Oxidative Modification of Proteins: From Damage to Catalysis,Signaling,and Beyond

1. Is protein carbonylation a regulatory mechanism?

At first glance, it is unlikely that protein carbonylation has a regulatory role since it has been considered an irreversible modification. However, data in the literature indicate a reversible mechanism of protein carbonylation, as evidenced by a decreased pool of carbonylated proteins, which cannot be attributed to protein degradation or de novo protein synthesis (579, 715). However, these data need to be confirmed with further investigations.

2. Specificity/susceptibility

Whether there are proteins more prone to carbonylation has been a matter of investigation. In several studies working with aged organisms, some proteins, in different organs of the same organism and different organisms, were found repetitively carbonylated (90). Among those proteins, 20% are involved in glucose metabolism, the tricarboxylic acid cycle, and the electron transport chain. Also, heat shock proteins (HSPs) and elongation factors of protein synthesis were carbonylated during aging, from bacteria to humans (90). Elegant studies from Nyström's group showed that increased mistranslation in Escherichia coli achieved utilizing specific mutations and drugs was paralleled by increased protein carbonylation (181). Moreover, mutants harboring hyperaccurate ribosomes exhibited drastically attenuated protein oxidation during growth arrest (34). As interpreted by the authors, verified protein carbonylation in aging cells might be a consequence of reduced transcriptional and translational fidelity independent of increased oxidant formation.

The crosstalk between protein carbonylation and the methionine sulfoxide reductase system has been proposed (448). There are data showing that methionine sulfoxide formation precedes carbonylation as organisms (yeast and mice) lacking the methionine sulfoxide reductase A enzyme (MsrA) accumulate carbonylated proteins. The hypothesis is that the structural change in the protein carrying the methionine oxidation increases the vulnerability of the protein to carbonylation. While this mechanism sounds attractive, no data demonstrate that the increased occurrence of carbonyl proteins (e.g., in aging and neurodegenerative diseases [NDs]) is associated with decreased methionine sulfoxide repair in humans.

3. What could explain specificity?

Protein location in specific subcellular compartments and their abundance would be essential parameters to preview protein carbonylation specificity. Nevertheless, classical work by the Sohal and Levine groups demonstrated that in the flying muscles of an aged population of Drosophila melanogaster, aconitase was the only carbonylated protein in the mitochondria (146). On the contrary, abundant proteins such as cytochrome c remained unchanged, as observed previously by Yan et al. (727). So, what could explain specificity? Cabiscol's group (90) identified a total of 179 proteins from different organisms that were increasingly carbonylated during aging and grouped them according to either their physiological function or location. They found that 9% and 2% of the total proteins listed were cytoplasmic and mitochondrial, respectively. Concerning their physiological function, 11% were HSPs, 11% were involved in amino acid and protein metabolism, and more than 20% in energy production. Another important group was the nucleotide-binding proteins because of the high probability of metal binding to the nucleotide (90).

In an attempt to determine if there are protein sites more prone to carbonylation, Temple et al. (635) showed that only two Lys residues, among 59, were carbonylated in the human serum albumin when the protein was challenged in vitro with ascorbate/Fe³⁺. However, when challenged with hypochlorous acid (HOCl), five Lys residues were modified. These data suggested that carbonylation is selective to some structural features and dependent on the oxidative species. In agreement, Maisonneuve et al. (399) reported that protein carbonylation should be a predictable process since Arg-Lys-Pro-Tyr-rich sequences are the main carbonylatable sites. Amazingly, the abundance of these sites is related to protein function since a high percentage of proteins containing carbonylatable sites were found to be involved in translation, ribosomal structure, energy production, and nucleotide transport.

The increased application of MS analysis and proteomic approaches for identifying oxidative modifications in proteins is expected to reveal additional information on the susceptibility and specificity of protein carbonylation.

4. Pathologies associated with protein carbonylation

Although the association of protein carbonylation with degenerative processes was proposed based on its prevalence during aging, very few reports are presented in the literature regarding human aging (4, 228, 459). Increased protein carbonylation in aging might be a marker as its production depends on the decreased antioxidant defense and decreased proteolysis. There have been discussions about whether protein carbonylation is the cause of senescence or merely an aging diagnostic marker. These possibilities came from some studies showing that protein carbonylation is (i) increased in species with short life expectancies [e.g., crawler; (597)]; (ii) increased in nonculturable cell populations of E. coli when compared with culturable cells (168); and (iii) decreased in the mitochondria of mice submitted to caloric restriction (356).

In the present decade, there are many reviews relating increased and/or specific protein carbonylation to a broad spectrum of human pathologies, including diabetes and obesity (270, 558), skeletal muscle dysfunction (38), obstructive pulmonary disease (765), cardiovascular diseases (641, 675), chronic kidney disease (658), NDs (239), and hepatocellular carcinoma (228). The majority of the determinations of protein carbonyl levels in those studies were performed in the plasma using different carbonyl detection techniques. Some of these studies established a correlation between the pathology and the pool of carbonylated proteins. For example, Bollineni et al. (65) examined the plasma of lean and obese individuals with or without type 2 diabetes (five individuals per group). They identified 18 carbonylated proteins in the plasma of the diabetic group. Those proteins were not resident plasma proteins but from other sources. Although the study demonstrated that protein oxidation underlies this pathological condition, the presence of the same nonoxidized protein in the plasma had already been linked to type 2 diabetes, obesity, and metabolic diseases. Therefore, their oxidation would not be a marker for these conditions unless it was a determinant for their occurrence in the plasma. Zinellu et al. (765) reviewed studies centered on determining plasma protein carbonyls in patients with chronic obstructive pulmonary disease (COPD). They concluded that one study, out of 16, found a positive correlation between increased carbonyl proteins and COPD. Some of the studies found that sometimes, but not always, plasma carbonyl proteins proceed in parallel with disease progression. An extended review by Tucker et al. (658) on chronic kidney diseases reports that protein carbonyls increase with disease progression. Since protein carbonyls increase in parallel with the severity of the diseases, it is plausible that protein carbonyls could be used as markers for tracking disease progression. In addition, the levels of carbonyl proteins decrease after renal transplantation and l-carnitine supplementation. However, none of these studies identified the proteins prone to carbonyl modification.

In summary, carbonylated proteins are present in pathological processes, but no specific proteins are associated with those conditions.

