Redox control of protein degradation

1Aerobic metabolism has the advantage of a better energy yield, but at the cost of generating reactive oxygen species (ROS). Indeed, the leakage of superoxide from mitochondrial respiratory chain complexes I and III constitutes one of the major sources of ROS production. Other sources of harmful ROS include unfolded protein response at the endoplasmatic reticulum (ER), and oxidant byproducts generated at peroxisomes. Moreover, evolution has used oxygen to modify certain proteins, now termed redox switches, as a cell signaling mechanism in survival and regeneration among other pathways. This clever use of ROS is best exemplified by NADPH oxidases (NOX), situated in the plasma membrane, whose main role is to generate superoxide and ultimately hydrogen peroxide (H2O2) as second messengers.

2Cells have efficient enzymatic and non-enzymatic strategies to modulate redox signaling and maintain redox homeostasis. In addition, antioxidants are also obtained from exogenous sources, with the diet as the main supplier. However, many pathological conditions or the normal decline in cell homeostasis related to ageing lead to a gradual imbalance between ROS formation and degradation and result in detrimental alterations of macromolecules. Sulfur containing amino acids, cysteine and methionine, are responsible for reversible and irreversible modification of proteins. In addition, proteins can form adducts with oxidizing byproducts. Fig. 1 summarizes oxidative modifications of sulfur containing amino acids.

3In this review, we summarize the most relevant findings about the degradation of oxidized proteins, here termed oxyproteins, and the impact of oxidative stress on proteolytic cell systems, and gene expression.

4The UPS participates in the degradation of soluble proteins in cytosol and nucleus. The central core of the UPS is the 20S proteasome, which is present in animals, plants and bacteria. This 700 kDa-multisubunit protease is highly effective in the proteolytic degradation of misfolded or otherwise altered proteins. In eukaryotes, the 20S proteasome binds one or two regulatory 19S complexes to give rise to the 26S proteasome, changing its activity and specificity towards native folded proteins.

5The main difference between ubiquitin (Ub)-dependent and Ub-independent proteasomal degradation is ATP-requirement. Based on this, the fact that ATP addition to cell lysate had no effect or even decreased proteasomal degradation of oxyproteins, led to the conclusion that these proteins are degraded by the 20S proteasome independently of both regulatory 19S complex and Ub. In accordance to this, the 26S proteasome and the ubiquitinating machinery are much more sensitive than its 20S core to oxidative stress. Thus, 20S proteasomal degradation is unaffected by H2O2 concentrations of up to 5 mM, while ATP-dependent degradation by the 26S proteasome begins to decline at 400 µM, and is completely abolished at 1 mM.

6On the other hand, some oxyproteins might require ubiquitin-dependent 26S proteasome degradation. Thus, Dudek and co-workers showed that carbonyl-containing proteins, a direct measure of protein oxidation, are selectively removed by ubiquitinating machinery. Furthermore, it has been reported that inhibition of USP14, deubiquitinating enzyme that associates with the proteasome, may enhance the clearance of oxyproteins, and thus, cellular resistance to oxidative challenges. Based on the extensive review of Aiken and co-workers and Shang and Taylor changes in proteasomal activity upon oxidative pressure are illustrated in Fig. 2.

7Apart from its role in the degradation of oxidatively damaged proteins, the proteasomal system is also involved in highly controlled degradation of proteins acting as redox switches. This is the case of certain transcription factors such as hypoxia-inducible factor-1 (HIF-1), Nuclear factor (erythroid-derived 2)-like 2 (NRF2) or nuclear factor kappa B (NFκB) as illustrated in Fig. 3.

