Serine phosphorylation facilitates protein degradation by the human mitochondrial ClpXP protease

Significance

ClpXP is a mitochondrial matrix protease that degrades damaged or misfolded proteins and helps maintain mitochondrial health and cellular function. Much like how ubiquitin tags proteins for degradation in the cytosolic ubiquitin–proteasome system, our study reveals that phosphorylation of serine residues serves as a key signal for recognition and degradation by ClpXP. This report is the first to identify a substrate degron for human ClpXP and its mode of interaction with ClpX. Uncovering this mechanism provides insights into mitochondrial protein quality control and highlights a strategy to develop ClpXP inhibitors for cancers such as acute myeloid leukemia, that depend on this protease survival.

Abstract

ClpXP is a two-component mitochondrial matrix protease. The caseinolytic mitochondrial matrix peptidase chaperone subunit X (ClpX) recognizes and translocates protein substrates into the degradation chamber of the caseinolytic protease P (ClpP) for proteolysis. ClpXP degrades damaged respiratory chain proteins and is necessary for cancer cell survival. Despite the critical role of ClpXP in mitochondrial protein quality control, the specific degrons, or modifications that tag substrate proteins for degradation by human ClpXP, are still unknown. We demonstrated that phosphorylated serine (pSer) targets substrates to ClpX and facilitates their degradation by ClpXP in biochemical assays. In contrast, ClpP hyperactivated by the small-molecule drug ONC201 lost the preference for phosphorylated substrates. Hydrogen deuterium exchange mass spectrometry combined with biochemical assays showed that pSer binds the RKL loop of ClpX. ClpX variants with substitutions in the RKL loop failed to recognize phosphorylated substrates. In intact cells, ClpXP also preferentially degraded substrates with pSer. Moreover, ClpX substrates with the pSer were selectively found in aggregated mitochondrial proteins. Our work uncovers a mechanism for substrate recognition by ClpXP, with implications for targeting acute myeloid leukemia and other disorders involving ClpXP dysfunction.

The ubiquitin proteasome system (UPS) is the primary regulator of cytosolic proteostasis. It controls most nonlysosomal protein degradation and tags proteins for proteolysis via ubiquitination (1). Mitochondria lack a UPS. Instead, several analogous proteases, including ClpXP protease, maintain mitochondrial protein quality (2). The ClpXP system is an evolutionarily conserved member of the ATPases associated with diverse cellular activities (AAA+) protein degradation family (3). ClpXP consists of the hexameric ClpX regulatory particle that caps each end of the tetradecameric ClpP protease (4). ClpX captures, unfolds, and translocates substrates into the proteolytic chamber of ClpP for degradation.

ClpXP is overexpressed in multiple malignancies (5–7), including acute myeloid leukemia (AML) where its abundance increased in 45% of patient samples compared to the 95th percentile expression in CD34+ hematopoietic progenitor cells from normal individuals (5). Inhibiting ClpP results in the accumulation of misfolded or damaged respiratory chain proteins, leading to impaired oxidative phosphorylation and mitochondrial metabolism, which selectively affects leukemic cells with elevated ClpP expression (5). In contrast, hyperactivation of ClpP by imipridones, such as ONC201, leads to uncontrolled degradation of respiratory chain proteins (8). ONC201 has demonstrated preclinical efficacy in both hematologic malignancies and solid tumors and is showing promising results in clinical trials (9). Taken together, both inhibition and hyperactivation of ClpP impair oxidative phosphorylation, but through distinct mechanisms (5, 8). Thus, targeting ClpXP could be a unique therapeutic strategy for AML and other malignancies.

Proteomic studies identified putative ClpXP substrates involved in key mitochondrial functions like the Krebs cycle, oxidative phosphorylation, and amino acid metabolism (5, 10, 11), but the specific degrons that target these substrates to human ClpXP are unknown. In many bacteria, the SsrA amino acid tag (AANDENYALAA-COO^– in Escherichia coli) is cotranslationally added to the C terminus of incompletely synthesized polypeptides and is recognized by the homologous bacterial ClpXP enzyme (12, 13). However, mitochondria do not have an equivalent SsrA tagging system. In some bacterial species, another AAA+ unfoldase, ClpC, recognizes and degrades proteins with phosphorylated arginine. Phospho-arginine (pArg) residues bind a pair of glutamic acid residues in the N-terminal domain (NTD) of ClpC (14). Notably, the NTD of ClpX has neither sequence nor structural similarity to that of ClpC.

Here, we explore the posttranslational modifications that tag substrate proteins for degradation by human ClpXP in cell-free assays and in intact cells. We found that phospho-Ser (pSer) binds the RKL loop of ClpX which in turn promotes mitochondrial ClpXP-mediated protein degradation, thus highlighting a mechanism of substrate recognition that is distinct from bacterial homologs.