1. Protein glycation

Protein glycation is intrinsically associated with high glucose levels in the blood and tissues. At first glance, protein glycation is not necessarily associated with oxidative mechanisms, although oxidatively modified sugars retain similar reactivity to render Amadori end-products with protein amino groups (Fig. 2). Nonenzymatic protein glycation occurs when sugars condense with protein amino groups forming a Schiff's base, which rearranges to form an Amadori product (Fig. 2). Amadori products further react with amino groups of other protein molecules to generate Maillard products, also known as AGEs. Among the reported AGEs, pentosidines that originate from Arg and Lys, hydroimidazolones from Arg, and other biomarkers derived from Arg, Lys, and Cys are of bioanalytical interest and can be quantitatively measured (529). Pentosidines are fluorescent Lys-Arg intermolecular protein crosslinks and sequentially oxidized, thus referred to as a glycoxidation process. They accumulate in bone and connective tissues (573) and are used as urine and blood biomarkers of age-related diseases such as osteoporosis and diabetes (560, 575) (Fig. 2). AGEs were first described in vivo as a modification of long-lived proteins, such as hemoglobin, lysozyme, collagen, elastase, and alkaline phosphatase (486). Indeed, glycated hemoglobin is a suitable marker for high glucose levels. AGEs are also formed from exogenous sources, such as cigarette smoking (477) and food (200).

The relationship between protein glycation and pathologies associated with protein-based oxidative stress relies on the fact that protein glycation is involved in protein aggregation and delayed oxidized protein degradation since AGEs may block the entry of the proteasomal core (486). AGEs are recognized by a cell surface type I receptor belonging to the immunoglobulin superfamily (471), which implies a set of metabolic interferences, including interactions with oxidative and nitrosoactive stress (486). Nevertheless, it is out of the scope of the present review to cover the rich literature on AGEs, as it is not a direct consequence of protein modification through by-products of oxidative mechanisms. However, their participation in oxidative stress should not be ruled out. A search in the PubMed platform (April/2020) using Protein glycation as the keyword returned 15,000 total publications that included 2800 reviews. For more information, we advise readers to search for specific literature on AGE-associated pathologies.

2. Adducts formed with by-products of lipid peroxidation

Intracellular intermediates of lipid peroxidation, such as lipid hydroperoxides, can be formed either by enzymatic or nonenzymatic mechanisms (105, 230, 479). Lipid hydroperoxides are formed by reactions involving lipoxygenases, cyclooxygenases, and cytochrome P450. They are efficiently reduced to the corresponding alcohols by the antioxidant enzymes, glutathione peroxidases (Gpxs), and peroxiredoxins (Prxs). While certain types of lipid oxidation products exert beneficial functions, acting as ligands and modulators of cellular signaling processes, the overproduction of lipid hydroperoxides and/or their inefficient reduction can promote deleterious effects by increasing the production of reactive lipid by-products.

The uncontrolled production of lipid hydroperoxides is associated with the induction of cell death (e.g., ferroptosis) (202), inflammation, and several pathologies, such as cardiovascular and NDs. Notably, lipid hydroperoxides can react with metal ions or other one-electron oxidants to produce peroxyl and/or alkoxyl radicals in a Fenton-like reaction (739). These radical intermediates can further propagate lipid peroxidation, yielding various secondary reactive lipid species known as lipid electrophiles.

Esterbauers's group characterized several short-chain aldehydes, including HNE, MDA, and ACR (186). Chemical structures of major lipid electrophiles are shown in Figure 4. Fatty acid-derived aldehydes can be grouped by size into short-chain (<6C), medium-chain (6–12C), and long-chain aldehydes (>14C, e.g., hexadecenal, cyclopentenone PGs, and isoprostanes [IsoPs], levulinaldehydes). In addition, lipid electrophiles can be categorized into truncated oxidized phospholipids (oxPL) and sterol-derived aldehydes (β-hydroxy-5-oxo-5,6-secocholestan-6-al, Seco-A and its aldol product, Seco-B).

OPEN IN VIEWER

FIG. 4.

Structure of short-, medium-, and long-chain aldehydes, truncated oxidized phospholipids, and sterol aldehydes. POVPC, 1-palmitoyl-2-(5-oxovaleroyl)-sn-glycero-phosphocholine.

The short-chain and medium-chain aldehydes include several alkanals and β-alkenals (α, β-unsaturated aldehydes) produced by polyunsaturated fatty acid (PUFA) lipid peroxidation. Moreover, various long-chain aldehydes (IsoPs and isolevuglandin [IsoLG]) are produced by fatty acid cyclization reactions (427, 565), as well as from ether glycerolipids (plasmalogens) and sphingosine-1-phosphate, by enzymatic and nonenzymatic oxidation mechanisms (554). These lipid electrophiles react with nucleophilic groups in biomolecules such as proteins to produce lipoxidation adducts, which is reviewed in the next sections.

a. Aldehydic protein modification

Although the modification of proteins by aldehydes and its interference with protein function have been reported since the 1940s (database: PubMed; keywords: aldehydes AND proteins), evidence that aldehydic products of lipid peroxidation could affect protein function was first reported by studies with rat liver microsomal fractions (45, 195). The first detected protein adduct with lipid peroxidation by-products was described in liver microsomal proteins (46) and upon the formation of low-density lipoprotein (LDL)-HNE adducts (317). Since then, there have been thousands of reports in the literature on this subject. In fact, there has been an almost 2.5-fold increase in the number of publications in the present decade, with many investigators currently dedicated to unraveling the meaning of protein adducts with aldehydic by-products of lipid peroxidation in physiological and pathological processes (224, 747).

(1) Protein-HNE adducts

HNE is probably the most studied lipid by-product [highlighted in a special issue dedicated to HNE (519)]. It is highly reactive toward nucleophilic residues in proteins. As shown in Figures 2 and 5, the mechanisms by which HNE modifies proteins are almost exclusively centered on Michael addition to Lys, Cys, and His protein residues. Besides the chemical possibility of forming Schiff's bases where the HNE aldehyde group reacts with the ɛ-amine group of protein Lys residues, it is estimated that more than 99% of HNE-modified proteins result from Michael addition. As mentioned above, since a carbonyl group in the HNE tail is preserved in proteins modified by the Michael addition, these products are usually classified as carbonyl-proteins.

OPEN IN VIEWER

FIG. 5.

Michael and Schiff base adducts formed by the reaction of HNE with nucleophilic residues in proteins. In Michael addition reaction, HNE readily reacts with the Cys thiol group and with the amine groups of His and Lys residues. Cys residues are the most reactive, followed by His and Lys. Michael adducts are stabilized through conversion to cyclic hemiacetals. HNE can also react with Lys-ɛ-amino groups forming Schiff base adducts that are dehydrated, yielding the 2-pentylpyrrole adduct. Cys, cysteine; HNE, 4-hydroxy-2-nonenal; IMI, His imidazole group. Color images are available online.

In recent years, proteomic approaches have been used to uncover HNE-modified proteins (732, 755). Data reported in these studies are based on cells incubated with HNE (732) and extracted cellular proteins incubated with 4-HNE (755). The primary importance of these studies was (i) to identify the proteins that are potentially prone to HNE modification, (ii) to specify the protein motif/sequences that are more likely to be modified residues, (iii) to investigate the stability of the modified protein, and (iv) to discover the most common residues undergoing modification. The main conclusions from these studies are that (i) 400 to 600 proteins were found modified by HNE, (ii) the modification was only through the Michael addition mechanism, (iii) Cys and His residues were the most common modification sites, and (iv) in one study, aspartic acid was found to be enriched around both modified His and Cys residues (755), while in another study, Lys was found in the vicinity of Cys-modified residues (732).