8The UPS itself is subjected to alterations derived from oxidative stress. As shown in Fig. 4, ubiquitinating enzymes are inactivated by disulfide bond formation, S-nitrosylation, and S-glutathionylation. The E3 ligase adapter kelch-like ECH-associated protein 1 (KEAP1) is an excellent example (Fig. 3B). The proteolytic activity of the proteasome requires intact sulfhydryl groups of Cys residues in the beta subunits of the calatytic 20S core for catalysis. These Cys are largely responsible for the susceptibility of the proteasome to oxidative insults. Moreover, both core 20S and regulatory 19S complexes are targets of oxidative modification, including 4-hydroxynonenal modification, carbonylation, S-glutathionylation, glycoxidation, as well as ADP-rybosilation and phosphorylation, which are indirect ROS-induced modifications (Fig. 4). Considering that the 26S proteasome is much more sensitive to oxidative stress than its 20S core, it is not surprising that almost all these modifications were detected on the 19S regulatory subunits leading to decreased activity of the 26S proteasome. Although the majority of these modifications result in decreased 20S proteasome activity, some of them show the opposite effect. Thus, it has been shown that S-glutathionylation of two Cys residues on α5 subunits of 20S in yeast, increases proteasome activity via opening of the annulus. Also, ADP-ribosylation of Glu, Asp, or Lys of nuclear 20S proteasome by poly[ADP-ribose] polymerase 1 (PARP1), increases its chymotrypsin-like activity. The PARP1 is activated in the response to single-strand and double-strand breaks of DNA. Besides ADP-ribosylation, phosphorylation seems to be important post-translation modification that may regulate proteasome activity upon oxidative stress, but this still needs to be experimentally confirmed.

9Up-regulation of proteasome genes is coupled with enhanced capacity of the cell to cope with detrimental effects of oxidative insult. Thus, it has been shown that increased expression of the UMP1, proteasome assembly protein, improves cell viability upon exposure to different oxidants, probably via up-regulation of the proteasome β subunits. It is interesting that NFκB and activator protein-1 (AP1) do not seem to participate in regulation of most proteasome genes while Nuclear factor (erythroid-derived 2)-like 1 (NRF1) and NRF2 appear to have a major role. Thus, it seems that regulation of the proteasomal and NRF2 activity is bidirectional, i.e. in the basal conditions the proteasome down-regulates NRF2 activity by degradation, while upon oxidative stress, released NRF2 up-regulates proteasomal activity and protects the cell from the accumulation of oxyproteins. This effect may explain why pre-treatment of cultured neurons with low doses of proteasome inhibitors lead to increased, not decreased, proteasomal activity.

10Autophagy refers to any intracellular process that leads to degradation of cytosolic components inside lysosomes. There are three different types of autophagy in mammals: macroautophagy, chaperone mediated autophagy (CMA) and microautophagy. Macroautophagy (often referred to as autophagy) is a process in which a portion of the cytosol is surrounded by a growing double membrane (autophagosome) which eventually fuses with the lysosome (autophagolysosome), where the content will be degraded. Cytosolic substrates with a KFERQ-like motif can be selectively recognized, translocated and degraded inside the lysosome by CMA. Finally, the direct invagination of the lysosomal membrane can introduce cytosolic portions into the lysosome in a type of autophagy called microautophagy.

11Mild oxidative stress conditions, redox-mediated signaling or oxidative modification of macromolecules up-regulate the autophagy flux, leading to elimination of non-functional and potentially damaging protein aggregates and affected organelles (Fig. 5A and B). Several groups have reported redox-modification of autophagy components with a general effect on autophagy induction, although some exceptions can be found (Table 1). Moreover, appropriate GSH levels are essential for basal autophagy. Indeed, GILT (gamma-interferon inducible lysosomal thiol reductase)-deficient fibroblasts show decreased GSH levels and increased autophagy, associated with the up-regulation of the ERK signaling pathway and with nuclear translocation of high-mobility group protein B (HMGB). Starvation-induced autophagy decreased levels of intracellular GSH, which correlated with increased autophagy flux in different carcinoma cell lines.

