Multitasking in the mitochondrion by the ATP-dependent Lon protease

Abstract

The AAA⁺ Lon protease is a soluble single-ringed homo-oligomer, which represents the most streamlined operational unit mediating ATP-dependent proteolysis. Despite its simplicity, the architecture of Lon proteases exhibits a species-specific diversity. Homology modeling provides insights into the structural features that distinguish bacterial and human Lon proteases as hexameric complexes from yeast Lon, which is uniquely heptameric. The best-understood functions of mitochondrial Lon are linked to maintaining proteostasis under normal metabolic conditions, and preventing proteotoxicity during environmental and cellular stress. An intriguing property of human Lon is its specific binding to G-quadruplex DNA, and its association with the mitochondrial genome in cultured cells. A fraction of Lon preferentially binds to the control region of mitochondrial DNA where transcription and replication are initiated. Here, we present an overview of the diverse functions of mitochondrial Lon, as well as speculative perspectives on its role in protein and mtDNA quality control.

1. Introduction

The ATP-dependent Lon protease is highly conserved throughout all phylogenetic kingdoms from archaea, eubacteria, fungi to mammalian mitochondria and peroxisomes. Our understanding of mitochondrial Lon is based largely on studies employing yeast as well as animal cell systems, which together have demonstrated that Lon supports cell viability during hypoxic, oxidative and endoplasmic reticulum (ER) stress [1–3]. In yeast, Lon and m-AAA are the only ATP-dependent proteases within the mitochondrial matrix, whereas in metazoans ClpXP is also present (Fig. 1). Together they provide an integrated surveillance system that monitors mitochondrial protein quality control. However, their respective functions are not necessarily redundant. Recent work suggests that Lon and ClpXP may operate in distinct pathways during adaptive responses to cellular stressors [1–6]. In addition, mitochondrial Lon is implicated in the maintenance and expression of mitochondrial DNA (mtDNA), as a yeast knockout results in large mtDNA deletions and the failure to process mitochondrially-synthesized RNAs that contain introns [7–9]. In mammals, a sub-population of Lon is associated with mitochondrial nucleoids [10–12] and interacts with mtDNA in a sequence-specific and strand-specific manner [13–17], binding to nucleic acid sequences with a propensity to form G-quadruplexes [16]. In cultured human cells, Lon preferentially binds to the control region of the mitochondrial genome where transcription and replication are initiated [15]. In Drosophila cells, Lon has been shown to degrade mitochondrial transcription factor A (TFAM), which is a regulator of mtDNA metabolism [18]. This review provides a current outlook on the biology of mitochondrial Lon.

Fig. 1. ATP-dependent proteolytic machines within metazoan cells.

The 26S proteasome is the only non-mitochondrial energy-dependent protease within animal cells that operates in the cytosol and nucleus. i-AAA and m-AAA are transmembrane proteases located within the mitochondrial inner membrane (IM) with their catalytic sites for ATP hydrolysis and proteolysis located within the intermembrane space (IMS) and matrix, respectively. Lon and ClpXP are soluble proteases within the matrix.

2. Lon as the simplest ATP-powered proteolytic machine

Lon is distinguished by its simplicity as a soluble, single-ringed homo-oligomer [19–22]. By contrast, other AAA⁺ proteases such as the 26S proteasome, HslUV, ClpP- and FtsH- like proteases are more complex as they possess multiple sub-assemblies or are membrane-embedded [23, 24]. Figure 1 shows the ATP-dependent proteolytic machines within metazoan cells. In the cytosol and nucleus of higher eukaryotes, the 26S proteasome is the most elaborate AAA⁺ protease, consisting of a 20S protease core particle and two 19S regulatory caps [25–28]. The 20S core is composed of four stacked heptameric rings that form a barrel-shaped proteolytic chamber, and both ends are bound to a respective 19S regulatory cap. Each 19S cap is further sub-assembled into a “lid,” which contains ubiquitin-binding subunits and a “base,” which contains ATPase subunits. Similar to the 26S proteasome, ClpXP in the mitochondrial matrix is a soluble enzyme in which the protease and ATPase activities are found on distinct structural sub-complexes. ClpP consists of identical subunits assembled into two stacked heptameric rings that form a compartmentalized proteolytic chamber, which is capped at either end with a single hexameric ring of ClpX ATPase subunits [29–32]. In the mitochondrial inner membrane, m-AAA and i-AAA are transmembrane complexes that possess catalytic domains with opposing orientations that face either the matrix or the inter-membrane space, respectively [33–35]. In humans, m-AAA is found as both a homo-oligomer of AFG3L2 subunits and as a hetero-oligomer of AFG3L2 and paraplegin subunits while i-AAA is composed of YMEL1 subunits. However, the precise arrangement and stoichiometry of subunits within these complexes are unknown.

Mitochondrial Lon is encoded by a nuclear chromosomal gene and is translated in the cytosol as a precursor polypeptide that carries an amino-terminal mitochondrial targeting sequence (MTS) (Fig. 2A) [36, 37]. This sequence directs the translocation of the Lon precursor across the mitochondrial outer and inner membranes. Upon protein import into the matrix, the targeting sequence is cleaved off, resulting in the mature processed protein. The yeast homolog of Lon (Pim1p) also undergoes an autocatalytic cleavage of its amino-terminus [38]. Each subunit contains three functional domains: the N-terminal domain (N-domain), the ATPase domain or AAA⁺ module, and the protease domain (P-domain) (Fig. 2A). The N-domain exhibits the highest degree of divergence in both amino acid composition and length, and is proposed to mediate substrate recognition. Within the AAA⁺ module, the Walker Box A and B motifs are responsible for binding and hydrolyzing ATP. The P-domain shows the highest evolutionary conservation surrounding the proteolytic active site serine residue [39].