Regarding the stability of the modified protein, when cells incubated with HNE were allowed to recover for 4 h, 87% of the quantifiable HNE adducts were reduced by twofold (732). However, the study found that among four modified-Cys residues in the FAM120A protein, one of them was still modified after the 4-h recovery, suggesting an unknown intracellular repair mechanism. Indeed, proteasomal inhibition did not affect the turnover of the HNE-modified proteins. The modified proteins found in those studies are localized in almost all intracellular compartments and the extracellular exosome. Moreover, they are functionally related to RNA processing, protein ubiquitination, and cell cycle, among others. On the contrary, when six liver cell lines without any treatment were analyzed for the pool of HNE-modified proteins, only a few proteins were found modified, 14 at Cys residues, 14 at His, and 5 at Lys (755).

(2) Protein-HNE adducts as markers of pathophysiological processes

Only 8% of the HNE produced inside cells bind to proteins (586). The proteins potentially modified by HNE found in human samples are mainly transcription factors of pathways involved in the antioxidant response, inflammation, cell cycle, protein synthesis, and apoptosis (104, 224). The concept of a biomarker as an indicator of diseases or their progress cannot yet be assigned to HNE-modified proteins. For this, many studies of epidemiological nature are necessary.

Nonetheless, there are many reports in the literature pointing in that direction. One important observation is that while protein-HNE adducts are relatively vulnerable to degradation, they are more stable than HNE itself. Moreover, proteins in the extracellular compartments are much less exposed to proteases. Notably, in recent years, many reviews highlighting protein-HNE adducts in human pathological processes have been published. Table 2 summarizes human diseases where protein-HNE adducts are reported either as a possible biomarker or a contributing factor in the development of the investigated pathological processes.

OPEN IN VIEWER

Table 2.

Human Pathological Conditions Associated with Protein-HNE Adducts

Disease	References
Aging	Reviewed in Barrera et al. (39)
Alcoholic cirrhosis and autoimmune liver diseases	(449); reviewed in Barrera et al. (39)
Atherosclerosis	(64, 219, 317, 350, 472)
Autoimmune diseases	(246, 648); reviewed in Barrera et al. (39) and Kurien et al. (352)
AD	(510, 542, 637)
Cardiovascular disease	Reviewed in Afonso and Spickett (3)
Chronic inflammatory diseases	Reviewed in Barrera et al. (39)
Type 2 diabetes	(650)
Gastrointestinal diseases	(465)
Neurodegenerative diseases	Reviewed in Barnham et al. (37), Barrera et al. (39), and Jomova et al. (312)
Obesity	Reviewed in Cohen et al. (126)
Tumors	Reviewed in Barrera et al. (39)

HNE, 4-hydroxy-2-nonenal; AD, Alzheimer's disease.

(3) Protein-HNE adducts and signaling

One crucial question about oxidatively modified proteins is whether they function as signaling or regulatory biomolecules. Herein, we define a regulatory role as a process where the protein is modified so that its function is changed to modulate or cope with the process it is engaged in, discussed in Section IV. On the contrary, a signaling process implies the participation of the modified protein in promoting a cellular response to a given stimulus (e.g., antioxidant response, gene transcription, and others). In the case of protein-HNE adducts, the most widely reported effect is the participation of these adducts in cell signaling involving the activation of kinases and transcription factors related to redox homeostasis (Nrf2, NF-κB, AP-1, PPAR) (28, 389, 751).

One of the most explored mechanisms is the modulation of the mitogen-activated protein kinase (MAPK) family, including the well-known ERK, JNK, and p38-MAPK subfamilies. The mechanism by which members of the MAPK signaling cascade are activated is not well known. Data show that protective or pathological functions are dependent on HNE concentrations used to promote cellular responses either to cope with oxidative stress or to induce apoptosis. Physiological HNE concentration is estimated to be in the range of 0.28–0.68 μM in normal human plasma, while concentrations up to ∼5 μM can be found in rat hepatocytes (751). Most of the signaling effects have been observed in studies using low physiological doses of HNE (<1 μM), whereas high supraphysiological doses (>10–20 μM) have been shown to induce processes such as autophagy, senescence, and cell cycle arrest (28, 751).

Other signaling cascades affected by protein-HNE adducts involve protein kinase C isoforms (PKCs), a family of serine/threonine kinases phosphorylating various targets and playing an essential role in cell proliferation, differentiation, and tumorigenesis processes (28). The selectivity of HNE with the various PKC isoforms (conventional: α, βI, βII, γ; novel: δ, ɛ, η, θ; and atypical: ζ, λ/τ) depends on HNE levels, which is indicative of differential effects based on the reactivity of the isoforms for HNE (114). Conventional and novel PKCs are lipid-sensitive (modulated by diacylglycerol levels) and calcium-dependent isoforms. HNE can activate PKC indirectly by inducing phospholipase C activation, which cleaves phosphatidylinositol-4,5-bisphosphate to generate inositol triphosphate and diacylglycerol. In contrast, HNE PKC inhibitory effects are thought to be mediated by direct adduction with the protein. However, the underlying mechanism of HNE selectivity toward different PKC isoforms remains unknown.

Among transcription factors directly or indirectly affected by HNE are those related to genes of the antioxidant response, protein folding, inflammation, and cell cycle (389). At low HNE concentrations, protein-HNE adduct formation modulates gene transcription, allowing cells to recover from moderate oxidative stress. An example is the Keap1-HNE adduct, which prevents the Nrf2-Keap1 complex from forming, thereby allowing the translocation of Nrf2 to the nucleus to induce the transcription of antioxidant and detoxification enzymes (heme oxygenase-1, Prx, thioredoxin [Trx], Gpx, glutathione S-transferase [GST], etc.). The regulation of Nrf2 signaling involves multiple signaling molecules, including Keap1, PKCs, and p21, among others. Moreover, Keap1 is equipped with multiple cysteine-based sensors that are modulated by various types of endogenous/exogenous stressors, including H₂O₂ and electrophiles such as HNE (625). Details on the sensing mechanisms of the Nrf2-Keap1 system toward different electrophiles and oxidants are still under investigation.