Results

ClpXP Preferentially Recognizes pSer Amino Acids and Peptides.

ClpXP plays a crucial role in preserving mitochondrial function by selectively degrading damaged and misfolded respiratory chain proteins (5, 6). However, the degrons that tag substrates for degradation by ClpXP are unknown. Despite the differences between human and bacterial systems, most notably the absence of ClpC in human mitochondria, we postulated that human mitochondrial ClpXP could preferentially recognize phosphorylated substrates.

α-casein has been widely used as a model substrate in proteolysis research due to its susceptibility to degradation by various proteases. However, ClpP alone is only active against short peptides and requires the ClpX unfoldase to degrade full-length proteins such as α-casein (SI Appendix, Fig. S1A) (15). α-casein has up to 13 phosphorylated sites on serine (Ser), threonine (Thr), tyrosine (Tyr), and aspartic acid (Asp) residues (16, 17). To determine whether α-casein phosphorylation influences its degradation by ClpXP, we employed a cell-free protease assay using recombinant human ClpXP with phosphorylated or dephosphorylated α-casein as substrate (SI Appendix, Fig. S1B). In cell-free assays, ClpXP degraded α-casein more effectively than dephosphorylated α-casein (Fig. 1A and SI Appendix, Fig. S1C). Of note, we observed no changes in levels of ClpX up to 6 h after incubation of recombinant ClpXP with α-casein (SI Appendix, Fig. S1 D and E).

Fig. 1.

ClpXP preferentially degrades pSer-containing substrates. (A) α-casein or dephosphorylated α-casein (5 µM) was incubated with recombinant human ClpXP (1 µM, 6 h), human ClpX–E. coli ClpP (6 µM, 48 h), E. coli ClpAP (1 µM, 0.5 h), or ONC201-human ClpP (1 µM, 0.5 h). Proteins were analyzed by sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE), Coomassie blue staining, and semiquantified by densitometry. Data represent the mean ± SD (n ≥ 3 biological replicates) residual full-length α-casein, expressed as a percent of control samples. Corresponding blots are shown in SI Appendix, Figs. S1C, and S2 A–C. P-values were derived from unpaired t tests. (B) Recombinant ClpXP (1.5 µM) protein was incubated with FITC-α-casein (5 µM) in the presence of increasing concentrations of free or phosphorylated amino acids (Ser, Thr, Tyr, or Arg), peptides or free phosphate. The rates of FITC-casein degradation were measured for 1 h at 37 °C. Data represent the mean ± SD (n = 3 technical replicates) rate of FITC-α-casein degradation expressed as a percent of control samples. A representative experiment from at least three biological replicates is shown. (C) Recombinant ClpXP (1.5 µM) protein was incubated with FITC-α-casein (5 µM) in the presence of increasing concentrations of acetylated serine (Ac-Ser) or monoubiquitin (Ub). The rates of FITC-α-casein degradation were measured for 1 h at 37 °C. Data represent the mean ± SD (n = 3 technical replicates) rate of FITC-casein degradation expressed as a percent of control samples. A representative experiment from three biological replicates is shown. (D) Recombinant LonP1 (0.5 µM) protein was incubated with FITC-α-casein (0.8 µM) in the presence of increasing concentrations of pSer. Degradation rates of FITC-α-casein were recorded for 1 h at 37 °C. Data represent the mean ± SD (n = 3 technical replicates) rate of FITC-casein degradation expressed as a percent of control samples. A representative experiment from three biological replicates is shown.

To investigate the origin of the preference for phosphorylated α-casein within the ClpXP complex, we took advantage of the ability of ClpP and ClpX to form heterologous complexes (4, 18). Human ClpX mixed with E. coli ClpP still favored phosphorylated α-casein (Fig. 1A and SI Appendix, Fig. S2A). However, when human ClpX was replaced with E. coli ClpA, an unfoldase common in Gram-negative bacteria, this preference was abolished and the E. coli ClpAP complex degraded phosphorylated and dephosphorylated α-casein indiscriminately (Fig. 1A and SI Appendix, Fig. S2B). ONC201 is an imipridone class anticancer drug that binds ClpP, displaces the ClpX unfoldase, and hyperactivates the protease, thereby leading to uncontrolled, but selective, mitochondrial protein degradation (8, 19). In contrast to ClpXP, ONC201-activated ClpP degraded both phosphorylated and dephosphorylated α-casein (Fig. 1A and SI Appendix, Fig. S2C). These data demonstrate that ClpX, but not ClpP, imparts specificity for phosphorylated α-casein and highlights the distinct substrate preferences that differentiate human ClpX from its bacterial homolog.

