 3, Figure 4B and 3D).

Figure 4.

Discovery of two distinct substrate regions within D2 that mediate grip.A, sequences of GFP^min variants containing an LYV cassette and different sections of the ssrA tag. B, degradation rates of 5 μM GFP^min variants corresponding to the sequences in (A). C, sequences of GFP^min variants that contain different variations of LYV and DEN cassettes to test effects of the elements in different locations while maintaining the relative spacing of the two sequence blocks. D, degradation rates of 5 μM GFP^min variants corresponding to the sequences in (C). The pink dotted line is the average degradation rate of our new “best substrate” ¹³LYV-²²DEN₂₅ from panel (B). (All) Sequences of each variant are shown in the chart in their location relative to D1 and D2 rings, and a dash represents a glycine residue. Large hydrophobic residues are colored in green. Orange residues are small hydrophobics. Purple denotes polar side chains, and pink denotes negatively charged residues. Gray dotted line marks the mean degradation rate of GFP-ssrA. LYV, Leu-Tyr-Val.

We made three additional substrates to address the importance of the position and the order of LYV and DEN elements in the linker (Fig. 4C). Each of these substrates was degraded ∼4-fold slower than the ¹³LYV-²²DEN₂₅ substrate and only slightly faster than the Gly₂₅ parent (compare Figure 2, Figure 4D and 2E). We conclude that the order of well-gripped substrate elements is important, as, for example, ¹³LYV-²²DEN₂₅ was degraded substantially faster than the ¹³DEN-²²LYV₂₅ substrate with flipped grip elements (Fig. 4D).

Side chain preferences in ClpP-distal and ClpP-proximal grip elements

In the context of linkers composed predominantly of glycines, the ¹³LYV-²²DEN₂₅ substrate was degraded fastest by ClpAP. To dissect and identify key features of the ClpP-distal-grip motif (initially, the ¹³LYV sequence), we made and assayed substrates with the sequence ¹³XXX-²²DEN₂₅ where the X residues were different classes of amino acids (Fig. 5A top, B). Changing the ¹³LYV tripeptide to ¹³GYV, ¹³LGV, ¹³VLY, ¹³AAA, ¹³SAS, ¹³III, or ¹³KRK had little deleterious effect on degradation by ClpAP compared with the ¹³LYV-²²DEN₂₅ substrate and were all degraded at rates >1.5 min^-1 ClpA₆^-1. However, acidic residues were suboptimal, with ¹³EDE-²²DEN₂₅ being degraded more slowly. Remarkably, the ¹³LYG-²²DEN₂₅ substrate was degraded with a ∼4-fold slower rate (∼0.6 min^-1 ClpA₆^-1) than ¹³LYV-²²DEN₂₅, just over our threshold level for poor grip (Fig. 5A top, B). These data suggest that a non-Gly amino acid may be especially important at position 15. However, the “grip-motif code” is not simple but rather this function can be fulfilled in multiple ways. Given the poor degradation of ¹³LYG-²²DEN₂₅, we were surprised by the robust degradation rate observed for the substrate ¹²LYV-²²DEN₂₅ (2.6 ± 0.29 min^-1 ClpA₆^-1) where position 15 was also Gly. This ¹²LYV-²²DEN₂₅ substrate was designed to probe the importance of the exact position of the ClpP-distal-grip motif and the spacing between the two motifs; therefore, “LYV” was shifted one residue distal, closer to the region predicted to be between the D1 and D2 rings. The result that this substrate was degraded as rapidly as some of our “best” substrates reveals that there is flexibility in the precise positioning of well-gripped sequences. Together with our analysis of multiple sequences at positions 13-14-15, these data establish that both the exact location and the chemical nature of well-gripped side chains in the ClpP-distal sequence are not strict in nature, although specific features contribute to superior motifs. One key feature (based on results to date) is that the distal motif requires two properly positioned “grippable” residues, as a single non-Gly in this region is insufficient to provide any detectable grip function.