HNE is metabolized through enzymatic detoxification systems, namely, Phase I and Phase II metabolic pathways. As a result of both metabolic pathways, the percentage of HNE found on proteins was reported to be around 5% of the total HNE in rat liver and brain (26, 433, 751). HNE is enzymatically modified through glutathione (GSH) conjugation as a substrate of GSTs (Phase II). Hence, GSTs function as significant determinants of cellular levels of HNE and play a role in the regulation of the HNE-protein adduct formation. Another important class of enzymes in HNE detoxification is the family of aldehyde dehydrogenases (ALDHs), which converts HNE to 4-hydroxy-2-nonenoic acid. The ALDHs have been implicated in a series of pathological conditions as protective players against HNE toxicity (433). Experimental approaches in animal models, by ALDH2 gene deletion or ALDH2 drug activation, have confirmed and given clues about the involvement of protein-HNE adducts in many pathologies, such as neurodegeneration, heart diseases, atherosclerosis and, nonalcoholic fatty liver diseases (433). NADH-dependent alcohol dehydrogenase and NAD(P)H-dependent aldo-keto reductase have also been shown to detoxify HNE by reducing it to 1,4-dihydroxy-2-nonene (751).

(4) MDA- and acrolein-protein adducts

MDA is generated as a by-product of nonenzymatic lipid peroxidation and a side product that arises during thromboxane A₂ biosynthesis (28). MDA concentrations in human plasma have been found in the range of 0.36 to ∼15 μM (599). MDA reactivity is based on its electrophilic nature, thereby reacting with nucleophiles such as the protein amine residues (Lys, Arg, His) to generate Schiff's base adducts (28). MDA reacts in vivo with primary protein amines to form the N-(2-propenal) Lys or Lys-Lys crosslinks (663). These adducts are also referred to as advanced lipoxidation end-products (ALEs; Section II.B; Fig. 2). MDA can also react with DNA and proteins, and in some cases, mediate the formation of DNA-protein crosslinks (682). The mouse apoB-100 fraction of oxidized LDL was one of the first protein-MDA adducts detected (488). These adducts are found in almost all the pathologies associated with HNE-protein adduct formation (Table 1) (28).

ACR is an unsaturated aldehyde and the strongest electrophile of all reactive aldehydes and consequently presenting the highest reactivity with nucleophilic residues in proteins (Cys, His, and Lys residues) (186). The FDP-Lys adduct (3-formyl-3,4-dehydropiperidino-lysine) is a stable adduct that was first detected in oxidized LDL (664). Studies by Esterbauer showed that ACR reacts 110–150 times faster with GSH than HNE (186). Conjugation to thiols can occur spontaneously or be catalyzed by GST isoenzymes (27). GSH conjugation and conversion to mercapturic acid metabolites (2-carboxymethylmercapturic acid and 3-hydroxypropylmercapturic acid, as a major urinary product) in the liver and kidney are the main routes for ACR elimination through urine. ACR is produced: (i) as an end product of lipoperoxidation, (ii) by myeloperoxidase (MPO) from threonine at sites of inflammation, (iii) from protamine oxidation, and also (iv) as a metabolite of the anticancer drug cyclophosphamide (613). Human exposure to ACR not only occurs through metabolism and endogenous lipid peroxidation but also by the oral ingestion of food (e.g., cooking oil) and water, as well as through respiratory (cigarette smoke, automobile exhaust) and dermal routes (432).

Increased protein-ACR adduct concentrations have been reported in many human pathological conditions, such as cardiovascular diseases, diabetes, spinal cord injury, alcoholic liver disease, photo-damaged skin, and others. Furthermore, protein-ACR adducts are suggested to play a role in the development of many pathological conditions (83, 432). However, the causality and mechanisms by which ACR adducts induce each pathological condition remain to be determined. The pathological association of ACR with various diseases has been based on detecting higher levels of ACR adducts in sera of patients (83). At the experimental level, ACR has shown to induce pro- or anti-inflammatory effects, depending on the ACR dose and exposure duration (83, 432). ACR (up to 30 μM) caused suppression of innate macrophage responses, and this effect was causally correlated to ACR adduction and inhibition of JNK2 and NF-κB (286). Moreover, ACR treatment has shown to induce endoplasmic reticulum (ER) stress, impair protein biosynthesis and gene transcription, modulate membrane permeability, and increase apoptosis (432). As described in a myocardial ischemic injury model, ACR forms adducts with the mitochondrial PKCɛ leading to mitochondrial dysfunction (691). Besides, ACR high reactivity with thiol proteins impairs critical antioxidant systems based on the protein thiol catalysis by causing GSH depletion (432).

(5) Oxysterol protein adducts

Oxysterols are oxygenated derivatives of steroids that are formed by enzymatic and nonenzymatic pathways. The most abundant steroid in mammalian cells is cholesterol, playing essential roles in membrane structure and signaling. The enzymatic pathway is responsible for cholesterol hydroxylation and its conversion to bile acids and steroidal hormones (458). Since it is unsaturated, cholesterol is oxidized nonenzymatically by free radicals, singlet oxygen, and ozone. These reactions produce oxysterol containing hydroperoxy, hydroxy, ketone, epoxide, and aldehydic groups (171, 595, 764). Importantly, secosterol aldehydes (SecoA and SecoB) are electrophilic oxysterols produced in the reaction of cholesterol with ozone (704), singlet oxygen (643, 667), and free radicals (764). Secosterol aldehydes have been detected in biological samples, including the human brain (66) and atherosclerotic tissues (703), and have been implicated in the pathogenesis of cardiovascular and NDs. Concentrations of secosterol aldehydes in systemic circulation were found to be ∼30 nM under basal conditions and ∼200 nM in patients with inflammatory artery disease (705).

Secosterol aldehydes can modify several proteins leading to misfolding and aggregation. They have been shown to accelerate the in vitro amyloidogenesis of β-amyloid peptides (Aβ 1–40 and 1–42), the α-synuclein fibrillation (66), and SOD1 aggregation (143, 144). These modifications were hypothesized to favor the formation of neurotoxic protein aggregates linked to Alzheimer's, Parkinson's, and amyotrophic lateral sclerosis. Aβ modification occurred by Schiff base formation with basic amino acid residues, specifically Lys 16 and 28, and the N-terminal group of Asp 1 (670). Similar adduct formation has been observed with bovine myelin basic protein (bMBP) (138). Conformational changes and agglomeration of this protein have been attributed to the covalent attachment of SecoB, leading to increased exposure of the peptide domain V86-T98, which is related to the immunological reaction, and decreased exposure of the F42 and F4 proteolytic regions, which lead to the cleavage of bMBP (138). Secosterol aldehydes also form Schiff base adducts with Lys residues located in loop VII (electrostatic loop) and nearby the dimer interface in apoSOD1, enhancing its propensity to aggregate (143). Notably, apoSOD1 aggregation was dramatically increased by the hydrophobic sterol aldehydes and not by less hydrophobic aldehydes, HNE or 4-hydroxy-2-hexenal (HHE), indicating that aldehyde hydrophobicity critically affects protein aggregation (144).