To elucidate the specific phosphorylated amino acids recognized by ClpXP, we performed the cell-free protease assay described above with fluorescein isothiocyanate-labeled α-casein (FITC-casein) in the presence of increasing concentrations of phosphorylated (p) or nonphosphorylated Ser, Thr, Tyr, and Arg amino acids. ClpXP degraded α-casein and FITC-casein similarly (SI Appendix, Fig. S3A). pSer and the structurally related pThr inhibited ClpXP-mediated degradation of FITC-casein, while pTyr, pArg, nonphosphorylated amino acids, and free inorganic phosphate had no effect (Fig. 1B). pSer also inhibited the degradation of α-casein by the heterologous human ClpX–E. coli ClpP complex (SI Appendix, Fig. S3B). We extended our analysis to phosphorylated peptides [Ala–pSer–Ala (ApSA), Ala–pThr–Ala, Ala–pTyr–Ala, and Arg–Arg–Ala–pSer–Val–Ala] and found that only peptides containing pSer and/or pThr inhibited ClpXP-mediated degradation of FITC-α-casein, while nonphosphorylated Ser and Thr did not (Fig. 1B). Of note, pSer also inhibited the degradation of dephosphorylated α-casein (SI Appendix, Fig. S3C). These results coupled with the mutagenesis studies below are consistent with pSer blocking substrate translocation through ClpX.

ClpXP also showed selectivity for other phosphorylated substrates. Human tau is a microtubule-associated protein that is natively unstructured and prone to aggregation (20, 21). Tau tubulin kinase 1 (TTBK1) phosphorylates tau protein on multiple Ser sites (22). Recombinant wild-type human tau protein purified from E. coli is not phosphorylated (23) and was not degraded by ClpXP (SI Appendix, Fig. S4A). In contrast, recombinant tau that was phosphorylated by TTBK1 (p-tau) was degraded by ClpXP and its degradation was inhibited by the addition of pSer to the reaction (SI Appendix, Fig. S4 A and B).

ClpP has peptidase activity independent of ClpX, and ClpX has ATPase activity independent of ClpP. Therefore, we assessed the effects of pSer and pThr on ClpP peptidase activity and ClpX ATPase activity. pSer or pThr did not affect the peptidase activity of ClpP against Ac–Trp–Leu–Ala–AMC as substrate (SI Appendix, Fig. S5A). pSer and pThr also did not inhibit ClpX-mediated hydrolysis of ATP (SI Appendix, Fig. S5B). Thus, pSer and pThr do not target the peptidase or ATPase activity of ClpP and ClpX, respectively.

In contrast to the effects seen with phosphorylation, other common posttranslational modifications such as monoubiquitination (Ub) and Ser acetylation did not inhibit ClpXP activity (Fig. 1C). Moreover, pSer did not inhibit the enzymatic activity of another AAA+ mitochondrial matrix protease, LonP1 (Fig. 1D).

To complement our activity-based assays, we used differential scanning fluorimetry to test whether pSer and pThr could bind to ClpX. pSer and pThr, but not Ser, Thr, pArg, or pTyr stabilized ClpX, consistent with pSer and pThr binding ClpX (SI Appendix, Fig. S6). Taken together, these results establish that pSer and pThr can directly bind ClpX, highlighting a mechanism of substrate recognition that is distinct from bacterial homologs.

pSer Binding Affects the Conformational Dynamics of ClpX.

To further characterize the interaction between ClpX and peptides containing pSer, we attempted to solve the structure of apo and peptide-bound ClpX using electron cryomicroscopy (cryo-EM). We were able to visualize single hexameric particles in the presence of ATPγS·Mg (SI Appendix, Fig. S7 A and B and Table S1). The severe preferential orientation of ClpX particles precluded our ability to reconstruct a high-resolution map to unambiguously locate the pSer binding sites (SI Appendix, Fig. S7C). Nonetheless, the addition of ApSA peptide did not affect the overall structure of ClpX (SI Appendix, Fig. S7 A and B) and our maps could be fit using a homology model of ClpX (SI Appendix, Fig. S7D). Thus, ClpX maintains its known hexameric form in the presence of phosphorylated peptides.