Figure 5.

Investigating the minimum side chain requirements for a superior grip motif.A, (Top) Sequences of GFP^min variants that contain variations of the ClpP-distal grip motif in positions 13-14-15 including combinations of nonpolar, polar, and charged motifs and an invariant ²²Asp-Glu-Asn block. (Bottom) Sequences of GFP^min variants containing the “gold standard” ClpP-distal grip sequence and different variations of the ClpP-proximal grip motif, a substitution of three alanines, or three positively charged residues. Residue colors are as in Fig. 4 with the addition that blue single letter abbreviations represent positively charged side chains. B, degradation rates of 5 μM GFP^min variants corresponding to the sequences in panel A, top. C, degradation rates of 5 μM GFP^min variants corresponding to the sequences in panel A, bottom. In (B) and (C), the gray dotted line marks the mean degradation rate of GFP-ssrA, and the pink dotted line is the average degradation rate of our newest “best substrate” ¹³LYV-²²DEN₂₅. LYV, Leu-Tyr-Val.

A similar analysis to elucidate good ClpP-proximal-grip motif(s) was carried out using substrates with the sequence ¹³LYV-²²XXX₂₅ (Fig. 5A bottom, C). When dissecting the determinants in the ²²DEN element, substituting one or two of the residues with glycine caused only small reductions in the degradation rate, and replacing the ²²DEN tripeptide with ²²AAA or ²²KRK had even smaller consequences (Fig. 5A bottom, C). Recall that large polar residues were considerably less effective in the ClpP-proximal grip region (e.g., ¹³LYV-²²LYV₂₄, degradation rate: ∼0.7 min^-1 ClpA₆^-1 Fig. 3D). Hence, we conclude that substantial variability is tolerated in both grip elements predicted to occupy the channel of the D2 ring, although in each case, specific positional and chemical preferences are evident. For example, in contrast to the ClpP-distal motif, a single non-Gly in the proximal motif does provide substantial grip function. Thus, our results establish that two very short regions of non-Gly residues spread across the D2 channel are critical for successful degradation by providing grippable sequences that allow ClpAP to work on and successively unfold the difficult GFP^min substrate.

The slip/grip linker principally affects unfolding rather than substrate recognition

The goal of this study was to analyze how sequences within the ClpA channel, as this unfoldase approaches a stably folded protein or domain, affect the enzyme’s ability to promote efficient unfolding—an effect referred to as “grip.” However, the degradation assays used to analyze our substrates did not distinguish between the effects of sequence changes on substrate recognition (i.e., the rates of forming and dissociating of the ClpA–substrate complex) versus protein unfolding. To help identify the reaction step(s) most affected by changes in the linker region, we determined the rate of degradation as a function of substrate concentration and fit these data to the Michaelis–Menten equation (Fig. 6). Although protein degradation by AAA+ proteases does not conform to the simple Michaelis–Menten formalism, this type of analysis is widely used to gain information on whether specific substrates are poorly recognized or poorly processed by the mechanical unfolding activity of the enzymes. The V_MAX parameter is especially informative as good versus poor recognition should not affect its magnitude. (Note, all V_MAX values were normalized to the concentration of ClpA hexamers). The K_M values also inform about which reaction step(s) are altered by changes in the substrate; during substrate recognition, dissociation of the enzyme–substrate complex is usually much more rapid than the unfolding rate, such that the K_M value is often similar to the K_D for the ClpAP–substrate complex (36).

Figure 6.