A study evaluating the reactivity of secosterol aldehydes in human atherosclerotic tissue observed that this aldehyde induces the formation of apolipoprotein C-II (apoC-II) amyloid fibers in vitro (615). Notably, macrophages secrete apoC-II in the atherosclerotic process, and its fibrillation is directly related to plaque formation in the disease (615). Of note, secosterol aldehydes also inhibited nitric oxide synthase (NOS), which may contribute to the development of vascular and NDs (355). The inhibition mechanism seems to be related to the blockade of the enzyme binding site with its cofactor calmodulin through the formation of Schiff bases with Lys residues present in this region (355). Furthermore, the role of secosterol aldehydes in cancer-related mechanisms was investigated by Nieva et al. (478). They have shown that secosterol aldehydes, HNE and HHE, form adducts with Lys residues on wild-type p53 protein. However, similarly to the apoSOD1 study (143, 144), only the hydrophobic secosterol aldehydes induced p53 amyloidogenesis, while the more polar aldehydes, HNE and HHE, were not able to induce protein aggregation (478).

Overall, studies reviewed in this section highlight the potential role of electrophilic oxysterols in protein modification. Being highly hydrophobic, sterol aldehydes derived from cholesterol dramatically affect the conformational stability of the protein to which they are bound. Further in vivo studies are required to clarify the importance of oxysterol-induced protein modifications and their potential role in cancer, and cardiovascular and NDs.

(6) Protein adducts with products of PUFA cyclization

Arachidonoyl-containing lipids are oxidized by free radical-catalyzed peroxidation producing a series of prostaglandin F2-like compounds known as IsoP (447), which are important modulators of inflammatory signaling (555). Although some classes of IsoPs undergo Michael addition reactions with protein thiol groups, no protein-IsoP adducts have been identified. Presently, the only evidence of protein-IsoP adducts comes from the direct inhibition of IKK in a cellular model, where the biosynthesis of PGD2 was induced, which subsequently increased IsoP, a metabolite of PGD2 levels. As a result, NF-κB was activated, and the proinflammatory response ensued (557).

Levuglandin (LG) and IsoLG derivatives are acyclic γ-ketoaldehydes derivatives (levulinaldehydes) produced from the spontaneous rearrangement of arachidonate endoperoxide intermediates, prostaglandin H2 and H2-IsoP, generated through cyclooxygenase and free radical pathways, respectively. Of note, the cyclooxygenase pathway produces two levulinaldehyde stereoisomers, the LGE₂ and LGD₂, while the free radical pathway generates several isoLG stereo and regio-isomers (563, 565). The levulinaldehydes produced by the free radical pathway are also referred to as “isoketals (IsoK)” to distinguish them from the enzymatic products. However, chemically they are the same group of compounds, and the IsoLG nomenclature is used to designate both of them. Of note, IsoLGs have never been isolated from biological sources, most likely because of their high reactivity toward biomolecules (half-life of ∼2 min). Notably, IsoLGs have been reported to react approximately two orders of magnitude faster than HNE or MDA (76).

Immune assays have detected protein-IsoLG adducts in human samples, such as plasma, brain, and meningeal vessels (566). Protein–IsoLG adducts have also been detected by liquid chromatography tandem mass spectrometry (LC-MS/MS) (154, 737). Using this approach, IsoLF-Lys adducts in CYP27A1, a sterol C27-hydroxylase, have been identified in the human retina. Protein-IsoLG adducts have been implicated in several pathological conditions, such as alcoholic liver disease, Alzheimer's disease (AD), age-related macular degeneration, atherosclerosis, cardiac arrhythmias, cancer, end-stage renal disease, glaucoma, multiple sclerosis, and thrombosis (565). As discussed below, protein-IsoLG-modified adducts are supposed to be important proteasomal inhibitors (565). The so-called neuroketals are formed analogously to IsoLG through the oxidation of docosahexanoic acid, a PUFA highly enriched in the brain (51).

IsoLGs have a common γ-ketoaldehyde core linked to two different hydrocarbon chains. The γ-ketoaldehyde core reacts with the ɛ-amine group of protein Lys residues and amino groups in nucleic acids and aminophospholipids (565). The reaction mechanism proceeds through the formation of an imine adduct (Schiff's base), which irreversibly cyclizes to a pyrrole adduct (76). In the presence of oxygen, the pyrrole adduct is converted over time into highly stable lactam adducts, as well as hydroxylactam adducts (76). Alternatively, the pyrrole adduct can also react with other nucleophiles such as thiols or other pyrroles to produce protein/protein and protein-DNA crosslinking (55). Two different structures for the protein/protein crosslinks have been proposed, the bis-aminal and the pyrrole/pyrrole crosslinks.

Membrane proteins are vulnerable targets for IsoLG modification, leading to the formation of phospholipid IsoLG-protein complexes, which can impair the function of ion channels, enzymes, and receptors (75). Sirtuin, a deacetylase enzyme located very close to membrane lipids, is prone to IsoLG modification, reducing its activity and increasing the overall acetylated protein levels. The imbalance of protein deacetylation affects protein/protein interaction as acetylation alters electrostatic interactions and hydrogen bond networks (476), ultimately interfering with signal transduction. The degradation of protein-IsoLG adducts by the proteasome was compromised because adducted proteins are not suitable substrates, and the tentative degradation results in proteasomal inhibition (155).

In conclusion, increased levels of protein-IsoLG adducts are observed in many human diseases, such as atherosclerosis, myocardial infarction, hypertension, and AD (224). However, whether IsoLG adducts contribute to the pathogenesis of these diseases remains to be determined. Evidence supporting a potential role for these adducts on the development of pathological conditions comes from data where cell treatment with exogenous IsoLG induces a variety of relevant responses, including increased macrophage uptake of LDL, activation of platelet aggregation, inhibition of sodium and potassium channels, and inhibition of the proteasome (153). Notably, the presence of IsoLG adducts is increased in patients with atherosclerosis (564). Moreover, IsoLG-protein adducts were found to be elevated in HDL derived from patients with hypercholesterolemia, shedding light on the potential role of HDL-IsoLG adducts in cardiovascular disease (417).

(7) Methods for detecting aldehydic protein adducts

MDA was the first by-product of lipid peroxidation to be measured in biological samples as a free aldehyde (342). Aldehydic protein-adducts formed with MDA, HNE, and ACR, and other aldehydes have been evaluated using different techniques, including immune labeling/immunostaining (747) and mass spectrometric methods (MS). Various polyclonal and monoclonal antibodies have been raised against aldehyde-protein adducts and are now commercially available (324, 326, 489, 649, 665, 685). These antibodies have been successfully applied for the detection, quantification, and tissue distribution of lipid aldehydes, such as HNE, in oxidatively modified LDL particles, ischemic heart, and neurodegenerative disorders [reviewed by (604)].