12The importance of redoxtasis for CMA was first reported with the discovery of the degradation of IκB by CMA, as this process was prevented upon treatment with antioxidants. Further studies confirmed that CMA can be activated under mild oxidative stress circumstances and that protein oxidation facilitates the degradation by CMA. Indeed, when known CMA substrates or a pool of cytosolic proteins where incubated with increasing amounts of oxidizing agents, their degradation by CMA was accelerated. One explanation is that oxidative modification of these proteins causes their partial unfolding, not only promoting the exposure of hidden recognition motifs to the chaperones but also facilitating its translocation. Another possibility is that oxidation of certain residues creates a previously inexistent KFERQ-like motif. Increased levels of the CMA mediators LAMP-2A, lys-hsc70, Hip and hsp90 have been reported in this context, suggesting that increased CMA-mediated protein degradation is due to greater binding and up-take of substrates into lysosomes. In accordance with this, CMA down-regulation (through silencing the expression of LAMP-2A) compromised cell viability upon exposure to oxidant and pro-oxidant factors (H2O2, paraquat or cadmium). Moreover, the ectopic expression of LAMP-2A in liver of aged rodents (aimed to prevent the age-dependent decrease in CMA) leads to reduced levels of intracellular oxidized and aggregated proteins, improves the response to stressors and preserves organ function.

13Recently, an oxidative stress-induced form of microautophagy was described, arising from mitochondria-derived vesicles (MDVs). MDVs transported to the lysosomes are enriched in oxidized proteins and are generated in response to oxidative stress. Protein degradation mediated by MDVs does not require activation of the autophagy key actors Atg5, Rab9 or Beclin-1, but it does require the protein kinase PINK1 and the ubiquitin E3 ligase Parkin. This suggests that similar mechanisms apply as in mitophagy, but at restricted patches of the mitochondrial surface. MDVs may thus be generated as a first order of defense to handle redox-damaged proteins in order to prevent functional failure of the organelle.

14Apart from an initial and rapid increase in the autophagy flux mediated by post-translational protein modifications, a delayed and extended autophagy response relies on the activation of specific transcription factors such as NRF2, NFκB, p53 or FOXO3a (Fig. 5D). NRF2, the master regulator of redox homeostasis, has been recently related to the modulation of autophagy. The autophagy-related protein p62/SQSTM1 binds to KEAP1 at the NRF2 binding site, thus promoting NRF2 release from KEAP1 and enabling NRF2-dependent gene expression. Binding of p62 to KEAP1 is favored upon phosphorylation of p62 in an mTORC1 dependent manner. One dimer of p62 can bind to both KEAP1 and LC3, resulting in its degradation. Moreover, the p62 gene has been shown to be a target of NRF2, creating a positive feedback loop where p62 modulates NRF2 protein levels, which in turn control p62 gene expression. Interestingly, NRF2 may modulate other autophagy players as it has been reported to reduce phospho-TAU levels in a mouse model of AD through inducing the expression of NDP-52. Regarding CMA, the levels of LAMP2A mRNA in the livers of rats treated with paraquat were significantly higher than in untreated rats, although degradation of LAMP2A was only slightly reduced and the distribution between matrix and membrane was not altered. The exact transcription factors involved in this up-regulation have not yet been elucidated, but ROS regulated T-cell receptor-induced LAMP2A expression may rely on RISP, DUOX1 or NFAT activities.

15The induction of autophagy by ROS can, in turn, modulate ROS levels. Thus, autophagy activation with trehalose was accompanied by the up-regulation of glutathione levels, supporting the antioxidant role of autophagy. Another example is the p53-inducible TIGAR protein, whose ability to inhibit autophagy has been correlated with the suppression of ROS, with no clear effects on the mTOR pathway and is p53 independent.

16Covalent cross-links, disulfide bonds, hydrophobic interactions and heavily oxidized aggregates are less effectively degraded by the proteasome and are redirected to degradation by the autophagy system. If these irreversible aggregates become resistant to hydrolases, they accumulate in the form of lipofuscin. Moreover, cross-linking of proteins at the lysosomal membrane can increase proton permeability and luminal pH. As a consequence of this, lysosomal hydrolases are less effective at degrading their substrates, which will favor the accumulation of lipofuscin. Lipofuscin-loaded human fibroblasts exhibit reduced autophagy.