Given the preferred orientation issue with cryo-EM analyses of ClpX, we turned to hydrogen deuterium exchange mass spectrometry (HDX-MS) to examine the influence of the binding of pSer-containing peptides on the structure and conformational dynamics of ClpX. HDX-MS is a powerful approach for investigating weak binding interactions and structurally heterogeneous protein systems (24–26). Backbone amides involved in binding interactions exhibit slow deuteration influenced by conformational fluctuations, whereas exposed or dynamically fluctuating sites undergo rapid hydrogen for deuterium exchange (27). HDX-MS is most informative when comparing the deuterium uptake of the unbound and ligand-bound states of proteins (28). Ligand interactions typically result in decreased deuterium uptake in residues involved in binding, and any allosteric changes might present as increased or decreased uptake (29). To assess the sensitivity of HDX-MS for probing ClpX–ligand interactions, we compared the deuterium uptake profile of apo ClpX with that of ClpX incubated with ATPγS·Mg. Our HDX workflow resulted in 151 peptides with a sequence coverage of 94% and a peptide redundancy level of 3.3 (SI Appendix, Fig. S8 and Table S2). Relative to the unbound protein, ATPγS·Mg binding considerably reduced deuteration in structural components essential for nucleotide binding, including the Walker-A and -B motifs, the Arg finger, and the sensor-II and box-II residues (Fig. 2 A and B). These results underscore the ability of HDX-MS to identify the structural alterations of ClpX due to small-molecule binding.

Fig. 2.

pSer binding affects the conformational dynamics of ClpX. (A) Hydrogen/deuterium exchange mass spectrometry heat map of ClpX in the presence of ATPγS·Mg and ApSA peptides. HDX was initiated by incubating apo ClpX equilibrated with various small molecules into a D₂O-based buffer. Deuterium uptakes were measured for peptides covering the ClpX sequence (arranged horizontally) at increasing D₂O exposure time (arranged vertically) and differential deuterium uptakes were calculated for ATPγS·Mg–bound ClpX relative to apo ClpX (Top heat map), and ATPγS·Mg:ApSA-bound ClpX relative to ATPγS·Mg-bound ClpX (Bottom heat map). A schematic of key structural elements of ClpX is displayed on top of the sequence. (B and C) The differential deuterium uptake profiles of Fig. 2A at the D₂O exposure time of 0.16 min was mapped onto the UniProt homology model of ClpX. Top, Bottom, and side views were shown. Same color scale as Fig. 2A.

We next evaluated the binding of ApSA to ClpX. The ATPγS·Mg–bound condition served as a baseline for comparing the Ala–Ser–Ala (ASA)- and ApSA-bound states of ClpX. Binding of ApSA resulted in notable reductions in deuterium uptake across both short and extended D₂O exposure times (Fig. 2A). The millimolar IC₅₀ values for ApSA inhibiting ClpX (Fig. 1B) imply a rapid interconversion rate between the unbound and ApSA-bound states. Therefore, any differences in deuterium uptake as a result of ApSA binding are expected to be restricted to the shortest deuterium exposure times. When mapped onto the ClpX domain structure, regions with decreased deuterium uptake at the 0.16-min exposure time correspond to the RKL, pore-2 (RDV), the LGF loops, the small AAA+ domain, and areas surrounding the box-II motif (Fig. 2 A and C). Based on the homology model of human ClpX, the RKL loop is upstream in sequence relative to a β-strand and a short helix that reinforce the positioning of the LGF loop. In our HDX-MS data, we observed a reduction in deuterium uptake in peptides that encompass the entire strand and helix (Fig. 2C). The RKL loop is homologous to the RKH loop in bacterial ClpX and plays a crucial role in substrate recognition (30). The LGF loop is homologous to the IGF loop in bacterial ClpX and docks onto ClpP during ClpXP complex formation (31, 32). In the bacterial homologs, the RKH and pore-2 loops are rich in positively charged residues and are known to play a critical role in binding and translocating substrate proteins with negatively charged degrons such as SsrA (18, 32).

In agreement with our functional and binding assays (Fig. 1B and SI Appendix, Fig. S6), introducing nonphosphorylated ASA to ATPγS-bound ClpX showed no discernible changes in deuterium uptake patterns (SI Appendix, Fig. S9). Collectively, our findings indicate that ApSA likely interacts with the RKL loop of ClpX as an initial contact site.

The RKL Loop of ClpX Interacts with pSer on Target Proteins.

To further assess the role of the RKL loop in recognizing phosphorylated substrates, we substituted its positively charged residues with alanine (R401A, K402A), generating a ClpX_AAL variant. The ClpX_AAL variant retained its ATPase activity (Fig. 3A). Compared to wildtype ClpXP (ClpXP_WT), the combination of wildtype ClpP and ClpX_AAL (ClpX_AALP) had reduced ability to cleave FITC-casein and lost its selectivity between phosphorylated and dephosphorylated α-casein (Fig. 3 B and C and SI Appendix, Fig. S10A). ClpX_AALP also failed to degrade p-tau substrate (SI Appendix, Fig. S10B). Moreover, pSer or ApSA peptides did not block the residual protease activity of ClpX_AALP (Fig. 3D). Finally, addition of pSer or ApSA did not increase the melting temperature of ClpX_AAL, indicating a loss of binding (Fig. 3E). In summary, pSer binds to the RKL loop of ClpX, which is crucial for ClpX’s recognition of phosphorylated substrates.

