Control of plastidial metabolism by the Clp protease complex 1

20 Plant metabolism is strongly dependent on plastids. Besides hosting the 21 photosynthetic machinery, these endosymbiotic organelles synthesize starch, fatty 22 acids, amino acids, nucleotides, tetrapyrroles, and isoprenoids. Virtually all 23 enzymes involved in plastid-localized metabolic pathways are encoded by the 24 nuclear genome and imported into plastids. Once there, protein quality control 25 systems ensure proper folding of the mature forms and remove irreversibly 26 damaged proteins. The Clp protease is the main machinery for protein degradation 27 in the plastid stroma. Recent work has unveiled an increasing number of client 28 proteins of this proteolytic complex in plants. Notably, a substantial proportion of 29 these substrates are required for normal chloroplast metabolism, including 30 enzymes involved in the production of essential tetrapyrroles and isoprenoids such 31 as chlorophylls and carotenoids. The Clp protease complex acts in coordination 32 with nuclear-encoded plastidial chaperones for the control of both enzyme levels 33 and proper folding (i.e. activity). This communication involves a retrograde 34 signaling pathway, similarly to the unfolded protein response previously 35 characterized in mitochondria and endoplasmic reticulum. Coordinated Clp 36 protease and chaperone activities appear to further influence other plastid 37 processes, such as the differentiation of chloroplasts into carotenoid-accumulating 38 chromoplasts during fruit ripening. 39


Introduction
Plastids are central metabolic factories for plant cells.Besides being the site of carbon, nitrogen, and sulfur assimilation, they synthesize starch, fatty acids, amino acids, nucleotides (purine and pyrimidine bases), tetrapyrroles (including heme and chlorophyll), and isoprenoids (such as carotenoids, and the side chain of chlorophylls).In fact, their classification into different plastid types largely depends on the metabolites that they store: chloroplasts contain chlorophylls, chromoplasts accumulate carotenoids, amyloplasts are starch-rich plastids, and elaioplasts are specialized in storing lipids (Jarvis and Lopez-Juez, 2013;Rolland et al., 2012;Sakamoto et al., 2008).
Similar to mitochondria, plastids are endosymbiotic organelles that retain their own genome (plastome).The Arabidopsis thaliana plastome contains 45 RNA-coding genes and 87 protein-coding genes, but only two of them are involved in processes other than plastome gene expression or photosynthesis: accD encodes the carboxytransferase beta subunit of the acetyl-CoA carboxylase complex (which catalyzes the first committed step in fatty acid biosynthesis) and clpP1 encodes a proteolytic subunit of the stromal Clp (Caseinolytic protease) complex (Sato et al., 1999).Therefore, virtually all proteins involved in plastidial metabolism (including enzymes) are encoded by the nuclear genome and imported into plastids.Most of these plastid-targeted proteins bear an N-terminal targeting signal called plastid transit peptide.They are synthesized in the cytosol and transported into plastids in an energy-dependent process mediated by molecular machines (translocons) located in the outer and inner envelope membranes (Sakamoto et al., 2008).At the inner membrane their transit peptide is cleaved and the mature forms of the imported proteins are properly folded, assembled, and sorted to their particular subplastidial destination by a separate targeting process.Inside plastids, protein quality control (PQC) systems formed by chaperones and proteases promote correct protein folding and remove irreversibly damaged proteins.To ensure an appropriate supply of functional proteins, retrograde (plastid-to-nucleus) signaling pathways adjust nuclear gene expression to particular plastid needs, e.g. in response to environmental challenges (Chan et al., 2016;Grimm et al., 2014).
The multimeric Clp complex is the main stromal protease Among the several proteases found in plant plastids, the serine-type Clp protease complex is the main machinery for processive degradation of proteins in the plastid stroma (Nishimura et al., 2016;Nishimura and van Wijk, 2015).One of the roles of this housekeeping protease appears to be the removal of proteins that do not fold properly or lose their native conformation and become misfolded in the stroma (Nishimura and van Wijk, 2015;Zybailov et al., 2009).Because misfolded proteins tend to aggregate, the Clp protease complex might also prevent the formation of potentially toxic protein aggregates (Llamas et al., 2017;Nishimura and van Wijk, 2015;Zybailov et al., 2009).Additional PQC roles suggested for the Clp protease include the survey of proteins translocated into plastids, clearing proteins that remain unprocessed or misfolded after import but also removing proteins that aggregate and become stuck in the import channel (Flores-Perez et al., 2016;Huang et al., 2016;Sjogren et al., 2014).
