Cargo engagement protects protease adaptors from degradation in a substrate-specific manner

Protein degradation in bacteria is a highly controlled process involving proteolytic adaptors that regulate protein degradation during cell cycle progression or during stress responses. Many adaptors work as scaffolds that selectively bind cargo and tether substrates to their cognate proteases to promote substrate destruction, whereas others primarily activate the target protease. Because adaptors must bind their cognate protease, all adaptors run the risk of being recognized by the protease as substrates themselves, a process that could limit their effectiveness. Here we use purified proteins in a reconstituted system and in vivo studies to show that adaptors of the ClpXP protease are readily degraded but that cargo binding inhibits this degradation. We found that this principle extends across several adaptor systems, including the hierarchical adaptors that drive the Caulobacter bacterial cell cycle and the quality control adaptor SspB. We also found that the ability of a cargo to protect its adaptor is adaptor substrate-specific, as adaptors with artificial degradation tags were not protected even though cargo binding is unaffected. Our work points to an optimization of inherent adaptor degradation and cargo binding that ensures that robust adaptor activity is maintained when high amounts of substrate must be delivered and that adaptors can be eliminated when their tasks have been completed.

Protein degradation is a biological phenomenon conserved across all domains of life. Because it is an irreversible process, degradation must be executed in a highly regulated manner. In eukaryotes, target proteins are tagged by ubiquitin through specific ubiquitin-conjugating enzymes to generate a signature on the target protein that is recognized by the proteasome (1). In contrast, bacteria lack such a general tagging system. Proteolysis in bacteria relies on ATPase associated with diverse cellular activities (AAAϩ) 2 proteases that often recognize degradation tags (or degrons), which are short peptide motifs mostly found at the N or C termini of target proteins (2)(3)(4). In some cases, the target substrate is directly recognized by the AAAϩ protease, whereas, in others, adaptor proteins are employed to selectively destroy the target protein (5,6). Adaptor proteins often bind directly to their targets and physically tether to a protease to drive the degradation of their cargo substrates (7,8). For example, the well characterized adaptor SspB binds and stimulates degradation of trans-translation-derived substrates by the AAAϩ ClpXP protease (9,10).
In the model organism Caulobacter crescentus, DNA replication and cell division are tightly coupled to developmental transitions between swarmer (SW) and stalked (ST) states (11,12). A hierarchy of adaptor-mediated protein degradation by ClpXP plays a critical role in this process, with many regulators selectively degraded at the onset of DNA replication, which is coincident with the SW-ST transition (13)(14)(15). The phosphorylation-dependent CpdR adaptor is dephosphorylated during the SW-ST transition (16). This dephosphorylated CpdR binds and primes the ClpXP protease for substrate recognition (17). This primed ClpXP complex can also recruit the RcdA adaptor to recognize a second class of substrates (15). Finally, the RcdA adaptor can form a complex with the cyclic di-GMP (cdGMP)responsive adaptor PopA to facilitate degradation of a third class of substrates (15,18,19).
These hierarchical adaptors must physically bind to the ClpXP protease, giving ClpXP an opportunity to degrade these adaptors-a problem that extends to all adaptors that bind proteases. Here we show that individual adaptors within this hierarchy are protected from degradation by ClpXP when they are actively delivering their substrates. We dissect the RcdA adaptor to show that protection from degradation stems from adaptor binding to cargo but does not depend on destruction of the cargo by the ClpXP protease. Cargo binding does not shield the adaptor from degradation by ClpXP when a non-native degron is used to replace the adaptor-encoded degron, suggesting that the adaptor is not simply inherently more stable when bound to its target. Rather, chimeric fusion studies support a model where cargo binding facilitates masking of the native adaptor degron. We find that this same mechanism applies to the quality control SspB adaptor, suggesting that cargo-dependent inhibition of adaptor degradation is not a limited feature. The fine-tuning of adaptor lifetime balanced with substrate delivery would provide for rapid responses to a punctuated need for selective protease activity while allowing for a reset of protease complexes after the need has passed.