Despite the usefulness of immunoassays, this technique does not yield information about the precise identity of the modified proteins or structural details about adduction sites. Thus, high-resolution tandem mass spectrometry (MS/MS) analysis has emerged as the gold standard method for characterizing protein posttranslational modifications (PTMs) (10, 673). Proteomic methods commonly used to characterize protein-aldehyde adducts can be divided into “gel-based” and “gel-free” approaches (397). In “gel-based” assays, proteins are separated by one- or bidimensional gel electrophoresis before digestion by proteases, and in the “gel-free” approach, proteins are digested directly in solution. The peptide mixture is then analyzed by MS and identified by peptide mass fingerprinting (PMF) or by MS/MS. In PMF, intact peptide masses are searched against a database containing in silico digested proteins. In MS/MS analysis, peptides are fragmented, and the collection of fragment ions is used for peptide sequencing. The MS/MS proteomic analysis is often coupled to nanoflow liquid chromatography (LC-MS/MS), which greatly increases the coverage and precision necessary for detection and identification of modified proteins

Butterfield et al. have pioneered gel-based redox proteomics to identify oxidatively modified proteins in NDs (88). The gel-free LC-MS/MS strategy is the most commonly used approach for identifying aldehyde-protein adducts (29). The identification of adducted or oxidized proteins is based on the analysis of peptide mass shifts that can be searched manually or using automated computational tools (756).

Detection and identification of modified proteins can be especially challenging in complex mixtures containing thousands of proteins. The analytical complexity is increased due to the great diversity of targets and the extremely low concentrations of the lipid electrophiles (pmol-nmol/mg protein) and their respective protein adducts (572). For this reason, enrichment strategies have been used before MS analysis (124, 678, 689, 732, 755). Such chemoproteomic methods include aldehyde analogues bearing terminal azide or alkyne functionalities, such as azido- or alkynyl-HNE, subsequently captured by click reactions and enriched over avidin-based matrices (302). Indeed, combinations of click chemistry, streptavidin-based enrichment, and MS analysis have been successfully applied to identify sets of target proteins for alkynyl-HNE (678) and alkynyl-ONE (622) in cell lines. Similar strategies based on alkynyl-labeled sterols (631, 709) have been used to characterize the oxysterol adductome. Using alkynyl-HNE and isotopically tagged (light ¹²C₆- and ¹³C₆-heavy labeled) photocleavable azido-biotin reagents, Yang et al. identified and quantified ∼98 alkylation sites (86 Cys and 12 His residues) in intact RKO cells (732). Similarly, Zhang identified and quantified 2257 HNE-modified peptides mapping 1121 proteins in six liver cell lines using an aminooxy labeling strategy (755).

In conclusion, quantitative analysis of protein modifications induced by lipid electrophiles and/or other oxidizing agents is still challenging. Some accurate and high-throughput identification/quantitation methods based on MS analysis (2) and the use of specific tags and isotopically labeled compounds is now available (124, 678, 689, 732, 755). However, there are still limitations regarding the synthesis and availability of these specific reagents, cellular or in vivo delivery routes, the sensitivity of the assay, and other factors. Indeed, further developments of highly specific, sensitive, and quantitative protein PTM analysis methods might help solve many outstanding questions regarding protein modification mechanisms.

(8) Aldehydic modification of serum proteins

Several serum proteins have been shown to form protein-aldehydic adducts (747). This section describes the structural characterization of blood protein modifications induced by aldehyde by-products derived from lipid peroxidation and discusses their potential use as biomarkers. Other oxidative modifications involving the direct reaction of proteins with hypohalous species (HOCl), singlet oxygen, and RNS (ONOO⁻, NO₂ ^•) are not discussed here. However, they can be found in other excellent reviews (152, 171, 196, 264, 505, 616).

Since the proposal of the oxidation hypothesis of atherogenesis, numerous studies provided ample evidence supporting the involvement of oxidized LDL in atherosclerosis (234, 470, 612, 616). Biochemical mechanisms involved in LDL oxidation have been proposed. However, the in vivo mechanisms and exact composition of oxidized LDL particles responsible for atherosclerosis initiation and progression remain unclear (612). Oxidants responsible for LDL oxidation have been the subject of extensive studies and debate. Various in vitro and in vivo experiments have implicated MPO and MPO-derived oxidants (HOCl), lipoxygenase, NADPH oxidases, NOS, metal ions, and heme proteins in the mechanism of atherogenic oxidized LDL particle generation (97, 269, 401, 636). Common features observed under these oxidative insults include the induction of lipid peroxidation, antioxidant depletion, and the modification of apoB-100, the main protein present in LDL particles (375).

Early studies by Esterbauer and collaborators showed that HNE and MDA were the primary aldehydes formed during LDL oxidation (186, 317). Antibodies raised against HNE or MDA-modified LDL recognized oxidized forms of LDL and epitopes in atherosclerotic lesions (250, 317, 666). Moreover, MS analyses have provided important qualitative and quantitative details on sites of modifications induced by electrophilic compounds derived from fatty acid, phospholipid, and cholesterol oxidation on apoproteins (3). Lys and His were the primary residues modified by the aldehydes during copper-catalyzed LDL apoB-100 oxidation (64, 584, 666). Quantitative analysis of hydrolyzed aldehyde-amino acid adducts in an oxidized LDL sample by an adductomic method developed by Uchida's group demonstrated the presence of 6 mol/mol of HNE-His Michael adducts and 6 mol/mol of N ^ɛ-(8-carboxyoctanoyl)-Lys adducts (584). The latter is suggested to be formed by the reaction of 9-oxononanoylphosphatidylcholine (also called PONPC) with Lys residues of LDL apoB-100 (584).

Apart from modifications induced by short-chain aldehydes, studies point toward the relevance of lipoprotein(a) [Lp(a)] modifications induced by oxPL in cardiovascular diseases (61, 672). Owing to genetic, epidemiological, and clinical studies indicating an association between elevated Lp(a) and the risk of developing cardiovascular disease, attention has been recently focused on this lipoprotein particle (62). Lp(a) is a lipoprotein very similar to LDL in terms of lipid composition, which contains an apo(a) glycoprotein covalently linked to apoB-100 via a single disulfide bond. Immunohistochemical analysis using EO6, a monoclonal antibody that specifically recognizes oxPL-modified proteins, demonstrated an accumulation of oxidized Lp(a) in human atherosclerotic lesions (297) and plasma of patients with cardiovascular diseases (657). MS analysis of the Lp(a) fraction isolated from patients indicated the presence of different oxPL species (672). OxPLs are covalently linked to apo(a), specifically through Lys residues located at a Kringle domain called KIV10 (367). Remarkably, studies over the past 15 years conducted mostly by the Tsimikas' group (317, 365, 628, 655, 670) have provided consistent experimental evidence demonstrating the importance of oxPL as a proinflammatory risk factor associated with the epidemiology of cardiovascular disease and for Lp(a) as a major carrier of oxidized lipids in the plasma. Moreover, epidemiological and clinical data indicate that elevated plasma concentrations Lp(a) in arterial lesions are probably a causal risk factor for the development of cardiovascular diseases (61). The mechanism by which Lp(a) induces pathological events is not fully known.