Fig. 2.

(A) Domain structure and functional motifs within the primary amino acid sequence of human mitochondrial Lon. MTS - the mitochondrial targeting sequence of the Lon precursor protein binds to the mitochondrial outer membrane translocation machinery, which mediates protein import into the matrix. N domain- a highly variable amino-terminal domain proposed to function in substrate recognition and binding. AAA⁺ module- composed of the α/β sub-domain, which contains the Walker Box A and B motifs for ATP -binding and –hydrolysis, in addition to the small α sub-domain. P domain- the protease domain containing the serine (S) and lysine (K) residues forming the catalytic dyad at the proteolytic active sites. S. cerevisiae Lon contains a 54 amino acid charged region that is present between the AAA⁺ and protease modules. (B) Alignment of the primary amino acid sequences that include the Lon AAA⁺ module and P-domain from H. sapiens, S. cerevisiae and B. subtilis. Alignment was performed using Jotun-Hein Method. The sequence similarity within this region between H. sapiens and S. cerevisiae and B. subtilis is 58% and 49%, respectively. Walker Box motifs A and B are indicated by a blue line; the highly charged region present in yeast Lon/Pim1p is highlighted in yellow; and the residues forming the serine-lysine dyad are outlined in red.

Insights into the structure and function of mitochondrial Lon are provided by crystal structures of the bacterial holoenzymes from Bacillus subtilis and Thermococcus onnurineus NA1 (Ton) [21, 22], the truncated protease complex of E. coli Lon [40] as well as the isolated AAA⁺ module, N and P domains [41–44]. The structure of the protease domain complex of E. coli Lon revealed a unique serine protease that employs a serine-lysine dyad at its active site [40]. Lon differs from typical serine proteases such as chymotrypsin, which have a canonical catalytic serine-aspartate-histidine triad at their active sites. The presence of a serine-lysine dyad is also observed in the crystal structures of intact holoenzymes of B. subtilis and Ton Lon [21]. Like other self-compartmentalizing proteases, the active sites of the six assembled Lon monomers are sequestered within an aqueous chamber and inaccessible from the exterior. Another common feature of AAA⁺ proteases is that ATP or ADP is bound at the interface between adjacent monomers in the hexamer [21, 22, 41, 45]. The spatial layout of the TonLon holoenzyme is described as a “bowl and lid,” in which the interior of TonLon forms one continuous chamber without apparent gating between the ATPase and protease domains [21].

Several architectural features of mitochondrial Lon proteases show distinct species-specific diversity. Saccharomyces cerevisiae Lon purified from mitochondria is a ring-shaped complex with seven flexible subunits as determined by analytic ultracentrifugation and cryo-electron microscopy [20]. When purified in the absence of added nucleotide, the yeast Lon holoenzyme exhibits a distinct asymmetry in which two adjacent subunits are in an extended conformation [20]. The addition of ATP or non-hydrolyzable AMP-PNP during purification, leads to a conformational change and an increased frequency of symmetric ring-shaped particles. These findings suggest that ATP-binding in the absence of hydrolysis promotes dynamic rearrangements within the holoenzyme. The asymmetry of yeast Lon as demonstrated by cryo-electron microscopy is significantly more pronounced than that observed for B. subtilis Lon or TonLon as shown by X-ray crystallography. For TonLon, the protease domains are arranged with a near perfect six-fold symmetry relative to the axial pore. However, the AAA⁺ domains show a nucleotide-dependent asymmetry of the α/β sub-domains that upon nucleotide exchange are proposed to cause fluctuations in the diameter and shape of the portal thereby supporting substrate translocation [21].

The finding that bacterial Lon proteases are hexameric whereas S. cerevisiae mitochondrial Lon is heptameric, raises the question as to the oligomeric status of the mammalian holoenzyme. Clues to the human and yeast complexes are provided by homology modeling using the B. subtilis Lon structure (PDB 3M6A) as a template (Fig. 3 and Legend). Primary sequence alignment modified by secondary structural conservation demonstrates that human Lon has 43.3% identity with B. subtilis Lon (Supplementary Fig. 1), whereas a significantly lower structural identity is observed with TonLon (36%). The homology model of human Lon suggests it is a hexameric complex that has an asymmetric, open-ring arrangement reminiscent of yeast Lon [20]. When the same homology modeling approach is applied to S. cerevisiae Lon, it cannot be modeled as a hexamer (Supplementary Fig. 2 and 3). This is most likely explained by the primary amino acid sequence of S. cerevisiae Lon, which is ~250 amino acids longer than most bacterial and metazoan Lon sequences. Specifically, yeast Lon possesses a highly charged insertion of ~50 amino acids between the AAA⁺ module and the P domain, which is absent in bacterial and mammalian Lon sequences (Fig. 2B). The homology model of yeast Lon monomers into a hexameric complex suggests that the charged region creates a strong steric hindrance between neighboring subunits and the non-rectifiable clashing of side chains (Supplementary Figs. 2 and 3). Future studies are required to test the homology model of human mitochondrial Lon and to solve the crystal structure of the mammalian holoenzyme.

Fig. 3. Molecular modeling of human mitochondrial Lon protease.

