Mitochondrial ClpX activates an essential metabolic enzyme through partial unfolding

Mitochondria can control the activity, quality, and lifetime of their proteins with their autonomous system of chaperones, but the signals that direct substrate-chaperone interaction and outcome are poorly understood. We previously discovered that the mitochondrial AAA+ protein unfoldase ClpX (mtClpX) activates the initiating enzyme for heme biosynthesis, 5-aminolevulinic acid synthase (ALAS), by promoting incorporation of cofactor. Here, we ask how unfolding by mtClpX directs activation. We identified sequence and structural features in ALAS that position mtClpX and provide a grip for acting on ALAS. Observation of ALAS undergoing remodeling by mtClpX revealed that unfolding was limited to a subdomain extending from the mtClpX-binding site to the active site. Unfolding along this path was required for mtClpX to gate cofactor access to the ALAS active site. This targeted unfolding contrasts with the global unfolding canonically executed by ClpX homologs and suggests how substrate-chaperone interactions can direct the outcome of remodeling.


INTRODUCTION
AAA+ protein unfoldases use the energy of ATP hydrolysis to pull apart the structures of their substrate proteins. In general, this is accomplished by gripping the substrate polypeptide with aromatic side chains of loops that protrude into the central pore of the unfoldase hexamer; ATP hydrolysis drives conformational changes in the hexamer that push these loops downward, pulling and thus mechanically unfolding the substrate from the site of engagement 1 .
Although some AAA+ unfoldases are specialized for non-proteolytic protein unfolding, this activity has been best characterized as part of protein degradation, in which the polypeptide is directly translocated into the proteolytic chamber of a cocomplexed peptidase. This basic architecture and mechanism is used by eukaryotic and archaeal proteasomes and by several unfoldase-protease complexes shared among bacteria, mitochondria, peroxisomes, and chloroplasts [1][2][3] . One such bacterial unfoldase, ClpX, with its partner peptidase ClpP, is particularly specialized for regulatory degradation of substrates rather than protein quality control, conditionally selecting a varied repertoire of substrates to execute stress responses and cell fate decisions 4 .
Accumulating evidence indicates that mitochondrial ClpX (mtClpX) acts in a similarly regulatory capacity, although only a few substrates of mtClpX have been verified and the specific regulatory consequences of its actions on mitochondrial physiology are for the most part undetermined [5][6][7][8][9][10] . How mtClpX recognizes its substrates has also not been characterized, although its sequence preferences for substrate recognition clearly diverges from that of bacterial ClpX 11 .
We previously discovered that mtClpX) promotes heme biosynthesis by acting on the first enyzme in heme biosynthesis, 5-aminolevulinate synthase (ALAS) 5 . In contrast to unfolding of a protein substrate for degradation, as ClpX homologs are best understood to do, mtClpX activates ALAS by accelerating incorporation of its cofactor, pyridoxal 5'phosphate (PLP). This activation is non-proteolytic and requires ATP hydrolysis and intact pore loops, implicating mtClpX unfoldase activity in this unconventional activity.
Here, we address how mtClpX is specifically deployed to direct activation of an enzyme.
Using a peptide array of the ALAS sequence combined with protein mutagenesis and engineering, we defined a coherent mtClpX-binding site that spans the dimer interface of ALAS. This site contains separable elements that recruit mtClpX and engage mtClpX for initiation of unfolding. Using hydrogen-deuterium exchange coupled with mass spectrometry (HX MS), we observed that mtClpX induces exposure of a limited region of ALAS that extends from the mtClpX-binding site to the active site, serving as a gate to the active site. Engineered crosslinks that obstruct this path demonstrated that this remodeling is necessary for ALAS activation. Our observations describe the mechanism by which a conserved mitochondrial unfoldase activates an essential biosynthetic enzyme and provide a model for how the interactions between protein unfoldases and their substrates can be directed to widely divergent outcomes, from restricted unfolding and activation to complete unfolding and degradation.