Michaelis–Menten analysis to compare different variations of the ClpP-distal D2 motifs. To determine steady-state kinetic parameters, we assayed initial rates of degradation of the GFP^min-linker-ssrA substrates by ClpAP using a fixed concentration of enzyme (0.2 μM ClpA₆, 0.4 μM ClpP₁₄) and increasing concentrations of the different GFP^min variants. These data were fit to the Michaelis–Menten equation using GraphPad Prism 10 (https://www.graphpad.com/guides/prism/latest/curve-fitting/reg_michaelis_menten_enzyme.htm) to determine V_MAX and K_M values. The experiments were performed in technical triplicates and repeated on multiple days. The plotted values are the averaged degradation rates across all experimental replicates. Michaelis–Menten analysis of ClpAP degradation of ¹³LYV-²²DEN₂₅¹³GYV-²²DEN₂₅¹³LGV-²²DEN₂₅ and ¹³LYG-²²DEN₂₅ yielded the following: V_MAX values (in min^-1 ClpA₆^-1) were 2.7 ± 0.043, 2.3 ± 0.17, 3.6 ± 0.023, and 0.71 ± 0.024, respectively. K_M values (in μM) were 3.2 ± 0.42, 1.3 ± 0.21, 2.8 ± 0.25, and 1.5 ± 0.29, respectively. All values the means ± 1 SD. LYV, Leu-Tyr-Val.

Figure 6 shows the rate of degradation versus substrate concentration for four substrates that carry variations of the ClpP-distal motif: ¹³LYV-²²DEN₂₅, ¹³LGV-²²DEN₂₅, ¹³GYV-²²DEN₂₅, and ¹³LYG-²²DEN₂₅. K_M values varied between 1.3 ± 0.21 μM to 3.2 ± 0.42 μM and all had relatively large errors, as is common for measured K_M values with the AAA+ proteases (Fig. 6, legend). Therefore, we judge these values as similar, largely within error of one another; furthermore, the values were all between ∼ 1 to 5 μM, the range previously reported for degradation of ssrA-tagged substrates by ClpAP (9, 28, 37, 38). Importantly, the two “best” substrates, with ClpP-distal motifs of ¹³LGV and ¹³LYV (V_MAX of 3.6 ± 0.02 min^-1 ClpA₆^-1 and 2.7 ± 0.04 min^-1 ClpA₆^-1, respectively), had the highest (weakest) K_M values, whereas the poorly degraded ¹³LYG substrate (V_MAX of 0.71 ± 0.024 min^-1 ClpA₆^-1) had a lower (tighter) K_M. An elevated K_M is the expected result for reactions where the catalytic step is faster when comparing similarly recognized substrates. Therefore, these results do not corroborate the hypothesis that poor versus superior recognition explains why some of our test substrates are degraded slowly and others rapidly by ClpAP. In contrast, this Michaelis–Menten analysis supports reaction schemes where differences in the sequence of the slip/grip linker affect the rate of unfolding, the rate-limiting step in degradation of GFP substrates. This difference between substrates is reflected in the observed low, medium, and high V_MAX values. Although we did not complete full curves for all of our substrates, several others were analyzed by determining rates over a range of substrate concentrations and these data were in agreement with the analysis presented here. We conclude that by adding specific sequence elements to an all-Gly linker sequence in the substrate between the C-terminal ssrA degron and the folded region of GFP^min, the ability of ClpA’s D2 ring to grip the substrate is altered, which in turn affects ClpAP's capacity to successfully unfold a well-folded domain in its path.

Discussion

The fundamental mechanical activity of the ATPase component of AAA+ proteases is powering polypeptide translocation. This translocation activity threads unfolded substrates through an axial channel and into a peptidase chamber for degradation. Importantly, the same power stroke responsible for translocation also appears to be the core mechanism that unfolds stable protein domains. Like other enzymes involved in protein homeostasis, AAA+ proteases have many substrates. Moreover, because the entire length of each protein substrate must be translocated for complete degradation, the axial channels of these translocases require a broad tolerance for different polypeptide sequences. Indeed, cryo-EM structures show that axial-pore loops in the channel typically contact the polypeptide backbone of the translocating sequence, explaining the lack of strong sequence specificity (7, 9, 39, 40). However, when AAA+ proteases actively attempt to unfold a stable domain by pulling on an unstructured sequence in the channel, the sequence does matter. Examples of domain-adjacent sequences that appear slippery versus well-gripped have emerged (16, 20, 21, 41, 42). Here, we establish an initial map of substrate features that contribute to grip during unfolding by the double-ring AAA+ ClpA enzyme and degradation by the ClpAP protease. Our results taken together with cryo-EM structures showing substrate polypeptides within the ClpA channel, related studies of the single-ring ClpXP protease, as well as functional analyses of individual pore-1 loops in the AAA+ unfoldase of the 26S proteasome (7, 9, 16, 43, 44, 45) provide new insights and raise fresh questions about the operating principles of these important unfoldase/translocase motors.