Similar to other prokaryotic-type ATP-dependent plastidial proteases (such as FtsH and Lon), the Clp complex contains two differentiated domains: a barrelshaped proteolytic core with the catalytic center positioned inside the chamber to prevent unspecific protein degradation and a chaperone ring that delivers the protein substrates.But unlike FtsH and Lon, in the Clp protease these two characteristic domains are separated into individual multiprotein subcomplexes (Figure 1).The subunit composition of the Clp protease is similar in all plastid types (D'Andrea et al., 2018;Peltier et al., 2004).The proteolytic core (ClpRP) is formed by two asymmetric rings (P and R) and stabilized by plant-specific ClpT1 and ClpT2 proteins (Kim et al., 2015;Sjogren and Clarke, 2011).The P ring is formed by a 1:2:3:1 ratio of proteolytically active subunits ClpP3, ClpP4, ClpP5 and ClpP6.
Proteins entering the catalytic ClpRP barrel are assumed to be degraded into peptide fragments (5-10 residues) that are ejected from the core cavity through lateral pores (Peltier et al., 2004).
Substrate access to the proteolytic core is controlled by a hexameric ring presumably formed by homooligomers of the Hsp100-type chaperones ClpC1, ClpC2, and ClpD (hence referred to it as ClpCD).Mutants defective in ClpC1 are pale-green but knock-out lines of ClpC2 or ClpD show a wild-type phenotype, supporting a major housekeeping role for the former.These chaperones (particularly ClpC1) can directly recognize substrates and unfold them for continuous translocation into the proteolytic chamber in an ATP-dependent manner (Nishimura et al., 2016;Nishimura and van Wijk, 2015;Peltier et al., 2004).
In other cases, substrates are recognized by adaptors and then delivered to the chaperones (Figure 1).Several pathways involving such adaptors are known.One of them involves the recognition of client proteins by a binary adaptor formed by ClpF and ClpS1 (Nishimura et al., 2015).The current model proposes that ClpF, ClpS1, or a ClpF-ClpS1 complex recognizes the protein client and forms a ClpF-ClpS1-substrate complex that then binds to the Hsp100 chaperone ring of the Clp protease complex.A second pathway involves another chaperone type, Hsp70, which can interact with ClpC1 to synergistically promote the unfolding process (Shi and Theg, 2010;Su and Li, 2010).Hsp70 chaperones use J-proteins as adaptors that guide them to protein clients and confer substrate specificity.After recognition of specific proteins by particular J-protein adaptors, the substrates are delivered to Hsp70, which then interacts with ClpC1 for unfolding and subsequent ClpRPmediated degradation (Pulido et al., 2016;Pulido et al., 2013).The possibility of additional ways of targeting proteins to ClpC1, ClpC2, and ClpD chaperones remains open (Pulido et al., 2017).It is generally assumed that specific signals for degradation (named degrons) might be exposed in protein substrates that need to be removed.The observation that the N-terminus of proteins is a major determinant of protein turnover in prokaryotic and eukaryotic organisms led to propose the N-end rule, which states that particular N-terminal residues act as triggers for whole protein degradation.However, the identification of degrons and the relevance of the N-end rule for the recognition of Clp protease targets by the ClpCD ring chaperones or their adaptors (J-proteins/Hsp70, ClpF/ClpS1) remains an open question (Rowland et al., 2015).

Several metabolic enzymes are clients of the Clp protease
Despite the central relevance of the Clp protease in PQC and plastidial proteostasis, only a few proteins have been conclusively shown to be degraded by this proteolytic complex.As true Clp protease clients, they (a) show increased protein levels (but not transcripts) in mutants defective in Clp protease activity, (b) interact with chaperones of the ClpCD ring or their adaptors, and (c) are degraded at a slower rate in mutants of the chaperone or/and catalytic domains of the Clp complex.Interestingly, most of the proteins experimentally demonstrated to meet these criteria have a role in plastidial metabolism.Besides the thylakoid membrane copper transporter PAA2/HMA8, a protein with stromal domains that is degraded by the Clp protease by mechanisms that are not yet fully understood (Tapken et al., 2015), four enzymes have been found to be Clp substrates in Arabidopsis (Figure 1).The identified enzymes and the pathways in which they participate are described below.We also discuss other plastidial enzymes that accumulate upon interference with Clp protease activity but are not yet confirmed to be direct clients of this proteolytic complex.

Glutamyl-tRNA reductase and tetrapyrrole biosynthesis.