Oxidized lipids and lipoproteins can exert protective or adverse effects depending on the type of reactive lipid species formed, location, tissue, cell type, and protein-adducts formed (426). Aldehydic-protein adducts, protein crowding, and hypoxic/anoxic environments may modulate immune responses (352, 431). Covalent adducts formed by lipid aldehydes (e.g., oxidized truncated phospholipid- and MDA-adducts with protein) on the surface of oxidized lipoproteins can act as epitopes, known as oxidation-specific epitopes (OSEs) (57). These epitopes constitute damage-associated molecular patterns that are recognized by pattern recognition receptors (e.g., scavenger receptors and toll-like receptors), enabling the immune system to mediate their clearance (57). The requirement of OSEs for oxidized LDL or apoptotic cell clearance was demonstrated by experiments showing that monoclonal antibodies that bound to the OSEs inhibited binding and degradation of oxidized LDL by up to 91% (284). Thus, OSEs present in oxidized LDL or apoptotic cells, either as free oxidized lipid or as lipid-protein adducts, act as essential ligands required for their uptake and phagocytosis. Importantly, under pathological conditions, the clearance capacity of available phagocytes is overwhelmed, and the accumulated OSEs in damaged cells or lipoproteins trigger a condition of chronic inflammation (57).

In addition to LDL modifications, studies have also reported oxidative modifications of high-density lipoproteins (HDL). Notably, it has been shown that HDL is a primary carrier of circulating plasma lipid hydroperoxides (74). Apolipoprotein A-I (apoA-I) is the major protein of HDL, comprising ∼75% of the protein content, modified by the lipid-derived aldehydes—ACR and MDA (580, 581, 651). Aldehyde-induced alterations in HDL components have been proposed to produce dysfunctional HDL particles that lack cardioprotective properties. ACR and MDA adduction to apoA1, one of the significant HDL apoproteins, has been shown to potently alter its capacity to remove cholesterol from macrophages by impairing two critical steps in the ABAC1 pathway (580, 581). MS analysis revealed that both aldehydes primarily modified Lys residues. More recently, apoA1 was reported to be extensively oxidatively modified within the human aorta (173). In particular, apoA1 served as a selective target for oxidative modifications by MPO-generated chlorinating and nitrating oxidants within the artery wall (49, 581, 758). Quantitative analysis in human atheroma showed that ∼20% of apoA1 within the lesion is oxidized specifically at Trp72, forming monohydroxylated Trp product (2-OH-Trp). (290). Remarkably, oxidized apoA1 was shown to be dysfunctional and highly enriched in atherosclerotic lesions (290). Moreover, Trp oxidation in apoA1 significantly inhibits the ABCA1-dependent cholesterol efflux acceptor activity (290, 746).

Furthermore, albumin, a highly abundant serum protein (5–55 mg/mL, ∼0.6 mM), reacts with many electrophilic metabolites (559). Albumin-HNE is increased in patients with type 2 diabetes and alcoholic cirrhotic patients (449, 650). Sites of Albumin-HNE adduction have been characterized ex vivo by reacting albumin with HNE and then submitting the modified protein to protease digestion and peptide sequencing by MS (11, 93, 626). Some discrepant results are reported with regard to the preferred sites of modifications. These differences are attributed to the fact that commercially available albumins often contain mixed disulfides at Cys34 and present considerable variability in terms of fatty acids and other ligands that can alter its conformation and reactivity with electrophiles (559). Serum albumin modifications, especially at Cys34 (pKa 6.55), are hypothesized to serve as a potential biomarker for monitoring human exposure to exogenous and endogenous electrophiles (461, 559) as well as a biomarker of oxidative stress (379). Improved immunoassays combined with MS-based analytical approaches have been used to monitor in vivo albumin-aldehyde adduction (93, 538). Analytical strategies include an untargeted LC-MS/MS adductomic pipeline for the global characterization of albumin Cys34 oxidation and conjugation to electrophiles (241). Interestingly, a study conducted with urate electrophiles showed an increase in albumin-urate adducts in the plasma and synovial fluid from individuals with gout and rheumatoid arthritis (660).

Together with albumin, hemoglobin-aldehyde adducts represent promising blood biomarkers. Serum proteins are herein emphasized because, based on our understanding, investigating the presence of oxidatively modified proteins in the serum might be a promising approach to establish biomarkers for pathophysiological conditions and the progress of pathologies based on epidemiological studies (see discussion below). Hemoglobin-aldehyde adducts have been studied using several electrophiles (96), including acetaldehyde (614), 4-oxo-2-nonenal (740), and 2-octenal (741).

Regarding the fate of protein-aldehydic adducts, they are easy substrates for proteasomal degradation (244). The proteasome is also a notable intracellular target of modification by aldehydes (e.g., HNE), resulting in its inhibition (203). It has been shown that mildly crosslinked HNE-modified proteins are preferentially degraded by the proteasome, especially by the 20S proteasome (20SPT). However, extensively modified proteins can contribute to the accumulation of modified proteins, such as in AD, in which HNE-modification of β-amyloid peptide generates a progressively more selective and efficient inhibition of the human 20SPT chymotrypsin-like activity (585). Proteasome activity inhibition by Aβ1–40 was increased from ∼2% (without HNE) to 25% with 5 μM HNE and ∼40% with 10 μM HNE (585). This inhibition was correlated with the increased crosslinking and formation of amyloid-beta oligomers induced by HNE (585). Interestingly, molecular analysis has shown that oligomeric forms of proteins involved in NDs (Aβ, α-synuclein, and mutant huntingtin) adopt a three-dimensional (3D) conformation that inhibits the 20SPT through allosteric impairment of the substrate gate in the 20S core particle, thereby blocking protein degradation (638). In addition to proteasomal degradation, accumulating evidence suggests that lysosome (410) and autophagy (277) may play essential roles in HNE-protein adduct degradation.

1. Detection and analysis of oxidative protein/protein crosslinks

For many years, the predominant detection methods, such as sodium dodecyl sulfate–polyacrylamide gel electrophoresis and light scattering, only revealed gross protein modifications such as dimerization and/or oligomerization. Later, more specific methods involving antibodies or total protein hydrolysis followed by gas chromatography coupled to MS or LC-MS were developed. Both of these methodologies and antibody development require knowledge of the amino acid residues involved in the crosslink and the nature of the covalent bond. In this context, the long-known Tyr-Tyr crosslink that possesses strong fluorescence (232) became, after the disulfide, the most detected oxidative crosslink in biological samples. However, the proteins in these samples that contain Tyr-Tyr crosslinks and the specific residues that participate in the linkage remain mostly uncharacterized (17, 452).