Gripped sequences are within ClpA's major unfoldase

Figure 7A summarizes key characteristics of peptide segments that score as well-gripped by ClpA. We find that glycines within the D1 ring of ClpA have little effect on unfolding as illustrated by the modest negative effect of the Gly₁₂ linker, which is predicted to be within the D1 portion of the ClpA channel. In contrast, glycines predicted to be within ClpA's D2 ring substantially inhibit grip (see Gly₁₅, and especially Gly₁₈ and Gly₂₅). These conclusions converge with the current understanding that the D1 ring is largely dedicated to substrate recognition and restarting stalled translocation, whereas the D2 ring is the major unfolding and translocating motor (46, 47, 48, 49). In further support of the idea that the substrate sequence bound within the D1 ring is not important for unfolding, none of the amino acid substitutions introduced within the N-terminal 12-glycine segment of our test sequences have substantial positive or negative effects on the efficiency of GFP^min degradation by ClpAP.

Figure 7.

Summary of key grip characteristics for ClpAP versus ClpXP.A, linker sequences that provide good grip (in green lines and text of amino acids in the motif using the single-letter code) or poor/no grip (in yellow lines and black text). The threshold activity is 14% activity in this format. Linker positions below the green line that have no associated letter at that position are occupied by glycine. The activity values represent degradation rates, as percentages, after normalizing the observed rate to that of ¹³LVY-²²DEN₂₅ (2.0 ± 0.06 min^-1 ClpA₆^-1). B, model and comparison of CIpAP–substrate and ClpXP–substrate complexes involved in cycles of attempted substrate unfolding, as in Fig. 1, step 3 (ClpP is not shown). The linker regions are shown in dark gray, and the pore-1 loops are labeled and colored by subunit (with A being the top subunit in the hexamer spiral). The boxes between ClpA and ClpX list the individual functions of the two enzymes and/or enzyme rings, as well as features shared between ClpA D2 and ClpX. Specific enzyme substrate interactions (including adaptor interactions) are listed in red text. The black lines on the far left and far right represent the approximate distances, in amino acids (aa), between the C terminus of the folded substrate and the closest channel-bound residue associated with “superior grip” for each enzyme. LYV, Leu-Tyr-Val.

Two substrate-contact regions within the ClpA D2 motor make a substrate “grippable”

Replacing three adjacent glycines in the linker with large hydrophobic residues reveals a key grip region predicted to contact ClpA pore loops (and/or other channel features) near the ClpP-distal part of the D2 ring (Fig. 7). However, this grip sequence does not support robust unfolding if the remaining linker consists exclusively of glycines. Rather, a second grip region, predicted to be near the pore loops in the ClpP-proximal portion of the D2 ring also contributes to unfolding activity. Characterization of multiple linker variants suggests that a very good ClpP-distal-grip motif can be formed of large hydrophobic, alanine, polar, and positively charged residues at positions 13, 14, and 15. Although ClpP-distal grip sequences shorter than three residues were not investigated extensively, our analysis reveals that a minimum of two grippable side chains (at positions 13 or 14 and 15) are sufficient, whereas a single non-Gly at any of these locations has no detectable grip function. The second grip element, the ClpP-proximal-grip motif, functions best when located at positions 22, 23, and 24; and (based on the sequences tested to date) most non-Gly residues support good grip in this second element, although three consecutive, large hydrophobics function poorly when compared to either three polar/charged residues or three alanines. In contrast to the distal motifs, a single well-gripped residue in position 22, 23, or 24 is sufficient for contributing very good grip in this ClpP-proximal region, provided it is paired with a functional ClpP-distal motif. If the two minimal motifs characterized here can be combined and retain function, then a 25-glycine slippery sequence filling the entire ClpA channel can be converted to a well-gripped sequence that allows ClpA to efficiently unfold a tightly packed native substrate by just three nonslippery residues appropriately positioned within (or very near) the D2 motor.