An N-terminal sequence directs mtClpX to activate ALAS
To determine how mtClpX induces ALAS to bind PLP, we first sought to identify how mtClpX recognizes ALAS and engages it for unfolding. As an unbiased strategy, we assayed binding of mtClpX to an array of peptides representing the linear sequence of ALAS. mtClpX bound six discrete peptide sequences within ALAS (Fig. 1A, blue; Fig.   S1C). Five of these sequences map to a structurally contiguous site in ALAS, consisting of an α-helix (α1) that is the most N-terminal structured element of ALAS, and a small region that is in direct hydrogen-bonding contact with α1 across the ALAS dimer interface (Fig. 1B). The sixth, most C-terminal peptide maps to a separate site about 40 Å away.
Because ClpX and other AAA+ unfoldases often initiate unfolding at or near protein termini, α1 of ALAS was an appealing candidate for an initial binding site for mtClpX. α1 is also immediately adjacent in structure and in sequence to a small β-sheet (β1-3) that abuts the active site. Unfolding initiated at α1 could restructure this region to allow PLP access to the active site. Indeed, we observed that β1-3 is conformationally responsive to the presence of PLP in the active site in our crystal structure of ALAS 12 : One protomer of the ALAS dimer lost PLP during crystallization in PLP-free solvent and β1-3 of the unoccupied protomer had weaker electron density and could not be modeled.
We previously also crystallized ALAS that had been incubated with hydroxylamine, which cleaves the covalent PLP-lysine bond in the active site and converts PLP to non-covalently bound and inactive PLP-oxime (Fig. S1A) 12 . Our crystal structure from this inactive, PLP-depleted preparation contains a PLP-derived species in the active site despite gel filtration to remove free cofactor. Because our previous assays for activation of ALAS by mtClpX were performed with this hydroxylamine-treated preparation of ALAS, we sought to determine whether mtClpX exchanges inactive cofactor, facilitates cofactor binding to unoccupied active sites, or both. We determined that hydroxylamine-treated and gel-filtered ALAS remained slightly more than halfoccupied with a species consistent with PLP-oxime (Fig. S1A, B). The activation of hydroxylamine-treated ALAS by mtClpX therefore represents a combination of PLP binding to unoccupied active sites and exchange of PLP-oxime for PLP. Both events are likely useful in supporting ALAS function in vivo, assisting de novo PLP binding and regenerating ALAS enzyme inhibited by damaged cofactor (for example, PMP resulting from decarboxylation-coupled transamination of PLP, or other natural errors in catalysis 13 ). Both events could be similarly facilitated by unfolding by mtClpX to expose the ALAS active site.
To test the importance of the putative N-terminal mtClpX binding site of ALAS, we engineered N-terminal fusions of ALAS with dihydrofolate reductase (DHFR) (Fig.   1C). N-terminal fusion of DHFR with ALAS (DHFR-ALAS) blocked stimulation of PLP binding by mtClpX, but appending the N-terminal sequence of ALAS through α1 to DHFR-ALAS (α1-DHFR-ALAS) restored the ability of mtClpX to stimulate PLP binding. Thus, mtClpX recognizes the N-terminus of ALAS to initiate activation. Upon binding its inhibitor methotrexate, DHFR becomes extremely mechanically stable and resistant to unfolding, thus providing a conditional block to mtClpX. Addition of methotrexate blocked mtClpX activation of α1-DHFR-ALAS (Fig. 1D) indicating that unfolding from the N-terminus is required for ALAS activation. Therefore, mtClpX requires the α1 site to recognize ALAS and unfolds from the N-terminus to induce activation.

Unfolding from the N-terminus is sufficient to activate ALAS
To test if unfolding from the N-terminus of ALAS is sufficient for activation, we attempted to direct E. coli ClpX to stimulate PLP binding. E. coli ClpX does not act on wildtype ALAS (Fig. 1E). To direct it to ALAS, we appended a recognition sequence for E. coli ClpX (λO 2-12 , from the E. coli ClpX substrate lambda phage O protein 14 ) to the Nterminus of ALAS (λO 2-12 -ALAS) (Fig. 1C). E. coli ClpX accelerated PLP binding to λO 2-12 -ALAS at a similar rate to mtClpX with wildtype ALAS (Fig. 1E, F). Therefore, an unfoldase acting from this N-terminal site is sufficient to stimulate ALAS activation. We also tested if the short, unstructured λO 2-12 tag might block mtClpX action at the Nterminus of ALAS as the folded DHFR domain did. We unexpectedly observed the opposite effect: λO 2-12 -ALAS was processed more rapidly than ALAS by mtClpX (Fig.   1F). This increased efficiency was not based on specific interaction of mtClpX with the λO 2-12 peptide (Fig. S1D), suggesting that mtClpX instead responded to the unstructured extension contributed by the λO 2-12 tag.