Fig. 3.

The RKL loop of ClpX is functionally important for the recognition of phosphorylated substrates. (A) Increasing concentrations of recombinant ClpX_WT or ClpX_AAL were incubated with ATP (5 mM) for 1 h at 37 °C. The release of free phosphate from ATP hydrolysis was measured using the Malachite Green Phosphate Assay. Data represent the mean ± SD rate of ATP hydrolysis expressed as a percent of wild type ClpX activity at 3 µM (n = 3 technical replicates). A representative experiment from at least three biological replicates is shown. (B) Increasing concentrations of recombinant ClpX_WT or ClpX_AAL in complex with wildtype ClpP protein were incubated with FITC-α-casein (5 µM). The rates of FITC-casein degradation were measured for 1 h at 37 °C. Data represent the mean ± SD (n = 3 technical replicates) rate of FITC-α-casein degradation expressed as a percent of wild type ClpXP at 3 µM. A representative experiment from at least three biological replicates is shown. (C) α-casein or dephosphorylated α-casein (5 µM) was incubated with recombinant human ClpX_AALP for 0, 1, 3, or 6 h. Proteins were analyzed by SDS-PAGE, Coomassie blue staining and semiquantified by densitometry. Data represent the mean percent ± SD (n = 3 biological replicates) residual full-length α-casein. Corresponding blots are shown in SI Appendix, Fig. S10A. (D) Recombinant ClpX_AALP protein was incubated with FITC-α-casein (5 µM) in the presence of increasing concentrations of pSer free amino acid or ApSA peptides. Rates of FITC-casein degradation were measured for 1 h at 37 °C. Data represent the mean percent ± SD (n = 3 technical replicates) rate of FITC-α-casein degradation. A representative experiment from at least three biological replicates is shown. (E) Recombinant ClpX_AAL (0.2 mg/mL) was preincubated with increasing concentrations of free or pSer amino acids or peptides for 10 min and the thermal unfolding of ClpX_AAL was monitored using SYPRO Orange, as described in the methods. Data represent mean ± SD thermal shift (∆Tm) of ClpX_AAL (°C). A representative experiment from at least three biological replicates is shown.

ClpXP Inhibition Enriches Phosphorylated Mitochondrial Proteins.

Next, we investigated whether pSer is important for ClpXP-mediated protein degradation in intact cells. First, we analyzed a published proteomics dataset which reported the posttranslational modifications in Jurkat cells with and without bortezomib treatment (1 µM, 4 h) (Fig. 4A) (33). Bortezomib inhibits the proteasome at low nanomolar concentrations, but at higher concentrations, including those used in this study, it also inhibits ClpXP (Fig. 4B) (34, 35). This dataset included 7,718 unique proteins with 5,588 proteins containing 27,129 phosphorylation sites, 1,454 proteins containing 3,152 acetylated-lysine sites, and 4,095 proteins containing 14,364 ubiquitination sites. Since ClpXP is primarily localized to the mitochondrial matrix, we focused our analysis on the phosphorylated mitochondrial matrix and inner membrane proteins. We identified 107 phosphorylated mitochondrial matrix and inner membrane proteins in this database, with 177 unique phosphorylation sites (152 pSer, 22 pThr, and 3 pTyr) (Fig. 4 C and D).

Fig. 4.

ClpXP inhibition enriches phosphorylated mitochondrial proteins. (A) Diagram of proteome and phospho-proteome (PTM-ome) profiling of Jurkat cells with and without bortezomib treatment as performed in ref. 33. (B) Recombinant ClpXP (1.5 µM) protein was incubated with FITC-casein (5 µM) in the presence of increasing concentrations of bortezomib. The rates of FITC casein degradation were measured for 1 h at 37 °C. Data represent the mean percent ± SD (n = 3 technical replicates) rate of FITC casein degradation. A representative experiment from three biological replicates is shown. (C) Number of mitochondrial matrix and inner membrane protein Ser/Thr/Tyr phosphorylation sites (pSTY) (count) that changed after bortezomib treatment in ≥2 biological replicates. The vertical dashed lines correspond to twofold changes in median site ratio, log2 (bortezomib/control). (D) Number of mitochondrial matrix and inner membrane protein serine (pSer) (Left), threonine (pThr) (Middle) or tyrosine (pTry) (Right) phosphorylation sites (count) that changed after bortezomib treatment in ≥2 biological replicates. The vertical dashed lines correspond to twofold changes in median site ratio, log2 (bortezomib/control). (E) Heat map of changes in phosphorylation site and protein level of ClpXP interacting mitochondrial matrix and inner membrane proteins (from SI Appendix, Table S3) after bortezomib treatment in n ≥2 biological replicates.