We note that glycine, our residue of choice for making slippery linker sequences, may contribute to poor grip in multiple ways. First, it has only a hydrogen side chain and larger side chains containing carbon, oxygen, and nitrogen provide extra functional groups and surfaces for potential enzyme contacts. In addition, glycine is unique in allowing a polypeptide to adopt conformations that are inaccessible to the other 19 natural amino acids. Therefore, the beneficial effects on substrate grip of adding nonglycine residues may be due to a mixture of (i) enhancing direct contacts between the ClpA channel and the substrate side chains and (ii) altering the conformation of the glycine-dominated backbone conformations of the linker that may affect grip indirectly.

Strategies of substrate-specific interactions in different AAA+ proteases

Our design of substrates and experiments were initially guided by an investigation of ClpX’s preferences for substrate-sequence elements that provide grip (16), allowing direct comparison of favored grip sequences for the ClpAP and ClpXP AAA+ proteases. Fig. 7B is a static diagram of plausible models for the “attempting-to-unfold” conformations (Fig. 1, step 3) of ClpAP and ClpXP and how the various enzyme functions are distributed within the two machines. The D2 ring of ClpA and the ClpX ring have nearly 50% sequence similarity, are HCLR-AAA+ -clade members, share identical “GYVG” sequences in their critical pore-1 loops, and overlays of recent enzyme-substrate structures demonstrate that the six ClpA-D2 modules and the six ClpX subunits align well (compare PDB: 7UIV for ClpA-D2 (9) and PDB: 8V9R for ClpX (50)). Based on the structures, we placed the linker segments within the channels of the diagram and highlighted positions (with residue one at the top of the channel) contributing to strong grip for each enzyme. As noted in (16), ClpXP degradation of GFP^min is strikingly slowed by the insertion of a 12-glycine linker between GFP^min and the ssrA tag. Remarkably, degradation is restored by substitutions of larger amino acids at linker positions three, four, or five, with large and branched hydrophobic residues at position four providing the best grip (16). These well-gripped substrate residues for ClpX are positioned to interact with the pore loops (and/or other channel features) in the top third of the ClpX spiral-shaped hexamer (Fig. 7B) (16). By contrast, ClpA needs at least three nonglycine residues in two nonadjacent regions to restore reasonable grip to a poly-glycine linker, and these residues are positioned to interact with the channel/pore loops near both the top (ClpP-distal) and bottom (ClpP-proximal) regions of the D2 ring. Interestingly, the ClpP-distal grip motif of ClpA and the critical grip region for ClpX both lie near the top of the channel within each enzyme's major unfoldase ring, and large/hydrophobic residues are, in both cases, very well-gripped at these positions. This comparison of ClpA with ClpX further underscores that substrate-grip sequences are positioned in the channel of the major unfoldase ring, even when, as with ClpA-D2, that unfoldase is relatively far from the protein to be unfolded. Intuitively, it might seem easier/more efficient, and thus advantageous, to grip the substrate linker directly adjacent to the region of the protein that must experience the “tug” of the mechanical power stroke as is observed with ClpX, but our results establish that having the “grippable” residues in the major unfoldase motor is most important.