The RcdA adaptor is degraded by a CpdR-activated ClpXP protease
We previously identified RcdA as an adaptor that delivers multiple cell cycle regulators exclusively to a CpdR-primed ClpXP (15). During additional characterization of RcdA-dependent degradation, we found that RcdA itself was degraded by the ClpXP protease in the presence of CpdR (Fig. 1, A and B). Consistent with this observation, cells lacking CpdR show higher steady-state levels of RcdA than wild-type cells (Fig. 1C). Translation shutoff experiments suggested that these increased levels are due to loss of RcdA degradation in ⌬cpdR strains (Fig.  1, D and E). Together, these data show that the RcdA adaptor is degraded by ClpXP in a CpdR-dependent manner both in vivo and in vitro. We next sought to determine which sequence elements were important for RcdA degradation.
Previous work has shown that the C terminus of RcdA binds to a CpdR-primed ClpXP to deliver substrates for degradation (15). As expected, RcdA⌬C was not degraded, even in the presence of CpdR, suggesting that the C terminus of RcdA is necessary for degradation (Fig. 2, B and C). We next used the fusion protein RcdA⌬CϳXB, where the substrate-binding domain of RcdA is appended with the ClpX-tethering motif of SspB ( Fig.  2A) (15). This construct can deliver substrates to ClpXP even in the absence of CpdR (supplemental Fig. S1A) (15) but was not robustly degraded (Fig. 2, D and E). Therefore, simply tethering the cargo-binding domain of RcdA to ClpXP is insufficient for RcdA adaptor degradation.
Our results suggest that the RcdA degron is encoded by the same C-terminal residues needed for binding the CpdR-primed ClpX. RcdA ends in GG, a dipeptide sequence shown previously to be able to be recognized by ClpX in appropriately presented substrates (20). We mutated these residues to make an RcdADD variant that was fully capable of delivering cargo substrates in a CpdR-dependent manner ( Fig. 2H and supplemental Fig. S1B) but was now substantially resistant to degradation compared with wild-type RcdA (Fig. 2, F and G). This allele supported CtrA degradation in vivo similar to wild-type RcdA (supplemental Fig. S1D) but was markedly stabilized (supplemental Fig. S1C), consistent with the in vitro results.

RcdA adaptor degradation is suppressed upon cargo binding
How is the RcdA adaptor able to effectively deliver its substrates to ClpXP given its own rather rapid degradation by the same protease? To address this question, we tested whether RcdA degradation was affected during delivery of its substrates. We found that addition of RcdA-dependent ClpXP substrates (TacA and CC2323) partially suppressed RcdA degradation (Fig. 3, A and B), suggesting that RcdA degradation is inhibited while it is actively operating as an adaptor. We considered two models for how this inhibition could come about: this suppression could be due to protection of the adaptor from the protease upon cargo binding, or suppression could arise from competition between the delivered substrates and RcdA degradation for the protease. To distinguish between these possibilities, we used substrate variants (TacADD and CC2323DD) that bind RcdA but are not recognized by the protease (supplemental Fig.  S2, A and B) (15). Interestingly, addition of these variants resulted in even stronger suppression of RcdA degradation (Fig. B, bands corresponding to RcdA and ClpP were quantified using ImageJ, and normalized intensities of RcdA over ClpP were plotted. Data represent mean Ϯ S.D. of two independent experiments. C, steady-state levels of RcdA in wild-type Caulobacter cells or cells lacking CpdR. Lysates from equal number of cells were loaded onto the SDS-PAGE gels. RcdA protein was detected by Western blotting using an anti-RcdA antibody. ClpP is shown as a loading control. Normalized mean levels are shown from two biological replicates Ϯ S.D. D, in vivo degradation of RcdA in wild-type or ⌬cpdR cells. After adding chloramphenicol to inhibit protein synthesis, lysates from equal volumes of cells were loaded onto SDS-PAGE gels. RcdA stability was monitored by Western blotting using an anti-RcdA antibody. ClpP was used as a loading control. E, bands corresponding to RcdA and ClpP were quantified, and normalized intensities of RcdA over ClpP were plotted. Data represent mean Ϯ S.D. of two independent experiments. 3, C and D). These results suggest that binding of cargo itself protects RcdA from protease recognition.