(A) The ribbon model was generated in PyMol. (B) The space filling model was generated in Schrödinger Suite. Models of the human Lon monomer containing the ATPase and protease domains were initially generated by the CPH server [121], (http://www.cbs.dtu.dk/services/CPHmodels/), which uses the SEGMOD tool of GeneMine (Molecular Application Group, GeneMine, Palo Alto, CA, formerly known as XLOOK) [122–124]. SEGMOD divides the target sequence into short segments followed by searches for homologous structural fragments in Protein Data Bank (PDB). The fragments are then fitted onto the template structure using secondary structure alignment and multiple independent models were generated of the entire structure. The optimal structure is the average of multiple models (GeneMine Manual, pp. 425 – 426). The resulting structures were submitted to the PDB to identify protein structures with highest homology. A sequence alignment was then generated to identify conserved and defined regions of secondary structure. The secondary structural alignment demonstrated that human Lon has 43.3% identity with Bacillus subtilis Lon (Supplementary Figure 1). By contrast, human Lon showed significantly lower structural identity (36%) with Thermococcus onnurineus NA1 (Ton) Lon. Therefore, the crystal structure of B. subtilis Lon (PDB file 3M6A) [22], was employed as the template structure to model human Lon. The loops joining the secondary structures were modeled by searching for homologous structures in PDB. The side chains were then added using energy constraints. The model obtained from the CPH server was refined using Schrödinger Suite (Schrödinger Inc. LLC, New York, NY). The ‘Prepare Protein’ protocol was used to assign bond orders and valences of each atom, to add the required hydrogen atoms and to optimize H-bond and side chain orientations. The resulting monomer structure was minimized using the ‘Impact’ utility with a OPLS2005 force field. The quaternary structure model of human Lon was generated by Sybyl (Version X; Tripos Associates, St. Louis, MO). A computer program in Sybyl Programming Language (SPL) was written to incorporate a C₆ symmetry related operation with energy constraints that limited the steric clash to a distance of 0.5 Å between two side chain atoms at the interface of adjacent monomers in the complex. In the event of steric clashes, energy minimization of the side chains was performed using AMBER force field and Gesteiger-Marshelili charges [125, 126]. The final model was further subject to ‘Protein Preparation’ Wizard’ tool of Schrödinger Suite. A similar protocol was used to generate the model for yeast mitochondrial Lon protease.

3. Enzymology of mitochondrial Lon

3.1 Assaying Lon-mediated proteolysis

The proteolytic activity of Lon in its purified form or in cell and mitochondrial extracts has been investigated primarily using fluorogenic peptide and protein substrates. Early studies with E. coli Lon (or La) employed fluorogenic tetrapeptides such as Glt-Ala-Ala-Phe-MNA or Suc-Ala-Ala-Phe-MNA [46]. In the absence of ATP, Lon cleaves these tetrapeptides, albeit weakly, whereas in the presence of ATP or its non-hydrolyzable analogs, peptide cleavage is stimulated substantially. This demonstrates that ATP hydrolysis is not required for degrading short unstructured peptides, and that ATP-binding alone is stimulatory. The degradation of larger protein substrates by E. coli Lon has been characterized using [³H]- or fluorescently labeled- casein, leading to the observation that ATP -binding as well as -hydrolysis are required for proteolysis [47–50]. Because these peptide and protein substrates are commercially available and convenient to use, they are currently being employed by various investigators to measure mitochondrial ATP-dependent proteolysis in cell or mitochondrial extracts, and in some cases proteolytic activity is attributed exclusively to Lon. However, these substrates are not Lon-specific and are degraded by other cellular proteases. For example, the fluorogenic Ala-Ala-Phe peptides are degraded by chymotrypsin as well as the 20S proteasome [51]. The lack of protease specificity combined with the relatively poor catalytic efficiencies of Lon for these peptide substrates (k_cat/K_m < 10² M⁻¹s⁻¹) limits their usefulness to biochemical assays with purified proteases. Similarly, casein is degraded by a wide variety of mitochondrial and non-mitochondrial proteases. Thus, ATP-stimulated or ATP-dependent proteolysis of these reporters in cell and mitochondrial extracts is by no means a diagnostic signature of Lon activity.

The need for specific reporter substrates to distinguish Lon-dependent protease activity led to the development of FRETN 89-98 to study the mechanism of E. coli Lon [53]. FRETN 89-98 consists of a peptide sequence from the bacteriophage λN protein (amino acid residues 89-98), which is an endogenous of bacterial substrate [54]. Intact FRETN 89-98 carries a carboxyl-terminal anthranilamide fluorophore (donor) that is internally quenched by an amino-terminal 3-nitrotyrosine (acceptor). Upon cleavage of FRETN 89-98, the donor and acceptor separate from one another leading to increased fluorescence. Human mitochondrial Lon also degrades FRETN 89-98 and has been used to monitor its proteolytic activity [52]. Although FRETN 89-98 is also degraded by the 20S proteasome, this activity is not stimulated by ATP and thus can be distinguished from Lon [52]. Additional peptide or chemical reporters are required to distinguish specific mitochondrial and cytosolic ATP-dependent protease activity in cellular extracts.