Bortezomib treatment did not significantly change the total protein levels of the mitochondrial matrix and inner membrane proteins detected (SI Appendix, Fig. S11A). However, it did result in a twofold or greater increase (log2 fold change ≥1) in abundance for 28 of the 176 (16%) phosphorylation sites of the mitochondrial matrix and inner membrane proteins compared to the DMSO control (Fig. 4C) (17% pSer sites, 5% pThr sites, and none of the pTyr sites were enriched, Fig. 3D). By contrast, acetylated counterparts within the mitochondrial matrix and inner membrane space remained unaffected (SI Appendix, Fig. S11B).

We then overlaid the 107 unique phosphorylated mitochondrial matrix and inner membrane proteins with a dataset of ClpXP interacting proteins as determined by ClpX BioID mass spectrometry (SI Appendix, Table S3) and our previously published dataset of ClpP interacting proteins (5). We identified nine phosphorylated mitochondrial proteins with 15 phosphorylation sites that interacted with ClpX and/or ClpP by BioID (Fig. 4E). The respiratory chain complex II subunit succinate dehydrogenase complex flavoprotein subunit A (SDHA) and the complex IV subunit NADH dehydrogenase 1 alpha subcomplex 4 (NDUFA4) were top hits, where pSer sites were enriched 4.6- and 3.8-fold, respectively, after bortezomib treatment (Fig. 4E). Notably, our prior research highlighted SDHA as a potential ClpP substrate (5). We confirmed that levels of serine-phosphorylated SDHA (pSer-SDHA) increased in OCI-AML2 leukemia cells after treatment with bortezomib (SI Appendix, Fig. S11C). Additionally, ClpX interacted with SDHA in OCI-AML2 cells as evidenced by a proximity ligation assay (SI Appendix, Fig. S11D).

ClpXP Knockdown Increases Levels of pSer-Substrates and Recombinant ClpXP Preferentially Degrades pSer-Substrates.

Given these results, we asked how genetic depletion of ClpXP would impact levels of phosphorylated substrates in intact cells. We transduced OCI-AML2 cells with short hairpin RNA (shRNA) targeting ClpX, ClpP, or control sequences in lentiviral vectors. We then isolated mitochondrial fractions by gradient density centrifugation and immunoprecipitated pSer-containing proteins from the mitochondrial fraction. We used immunoblotting to measure the level of pSer-SDHA and pSer-NDUFA4 in the immunoprecipitated fraction. Knockdown of both ClpX and ClpP increased levels of pSer-SDHA (Fig. 5 A and B) with no change in total SDHA. Similar increases of pSer-NDUFA4 without changing total levels of NDUFA4 were also seen after ClpP knockdown (Fig. 5C).

Fig. 5.

ClpXP knockdown increases levels of pSer-substrates in cells and recombinant ClpXP preferentially degrades pSer-substrates. (A–C) Mitochondrial lysates were collected from OCI-AML2 cells transduced with shRNA targeting ClpX (A) or ClpP (B and C) 9 d posttransduction. Serine phosphorylated proteins were enriched from the lysates by immunoprecipitation. Levels of SDHA, NDUFA4, ClpX, ClpP, and the heavy chain were measured in the total mitochondrial lysate (input) and pSer immunoprecipitated fractions by immunoblotting and semiquantified by densitometry. Representative immunoblots are shown. Densitometries represent mean ± SD relative fold changes of pSer-SDHA and pSer-NDUFA4 in comparison to control samples from ≥3 biological replicates. P-values were calculated using unpaired t tests. (D) pSer immunoprecipitated fractions of OCI-AML2 cells were incubated with 0, 1.5, or 3 µM recombinant ClpXP with or without 5 mM ApSA peptide for 3 h at 37 °C. After incubation, levels of SDHA, and heavy chain were measured by immunoblotting. Representative immunoblots from three biological replicates are shown.

Then, we asked whether ClpXP preferentially degrades pSer-SDHA. We added recombinant ClpXP to total mitochondrial lysates and immunoprecipitated the pSer fraction from OCI-AML2 cells. Adding recombinant ClpXP depleted pSer-SDHA without changing levels of nonphosphorylated SDHA (Fig. 5D and SI Appendix, Fig. S12). Moreover, ClpXP-mediated degradation of pSer-SDHA was blocked by adding ApSA peptide (Fig. 5D). Overall, these results are consistent with the preference of ClpXP for pSer-tagged substrates in intact cells.

pSer-Substrates Are Enriched in the Insoluble Mitochondrial Fraction and Are Increased upon Proteolytic Stress.