Cargo binding protects adaptor degradation
The cdGMP binding protein PopA is an upstream adaptor of RcdA during the hierarchical delivery of cell cycle substrates (15). Our above work suggests that cargo binding to RcdA alone is sufficient for inhibition of degradation. Because PopA directly interacts with RcdA (18,19), we speculated that PopA would also protect RcdA from degradation. Consistent with this hypothesis, we found that PopA addition inhibited RcdA degradation (Fig. 3, E and F). Addition of cdGMP did not further affect this suppression (Fig. 3, E and F), validating prior observations that PopA binds RcdA regardless of cdGMP (18,19). Similar to its stabilizing effect in vitro, overexpression of PopA or a cdGMP-insensitive variant (PopA-R357G) (18) reduces RcdA degradation in vivo (Fig. 3, G and H). Importantly, PopA was not degraded either in vivo or in vitro ( Fig. 3E and supplemental Fig. S2, D-F), further supporting our working model that cargo binding alone is sufficient for RcdA stabi-lization even when that cargo is not delivered to the ClpXP protease.

Cargo-mediated suppression of adaptor degradation is not a limited phenomenon
To determine whether suppression of adaptor degradation upon binding of cargo is a general phenomenon, we monitored the stability of other known adaptors from Caulobacter. The adaptor SspB facilitates degradation of ssrA-tagged substrates by directly tethering them to the ClpXP protease (9, 10). The Caulobacter SspB was shown previously to be a ClpXP substrate in vitro (21), and we found that SspB degradation is suppressed in the presence of GFP-ssrA (Fig. 4, A and B). Similar to our work with RcdA, this inhibition does not depend on substrate degradation, as addition of the nondegraded GFP-ssrADD variant (which binds SspB as well as wild-type (23)) also stabilizes SspB degradation (Fig. 4, A  and B).

Cargo binding protects adaptor degradation
CpdR is an adaptor that binds ClpXP to prime it for cell cycle-dependent activity (17). CpdR is degraded by the ClpXP protease in vivo (22), and we found it to be degraded on its own by ClpXP in vitro (Fig. 4, C and D) (17). When we monitored CpdR while it was delivering its substrate PdeA to ClpXP, we found that CpdR degradation was reduced (Fig. 4, C and D).
Addition of the nondegradable PdeADD also suppresses CpdR degradation (Fig. 4, C and D). CpdR degradation was not strongly affected by RcdA alone (supplemental Fig. S3, A and B) or by RcdA-cargo complexes (supplemental Fig. S3, C and D), which we attribute to a relatively weak interaction between RcdA alone and CpdR-ClpX. In sum, we find that suppression of adaptor degradation upon sufficiently robust cargo binding is a conserved phenomenon in many adaptors.