3.2 Functional mechanics of substrate recognition and catalysis by Lon

Our current understanding of substrate selection and cleavage by Lon proteases is based principally on studies employing the bacterial enzyme. However, insights into the functional mechanics of mitochondrial Lon can be drawn from the bacterial system as human Lon is able to degrade some bacterial Lon substrates in vitro [52], and yeast Lon/Pim1p can be functionally replaced by E. coli Lon [55]. Comparing the degradation profiles of several endogenous or heterologous substrates of E. coli Lon led to the postulate that solvent exposed hydrophobic peptide patches in an unfolded protein function as recognition tags for the protease [56–59]. In general, deletion of such tags stabilize proteins that are otherwise degraded by Lon [57, 59–61] and in some cases, protein degradation is initiated by incorporating a recognition tag into a stable heterologous reporter protein [57–59]. Some stably folded proteins with accessible recognition tags are unfolded and degraded by E. coli Lon [58]. However, other signals or features of substrates that are not hydrophobic in nature can also target proteins for degradation by bacterial Lon. Analysis of the peptide sequences within the transcriptional activator SoxS that are required for Lon-mediated proteolysis show that hydrophobicity does not appear to be an essential feature of the sequence elements identified [59]. The presence of a single histidine residue at the extreme C-terminus of the cell division inhibitor SulA signals its efficient degradation by E. coli Lon [60]. Studies examining the mechanism by which E. coli Lon degrades the DNA binding protein HUβ show that the recognition and binding sites within this substrate are independent of the initial cleavage site [62]. Although HUα and HUβ are both recognized and bound by Lon to the same extent as determined by rupture force spectroscopy and competition experiments, only HUβ contains an initiating Lon cleavage site. The rules governing the recognition and cleavage of physiological substrates by E. coli Lon are apparently less clear than those identified for heterologous reporter substrates and show more variability and complexity.

Studies with human mitochondrial Lon demonstrate that certain endogenous protein substrates are recognized only when they are in a folded functional state and not when they are unfolded [63]. For example, the unassembled α subunit of the mitochondrial processing peptidase (MPPα) is targeted by Lon only when it is in a native conformation that is trypsin resistant and capable of forming an active enzyme with its partner MPPβ. The initial cleavage sites within MPPα are at hydrophobic residues within structured regions that are exposed at the surface of the folded protein and surrounded by a highly charged environment. Interestingly, these initiating cleavage sites are not masked at the interface where MPPα is complexed with MPPβ. These findings suggest that substrate recognition by human mitochondrial Lon is not restricted to abnormal proteins that are misfolded or damaged.

The potential cleavage of a folded protein or structured region by mitochondrial Lon raises the question of how such substrates gain access to the active site of the protease where they can be initially clipped thereby leading to subsequent unfolding and degradation. One possibility offered by the structures of bacterial and yeast Lon is that a ring-opening mechanism within the enzyme permits access of structured regions or small folded proteins to the proteolytic chamber [20–22]. However, other possibilities are that the substrate is locally unfolded before translocation to the active site of the protease, and that multiple polypeptide chains may pass through the pore at once as has been observed for bacterial ClpXP [64].

The general paradigm for ATP-dependent proteolysis likely applies to the mitochondrial Lon protease (Fig. 4) [65–69]. The N-domain and AAA⁺ module of the holoenzyme first recognize and bind a specific recognition determinant or site within a protein substrate in an ATP-independent manner (step 1). ATP -binding and -hydrolysis drive conformational changes within the enzyme complex that unfold substrate polypeptides (step 2), and also power the translocation of the denatured substrates into the proteolytic chamber, which is sequestered from the surrounding aqueous environment (step 3). Once the unfolded substrate is accessible to the proteolytic active sites, peptide bond cleavage occurs (step 4). Substrate unfolding and translocation are proposed to constitute the rate-limiting steps of the reaction [70] through repetitive cycles of ATP hydrolysis that induce conformational changes within the protease [71, 72]. Translocation of substrate is hypothesized to occur by a threading mechanism [68, 73]. For certain substrates of ATP-dependent proteases, proteolytic digestion generates peptides ranging from ~5 to 20 amino acids and not a variety of partially digested polypeptide intermediates, leading to the conclusion that degradation proceeds processively [26, 54, 63, 65, 74].

Fig. 4. General paradigm for the recognition and degradation of protein substrates by the Lon protease.

Step 1 - recognition and binding of a structural feature or sequence within a protein substrate. Step 2- unfolding of structured protein substrates by the AAA⁺ domains of the holoenzyme, which requires ATP –binding and -hydrolysis. Unfolded substrates or peptides bypass this step. Step 3- translocation of unfolded polypeptides or short peptide sequences into the degradation chamber, which also requires ATP –binding and -hydrolysis. Step 4- peptide bond cleavage resulting in the generation of small peptide products. Reiterative rounds of substrate unfolding and translocation are required to degrade substrates completely.

Transient kinetic techniques have been employed to elucidate the timing mechanism of bacterial Lon-mediated catalysis using the ATP-dependent peptide bond cleavage of FRETN 89-98 and its dansylated analog [75, 76]. Such studies have yet to be performed with human Lon. Although FRETN 89-98 is an unstructured decapeptide, its cleavage by Lon exhibits comparable ATP-dependency as the full-length native λN protein. Kinetic studies have been performed by monitoring peptide bond cleavage by fluorimetry and the first turnover of ATP hydrolysis by radiometry under identical reaction conditions, thereby permitting the determination of catalytic rate constants. ATP hydrolysis occurred within the same time frame as substrate translocation prior to peptide bond cleavage. Thus, substrate translocation is the rate-limiting step in FRETN 89-98 cleavage, which is coupled to ATP hydrolysis. As substrate translocation and peptide bond cleavage is sustained by the non-hydrolyzable analog AMP-PNP (albeit with reduced efficiency), and inhibited by ADP, it is proposed that Lon undergoes conformational changes when bound to different adenine nucleotides, which power the movement of substrate into the proteolytic chamber [75].