17Oxidative damage to the lysosomal membrane may lead to the leakage of hydrolases into the cytosol, resulting in further cellular damage (Fig. 5C). The direct damage of the lysosomal membrane by specific oxyproteins, such as ferritin or low density lipoproteins (LDL) in macrophages, has been reported. Moreover, lysosomes themselves can be a source of ROS. Thus, inhibition of autophagy with methylamine, chloroquine or 3-methyladenine prevented ROS formation. Autophagy degradation of catalase under caspase inhibition conditions has also been proposed as a mechanism contributing to the accumulation of ROS. The degradation of iron-containing macromolecules or organelles (ferritin or mitochondrial electron transport complexes for example) leads to the intra-lysosomal accumulation of Fe2+. Besides, H2O2 resulting from different cellular processes can either enter the lysosome or be generated in its lumen as a result of degraded mitochondria. Both species may react (Fenton reaction), resulting in the formation of the extremely reactive hydroxyl radicals.

18The UPR is an adaptive response to ER stress meant to reduce the protein folding load in the ER and increase ER folding capacity. This is achieved by a general suppression of translation, retrotranslocation of proteins and their degradation by ERAD (ER associated protein degradation), degradation of ER-associated mRNAs, expansion in ER volume and increased synthesis of ER chaperones (reviewed in ). Oxidative stress leads to protein unfolding or misfolding in the endoplasmic reticulum (ER) which in turn produce ER stress and activate the UPR.

19Three membrane-associated proteins have been identified for sensing ER stress in eukaryotes: activating transcription factor 6 (ATF6), pancreatic ER eIF2α kinase (PERK, also double-stranded RNA-activated protein kinase-like ER kinase), and inositol-requiring kinase 1 (IRE1). The luminal domain of each sensor is bound to the chaperone 78 kDa glucose-regulated protein (GRP78/BiP) in the resting state. GRP78/BiP dissociates upon ER stress to bind unfolded proteins, leading to the activation of the three sensors.

20Upon release from GRP78/BiP, ATF6 reaches the Golgi apparatus, where it is sequentially cleaved and the fragment generated (ATF6f) is capable to enter the nucleus and induce transcription of UPR genes that collectively may reduce ER stress, such as XBPI, BiP or CHOP. ATF6 contains two conserved Cys residues that can form intra- and intermolecular disulfide bonds. Upon ER stress, only the reduced monomer form can reach the Golgi apparatus to be cleaved and act as a transcription factor, and therefore redox control of the disulfide bonds in ATF6 is crucial for its export.

21Most of the PERK signaling is mediated through phosphorylation of the alpha subunit of the eukaryotic initiation translation factor 2 (eIF2α), which transiently inhibits protein translation. However, selective translation of the transcription factor ATF4 and subsequent expression of its target genes is induced under these circumstances (Fig. 6A). ATF4 induces expression of genes like HO-1 and p62/SQSTM1 that are known to be induced upon oxidative stress and can have a role in the modulation of redoxtasis. Moreover, ATF4-induced CHOP expression results in the expression of Ero1α, an enzyme that contributes to oxidation in the ER (Fig. 6A). Ero1α causes Ca2+ leakage from the ER, activating CaMKII in the cytosol, which in turn induces NOX2 and causes oxidative stress. Moreover, in cooperation with FoxO3, CHOP has been implicated in the transcriptional repression of Bcl2 and transactivation of BIM and PUMA that leads to enhanced oxidant injury and apoptosis (Fig. 6A).

22PERK phosphorylates NRF2 promoting its dissociation from KEAP1, translocation to the nucleus and expression of its target genes. Different UPR genes have been reported to contain AREs in their promoter regions. Since NRF2 also promotes the expression of proteasome and autophagy-genes it is not surprising that tunicamycin-induced ER stress significantly up-regulates proteasomal activity.

23The transcription factor NRF1, also involved in the antioxidant response, localizes to the ER membrane and may undergo nuclear translocation upon deglycosilation or intra-membrane/proteasomal cleavage. NRF1 localization along with its ability to up-regulate proteasomal subunits after proteasome inhibition makes it reasonable to consider an NRF1 up-regulatory function of ERAD during ER stress. Tunicamycin treatment of myc-tagged NRF1 transfected cells resulted in higher levels of the 110 kDa Nrf1 fragment in the nucleus when compared with untreated cells. However, another group later reported that this isoform does not have the ability to transactivate ARE-containing genes.