mtClpX relies on an unstructured N-terminal extension for rapid activation of ALAS
The observation that an unstructured extension to the N-terminus of ALAS stimulated its activation by mtClpX led us to reexamine the native N-terminus of ALAS in the mitochondrion. Many mitochondrial proteins, including ALAS, are translated with an N-terminal targeting sequence that is cleaved after import into the mitochondrial matrix. The N-terminus of the ALAS constructs we used in our previous experiments (position 58 in the coding sequence) was based on a previous proteomic survey of mature N-termini of mitochondrial proteins in yeast 15 . To specifically observe the mature Nterminus of ALAS, we performed Edman degradation on ALAS purified from yeast cell extracts. ALAS appeared as a single processed species by Western blot (with possible trace uncleaved preprotein remaining, Fig. S2A). The N-terminal sequence of this species corresponded to cleavage after residue 34 of the preprotein (∆34-ALAS), indicating that an additional 23 amino acids are retained by the mature protein beyond the previously detected species (∆57-ALAS) ( Fig. 2A, Fig. S2B). One possible source of this discrepancy is that the slower isolation procedure used to globally identify mitochondrial N-termini may have resulted in additional processing to yield Δ57-ALAS.
We tested the effect of this natural N-terminal extension of ALAS on mtClpX activation of ALAS and found that mtClpX activated ∆34-ALAS even more rapidly than  (Fig. S2B), suggesting that the role of this extension may be sequence independent. We did not detect any additional binding sequences in a peptide array including the entire preprotein sequence of ALAS, additionally supporting a sequence-independent interaction (Fig. S1D). This extended sequence has a dynamic or disordered structure; the N-terminus of Δ34-ALAS to the beginning of α1 (residue 71) exhibited near-instantaneous completion of deuterium uptake in hydrogen-deuterium exchange experiments ( Fig. S2C; discussed further below). Therefore, mtClpX is a more potent activator of ALAS than we had previously appreciated, facilitated by a flexible element at the natural mature N-terminus of ALAS.

Multiple sequence-specific contacts direct mtClpX action on ALAS
To identify residues that recruit mtClpX and position it on ALAS for unfolding and activation, we scanned the peptide sequences that bound mtClpX ( Fig. 1A) with alanine and aspartate on an additional peptide array probed with mtClpX ( Fig S3). In this array, we identified point mutations in each peptide that caused near-complete loss of mtClpX binding, with the exception of the most C-terminal peptide (Fig. S3). Most residues sensitive to mutation were perturbed by both alanine and aspartate. Several bulky hydrophobic amino acids (leucine, phenylalanine, and tyrosine) were highly represented in positions important for mtClpX interaction. This side-chain preference differed from the characterized preference of E. coli ClpX, for which basic residues are often important in substrate recognition 16 . The mtClpX-preferred residues we observe are also notable in that they comprise the set of signal residues described in the mitochondrial N-end rule 15 .
To test the role of these binding sequences in ALAS binding and activation, we mutated several residues that were important for mtClpX binding to ALAS-derived peptides in ∆34-ALAS. Alanine substitution at positions in α1 (F71A, Y73A, and the double substitution F71A/Y73A) reduced coprecipitation with mtClpX ( Fig. 3A,B), and increased the K M for activation by mtClpX ( Fig. 3C-E), consistent with α1 functioning as a sequence-specific tag that recruits mtClpX to begin pulling on ALAS.
The double substitution variant (ALAS F71A/Y73A ) additionally had an increased rate of spontaneous PLP binding ( Fig. S4A) and had an increased V max for mtClpX-stimulated PLP binding (Fig. 3D,E). These residues are positioned at the dimer interface: Y73 directly contacts the other protomer through hydrogen bonding, and F71A is positioned for a CH-pi interaction with P246 (Fig. S4B). Although F71A and Y73A single variants retained synthase activity similar to wildtype, the double variant ALAS F71A/Y73A retained only ~10-20% of wildtype activity (Fig. S4C). Removing contacts between the protomers through these mutations likely increases the solvent accessibility of the active site, thus increasing the spontaneous PLP binding rate of ALAS at the expense of active site function. Because the rate of unfolding of protein substrates by ClpX homologs is proportional to the mechanical stability of their fold, this reduced contact may be responsible for the increased V max of mtClpX action on ALAS F71A/Y73A . These residues in α1 thus participate both in mtClpX recognition and in stabilizing contacts across the dimer interface that prevent spontaneous PLP exchange and promote activity.
In contrast to mutations in the N-terminal helix, a mutation at the surface of the mtClpX-interacting sequence cluster across the dimer interface from this helix, Y274A, did not perturb the K M for mtClpX activation of ALAS (although it did reduce coprecipitation (Fig. 3B)). The Y274A mutation instead reduced the V max of mtClpX activation (Fig. 3D, E). Y274 and other residues in this peptide cluster therefore may primarily interact with mtClpX to promote its rapid processing of ALAS.
To test if the contacts between ALAS and mtClpX we had identified were important for mtClpX to support ALA production in vivo, we introduced alanine mutations of the contact residues into the single genomic copy of ALAS in yeast (HEM1) in a wildtype or mtClpX-null (mcx1Δ) background. Within α1, F71A reduced and Y73A abolished mtClpX stimulation of ALA production in vivo ( In the presence of mtClpX, deuteration was increased in a defined region of ALAS over time (Fig. 4A, B; Supplemental File 1), consistent with mtClpX enacting a limited, specific conformational change rather than globally unfolding ALAS. mtClpXinduced deuterium uptake was localized to α1, the β sheet that connects α1 to the active site (β1-3) and the following active-site loop and α2, and several non-contiguous peptides surrounding the active site of ALAS (Fig. 4A, B). The mass spectra of most of these peptides exhibited clear EX1 kinetics 17  Two non-homogeneous features in the linear path of mtClpX-induced exchange suggest that mtClpX may not unfold along the entire length of this path. First, the unfolding half-life (t 1/2 ) 18 for mtClpX-induced exchange in these elements is not uniform: exchange in α1 and β1-3 has a shorter t 1/2 than the exchange further along this path (as-α2 indicated in Fig. 1A, C) and in the non-contiguous peptides surrounding the active site ( Fig. 4A right panel; Fig. 4C). These regions therefore may be exposed as a secondary and/or lower-probability consequence of mtClpX remodeling of α1 and β1-3, giving rise to a longer t 1/2 for deuterium uptake. Second, a short sequence at the junction between α1 and β1 (82-89) exhibited no mtClpX-induced exchange (Fig. 4D,E). A possible explanation for this observation is that mtClpX arrests when its pore reaches this site.
Occupancy by mtClpX could confer similar protection from solvent as the native structure of ALAS while regions further into the protein are deprotected by conformational change propagating from this site.
Together with our mapping of mtClpX binding sites, these data indicate a path of unfolding from the N-terminus of ALAS through some portion of the first β sheet.
Unfolding along this path could thus gate access of PLP to the active site.