4. Selective protein degradation by mitochondrial Lon in yeast and mammals

4.1. Degradation of misfolded, unassembled and damaged proteins

Lon-mediated proteolysis is essential for mitochondrial function and cellular homeostasis. Studies in S. cerevisiae demonstrate that yeast strains lacking the homolog of Lon referred to as Pim1p fail to grow on non-fermentable carbon sources, suffer large mtDNA deletions, exhibit reduced ATP-dependent proteolysis and accumulate electron dense aggregates in the mitochondrial matrix [7, 8]. Some of these phenotypes are also observed in S. pombe strains lacking the Lon protease [77]. Early clues that Pim1p participates in mitochondrial protein quality control arose form the finding that unassembled subunits of the F₁ ATPase and mitochondrial ribosomal proteins (MRPs) are degraded in control yeast strains expressing Pim1p but not in Δpim1 strains lacking the protease [7, 78]. Like the Δpim1 mutant, the matched control strains carry large mtDNA deletions, and therefore do not produce mitochondrially-encoded subunits of respiratory complexes or mitochondrial ribosomes. In the absence of these proteins, partner subunits imported into the mitochondrial matrix fail to assemble properly and are selectively degraded by Pim1p. For some misfolded heterologous proteins, Pim1p-mediated proteolysis requires the cooperation of the mitochondrial chaperones Ssc1p (mtHsp70) and Mdj1p [79], or Hsp78p [80–82]. These mitochondrial chaperones mediate protein disaggregation and maintain substrates in an unfolded state, thereby promoting Pim1p-mediated proteolysis.

In S. cerevisiae and the filamentous fungi Podospora anserina, Pim1p/Lon plays an important role in the defense against environmental stressors such as oxidative stress and elevated temperature [77, 84, 86, 89]. Steady state levels of Pim1p are increased by the oxidizing agent H₂O₂ or by heat shock at 42°C, whereas the mitochondrial inner membrane m-AAA proteases, Yme1p and Yta10p show little or no change in protein expression under these conditions [86]. A Δpim1 strain is growth inhibited when grown in the presence of H₂O₂ or menadione, which induce reactive oxygen species (ROS) [84, 86]. When Pim1p is overexpressed, the aggregation of mitochondrial aconitase (Aco1p) is prevented [86]. In P. anserina, the constitutive overexpression of Podospora Lon (PaLon) leads to lower levels of carbonylated aconitase and carboxymethylated proteins, reduced secretion of hydrogen peroxide and increased resistance to oxidative stress. An interesting observation in P. anserina is that overexpressing PaLon extends lifespan with no apparent defects in respiration, growth or fertility [89].

S. cerevisiae provides a simplified system for studying Pim1p/Lon function as this organism lacks ClpP. Global proteomic analyses using mitochondrial extracts isolated from Δpim1 yeast and control strains have identified protein substrates of Pim1p [83, 84]. The first proteomics studies identified seven metabolic enzymes as Pim1p substrates- Ilv1, Ilv2, Lsc1, Lys4 and Yj1200c [83, 85], Prx1 and Ilv3 [86]. The import of radiolabeled precursor proteins into isolated control or Δpim1 mitochondria was performed to determine the kinetics of substrate degradation. In control mitochondria containing Pim1p, ~50–90% of the newly imported precursor proteins are degraded after 4 hours in an ATP-dependent manner, whereas in Δpim1 mitochondria only ~20–50% of imported proteins are degraded [83, 85, 86]. Several of the Pim1p substrates identified contain a prosthetic group or cofactor. Lys4, Ilv3 and Yj1200c are known Fe/S cluster-containing proteins; Ilv1 requires pyroxidal pyrophosphate as a cofactor, and Ilv2 requires thiamine pyrophosphate. Although Lsc1 contains neither a prosthetic group nor cofactor, it is a subunit of a heterodimeric enzyme complex. One possibility is that these proteins are susceptible to Pim1p-mediated proteolysis if they are improperly associated with their respective cofactors or partner proteins. Another possibility is that some cofactors such as Fe/S clusters are not only vulnerable to oxidative damage but they can also promote oxidative stress when destabilized, leading to the generation of oxygen radicals through Fenton chemistry. Notably however, the Fe/S cluster protein mitochondrial aconitase (Aco1p) does not accumulate in Δpim1 mitochondria [83, 84], nor is its turnover significantly altered upon import into isolated Δpim1 mitochondria [83]. Consistent with these findings, a knockout of Lon in S. pombe does not exhibit altered aconitase activity or increased levels of reactive oxygen species (ROS) [77]. Another proteomics study was undertaken to identify oxidized proteins that accumulate in Δpim1 mitochondria [84]. An oxidation index value was assigned to the various accumulated proteins, and was defined as fold increase in oxidized protein/total protein in Δpim1 versus control mitochondria. The proteins with values >1.15, were considered Pim1p substrates: Ald4 (1.15) Atp2 (1.29), Ldp1 (1.54), Sod2 (1.19), Pdb1 (1.99), Hsp60 (2.09), Ilv5 (5.22). Ilv5 was shown previously to be a substrate of Pim1p [80]. In this study, oxidized Aco1p was not amongst the proteins accumulating in Δpim1 mitochondria.

In mammalian cells, Lon-mediated proteolysis has been implicated in the selective degradation of oxidized aconitase [87, 90]. In one study, partially purified mitochondrial aconitase was tritium-radiolabeled (³H-aconitase) and treated with or without H₂O₂ at millimolar concentrations and then dialyzed before reacting with mitochondrial extracts or purified Lon. Results suggest that Lon selectively degrades the ³H-aconitase preparation exposed to 2.5 – 5 mM H₂O₂. Resistance to Lon-dependent proteolysis of ³H-aconitase treated with ≥10 mM H₂O₂ is observed, possibly as a result of increased protein aggregation and hydrophobicity [87]. In addition, the down-regulation of Lon in human fibroblasts correlates with decreased breakdown of the ³H-aconitase preparation [87]. The relationship between Lon protein levels and oxidized aconitase was examined in skeletal muscle of young versus old mice that were either wild-type (Sod2^+/+) or heterozygous (Sod2^−/+) for the mitochondrial manganese superoxide dismutase. Lon protein levels are lower in tissue extracts from old animals regardless of Sod genotype, as well as from young animals that are Sod2^−/+; decreased Lon protein is associated with increased levels of carbonylated protein presumed to be aconitase [90].