Genetic depletion of ClpP leads to the accumulation of faster migrating bands of substrates on native gels, with no change in levels of total substrate (5). This led us to hypothesize that pSer-substrates might represent damaged or aggregated proteins that preferentially partition in the insoluble fraction of mitochondrial proteins. To test this hypothesis, we measured total and pSer-SDHA in the soluble and insoluble fractions of mitochondrial proteins. Knockdown of ClpX increased levels of pSer-SDHA only in the insoluble fraction, and we did not detect pSer-SDHA in the soluble fraction, either under basal conditions or after ClpX knockdown (Fig. 6A).

Fig. 6.

pSer-substrates are enriched in the insoluble mitochondrial fraction and are increased upon proteolytic stress. (A) Mitochondria were isolated from OCI-AML2 cells 9 d after transduction with shRNA targeting ClpX and then treated with digitonin (8 g/g). Digitonin soluble and insoluble fractions were collected and pSer proteins were enriched by immunoprecipitation. Levels of SDHA and ClpX were measured by immunoblotting. A representative blot from three biological replicates is shown. (B) OCI-AML2 cells were treated with 0, 1, or 10 µM of antimycin A for 2 h or 24 h. After treatment, mitochondria were isolated from the cells, lysates were prepared and the pSer proteins were enriched by immunoprecipitation. Levels of SDHA and heavy chain were measured by immunoblotting. A representative blot from three biological replicates is shown. (C) Mitochondria were isolated from OCI-AML2 cells 24 h after treatment with increasing concentrations of antimycin A and then treated with digitonin (8 g/g). Digitonin soluble and insoluble fractions were collected and pSer proteins were enriched by immunoprecipitation. Levels of SDHA and heavy chain were measured by immunoblotting. A representative blot from three biological replicates is shown.

As an alternate approach to determine whether pSer-SDHA represents damaged or aggregated protein, we induced mitochondrial proteolytic stress by treating OCI-AML2 cells with antimycin A to increase mitochondrial reactive oxygen species (ROS) or by inducing heat shock by culturing cells at 42 °C for 1, 10, or 20 h (Fig. 6B and SI Appendix, Fig. S13A). Antimycin A disrupts oxidative phosphorylation by inhibiting cytochrome c reductase (36). Treatments with antimycin A increased levels of pSer-SDHA exclusively in the insoluble fraction (Fig. 6C). To generalize these findings, we demonstrated that antimycin also increased levels of pSer-NDUFA4 (SI Appendix, Fig. S13B). Taken together, these data support the mechanism that ClpXP preferentially recognizes pSer-substrates, particularly in aggregated proteins and that pSer-substrates comprise only a minor fraction of total protein levels under normal conditions with functional ClpXP. However, this fraction increases significantly upon ClpXP inhibition or in response to proteolytic stress induced by external stimuli.

Discussion

ClpXP maintains protein quality within the mitochondrial matrix by degrading damaged or misfolded proteins (5, 6). Targeting human ClpXP through inhibition or hyperactivation is cytotoxic to malignant cells and ligands targeting ClpXP are being developed as anticancer agents (5, 8). To date, little is known about how proteins are recognized for degradation by human ClpXP. Here, we found that pSer facilitates ClpXP-mediated degradation of substrate proteins by binding to the RKL loop of ClpX.

In bacteria, AAA+ unfoldases such as ClpA, ClpC, and ClpX form a complex with ClpP to degrade select substrates in the cytoplasm (37). These unfoldases recognize degrons on the N or C termini of substrates which direct the substrates to the protease for degradation (38–40). In Bacillus subtilis, pArg marks proteins for degradation by ClpCP (14). Upon heat shock, damaged/aggregated proteins are phosphorylated on Arg residues by McsBA kinase. The pArg degron then interacts with the NTD of ClpC and initiates the substrate translocation into the ClpP protease lumen (14). In E. coli, ClpXP degrades polypeptides with a SsrA tag (41, 42). Interestingly, despite substantial sequence conservation with E. coli ClpX (41% identity, 63% similarity), human ClpX does not recognize SsrA-tagged substrates (18).

Instead, our work demonstrates that in human ClpXP, the RKL loop of ClpX preferentially recognizes substrate proteins containing pSer and pThr, facilitating their ClpXP-mediated degradation. The difference in specificity between human and bacterial ClpX homologs may be due to the fact that the RKH loop in bacterial ClpX, which plays an important role in determining substrate specificity, is more positively charged compared to the homologous RKL loop in human ClpX. Bacterial ClpX, akin to other AAA+ unfoldases, adopts a spiral-staircase arrangement of pore-loop residues that encircle the substrate (43). Sequential ATP hydrolysis leads to dynamic movements of its loops and directional translocation of substrates via a conserved “hand-over-hand” mechanism (31, 42). While no experimental structure of human ClpXP is available, human ClpX may use a similar mechanism to engage pSer-containing substrates. In addition to interactions between pSer and the RKL and pore-2 loops, our HDX-MS data also revealed reduced deuterium uptake in the box-II motif and small AAA+ domains of ClpX. In bacteria, the conformation of these elements is crucial for ClpX ATPase activity (42). Since pSer did not affect ClpX ATPase activity, we propose that the binding or conformational changes detected in these regions are not critical for substrate recognition and may reflect the allosteric nature of ClpX. High-resolution structural studies are needed to visualize the binding of pSer to ClpX at atomic resolution.