The enzyme glutamyl-tRNA (GluT) reductase (GluTR) catalyzes an early step in the tetrapyrrole biosynthesis pathway (Figure 1).GluT, produced after ligation of Glu to tRNA(Glu), is reduced to Glu-1-semialdehyde by GluTR and then transaminated to 5-aminolevulinic acid (ALA), the first committed precursor of the pathway.Tetrapyrrole products are central for chloroplast metabolism as they are required for photosynthesis (chlorophylls), nitrogen and sulfur assimilation (siroheme), redox reactions (heme), photoperception (phytochromobilin), and even retrograde signaling (Chan et al., 2016;Grimm et al., 2014).Tetrapyrrole biosynthesis is tightly regulated to balance the production of chlorophylls and heme and hence prevent the accumulation of metabolic intermediates that can potentially cause severe photooxidative damage.Thus, chlorophyll biosynthesis needs to be downregulated in darkness to prevent excessive accumulation of protochlorophyllide in the dark (Meskauskiene et al., 2001).
As a key regulatory enzyme of the tetrapyrrole pathway, GluTR is controlled at multiple levels.Among them, interaction of the N-terminal region of GluTR with the GluTR binding protein (GBP) results in association of part of the stromal enzyme to the thylakoid membrane (Czarnecki et al., 2011).It is proposed that this mechanism allows two separate pools of GluTR for the separate production of ALA for heme in the thylakoid membrane and for chlorophylls in the stroma.The Nterminal domain of GluTR harbors aggregation-prone motifs, but interaction with the plastidial chaperone SRP43 efficiently prevents aggregation, thus enhancing the stability of the protein and promoting its localization to the thylakoid membrane (Wang et al., 2018).Furthermore, this N-terminal region can also interact with the ClpF and ClpS1 adaptors and the ClpC1 chaperone for Clp protease-mediated degradation of the enzyme (Apitz et al., 2016;Nishimura et al., 2015;Nishimura et al., 2013).A truncated GluTR protein lacking the N-terminal region remains in the stroma and is more stable during prolonged darkness, causing an enhanced accumulation of protochlorophyllide and a necrotic phenotype upon reillumination (Apitz et al., 2016).Consistent with the role of the N-terminal domain as a degron, the rate of GluTR proteolysis in the dark is increased in the absence of GBP (i.e. when the degron is exposed) but decreased in mutants impaired in ClpC1 or the proteolytic core subunit ClpR2 (Apitz et al., 2016).While GluTR accumulates at higher levels in ClpC1 and ClpS1 mutants, the degradation rate of the enzyme in the dark appears not to be altered when ClpS1 is missing.It is therefore concluded that ClpC1 might have a more relevant role than ClpS1 (or ClpF) to recognize and direct GluTR into the Clp proteolytic core for degradation (Apitz et al., 2016).A proposed model (Wang et al., 2018) suggests that the interaction of SRP43 with unfolded GluTR after plastid import might ensure proper enzyme folding and prevent aggregation.Folded GluTR could then interact with GBP for thylakoid membrane targeting.It is likely that competition for the N-terminal domain between GluTR interactors (SPR43 and GBP) and Clp adaptors or/and chaperones, but also structural changes triggered by metabolite binding, misfolding or aggregation, might play a role in determining the degradation rate of this important enzyme.
Strikingly, the N-terminus is relevant for the Clp-mediated degradation of GluTR not only in plants but also in bacteria (Wang et al., 1999), suggesting a remarkable evolutionary conservation of the mechanisms controlling GluTR turnover.

Chlorophyllide a oxygenase and chlorophyll biosynthesis.
The final steps of chlorophyll biosynthesis (Figure 1) involve the incorporation of molecular oxygen into the C7-methyl group of the tetrapyrrole ring (porphyrin) of chlorophyllide a to produce chlorophyllide b followed by the esterification of isoprenoid-derived phytyl diphosphate to both chlorophyllides to form chlorophylls a and b, respectively (Kim et al., 2013b).Chlorophyll a occurs in antenna complexes and reaction centers, whereas chlorophyll b is absent from reaction centers.Chlorophyll b levels influence the antenna size of photosystem II and are controlled by the stability of chlorophyllide a oxygenase (CAO), the enzyme that catalyzes the conversion of chlorophyllide a to b (Sakuraba et al., 2007;Yamasato et al., 2005).Degradation of CAO when chlorophyll b is present requires the Nterminal domain (or A domain) of the enzyme (Yamasato et al., 2005) and is impaired in ClpC1-defective mutants (Nakagawara et al., 2007).The current model is that chlorophyll b may modulate the structure of the A domain either by direct interaction or via other factors.The induced change of conformation would then expose a degron for ClpC1-mediated unfolding and eventual ClpRP-mediated degradation of the whole enzyme (Sakuraba et al., 2009).CAO localizes in thylakoid and envelope membranes but lacks obvious membrane-spanning domain (Eggink et al., 2004;Yamasato et al., 2005).It is therefore possible that interaction of the N-terminal domain of CAO with ClpC1 (or an adaptor) occurs at the periphery of these plastidial membranes.