Although human Lon may degrade oxidatively damaged proteins, results also show that Lon itself is vulnerable to oxidative inactivation. In isolated non-synaptic rat brain mitochondria, peroxynitrite (ONOO⁻) blocks the ATP-stimulated protease activity of Lon by 75%, whereas ATP-independent activity is inhibited by 45% [91]. In addition, purified Lon is inactivated by H₂O₂ treatment [15]. The ATPase activity of purified Lon is blocked by 80% at 100 μM H₂O₂, and by 100% at >500 μM. In addition, yeast mitochondrial Lon is also strongly inhibited by the alkylating agent N-ethylmaleimide (NEM) [92]. Inhibition of mitochondrial Lon by NEM likely occurs by alkylation of cysteine residues resulting in enzyme inactivation. Thus, more thorough studies are required to evaluate the redox regulation of Lon and its quality control function in response to oxidative stress.

4.2. Protein quality control in the mitochondrial matrix- a division of labor in metazoans?

Although the role of Lon in protein quality control is one of its best demonstrated physiological functions, interestingly, it does not appear to be required for mediating the unfolded protein response in mitochondria (UPR^MT), nor to be up-regulated by this pathway. In cultured mammalian cells and worms, ClpXP plays a central role in controlling and responding to the UPR^MT (Fig. 5) [4–6, 93, 94]. By contrast, Lon does not appear to be an essential participant in this stress response pathway [93, 95, 96]. In cultured mammalian cells, the accumulation of an aggregation-prone protein within the mitochondrial matrix leads to the transcriptional up-regulation of ClpP [93, 94, 96]. The ClpP gene carries mitochondrial unfolded protein response elements (MURE1 and MURE2) in its promoter region, whereas the LONP1 gene lacks these promoter sequences. In C. elegans, studies show that ClpXP is required to initiate the UPR^MT stress response pathway leading to the transcriptional up-regulation of UPR^MT genes in the nucleus such as ClpP and the mitochondrial DnaJ-like protein Tid1 [4–6] (see Fig. 5 and legend). By contrast, Lon does not play a notable role in UPR^MT, as knocking down the worm homolog has no effect on this cell stress response pathway [95].

Fig. 5. Regulation of Lon and ClpXP in the ER and mitochondrial unfolded protein responses.

(1) Consequences of the unfolded protein response in the ER (UPR^ER)(1a) The accumulation of unfolded or misfolded proteins in the ER lumen activates PERK (protein kinase RNA-like endoplasmic reticulum kinase), which phosphorylates and thereby inhibits the translation initiation factor eIF2α. (1b) A block in eIF2α-dependent translation leads to the selective reduction of cytosolic protein synthesis, which includes proteins imported into mitochondria. (1c) The reduction in proteins imported from the cytosol may result in a stoichiometric imbalance of mitochondrial proteins that must assemble to form functional complexes. (1d) Signals originating from the cytosol or the mitochondrion may transcriptionally up-regulate the LONP1 gene by an unknown mechanism. (1e) Alternatively, accumulation of unfolded proteins in the ER directly activates the translation of the transcription factor XBP1 (X-box binding protein 1), which translocates to the nucleus and may activate LONP1 expression by binding to a putative UPRE (unfolded protein response element), which has yet to be identified. (1f) Hypoxia is a physiological inducer of ER stress that has been shown to up-regulate Lon expression by the activation of HIF-1α (hypoxia-inducible factor 1α), which binds to HREs (hypoxia response element) within the promoter of LONP1. (1g) The resulting increase of Lon may function to re-establish mitochondrial homeostasis.

(2) Consequences of the unfolded protein response in the mitochondria (UPR^MT). (2a) Accumulation of unfolded or misfolded proteins that tax the protein quality control capacity of mitochondria lead to the UPR^MT, which is initiated by ClpXP-dependent degradation of protein substrates thereby generating peptides that are effluxed from the matrix to the cytosol by the HAF-1 transporter located in the mitochondrial inner membrane. (2b) The effluxed peptides activate the transcription factors (ZC376.7 and possibly the CHOP/C/EBPβ heterodimer), which then translocate into the nucleus, binding to MURE and CHOP sequences up-regulating the expression of UPR^MT related genes encoding ClpP, mtHsp70, mtHsp60, mtHsp10, and mtDnaJ. (2c) The increased expression of ClpP relieves mitochondrial stress and re-establishes mitochondrial homeostasis.

Although Lon does not appear to have a major function in the UPR^MT, its role in the UPR^ER has been suggested. Lon expression is up-regulated in response to protein misfolding or increased protein burden in the ER, which is induced by either tunicamycin- a glycosylation inhibitor, thapsigarin- an ER Ca²⁺ ATPase inhibitor, or brefeldin A- an inducer of retrograde traffic of Golgi proteins to the ER; all of these agents activate UPR^ER [3]. Results show that UPR^ER-stimulated Lon overexpression is dependent on the PKR-like ER-localized eIF2α kinase (PERK), which is specifically activated by UPR^ER [3]. In addition, hypoxia, which also induces ER-stress, has been shown to up-regulate Lon (see next section). One potential reason why the UPR^ER may up-regulate Lon in mitochondria is that the PERK-dependent block in protein synthesis decreases protein import into mitochondria, leading to the stoichiometric imbalance of proteins produced in the cytosol versus the mitochondrion, thereby causing protein misfolding, misassembly and aggregation. Lon-mediated degradation may function to relieve the load of abnormal mitochondrial proteins caused by the UPR^ER.