Opening the β1-3 gate is required for ALAS activation
To test if mtClpX unfolding of ALAS from the N-terminus through α1 and some or all of the following β sheet is required to stimulate cofactor binding, we designed cysteine pairs flanking this path, such that formation of a disulfide bond between these pairs would block the linear path for ClpX unfolding (Fig. 5A). No disulfide bonds between the native cysteines in ALAS were predicted; in agreement with this prediction, nearly all wildtype ALAS protein remained as a monomer after oxidation (Fig. 5B). mtClpX stimulated oxidized and reduced ALAS to bind PLP at a similar rate (Fig. 5C, D), providing a clean background against which to observe perturbation by introduced disulfide pairs. We designed one cysteine pair (positions 68 and 243) to link the Nterminus of α1 to an adjacent site in the other protomer and a second pair (positions 88 and 427) to link the α1-β2 junction with an adjacent site in the same protomer (Fig. 5A). ALAS 68x243 and ALAS 88x427 both ran as monomers when reduced, but upon oxidation exhibited a near-complete shift in mobility. Oxidized ALAS 68x243 shifted to an apparent higher molecular weight, approximately twice that of ALAS monomer, consistent with the cross-protomer disulfide bond. Oxidized ALAS 88x427 shifted to a slightly faster-migrating species, consistent with the intramolecular crosslink between these residues ( Fig. 5B). mtClpX action upon both oxidized variants was strongly attenuated compared to wildtype (Fig. 5C, D). Upon re-reduction of the crosslink, mtClpX induced both cysteinepair variants of ALAS to bind PLP at a similar rate as for wildtype ALAS, indicating that loss of mtClpX action on these variants was specifically perturbed by disulfide bond formation (Fig. 5C, D). Therefore, mtClpX must unfold from its N-terminal engagement site on ALAS and extract at least the first β strand of ALAS in order to activate ALAS.