The isoprenoid phytol chain of chlorophylls is synthesized from precursors derived from the stromal methylerythritol 4-phosphate (MEP) pathway.In fact, a strict coordination between the tetrapyrrole and the MEP pathways is essential to prevent the formation of oxidative metabolic intermediates (Kim et al., 2013b).MEP-derived prenyl diphosphates isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) are also used for the production of other plastidial isoprenoids involved in photosynthesis (tocopherols and carotenoids as photoprotectants, and prenylquinones such as plastoquinone and phylloquinones for electron transport) and signaling (including several types of retrograde signals, hormones, and volatiles for environmental interactions) (Rodriguez-Concepcion and Boronat, 2015).The control of protein turnover has a major impact on MEP pathway enzymes including deoxyxylulose 5-phosphate (DXP) synthase (DXS), DXP reductoisomerase (DXR), hydroxymethylbutenyl 4-diphosphate (HMBPP) synthase (HDS), and HMBPP reductase (HDR) (Figure 1).All these enzymes accumulate at higher levels in Arabidopsis plants with decreased Clp proteolytic activity such as mutants defective in either ClpC1 or ClpRP subunits (Table 1) (Flores-Pérez et al., 2008;Kim et al., 2013a;Kim et al., 2009;Nishimura et al., 2013;Rodriguez-Villalon et al., 2009;Rudella et al., 2006;Sauret-Güeto et al., 2006;Welsch et al., 2018;Zybailov et al., 2009).However, only DXS has been experimentally shown to be a direct Clp substrate based on the three criteria described above (Figure 1).DXS can either directly interact with ClpC1 or be delivered to this chaperone via J20, a J-protein adaptor of Hsp70 (Pulido et al., 2016), as detailed below.DXR is more stable and does not rely on J20 for degradation (Perello et al., 2016;Pulido et al., 2016).Instead, removal of excess DXR takes place by packaging the enzyme in vesicles formed by engulfing stromal fractions that likely retain Clp protease activity (Perello et al., 2016).The accumulation of MEP pathway enzymes does not change in mutants defective in ClpF or ClpS1 adaptors with only one exception: HDS (Nishimura et al., 2015;Pulido et al., 2016).However, HDS levels do not increase but decrease in the mutants (Table 1) (Nishimura et al., 2015), i.e. the opposite that it would be expected if the ClpF/ClpS1 pathway delivers this enzyme to the Clp complex.
Beyond Arabidopsis, work in other systems has shown that the accumulation of DXS and other MEP pathway enzymes is directly or indirectly controlled by the Clp protease in tobacco (Nicotiana tabacum) chloroplasts (Moreno et al., 2018), tomato (Solanum lycopersicum) chromoplasts (D'Andrea et al., 2018), the plastidlike apicoplasts of the malaria parasite Plasmodium falciparum (Florentin et al., 2017) and even bacterial cells (Ninnis et al., 2009).
Consistent with the major contribution of DXS to the control of the MEP pathway flux (Pokhilko et al., 2015;Wright et al., 2014), its activity is regulated by several post-translational mechanisms (Rodriguez-Concepcion and Boronat, 2015), including Clp-mediated degradation (Flores-Perez et al., 2008;Llamas et al., 2017;Perello et al., 2016;Pulido et al., 2016;Welsch et al., 2018).A shortage of MEP pathway products upregulates DXS enzymatic activity and promotes the accumulation of higher protein levels (Ghirardo et al., 2014;Han et al., 2013;Pokhilko et al., 2015).It is possible that binding of IPP or DMAPP to DXS causes a conformational change that inactivates the enzyme and renders it more susceptible to degradation (Pokhilko et al., 2015).Stress episodes can also disrupt the native structure of DXS, resulting in protein misfolding, and eventually aggregation in chloroplasts.The high aggregation propensity of DXS was also observed in cyanobacteria, the evolutionary ancestors of chloroplasts (Kudoh et al., 2017).Work in Arabidopsis has unveiled a PQC mechanism to deal with the unwanted accumulation of inactive DXS proteins (Figure 1).The J-protein adaptor J20 specifically binds misfolded forms of DXS and delivers them to the Hsp70 chaperone for either proper folding or degradation (Pulido et al., 2013).The fate of Hsp70-bound enzyme appears to depend on the relative abundance of Hsp100 chaperones that synergistically contribute to the unfolding process.Under normal growth conditions, inactive Hsp70-bound DXS proteins are delivered to ClpC1 and unfolded prior entrance into the ClpRP proteolytic chamber to be degraded.