Inhibition of adaptor degradation by cargo binding is not due to global stabilization
We hypothesized that the cargo binding could affect adaptor degradation in one of two nonexclusive mechanisms. The first possibility is that cargo binding could cause adaptors to be generally more resistant to any forced unfolding. The second is that the adaptor degron itself could be masked upon cargo binding (Fig. 5A). We rationalized that if the first model was correct, then binding of cargo would stabilize the adaptor regardless of how it was recognized by ClpX. To test this, we generated a chimeric fusion protein, RcdA⌬CϳCtrA 15 , where we used the CtrA degron to drive degradation of an RcdA construct lacking its normal degron/tethering motif (RcdA⌬CϳCtrA 15 ) (Fig.  5B). This chimera was slowly degraded by ClpXP alone (Fig. 5, C and D), but we did not observe any changes in degradation with the CC2323 cargo (Fig. 5, C and D). Importantly, RcdA⌬CϳCtrA 15 still binds CC2323 strongly (supplemental Fig. S4A), suggesting that cargo binding alone is insufficient to generally protect an adaptor from degradation. To address whether this result was a consequence of the slower degradation of RcdA⌬CϳCtrA 15 , we also used the well characterized ssrA degron, which is strongly recognized by ClpXP. RcdA⌬CϳssrA is degraded rapidly by ClpXP alone, and, similar to our other results, degradation is not affected by the presence of the CC2323 cargo (Fig. 5, E and F). Taken together, these results disfavor a model where cargo binding alone is sufficient to generally protect adaptors from ClpX-mediated unfolding.
Our working hypothesis is that protection of RcdA degradation upon cargo binding arises from masking of its inherent degron rather than general stabilization (Fig. 5A). To explore this model, we fused the CtrA degron to the full-length RcdA protein (Fig. 5B). This RcdAϳCtrA 15 construct was degraded by ClpXP without CpdR, but cargo binding did not suppress degradation, although this construct was fully competent for cargo binding (Fig. 5, G and H, and supplemental Fig. S5A). Addition of CpdR increased degradation of this chimera (Fig. 5,  G and H), presumably because of recognition of the combined degradation/tethering motif of RcdA by the CpdR-primed ClpX (15). Finally, addition of the non-degradable RcdA cargo CC2323DD substantially inhibited degradation of the fusion protein RcdAϳCtrA 15 (Fig. 5, G and H). Our conclusion is that the protective effect of cargo binding is specific for the recognition of the native degron of the adaptor that, in the case of RcdA, requires CpdR for its recognition. We next tested whether this paradigm could be applied to other cases of adaptor degradation.
The Caulobacter SspB protein contains a ClpX-binding motif at its C terminus (15,23) and an N-terminal motif required for its degradation (supplemental Fig. S4B) (17,24). Because we found that suppression of RcdA degradation upon cargo binding requires recognition of its native degron, we hypothesized that a similar mechanism might exist for SspB degradation. To test our hypothesis, we appended the CtrA degron to the C terminus of a minimized SspB variant that lacks its native degron but still delivers cargo ( Fig. 6A and supplemental Fig. S4C). This fusion protein (⌬N9SspBϳCtrA 15 ) is degraded robustly by ClpXP, but addition of cargo does not protect it from degradation (Fig. 6, B and C). Similar to our above results, fusing the CtrA degron to the full-length SspB resulted in degradation by ClpXP and partial suppression of this degradation in the presence of cargo (Fig. 6, D and E). We note that, in cases where both the artificial and natural degrons are present, binding of cargo appears to only limit the protease recognition of the natural degron, leaving the artificial degron free to be engaged. Finally, it is possible that binding of a protein cargo results in steric clashes that prevent recognition of the adaptor degron by the protease. To address this, we asked whether an adaptor binding peptide is sufficient to protect adaptors from degradation. Specifically, we used an ssrA-derived peptide (AANDN-NYA) that retains the SspB binding site but lacks the ClpXP recognition determinant (23). We found that addition of this peptide inhibits SspB degradation by ClpXP (Fig. 6, F and G). Thus, our data support a model where cargo binding leads to masking of the native degron of adaptors, likely through conformational changes.

Discussion
Bacterial proteolytic adaptors facilitate degradation of regulatory proteins or poor-quality proteins. These auxiliary factors work in a variety of ways, such as directly tethering substrates to their cognate protease or priming either the protease or substrate (9,17,25). Because adaptors must be in close physical proximity to the protease, understanding how adaptors avoid being destroyed impacts our general understanding of adaptormediated proteolysis. Here we highlight cases from the model bacterium C. crescentus to show how ClpXP adaptors are selec- tively protected from degradation upon cargo binding. Specifically, we show that the RcdA, CpdR, and SspB adaptors are degraded by ClpXP but are shielded from degradation upon cargo binding. Using chimeric adaptors, we find that the protective effect we observe is most consistent with a model where binding of cargo elicits a masking of the adaptor degron (Fig. 7). Our work shows that adaptor degradation is modulated by cargo binding and that this regulation depends on the specific adaptor-substrate pairing.
Adaptor proteolysis has been observed in other protease systems. For example, the MecA adaptor activates the ClpC unfoldase to target the ComK protein for degradation by the ClpCP protease in Bacillus subtilis and is itself degraded (25,26). The cyanobacterial adaptor NblA delivers phycobilisomes to the Clp protease for degradation, and purified NblA is degraded in vitro (27,28). By contrast, some adaptors remain stable regardless of delivery of their substrates. For example, the RssB adaptor delivers RpoS to ClpXP in Escherichia coli, but RssB itself is not degraded (29). The ClpS adaptor delivers N-end rule sub-strates to ClpAP and is itself not degraded, although ClpS contains an N-terminal determinant sufficient for ClpAP recognition (30). In this case, the current model is that the core of ClpS contains a tightly folded domain that prevents its degradation during delivery even when ClpA engages this adaptor (30).
This work yields molecular insights into adaptor degradation and its control. A clear next step of study is to understand the physiological relevance of adaptor degradation. When CpdR was first identified, the degradation of CpdR was speculated to serve as a mechanism to release the ClpXP protease from its subcellular polar location (22). Similarly, MecA degradation was suggested to be important for the dynamic assembly and disassembly of the protease complex ClpCP (31). By contrast, the stability of the RssB adaptor is thought to enable its recycling for repeated rounds of RpoS delivery (28,32). Collectively, our work shows that adaptors are more resistant to proteolysis when loaded with cargo. This suggests that, regardless of the cellular outcome of adaptor degradation (changes in protease location or protease assembly), adaptor degradation is highly balanced with cargo occupancy, an ideal mechanism for ensuring protease activity when activity is truly needed.