DISCUSSION
We sought to define the mechanism by which mitochondrial ClpX deploys its unfoldase activity to activate ALAS, the initiating enzyme in heme biosynthesis. Here, we have defined sequence and structural elements that direct mitochondrial ClpX to bind ALAS and unfold a substructure of the protein, gating PLP cofactor to the active site. Our findings describe the mechanism of biogenesis and repair of an essential mitochondrial enzyme. This mechanism involves a diversion of mtClpX from the complete, global unfolding that ClpX homologs are best understood to execute. Our data indicate that mtClpX stalls in a region that lacks the few features previously described to stall unfoldases, suggesting that a greater variety of signals between unfoldase and substrate can direct alternative outcomes. This non-canonical action of a ClpX homolog also suggests that partial unfolding may be deployed to maintain and regulate a larger and more diverse set of cellular proteins than previously appreciated.
The elements comprising the multivalent ALAS recognition site affect different parameters of mtClpX action, suggesting a model for how they cooperate to recruit mtClpX and promote unfolding of ALAS. First, a composite site that includes α1 and a region immediately across the ALAS dimer interface recruits mtClpX using sequencespecific contacts. Mutation of residues in this site decreased the avidity of the ALAS-mtClpX interaction and insertion of an unfolding-resistant domain between this site and the body of ALAS blocked activation by mtClpX. We propose that these contacts recruit mtClpX and position it to pull efficiently on the ALAS structure. Once recruited, mtClpX initiates unfolding from the unstructured N-terminal tail of ALAS that precedes α1.
Initiation of unfolding did not depend on the sequence of this tail and its deletion primarily increased the V max of mtClpX activation rather than the avidity of interaction. This suggests that the N-terminal tail of ALAS provides a readily available grip for the pore loops of mtClpX to initiate unfolding. This composite signal, a sequence-specific recognition site and an sequence-independent unstructured element, is similar to a code that has been observed for protein unfolding by the proteasomal unfoldases: ubiquitin conjugation confers recognition of a protein substrate, but a nearby unstructured element is important for efficient unfolding 19 .  22,23 . For these substrates, as well as some engineered model substrates, the truncation point is between independent domains; the unfoldase is limited by encountering a mechanically stable domain and an immediately preceding sequence that appears to stall the unfoldase. These sequences are often (but not always) low complexity in some way; how they interact with an unfoldase to promote stalling is not known. The partial unfolding of ALAS we observe does not obviously follow this code. The stopping point for mtClpX appears to be in the middle of a compact fold, rather than in a linker between domains. Although ALAS contains some alanine repeats in its natural Nterminus, replacement with the non-repetitive λO 2-12 sequence still supported robust stimulation of cofactor binding by mtClpX.
Unlike ClpX, the proteasomal unfoldases, and other AAA+ unfoldase generalists that act on a wide variety of substrates, some AAA+ unfoldases act as dedicated modulators of one client protein. Several of these specialists, such as Rca, which deinhibits Rubisco, or Trip13, which inactivates the mitotic checkpoint by remodeling Mad2, carry out a limited unfolding of their client that, like mtClpX with ALAS, does not appear to follow the stalling sequence/stable domain code 24,25 . Other substrate features and their interactions with the unfoldase may similarly direct both specialist AAA+ unfoldases and generalist AAA+ unfoldases with specialized substrates. The close apposition of the mtClpX binding site and its apparent stopping point suggests that the multivalent interactions that we have determined to recruit mtClpX to ALAS may themselves contribute to limiting its translocation of ALAS as well. Future probing of the arrest points of ALAS-mtClpX and these dedicated client-unfoldase pairs will illuminate signals that allow protein unfoldases to direct precise and varied outcomes for their substrates.

Protein purification
S. cerevisiae ALAS (Hem1) and mtClpX (Mcx1) and related variants were expressed and purified as described previously 5

ALA measurement
Overnight cultures of yeast grown at 30°C in synthetic complete medium (CSM + YNB, Sunrise Science Products) + 2% glucose were used to inoculate cultures to OD 600 0.1-0.15 in the same media formulation. All cultures were grown to OD 600 1.0-1.25 at 30°C with shaking at 220 rpm. The equivalent of 10 mL at OD 600 1.0 was harvested by filtration.
Cell extracts were prepared and ALA was measured by colorimetric assay using modified Ehrlich's reagent as previously described 5 , except yeast extracts were cleared by centrifugation at 21,000 g rather than 2000 g.

Peptide arrays
SPOT arrays of 15-amino-acid peptides, C-terminally linked to a cellulose membrane, were synthesized by standard Fmoc techniques using a ResPep SL peptide synthesizer Edman sequencing/protein isolation S. cerevisiae ALAS was isolated from yeast cell extract by means of a C-terminal His 7 tag (strain JKY220). 1 L YPD was inoculated to OD 600 0.05 from a saturated overnight culture and grown to OD 600 1.0 (30°C, 220 rpm shaking). Cells were harvested by centrifugation (3500 g, 5 min) and washed once by centrifugation in lysis buffer (50 mM Tris, 500 mM NaCl, 10 mM imidazole, 10% glycerol). The cell pellet was snap-frozen,