Environmental episodes causing protein folding stress (e.g.heat) do not cause changes in the levels of ClpC1 but induce the accumulation of ClpB3, another plastidial member of the Hsp100 family.Unlike ClpC1 and the other Hsp100 chaperones associated to the ClpCD complex, ClpB3 lacks the IG(F/L) tripeptide (also named P-loop motif) required for interaction with the ClpRP core subunits but harbors a domain that allows efficient interaction with Hsp70 chaperones (Flores-Perez et al., 2016;Kim et al., 2001;Peltier et al., 2004;Pulido et al., 2016).
Therefore, stress episodes activate the ClpB3-dependent pathway to unfold Hsp70-bound DXS into the stroma for spontaneous refolding and reactivation (Pulido et al., 2016).This model implies that although the refolding strategy (via ClpB3) is likely favored, Clp-mediated degradation (via ClpC1) is the housekeeping pathway to remove inactive DXS.Most likely, both share the same goal, i.e. to prevent protein aggregation in the plastid and maintain the enzymes in a catalytically active form.The observation that double mutants defective in ClpB3 and ClpR2 were unable to survive under autotrophic conditions (Zybailov et al., 2009) illustrates the key relevance of the seemingly antagonistic refolding and degradation pathways for plant life.The common regulation of both the MEP and the tetrapyrrole pathways by the Clp protease (Figure 1) likely contributes to finely balance the supply of metabolic precursors for the production of chlorophylls (Kim et al., 2013b).

Phytoene synthase and carotenoid biosynthesis.
The activity of the MEP pathway enzymes generates prenyl diphosphates used by the enzyme phytoene synthase (PSY) to produce phytoene, the first committed intermediate of the carotenoid pathway.Carotenoids are essential for photoprotection of the photosynthetic apparatus against excess light, act as precursors of hormones and retrograde signals, and function as natural pigments in non-photosynthetic organs such as flowers and fruits (Rodriguez-Concepcion et al., 2018).Yeast two-hybrid experiments followed by bimolecular fluorescence complementation assays in plant cells showed that PSY, the main rate-determining enzyme of the pathway (Fraser et al., 2002), can physically interact with ClpS1 and several Clp protease subunits, including ClpC1 and ClpD (Welsch et al., 2018).
However, only the absence of ClpC1 results in increased enzyme levels as a consequence of reduced PSY turnover rate, suggesting a dispensable contribution of ClpS1 or ClpD in the delivery of the enzyme to ClpRP-mediated degradation (Figure 1).As expected, PSY enzyme levels are also increased in mutants defective in individual subunits of the ClpRP proteolytic core such as ClpP4 and ClpR1.PSY stability further increases in the presence of ORANGE (OR), a J-likeprotein with chaperone activity that also promotes PSY enzymatic activity (Park et al., 2016;Pulido and Leister, 2018;Welsch et al., 2018;Zhou et al., 2015).OR might play a similar role as ClpB3 in the case of DXS, i.e. promote correct PSY folding (and hence enzymatic activity) and prevent its misfolding and eventual aggregation by the Clp protease (Figure 1).Similar to PSY, other carotenoid biosynthetic enzymes overaccumulate in mutants defective in ClpC1 or ClpRP core subunits (Figure 1) (Welsch et al., 2018).While further experiments will be required to conclude whether these enzymes are true Clp protease targets, the available evidence suggests that Clp-dependent mechanisms contribute to coordinate the MEP pathway and the carotenoid pathway by influencing the accumulation of multiple enzymes of both pathways.
Besides possible secondary effects eventually impacting enzyme levels, this control mechanism involves the direct degradation of some enzymes, including those catalyzing the main rate-determining steps (DXS and PSY).Because the Clp protease also regulates the production of chlorophylls via direct control of GluTR and CAO turnover (Figure 1), this proteolytic complex might further coordinate chlorophyll and carotenoid biosynthesis for an efficient photosynthetic function in chloroplasts.Recent results further suggest that the Clp protease might ensure a proper supply of MEP-derived precursors for carotenoid biosynthesis in different plastid types.Downregulation of Clp protease activity during tomato ripening led to increased levels of both DXS and PSY enzymes in the carotenoid-accumulating chromoplasts of ripe fruit (D'Andrea et al., 2018).Importantly, Clp-defective fruits developed aberrant chromoplasts with disorganized membranes and displayed an orange color (instead of red) because they accumulated enhanced levels of the pro-vitamin A carotenoid β-carotene (D'Andrea et al., 2018).These data indicate that Clp protease activity controls not only carotenoid biosynthesis (via DXS and PSY stability) but also storage (via plastidial ultrastructure).It is conceivable that during normal fruit ripening Clp protease function might facilitate the turnover and removal of many other enzymes and structural proteins that lose their functionality or that are not further required as chloroplasts differentiate into chromoplasts.