Bacterial strains and culture conditions
The Caulobacter and E. coli strains used in the study are tabulated in supplemental Table S1. E. coli strains were grown in lysogeny broth liquid medium or lysogeny broth agar plates at 37°C with the appropriate antibiotic (100 g/ml ampicillin, 50 g/ml kanamycin, or 50 g/ml spectinomycin). Caulobacter strains were grown in peptone yeast extract (PYE) liquid medium or PYE agar plates at 30°C with the appropriate antibiotic (25 g/ml spectinomycin or 5 g/ml kanamycin). Caulobacter strains grown in PYE liquid medium were supplemented with 0.2% xylose to induce gene expression wherever required.

Molecular cloning and generation of chimeric constructs
PopA and PopA-R357G were PCR-amplified and cloned into pENTR plasmids. The constructs were then moved into xylose-inducible expression plasmids, which appends an M2 epitope tag on the N terminus of the protein using Gatewaybased cloning (33). RcdA and RcdADD were PCR-amplified and cloned into pET23SUMO and pRXMCS2 vectors by the Gibson assembly method (34). RcdA⌬C, RcdA⌬CϳXB, and CC2323DD were PCR-amplified with the appropriate primers and then cloned into the pET23SUMO expression plasmid by the Gibson assembly method. Chimeric variants of RcdA and SspB were generated by round-the-horn site-directed mutagenesis. RcdA⌬CϳCtrA 15 and RcdA⌬CϳssrA constructs were generated in the pET23SUMO-RcdA⌬C vector by appending the 15 C-terminal residues of CtrA or the 14 residues that constitute the ssrA tag, respectively. The RcdAϳC-trA 15 fusion construct was generated in the pET23SUMO-RcdA vector by designing appropriate primers to append the last 15 residues of CtrA onto the C terminus of full-length RcdA. SspBϳCtrA 15 and ⌬N9SspBϳCtrA 15 were created by appending the 15 C-terminal residues of CtrA to a SspB variant that either contained or lacked the N-terminal nine residues, respectively. All chimeric constructs and mutant proteins were confirmed by sequencing. Oligonucleotide sequences are available upon request.

Protein expression and purification
All proteins were expressed in BL21(DE3)pLysS E. coli strains. The cells were grown to an optical density (A 600 ) between 0.4 and 0.8, induced with 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside for 3-5 h at 37°C with shaking and then centrifuged at 5000 rpm for 10 min. The pellets were resuspended in lysis buffer containing 50 mM Tris (pH 8.0), 300 mM NaCl, 10 mM imidazole, 10% glycerol, and 5 mM ␤-mercaptoethanol and frozen at Ϫ80°C. The cell suspension containing 1 mM PMSF was lysed using a Microfluidizer system (Microfluidics, Newton, MA). The clarified lysate was loaded onto a preequilibrated Ni-NTA column. The SUMO tag was cleaved by Ulp1-his protease, whereas the hexahistidine tag was cleaved by thrombin protease. The tags were then further removed by Ni-NTA affinity chromatography. Proteins were further purified by size exclusion chromatography using a Sephacryl 200 16/60 column. The protease components ClpX and ClpP were purified as described previously (21).