4.3. Lon-mediated proteolysis as a regulatory mechanism

In addition to protein quality control, Lon-dependent degradation also operates to regulate mitochondrial pathways by terminating the activity of specific proteins, or by modulating the activity of protein complexes by selective subunit degradation. During hypoxia, Lon participates in remodeling cytochrome c oxidase (COX) holoenzyme by selectively degrading the Cox4-1 subunit [2]. When oxygen availability is low, the hypoxia-inducible transcription factor HIF-1α binds to hypoxia response elements (HRE) in the promoter of the LONP1 gene leading to Lon up-regulation and Cox4-1 degradation. At the same time, HIF-1α also up-regulates an alternate isoform- Cox4-2, which is assembled into the COX complex replacing Cox4-1. Results demonstrate that Cox4-2 containing complexes are better optimized for transferring electrons and increasing the efficiency of respiration in hypoxic cells [1–3]. In yeast and mammalian cells, both wild type Lon and a protease-deficient mutant of Lon promote the assembly of COX complexes [3, 78], suggesting that Lon may have a chaperone-like function in the assembly and perhaps disassembly of proteins. It is tempting to speculate that this function of Lon is exploited in the selective degradation of Cox4-1 and coordinate assembly of Cox4-2, which occurs during the hypoxia-induced remodeling of the COX holoenzyme. In cells of the adrenal cortex, placenta and gonads, Lon degrades the steroidogenic acute regulatory protein (StAR), which mediates the rate limiting steps in steroid hormone synthesis [97]. StAR is required for the transfer of cholesterol from the mitochondrial outer membrane to the first steroidogenic enzyme, which is localized on the matrix side of the inner membrane [97]. The precise mechanism underlying StAR-dependent cholesterol transfer remains unclear. However, when its mitochondrial matrix targeting sequence is deleted, StAR can still carry out its essential function, which appears paradoxically to occur outside of mitochondria [98, 99], and its translocation into the mitochondrial matrix terminates its activity [97]. Within mitochondria, Lon is principally responsible for the degradation StAR [97] and may thus provide a mechanism for regulating the synthesis of steroid hormones. In addition, recent work demonstrates that Lon degrades alpha aminolevulinic acid synthase 1 (ALAS-1), which is the rate-limiting enzyme in the heme biosynthesis pathway. Accumulation of intracellular heme down-regulates ALAS-1 by a negative feedback loop that involves transcriptional repression, mRNA degradation, and blockade of mitochondrial ALAS-1 protein import [100–102]. Increased levels of heme also promote the Lon-dependent degradation of mature ALAS-1 in the mitochondrial matrix [103], and may thereby contribute to regulating heme biosynthesis. In Drosophila cells, Lon degrades transcription factor A of mitochondria (TFAM) such that knocking down Lon coordinately increases TFAM protein levels and mtDNA copy number. Conversely, Lon overexpression reduces TFAM levels as well as mtDNA copy number [18]. In Lon knockdown cells the levels of transcription factor B2 of mitochondria (TFB2M) were also elevated, suggesting that it may also be a substrate of Lon [18]. These results support the concept that Lon-mediated proteolysis functions as a mechanism for regulating essential rate-limiting and metabolic processes within mitochondria. However, the substrate recognition mechanisms that specify Lon-mediated degradation of Cox4-1, StAR, ALAS-1 and TFAM have yet to be determined.

5. Lon as a mitochondrial DNA binding protein

The association of Lon with DNA is a property of the protease that is conserved from bacteria to mammalian mitochondria [13–17, 104–111]. Purified bacterial Lon binds both single-stranded DNA (ssDNA) and double- stranded DNA (dsDNA), with sequence-specificity as shown in one study [110] and non-specifically in other studies [106–108, 111, 112]. By contrast, purified mammalian mitochondrial Lon interacts only with ssDNA, and binds to nucleotide sequences within the mitochondrial genome in a sequence-specific and strand-specific manner [13, 14, 16, 17]. Human Lon associates with mitochondrial DNA (mtDNA) sequences bearing at least 4 contiguous guanine residues that have a propensity for forming G-quadruplexes, which are four-stranded intra- or inter- molecular structures with a tetrad arrangement of guanines [13, 14, 16]. Every ~150 nucleotides within the heavy-strand of mtDNA there are 4–6 adjacent guanine residues that can potentially interact to form such higher ordered structures. Biochemical and biophysical analyses demonstrate that human Lon binds to mtDNA sequences that form parallel G-quadruplexes, and that this interaction exhibits a large negative heat capacity change, which is a thermodynamic signature of sequence-specificity [16].

The binding of human Lon to mtDNA is a physiological function of the protease as demonstrated by mtDNA immunoprecipitation (mIP) experiments in cultured cells [15]. One principal advantage of mIP is that it provides a snapshot of protein interaction to mtDNA in a physiological setting. Human Lon binds preferentially to the displacement loop (D-loop) or control region of mtDNA [15] (Fig. 6), which is the only non-coding region of the mitochondrial genome containing the promoters for heavy- and light- strand transcription and one origin of replication. Lon binding is also observed at other regions of mtDNA (Fig. 6). Bioinformatic analyses of the mtDNA sequences pulled down by Lon in mIP assays reveals a G-rich consensus sequence consistent with DNA foot-printing of purified Lon bound to a ssDNA oligonucleotide [15]. Consistent with biochemical data, the regions of mtDNA associated with Lon correspond to regions bearing predicted G-quadruplex DNA (Fig. 6). Not all regions of mtDNA predicted to adopt G-quartet structures were found to co-purifiy with Lon. At least four variables influence Lon binding to G-rich sites in the mIP assay: 1) G-quadruplex formation is influenced by the single- or double- stranded state of mtDNA, 2) G-quadruplex binding by other proteins may block Lon interaction, 3) Lon binding to specific G-quartets may depend on flanking sequences, and 4) the affinity or total number of Lon binding sites with a specific mIP amplification region is variable.