Other enzymes and metabolic pathways.
Genetic interference with the accumulation of Clp subunits perturbs the stoichiometry in the complex and negatively impacts its protein degradation activity.Based on the premise that decreased Clp proteolytic activity should lead to increased accumulation of its protein targets, quantitative proteomic approaches have been employed for the identification of potential Clp protease clients.Initially, the steady state proteome of wild-type Arabidopsis plants was compared with that of mutants defective in particular subunits of the complex (Kim et al., 2013a;Kim et al., 2009;Rudella et al., 2006;Sjogren et al., 2006;Stanne et al., 2009;Zybailov et al., 2009).While components of metabolic pathways impacting the production of starch, fatty acids, amino acids, isoprenoids, and thiamine were found to be upregulated in the mutants, this strategy per se could not distinguish between primary and downstream effects (e.g.loss of photosynthetic activity and downregulation of Calvin cycle enzymes) associated with the reduction of Clp protease activity (Nishimura and van Wijk, 2015).Further biochemical approaches using mutants defective in the adaptors ClpF and ClpS1 went a step forward and identified GluTR as a target of the complex (Nishimura et al., 2015;Nishimura et al., 2013).Other potential clients identified in these works are the enzymes deoxyarabinoheptulosonate 7-phosphate synthase (DHS) and chorismate synthase (CS), which catalyze the first and last reaction of the shikimate pathway, respectively.While DHS isoforms were consistently found to be upregulated in mutants defective in ClpC1 and ClpRP core subunits, the results for CS are much less clear (Table 1).The product of the shikimate pathway, chorismate, is converted into aromatic amino acids such as Phe or Tyr, which are also precursors for the biosynthesis of phenylpropanoids (e.g.lignins, flavonols, and anthocyanins) and isoprenoids such as tocopherols, plastoquinone, and phylloquinones (Figure 1).
More recently, induced downregulation of Clp complex subunits in tobacco led to propose additional enzymes as potential Clp protease clients based on their overaccumulation at early time points after triggering repression of Clp proteolytic activity (Moreno et al., 2018).Among them, two enzymes involved in thiamine biosynthesis appear as good candidates: THI1 and THIC (Figure 1).These two enzymes catalyze the first steps in the biosynthesis of the two moieties required for the production of thiamine, thiazole (THI1) and pyrimidine (THIC) (Goyer, 2010).Supporting the conclusion that they might be true Clp protease targets, the Arabidopsis THI1 and THIC proteins are accumulated at higher levels in mutants defective in ClpC1 or ClpR2 (Table 1) (Nishimura et al., 2013;Zybailov et al., 2009).In the case of THI1, ClpS1 mutants contained increased protein levels in one study (Nishimura et al., 2013) but decreased in another (Nishimura et al., 2015), preventing to reach a conclusion on the possible role of ClpS1 in delivering this enzyme to degradation.In any case, it is interesting to note that thiamine diphosphate (vitamin B1) serves as a cofactor for many enzymes (including DXS; Figure 1) and hence represents an important node in the control of the plant cell metabolism.

Plastidial proteostasis requires communication with the nucleus
Highly reproducible patterns deduced from quantitative proteomics analyses of Clp-defective mutants include a strong reduction of photosynthetic capacity (in part due to loss of thylakoid membrane homeostasis) and a systematic upregulation of all stromal chaperone systems, with most consistent increases typically observed for the stromal Hsp100 chaperone ClpB3 (Table 1).It was proposed that accumulation of ClpB3 and stromal protein folding machineries in the mutants was a response to deal with protein aggregation likely occurring when Clp protease activity was defective (Nishimura and van Wijk, 2015;Zybailov et al., 2009).By downregulating the expression of the plastome-encoded ClpP1 subunit of the ClpRP core in the unicellular alga Chlamydomonas reinhardtii, it was observed that these chaperones and small heat shock proteins such as Hsp21 increased at both transcript and protein levels (Ramundo et al., 2014).A role for retrograde signaling was later suggested in plants to increase the supply of nuclear-encoded plastidtargeted chaperones when loss of Clp protease activity caused protein folding stress in the plastid (Nishimura and van Wijk, 2015).More recently, the existence of a mechanism upregulating the expression of nuclear genes encoding plastidtargeted chaperones such as Hsp21, Hsp70, and ClpB3 was demonstrated in Arabidopsis (Llamas et al., 2017).The mechanism was designated chloroplast unfolded protein response (cpUPR), based on its conceptual similarity to those previously reported in mitochondria and endoplasmic reticulum (Hwang and Qi, 2018;Lin and Haynes, 2016).Although the term "unfolded" might be misleading and it would be most appropriate to name it protein folding stress response, we will keep using the term cpUPR to be consistent with the literature.