In vivo and in vitro protein degradation assays
The synchronization experiment was performed as described previously (15). Protein stability was measured by blocking the synthesis of proteins. Different strains of Caulobacter were grown to A 600 ϳ0.3 in PYE medium with the appropriate antibiotics. Protein expression was induced with xylose for the times indicated in the figure legends, and then translation was blocked by addition of 30 g/ml chloramphenicol. Equal volumes of samples were collected at different time points, as indicated in figures for Western blot analysis. Steadystate protein levels were measured by growing Caulobacter wild-type and ⌬cpdR strains to exponential phase. Equal numbers of cells were then used for Western blot analysis. In vitro degradation assays were carried out at 30°C and monitored by the loss of the protein of interest over time using SDS-PAGE gel analysis. Gels were scanned by an Odyssey CLx imaging system (LI-COR Biosciences) and quantified using ImageJ software (National Institutes of Health) to measure the change in band intensity over time. The concentrations used in the final reaction volume otherwise mentioned separately below were as follows: 2 M each TacA, TacADD, CC2323, CC2323DD, RcdA, RcdA⌬C, RcdA⌬CϳXB, RcdA⌬CϳCtrA 15 Fig. S4E. The reaction was initiated by addition of ATP regeneration mixture, which consisted of 4 mM ATP, 5 mM creatine phosphate, and 75 g/ml creatine kinase. The buffer used for the reaction was H buffer, which consisted of 20 mM HEPES-KOH (pH 7.5), 100 mM KCl, 10 mM MgCl 2, 10% glycerol, and 5 mM ␤-mercaptoethanol.

Immunoblot analysis
Cultures samples withdrawn at the different time points indicated in the figures were centrifuged at 15,000 rpm for 2 min. After removal of the supernatant, the pellets were resuspended in 2ϫ SDS-PAGE sample buffer containing 40 mM DTT. The samples were boiled at 95°C for 10 min and then centrifuged to remove cellular debris. After centrifugation, the extracts were resolved on 10 -15% SDS-PAGE gels. Proteins from the gel were then transferred to a PVDF membrane. After blocking the membrane with 3% milk-TBST (Tris-buffered saline (50 mM Tris, 150 mM NaCl, pH 7.5) ϩ 0.1 % Tween-20) buffer, the membranes were probed with polyclonal rabbit anti-CtrA (1:5000 dilution), anti-McpA (1:10,000 dilution), anti-RcdA (1:5000 dilution), anti-ClpP (1:5000 dilution), anti-SspB (1:20,000 dilution), or monoclonal mouse anti-FLAG M2 (1:5000 dilution, Sigma) antibodies overnight at 4°C. After washing off the excess primary antibody, the membrane was probed with goat anti-rabbit or goat anti-mouse (Millipore) secondary antibodies conjugated to HRP enzyme. Proteins on the membrane were then visualized by the luminescence from the HRP substrate using a chemiluminescence detection system (Syngene).

In vitro pulldown assays
Ni-NTA affinity resin was used to pull down the protein that binds to hexahistidine-tagged bait protein. The resin was preequilibrated with H buffer supplemented with 20 mM imidazole. His 6 -CC2323 (5 M) was incubated with RcdA (10 M), RcdA⌬CϳCtrA 15 (10 M), RcdA⌬CϳssrA (10 M), or RcdAϳCtrA 15 (10 M), and his 6 -RcdA (5 M) was incubated with CC2323DD (10 M) either alone or together in a 250-l final volume of H buffer containing 20 mM imidazole and 50 l of pre-equilibrated Ni-NTA resin at 4°C for 1 h. After 1 h of incubation, the resin was spun down at 700 ϫ g for 2 min to collect the flow-through. The resin was washed twice with H buffer supplemented with 20 mM imidazole at 350 ϫ g for 1 min. The bound complex was eluted with H buffer containing 200 mM imidazole by spinning at 700 ϫ g for 5 min. The fractions collected were then analyzed by SDS-PAGE gels.
Author contributions-K. K. J. and P. C. were involved in conceptualization, design, data analysis, and interpretations of the results. K. K. J. and M. S. performed the experiments. K. K. J. and P. C. wrote the manuscript.