24Activated IRE1 contains both a kinase and a RNAse domain. Its activation leads to mRNA splicing of an intron from transcription factor X box binding protein 1 (XBP-1) to trigger its translation. XBP-1 modulates expression of UPR target genes, including ER chaperons, glycosylation enzymes or ERAD components. Other RNAs are targeted through a process called regulated IRE1-dependent decay (RIDD) that reduces the amount of proteins in the ER. IRE1 kinase domain leads to the activation of the IRE1-TRAF2-JNK axis. Activation of NRF2 by MAPK is a controversial issue that appears to be context-dependent. However, in JNK-activated ER stress, the inhibition of JNK leads to NRF2 over-activation. These results point to NRF2 as a potential substrate of the IRE1-TRAF2-JNK pathway.

25The ER environment is highly oxidative because the ratio GSH/GSSH is 1:1–1:3 compared to 30:1–100:1 in the cytosol. This environment favors the formation of disulfide bonds necessary for correct protein folding in an ezymatic reaction catalized by protein disulfide isomerases (PDIs) (Fig. 6B). These enzymes also reduce improperly formed disulfide bonds in a process accompanied by glutathione consumption. Most PDIs contain at least one thioredoxin (Trx)-like catalytic domain. PDIs are oxidized by oxidoreductin proteins 1 (Ero1α and Ero1β in mammals), which regenerate themselves by transfering electrons to oxygen to produce H2O2. Peroxiredoxin IV metabolizes H2O2 into H2O. Disruption of redox balance in the ER can result in incorrect disulfide bond formation during protein folding, redox imbalance and oxidative stress.

26ROS production and oxidative stress can be considered an integral component of the UPR, triggering both transcriptional and post-translational responses which in turn lead to cell adaptation and survival or cell death by apoptosis.

27Although the physiological role of mammalian prion protein (PrPc) is not known, there is some evidence suggesting that it could play a role as endogenous ROS scavenger, protecting other structural and signaling proteins, because it has a high number of sulfhydryl groups in methionine residues. In addition, PrPc could be a redox switch since methionine sulfoxidation is involved in cell signaling.With ageing, there is a progressive accumulation of oxidized methionine residues in PrPc that contribute to protein misfolding, and participate in the transition from a monomeric globular form with α-helical content to a self-aggregating form with extended β-strand-rich structure with “infective” capacity to transmit the corrupted conformation to other native prion proteins. The transition of PrPc N-terminal region, from a random coil to a β-sheet structure transforms the soluble and protease-sensitive PrPc into the oxidized, insoluble and relatively protease-resistant PrPsc. PrPsc monomers and small oligomers induce nerve cell death after internalization and accumulation into the endolysosomal compartment where they cause lysosomal damage with subsequent proteolytic enzyme leakage and activation of caspase-dependent apoptosis. Autophagy efficiency is also attenuated, due to PrPsc protease-resistance, and aggregation of misfolded PrPsc also leads to mitochondrial failure. Assuming that the primary role of PrPc is to act as a global cell antioxidant, regulating the oxidative state of structural and signaling proteins, the conversion of soluble PrPc proteins to less soluble and aggregation prone oxidized PrPsc proteins could lead to a major depletion of the antioxidant PrPc pool thus leading to neuronal death.

28The key pathological features of Alzheimer’s disease are the presence of soluble amyloid β-peptides, for example Aβ(1–42), that accumulate in the intracellular and extracellular space and neurofibrillary tangles made of TAU protein aggregates. The amyloid β-peptides can be further modified into misfolded Aβ monomers, dimers, oligomers and intermediate products that are toxic, leading to cell death. The oxidation of Aβ(1–42) at the 35 methionine residue promoted by for example H2O2 or Cu2+, accelerates the production of toxic Aβ(1–42) products and further protein oxidation and lipid peroxidation. Compared to reduced Aβ(1–42), oxidized Aβ(1–42) is more resistant to degradation by autophagy and endosome–lysosome fusion which further contributes to its accumulation and toxic effects due to release of undigested contents into the cytosol. The toxicity of amyloid β-peptide derivatives decreases with protofibril and fibril formation and terminates in the formation of stable and inert amyloid plaques. Increased ROS formation, inside or outside the cell, favors the transition of Aβ monomers into toxic forms and also stimulates the breakdown of microtubule cytoskeleton by promoting zinc- or H2O2-induced TAU phosphorylation. Hyperphosphorylated TAU, together with the oxidized form of Aβ(1–42), leads to mitochondrial damage, subsequent reduction in ATP production, decreased mitochondrial potential, production of more ROS and finally cell death.