Studies using DXS as a model showed that plastid proteostasis involves a dynamic balance between degradation and repair of structurally compromised proteins.In non-stressed conditions, the Clp protease (via ClpC1) removes DXS proteins that become non-functional (e.g. when losing their native conformation).However, protein folding stress can occur when Clp protease activity becomes compromised and cannot deal with the accumulation of misfolded proteins.For example, malfunctioning of the plastome gene expression system can lead to altered levels of the catalytic ClpP1 subunit (Flores-Pérez et al., 2008), causing reductions in Clp protease activity (because of distortions in the stoichiometry of the complex) that eventually lead to protein aggregation (Llamas et al., 2017).Protein folding stress in the plastid somehow sends an unknown retrograde signal to upregulate the expression of the nuclear gene encoding the transcription factor HsfA2 (Llamas et al., 2017).HsfA2 directly binds the promoters of genes encoding ClpB3 and other plastidial chaperones, inducing their expression.As a result, more of these chaperones are made and sent to plastids to alleviate protein folding stress in this organelle.Higher levels of plastidial ClpB3 then promote the disaggregation of DXS, i.e. the recovery pathway.It is expected that upregulating the supply of plastidial chaperones would also increase the global folding capacity of the plastid, which should contribute to effectively remove protein aggregates, promote correct folding, and restore proteostasis.
Consistent with the conclusion that the basic PQC mechanisms are conserved among different plastid types, the downregulation of Clp protease activity during tomato fruit ripening also resulted in increased accumulation of chaperones in the chromoplasts of ripe fruit (D'Andrea et al., 2018).Specifically, genes encoding tomato homologues of ClpB3 and OR were upregulated in Clp-defective fruit, leading to increased levels of the corresponding proteins and subsequent protection of DXS and PSY enzymes, respectively, against degradation.Interestingly, OR chaperones not only promote PSY stability and enzymatic activity but also act as positive regulators of chromoplast differentiation and repressors of carotenoid degradation (Chayut et al., 2017;Park et al., 2016;Zhou et al., 2015).
It is proposed that Clp protease-regulated expression of nuclear genes encoding chaperones such as OR and ClpB3 might be a compensatory mechanism triggered when Clp protease activity is compromised during fruit ripening (D'Andrea et al., 2018).When Clp protease activity becomes limiting, the resulting defective protein turnover might impair chromoplast differentiation.The accumulation of non-functional proteins, however, would trigger a cpUPR to eventually mitigate protein folding stress (e.g. via ClpB3) and restore chromoplast development (e.g. via OR).The observation that, similar to that described in Arabidopsis (Llamas et al., 2017), tomato HsfA2 regulates the expression of plastidial chaperones to increase the protection against protein folding stress (Fragkostefanakis et al., 2015), suggests that several components of the organelle UPR might be conserved among different species.A better understanding of how the Clp protease coordinates with chaperones such as ClpB3 and OR in different plastid types should contribute to more rational engineering approaches for improved production of carotenoids and other plastidial metabolites.

Open questions
While significant progress has been achieved in the last few years on the identification of Clp protease protein substrates, we still have an incomplete map of the plastidial pathways impacted by this major proteolytic machinery.In addition, substrates involved in the storage of plastidial metabolites remain to be identified.
For example, Clp-defective mutants typically show increased levels of proteins associated to plastoglobules, which are thylakoid-derived compartments that store isoprenoids and other lipids (van Wijk and Kessler, 2017).However, it is unclear whether this is a primary effect of reducing the turnover of Clp protein substrates or a secondary consequence derived from interferences with thylakoid membrane homeostasis or stress responses (Nishimura and van Wijk, 2015).Recent approaches using inducible lines have been useful to distinguish early from late accumulating proteins (Moreno et al., 2018;Ramundo et al., 2014).Clp-trapping strategies similar to those used in bacteria (Flynn et al., 2003;Trentini et al., 2016) were expected to also be useful to identify bona fide plant targets, despite the challenges derived from the much higher complexity and essential nature of the plastidial Clp protease.Recent experiments using in vivo tagged core complexes containing inactive ClpP3 and ClpP5 variants in Arabidopsis could not identify new Clp protease clients, proposedly because the bottleneck for degradation might not be the catalytic activity of the ClpRP core but substrate recognition and unfolding (Liao et al., 2018).In vivo substrate trapping through partial inactivation of ClpC1, ClpC2 or ClpD might therefore provide an alternative, more efficient strategy to identify protein clients.