Fig. 6. Relative locations within the mitochondrial genome of predicted G-quadruplexes and Lon binding.

The positions of G-quadruplexes within the heavy-strand of human mtDNA (NC_001807) were predicted using Quadfinder Version 1 (http://miracle.igib.res.in/quadfinder/) [16]. The regions of mtDNA bound by Lon in cultured HeLa cells was determined by mtDNA immunoprecipitation (mIP) [15].

The importance of G-quadruplexes within the mitochondrial genome has not been studied. However, G-quadruplexes have been demonstrated in the promoter regions of nuclear genes in humans, and may either enhance or block transcription [113, 114]. Such G-quartets are also predicted within the E. coli genome, showing predominance within promoters of genes linked to transcription, secondary metabolite biosynthesis, and signal transduction [115]. One can envisage that association of Lon with G-quadruplexes within the mitochondrial genome provides a mechanism for recruiting the protease to specific sites where it can degrade or process proteins involved in mtDNA and mtRNA metabolism. The association of Lon with these higher order DNA structures may impart pause sites that function as checkpoints during replication, transcription or translation. In addition, the binding of mtDNA by Lon may directly inhibit or promote the processivity or timing of replication and/or transcription.

Human Lon has been identified as a protein component of mitochondrial nucleoids [12], which are multi-protein complexes of mtDNA maintenance factors that support genomic stability, inheritance and expression. Lon may interact with the mitochondrial genome through direct as well as indirect interactions. Lon is thus poised to degrade and regulate proteins mediating mtDNA replication and expression, and to remodel mitochondrial nucleoids. Although mtDNA binding by Lon in cultured cells has been established, the function of Lon at the mitochondrial genome remains unclear. Results show that Lon degrades transcription factor A of mitochondria (mt-TFA, TFAM), and modulates mtDNA copy number [18]. Lon-mediated degradation of TFAM may be a protein quality control mechanism for eliminating misfolded or damaged TFAM, and/or a regulatory mechanism for controlling mtDNA metabolism.

6. Mitochondrial Lon and links to disease

Protein misfolding and aggregation are associated with a variety of human diseases and aging [116–120], which have been linked to the increased generation of reactive oxygen species (ROS) and perturbations in cellular redox status. Mitochondrial proteins are at an increased risk of oxidative damage and conformational defects as they are located in close proximity to the respiratory chain, which is a powerful source of ROS. In addition, genetic mutations that cause the misfolding of mitochondrial proteins likely induce the expression or activity of mitochondrial AAA⁺ proteases. Mitochondrial Lon is implicated in various diseases or disease models where its expression levels are either up- or down- regulated. Table 1 summarizes some of these experimental findings.

Table 1.

Expression of mitochondrial Lon in various human diseases

Disease	Cause	Fold change	Method of analysis	References
Myoclonic epilepsy and ragged-red fibers (MERRF) syndrome	mtDNA mutation A8344G in tRNALys	2 fold increase	2D gel, mass spectrometry	[127]
Myopathy, encephalopathy, lactic acidosis, stroke-like episodes syndrome (MELAS)	mtDNA mutation A3243G in tRNALeu(UUR)	2–3 fold increase	mass spectrometry and immunoblotting	[128]
Friedreich ataxia (FRDA)	defect in mitochondrial frataxin	>2.5 fold increase	immunoblotting and RT-PCR	[88]
Non-small-cell lung cancer (NSCLC	multiple (tobacco, asbestos, pollution, gene mutation etc.)	2–5 fold increase	immunoblotting	[129]
Lipodystrophy	side effect of HAART	increase	microarray and immunoblotting	[130]
Familial amytrophic lateral sclerosis (fALS)	Mutations in copper- zinc superoxide dismutase (SOD1) gene	3-fold decrease	2D gel, mass spectrometry	[131]
Hereditary spastic paraplegia (SPG13)	axon degeneration in specific motor neurons, gene mutations	~40–50% decrease	immunoblotting and RT-PCR	[132]

Outlook

Understanding the structural dynamics and functional versatility of Lon and other mitochondrial AAA⁺ proteases are broad challenges for the future. The importance of energy-dependent proteases is not restricted to protein quality control but also extends to regulating constitutive metabolism and adaptive responses to cellular stress and transformation. The development of pharmacological tools as well as cell culture and animal models to study the biochemical and biological mechanisms of mitochondrial AAA⁺ proteolysis is a high priority in elucidating how these molecular machines support the integrated network of metabolic reactions at the cellular and organismal level.

Highlights.

• Homology modeling of human Lon predicts a hexameric complex.

• Role of Lon as a protein quality control protease.

• Lon-mediated proteolysis in the regulation of cellular metabolism.

• Potential role of Lon in mitochondrial DNA maintenance and expression.

• Lon expression changes in human disease.

Abbreviations

AAA⁺

ATPases associated with diverse cellular activities

ER

endoplasmic reticulum

ATP

adenosine triphosphate

ADP

adenosine diphosphate

MPPα

mitochondrial processing peptidase α subunit

TFAM

transcription factor A mitochondria

ALAS-1

alpha aminolevulinic acid synthase 1

StAR

steroidogenic acute regulatory protein

UPR

unfolded protein response