Our understanding of the types of side chains that can contribute positively to grip is expanded by adding an analysis of ClpAP to the previous work on ClpXP. A clear example is that negatively charged amino acids and alanines (∼100% normalized activity (Fig. 7A)) function best in the ClpP-proximal motif of ClpA. Positively charged residues are also good (85%), but the large hydrophobics tested were considerably less active (⁴LYV-¹³LYV₂₄ scores as 36%, Fig. 7A). These observations are in marked contrast to the types of amino acids that provide grip to ClpX, where small and negatively charged residues provide no, or only marginal grip function (16). As an aside however, it is premature to rule out that there might also be a second helpful substrate “grip element” that interacts lower in the ClpX ring; only a few experiments were done to probe the contribution to grip by substrate residues 8 to 12 in the ClpX test linkers (16).

Three functions of AAA+ proteases require at least a degree of sequence-specific substrate (or substrate–adaptor complex) interaction (Fig. 7B, red text): (i) degron recognition, (ii) adaptor interaction, and (iii) grip contacts. When considering these requirements, ClpX must be very economical to pack all these contact regions into its single, homo-hexameric ring. In contrast, because ClpA evolved to have two AAA+ motors, each ring can be specialized for different interactions. Therefore, the D2 unfoldase/translocase motor need only optimize substrate grip, which may explain why we found more substrate regions, and a wider range of side chain chemistries that assist in grip for ClpA than are observed for ClpX (16). The grip sequences needed by ClpA also suggest that substrates for double-ring unfoldases with a bottom primary motor, like ClpA and numerous other two-ring enzymes, a longer unstructured region prior to any difficult-to-unfold domain would be needed for successful unfolding.

In thinking about how these grip data may relate to the enzyme's unfolding process, it is intriguing to note that in most of our fastest degraded substrates there are six amino acids between the lowest residue of the ClpP-distal grip motif (residue 15) and the highest of the ClpP-proximal motif (residue 22) and six residues is also the most abundant step size for ClpAP observed by single-molecule optical trapping (18, 49). This observation might indicate a mechanistic relationship between how ClpA-D2 grips its substrate, and the length of polypeptide chain translocated down the channel during an enzymatic power stroke. However, it is important to consider that this study also demonstrates flexibility in the precise positioning of well-gripped residues. For example, one substrate where the lowest residue of the ClpP-distal grip motif is residue 14, combined with a ClpP-proximal motif starting residue 22, generates one of our most rapidly degraded substrates but has a seven-residue spacing. Similarly, the ¹³LYV ClpP-distal motif ending at residue 15 functions well with ²³E as its ClpP-proximal element (Fig. 5C) (and also ²⁴N, although perhaps not as well as with ²²D), again giving motif spacing of greater than six residues. As with the imprecise grip-residue spacing however, the distribution of step sizes for ClpAP is not tightly clustered around six residues, but relatively broad, such that it could easily include five and seven (or eight) residue steps (18).

Another observation related to the discovery that two distinct regions of the channel-bound substrate are required for ClpAP to unfold/maintain grip on a stable substrate is a recent finding with the AAA+ unfoldase ring of the 26S proteasome that reveals interesting potential parallels. This AAA+ unfoldase motor is composed of six related but genetically distinct subunits (Rpt1 through Rpt6) that form a hetero-hexameric shallow spiral, structurally similar to the ClpX and ClpA homo-hexameric subunits arrangements, that is responsible for unfolding and translocating substrates. Studies of pore-1 loop mutants of each subunit reveal that the loops of two of the six subunits are especially important (Rpt6 and Rpt4). Notably, in commonly observed structures, one of these subunits is near the top of the ring, whereas the other is near the bottom (45, 51, 52). Perhaps this observation is the “enzyme side” of the substrate grip regions we have identified for ClpA.