We observed greater degradation of phosphorylated proteins compared to dephosphorylated substrates in cell-free enzymatic assays. However, some residual degradation of dephosphorylated proteins was also observed, especially with dephosphorylated α-casein. While some phosphorylation sites remained on the commercially available dephosphorylated α-casein used in our assays, we concede that serine phosphorylation is likely not the only mechanism by which ClpXP recognizes substrates for degradation and additional mechanisms likely exist. For example, casein is an unfolded substrate and the unfolded nature of the protein may also contribute to its recognition by ClpXP. Future studies will elucidate these additional mechanisms.

In intact cells, ClpXP selectively degraded pSer-SDHA but not its nonphosphorylated form. Given that pSer-SDHA was enriched in the aggregated mitochondrial protein fraction, our data are consistent with a model where ClpXP recognizes pSer motifs on aggregated, damaged, or misfolded proteins, and proteolyzes them. Notably, as these experiments were performed in cells or cell lysates, it is possible that ClpXP-mediated degradation of pSer substrates is facilitated by adaptor proteins working in concert with ClpX.

The mechanism for tagging ClpXP substrates for degradation remains elusive. It is possible that damaged or misfolded proteins are actively phosphorylated by mitochondrial kinases upon becoming damaged. Over 25 kinases are reported to localize to the mitochondria (44–46) where they regulate mitochondrial metabolism, apoptosis, autophagy, fission, and fusion (45, 47, 48). For example, Ser/Thr kinases PKA, PINK1, and PKC phosphorylate respiratory chain complex in response to cellular stress (48–51). Identifying kinases responsible for the phosphorylation of ClpXP substrate could be achieved by investigating kinase interactions with SDHA under both basal and stress conditions. Alternatively, protein damage or misfolding may expose phosphorylation sites that are originally less accessible in the correctly folded proteins. Respiratory chain subunits, including SDHA and NDUFA4, are highly phosphorylated under basal conditions, and these phosphorylations contribute to proper respiratory complex formation (46). Proteomic studies identified at least 6 pSer, 2 pThr, and 6 pTyr sites on SDHA, and 3 pTyr and 1 pSer sites for NDUFA4 (52). In response to stress, pSer-SDHA and pSer-NDUFA4 rapidly accumulate, suggesting that mitochondrial damage may expose existing phospho-sites, tagging these proteins for ClpXP-mediated degradation. Future work will explore the kinases responsible for serine phosphorylation of SDHA and how this modification affects its function. While c-Src and Fgr kinases phosphorylate SDHA at Tyr215 and Tyr604 (53), the kinases targeting SDHA serine residues remain unidentified.

Small-molecule inhibitors of mitochondrial and bacterial ClpXP have been reported, but they primarily target the ClpP protease active site (5, 54–56). Compounds like phenyl esters and β-lactones inhibit ClpP in isolation, but their effects are reversed in the presence of ClpX, thereby failing to inhibit the ClpXP holoenzyme (56). Our findings suggest an alternative approach to targeting ClpXP by interfering with the substrate engagement and tagging mechanism of ClpXP.

In summary, we found a mechanism by which human ClpXP recognizes and degrades aggregated mitochondrial proteins. This process is driven by Ser phosphorylation, which promotes substrate binding to ClpX, primarily via electrostatic interactions with the RKL loop. As such, this work uncovers a degradation marker for human ClpXP and highlights a strategy to develop ClpXP inhibitors by disrupting substrate engagement.

Materials and Methods

Detailed description of recombinant protein expression and purification, cell-free degradation assays, enzyme activity assays, α-casein mass spectrometry PTM identification, differential scanning fluorometry, hydrogen/deuterium exchange mass spectrometry, electroncryomicroscopy, proteomic data analysis, BioID liquid chromatography-mass spectrometry, cell culture, shRNA knockdown, proximity ligation assay, confocal microscopy, whole-cell lysate preparation, mitochondria isolation, immunoprecipitation, isolation of soluble and insoluble protein fractions, and immunoblotting are provided in SI Appendix. All mass spectrometry data are available from the MassIVE database as entry MSV000096217 (57) (α-casein PTM), MSV000096143 (58) (HDX-MS), MSV000093014 (59) (ClpX BioID).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Proteomic data have been deposited in MassIVE (MSV000096217 (57), MSV000096143 (58), MSV000093014 (59)). All other data are included in the article and/or SI Appendix.