29Monomeric TAU is a natively unfolded and short-lived protein, thus being a good substrate for the 20S proteasome under normal conditions. Under stressing conditions it has been reported that the E3 ligase CHIP (Carboxy terminus Hsp70 interacting Protein), participates in TAU ubiquitlation leading to its degradation by the 26S proteasome. However, as indicated before, the 26S proteasome is sensitive to oxidative stress and loses efficacy in degradation of TAU, particularly when it has undergone posttranslational modifications (phosphorylation and oxidation among other modifications) and led to formation of toxic oligomers and aggregation. In fact, it has been reported that the proteasome activity is impaired in AD brains and that TAU-containing paired helical filaments inhibit the proteasome. Under these conditions TAU is degraded by p62-driven macroautophagy and LAMP2A-driven chaperone mediated autophagy. Of note, chaperone mediated autophagy is up-regulated under oxidative stress conditions, further suggesting its role in degradation of oxidized and aggregated TAU. A new mechanism of redox-dependent degradation of TAU has been described. The gene coding the autophagy adapter protein NDP52, contains several antioxidant response elements regulated by NRF2. In fact, in Nrf2-knockout mice, phosphorylated TAU accumulates in the brains concurrent with decreased levels of NDP52. In summary, TAU is a substrate of both the proteasome and the autophagy systems and both processes appear to be impaired in AD.

30α-Synuclein (α-SYN) is localized to presynaptic terminals in the central nervous system and modulates vesicular release of dopamine by attenuating its release under circumstances of repeated firing. In vivo, α-SYN can be present as a soluble unfolded protein that can aggregate into progressively less soluble oligomers, protofibrils and insoluble amyloid fibril form. This protein is normally degraded through the UPS, macroautophagy or CMA; but nitration and oxidative modifications slow down or even inhibit its proteasomal degradation. Besides, partial degradation can lead to its C-terminal truncation, favoring further aggregation. In addition, α-syn contains a KFERQ consensus sequence and oxidation of this protein (for instance, dopamine-modified α-syn) prevents its translocation into the lysosome lumen, acting as uptake blockers and preventing degradation of other CMA substrates. Attenuation of chaperone-mediated autophagy degradation leads to a compensatory increase in macroautophagy, accumulation of autophagosomes, and ultimately to cell death due to the release of undigested contents into the cytosol. Posttranslational modification of α-SYN, for example by phosphorylation, ubiquitination, nitration or oxidation, also reduces autophagosome degradation and promotes α-SYN oligomerisation. The toxic effect of these posttranslational modifications can be compounded by the propensity of α-SYN to bind to various molecules in neuronal cells. For example, Fe2+, dopamine or H2O2 oxidize methionine residues in the α-SYN monomer, that is the predominant form of oxidized α-SYN, and also promote the formation of stable oligomers with a resistance to fibrilisation that is proportional to the number of oxidized methionine residues. The oxidized α-SYN monomer can interact with lipids, as well as with proteins, changing their redox state and function thus sharing similarities with the oxidized Aβ(1–42). For example, the oxidized α-SYN monomer disrupts autophagy and also disrupts mitochondrial function.

31Recently, it has been shown that the Parkinson's associated protein DJ-1 binds to and inhibits the 20S proteasome and thus prevents the degradation of substrates such as a-syn or p53. Moreover, under oxidative stress conditions, DJ-1 induces the NRF2-dependent antioxidant response. Among the induced genes, NQO1 reinforces DJ-1 function by also inhibiting the 20S proteasome. However, NRF2 also induces the expression of 20S proteasome subunits. This robust regulation may be necessary to maintain proper 20S proteasome activity.

32Oxidative modification of proteins and proteolytic pathways compromise protein quality and cell viability. These events may be among the most relevant in driving protein toxicity in several pathologies as exemplified here for neurodegenerative diseases. It is now essential to find ways to prevent these effects through either reinforcing redox homeostasis or increasing the capacity of proteolytic systems.