Our results identifying how a very slippery poly-glycine sequence within ClpA's central channel can be converted to “well gripped” by changing a minimum of three amino acids if they are specifically positioned within the enzyme's major motor is an important addition to similar studies with ClpX, as well as to the numerous grip experiments done with the much more complex 26S proteasome. These substrate-focused data provide information about the inner workings of unfoldases that are complementary to the ongoing mutagenesis, enzymatic/biophysical, and structural work with these enzymes. Continued probing of the consequences of distinct sequences bound within these unfolding and destruction machines during distinct reaction stages therefore holds promise for aiding our fundamental understanding of the operating principles of this ubiquitous and biologically important enzyme family.

Experimental procedures

Expression and purification of ClpA and ClpP

ClpA was expressed from pET23b His₇-SUMO-ClpA(M169T)ΔC9 (48). His₇-SUMO is a solubility tag; M169T helps overexpression (53); ΔC9 deletion of the 9 C-terminal amino acids helps prevent ClpAP autodegradation (54). The plasmid was expressed in E. coli strain BL21 DE3 (New England Biolabs) and ClpA protein was purified as described in (47). Briefly, the cell pellet is lysed and centrifuged. The cleared lysate is incubated with ammonium sulfide to 40% total weight per volume. Next, the His₇-SUMO tag is cleaved by Ulp1 protease and the ClpA-containing fractions with the SUMO domain removed are buffer exchanged into the cation exchange buffer overnight. Finally, the protein solution is subjected to ion exchange chromatography, exchanged into HO buffer (50 mM Hepes-KOH (pH 7.5), 300 mM NaCl, 20 mM MgCl₂, 10% glycerol, 1 mM DTT), flash-frozen, and stored at −80 °C. ClpP-His₆ was expressed in E. coli strain JK10 (55), purified as described in (56), and stored in 50 mM Tris–HCl (pH 8), 150 mM KCl, 10% glycerol, 0.5 mM EDTA, and 1 mM DTT.

Plasmid construction and protein purification of GFP-linker-ssrA substrates

The full-length E. coli ssrA tag AANDENYALAA (6, 29) was cloned into the GFP^min-Gly₁₂ sequence (16) and then GFP^min-Gly₁₂-ssrA was cloned into pET23b plasmid–containing His₆-SUMO (gift from K. Zuromski, MIT). Cloning was done using Q5 site-directed mutagenesis (New England Biolabs) and expressed in E. coli strain NEB5a (New England Biolabs). Subsequent GFP^min variants were made using Gibson Assembly according to the manufacturer’s protocol (New England Biolabs). Substrate plasmids were expressed in E. coli strain BL21 (DE3) (New England Biolabs) and purified by nickel-nitrilotriacetic acid, Ulp1cleavage, MonoQ anion exchange, and Superdex75 size-exclusion chromatography. Purified substrates were assessed by SDS-PAGE stained with Coomassie Blue and stored in 25 mM Hepes-KOH (pH 7.5), 150 mM NaCl, 10% glycerol, and 500 mM DTT.

Degradation assays in vitro

Degradation assays were performed at 30 °C in HO buffer, with 5 mM substrate, 0.2 μM ClpA₆, 0.4 μM ClpP₁₄, 5 mM ATP, 32 mM creatine phosphate (Roche), and 0.08 mg/ml creatine kinase (MilliporeSigma). Degradation of the GFP variants was measured by a decrease in fluorescence (λex: 457 nm, λem: 511 nm) on a SpectraMax M5 plate reader (Molecular Devices). A standard curve was constructed for each variant of relative fluorescence units versus varying concentrations of substrate. The slope of this line was the normalization factor to convert relative fluorescence units to substrate concentrations. Degradation rates were calculated by dividing the slope of each GFP variant during the initial linear range of its degradation reaction by the normalization factor to determine the substrate concentration and by the enzyme concentration (0.2 μM ClpA₆). Each substrate was assayed in triplicate and multiple independently conducted biological replicates were performed to confirm results.

Data availability

All data are contained within the article, although in several cases only the averages ± 1 SD are presented. Individual values or raw data traces will be shared upon request by contacting TAB: tabaker@mit.edu.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Karen Fleming