To date, only ClpC1-dependent substrates have been described, suggesting that the other chaperones of the ClpCD ring (ClpC2 and ClpD) are partially redundant (Kovacheva et al., 2007;Nishimura et al., 2015).Strikingly, ClpD is found exclusively in stroma, whereas ClpC1 and ClpC2 are also present in association with envelope membranes for PQC of imported proteins (Sjogren et al., 2014).A related question is how the substrates of the Clp protease are selected for degradation, while preventing the removal of functional proteins and off-targets.
While the relevance of the N-end rule for chloroplast protein turnover remains unclear (Rowland et al., 2015), it is likely that the first step in the degradation pathway could be exposure of a degron after a conformational change caused by metabolite binding (as in the case of CAO) or by misfolding (as in the case of GluTR, DXS and PSY).Degrons could also be exposed after removal of proteins bound to the motifs required for degradation (as in the case of GluTR).Then, the degron could be recognized by the ClpF/ClpS1 tandem or the J-protein/Hsp70 system and delivered to ClpC1 or any other Hsp100 component of the Clp chaperone ring for unfolding and ClpRP-mediated proteolysis.Rather than a particular amino acid sequence, degrons might comprise a secondary structure and/or a post-translational modification (phosphorylation, acetylation, etc), which enormously complicate their identification.For instance, arginine phosphorylation was found to direct substrates for degradation by the Clp protease in bacteria (Trentini et al., 2016).Furthermore, it is likely that some enzymes could be degraded by different proteases in the plastid (Nishimura et al., 2016).Considering the protease network as a whole (Majsec et al., 2017) would also be required to fully understand how plastidial metabolism is regulated inside plastids in response to developmental and environmental cues.
Finally, the molecular pathway connecting Clp protease defects with enhanced expression of genes for plastid-targeted chaperones remains unknown.Recent results have shown that GUN1, a central regulator of plastid to nucleus communication, is degraded by the Clp protease via ClpC1 (Wu et al., 2018).
Another protein involved in tetrapyrrole-dependent retrograde signaling, GUN5, is also upregulated in Arabidopsis mutants defective in ClpC1 and Clp protease activity (Table 1).A role for the Clp protease in controlling the turnover of retrograde signaling mediators such as GUN1 and GUN5 is expected to represent a powerful tool in the communication of plastidial protein folding stress to the nucleus.However, neither GUN1 nor GUN5 appear to be required for the upregulation of genes encoding plastidial chaperones such as Hsp21, Hsp70, and ClpB3 that occurs soon after triggering protein aggregation in chloroplasts by interfering with Clp protease activity (Llamas et al., 2017).Alternative good candidates to participate in this process are the DNA-interacting factors WHIRLY (pTAC1/WHY1 and pTAC11/WHY3) and pTAC12/HEMERA, which have been located in plastidial transcriptionally active chromosomes (pTACs) or nucleoids but also in the nucleus (Melonek et al., 2016).These proteins are more abundant in Clp-defective plants (Table 1) (Moreno et al., 2018) and display the same size in chloroplasts and nuclei, suggesting that their accumulation in plastids with a compromised Clp proteolytic activity might cause their translocation to the nucleus to regulate nuclear gene expression.Future work should contribute to confirm this hypothesis and identify other dual-localized factors potentially involved in the cpUPR-associated retrograde signaling pathway that allows plants to overcome protein folding stress in the plastid.et al., 2009Kim et al., 2009Kim et al., 2013Nishimura et al., 2013 Nishimura et al.,  Numbers indicate the ratio of protein levels in the indicated Arabidopsis mutant relative to the wild-type (fold-change)

Zybailov
Colors indicate when protein levels in the mutant are statistically higher (blue) or lower (red) compared to the wild-type higher, proteins detected only or preferentially in the mutant nd, not detected

Figure 1 .
Figure 1.Degradation of plastidial enzymes by the Clp protease complex.Enzymes are indicated in green, adaptor systems are represented in purple, and proteins that prevent degradation are shown in black.Green arrows represent metabolic pathways.Solid black arrows mark confirmed interactions and dashed black arrows indicate potential interactions.The structure of the Clp protease complex and its components is also shown. 2015