The scenario started to change with the development of strategies to analyze enzymatic protein hydrolysates by LC-MS/MS strategies. These methodologies not only confirmed the formation of Tyr-Tyr crosslinks in proteins submitted to different oxidants or light in the presence of photosensitizers but also enabled the identification of the Tyr residues involved in the linkage. They also enabled the characterization of novel specific crosslinks in oxidized proteins, such as Trp-Trp (419, 503, 582), Trp-Tyr (209, 370), as well as Tyr-Lys and His-Lys (409). Although Trp-Tyr crosslinks were characterized before in the active sites of enzymes with peroxidase activity, these characterizations were dependent on X-ray crystallography (52, 724). Concerning the mechanisms for the formation of these novel crosslinks, the production of Tyr-Lys and His-Lys crosslinks by oxidation reactions remains under investigation (409). Conversely, there is a consensus in the literature that the formation of Trp-Trp and Trp-Tyr crosslinks occurs through radical-mediated mechanisms, as is the case with Tyr-Tyr crosslinks. The one-electron oxidation of Trp and/or Tyr residues leads to the corresponding protein-derived radicals (protein-Trp^•/protein-Tyr^•), which rapidly (k ∼ 5 × 10⁸ M ⁻¹·s⁻¹) recombine with itself or with the other to produce the crosslink (Trp-Trp, Tyr-Tyr or Trp-Tyr) (252, 502, 503). The yields of these products are likely to be low under most physiological conditions since high yields of protein-Trp^• and -Tyr^• radicals are required to favor their recombination reactions. In addition, these radicals react relatively slowly with O₂ (k ∼ 10⁵ M ⁻¹·s⁻¹) (189, 292) and rapidly with O₂ ^•− (k ∼ 10⁹ M ⁻¹·s⁻¹) (99, 145, 435), which are biologically ubiquitous. Still, protein crowding and hypoxic/anoxic environments may favor the formation of protein crosslinks by radical recombination (502).

Up to this point, most of the performed LC-MS/MS studies to investigate posttranslational crosslink modifications were limited to purified proteins oxidized in vitro. However, recent improvements, such as performing enzymatic hydrolysis of oxidized proteins in water labeled with ¹⁸O, optimization of MS/MS fragmentation methods, and the use of software packages to search for crosslinks opened up new avenues to study crosslinks present in oxidized proteins from biological samples (409). As a proof of concept, these authors used the developed strategies to analyze a protein extract from the gram-positive lactic acid bacterium Lactococcus lactis exposed to peroxyl radicals generated from the decomposition of 2,2-azobis (2-amidinopropane) dihydrochloride (AAPH). They were able to characterize 24 Tyr-Tyr, 4 Tyr-Trp, and 3 Trp-Trp specific crosslinks in specific proteins of the extracts (409). More recently, we reported the presence of Trp-Trp and Trp-Tyr crosslinks in crystallin proteins of human lenses with advanced nuclear cataract (502) (Fig. 6A). Therefore, we can anticipate that as instrumentation, software, and the understanding of oxidant chemistry advance, it will be possible to characterize crosslinks in biological samples more frequently and to discover novel protein crosslinks (252).

OPEN IN VIEWER

FIG. 6.

Protein crosslink. (A) Schematic representation of the pathways for the formation of Trp-Tyr, Trp-Trp, and Tyr-Tyr crosslinks in crystallin proteins of cataractous lenses. PS and PS* abbreviations stand for endogenous photosensitizers in the ground state and excited form, respectively. The crystalline structure shown corresponds to that of the human β-crystallin B2 chain (PDB 1YTQ). The figure shown was modified from Paviani et al. (502). (B) Schematic representation of a hypothetical pathway for the role of metal ions in Tyr-Tyr crosslink formation and protein aggregation in the brains of AD or PD patients. Tyr-Tyr crosslinks are present in Aβ-amyloid and α-synuclein aggregates in the postmortem brains of patients with AD and PD, respectively. In addition, the levels of free metal ions (Fe²⁺ and Cu²⁺) and oxidants increase in the central nervous system with aging, facilitating protein oxidation, crosslink formation, and protein aggregation (see Section II.C). Aβ or α-syn abbreviations stand for Aβ-amyloid and α-synuclein, respectively; reaction intermediates are omitted for clarity. AD, Alzheimer's disease; PD, Parkinson's disease; Tyr, tyrosine. Color images are available online.

2. Roles in pathophysiology

As stated above, the characterization of many oxidative crosslinks is recent, and there is a lack of data on their occurrence in biological samples. The exception is the Tyr-Tyr crosslink identified from the oxidized amino acid and protein-Tyr residues after total protein hydrolysis (enzymatic or chemical) (232). Therefore, the detection and quantification of Tyr-Tyr in biological samples are a useful biomarker of oxidative imbalance. Furthermore, increased Tyr-Tyr levels suggest risk for some diseases when accumulated in specific tissues (176, 232).

Indeed, several pathologies present increased Tyr-Tyr levels. For example, elevated levels have been reported in the plasma of patients with hyperlipidemia, chronic renal failure, and uremia undergoing hemodialysis (323, 378, 721), myocardium tissue after acute infarction, lenses of patients with cataracts, and urine of diabetic patients. The levels of Tyr-Tyr increase during aging (194, 207, 702) and acute inflammation (53). During the inflammatory response, the activity of the heme-containing enzyme MPO localized in phagosome membranes of immune cells, mainly neutrophils, produces large quantities of HOCl from hydrogen peroxide and chloride ions (20). HOCl promotes the formation of Tyr-Tyr (129, 273, 428), which can also be formed by peroxynitrite (391, 568) and by peroxidases/H₂O₂, including MPO in the absence of chloride (272). In some clinical conditions, MPO plasma levels rise above those of the healthy population (33, 78, 285). Higher Tyr-Tyr levels in patients undergoing hemodialysis correlated with increased MPO activity in response to blood contact with dialysis membranes (337).

Several studies also indicate an association between NDs and Tyr-Tyr crosslinks. The presence of these crosslinks in the β-amyloid and the α-synuclein from postmortem brain samples from AD and Parkinson's disease (PD) patients, respectively, suggests that Tyr-Tyr crosslinks may have an important role in the assembly of the β-amyloid fibrils in the neuropil and of α-synuclein assemblies in the Lewy bodies. The amyloid fibrils are key characteristics of the aggregates in these NDs (12, 13). The high concentrations of copper and iron ions found in amyloid plaques and substantia nigra are likely catalysts for in situ formation of protein-Tyr^•, which are Tyr-Tyr crosslink precursors (86, 110, 170, 551, 567) (Fig. 6B). Copper ions greatly enhance Tyr-Tyr levels present in the β-amyloid (12, 13). Similarly, Tyr-Tyr crosslinks form in vitro via metal ion-catalyzed oxidation. In addition, Tyr-Tyr crosslinks are biomarkers of oxidative stress in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) mouse model of PD (12, 508, 509). Of relevance, the levels of free metal ions and oxidants increase in the central nervous system with aging, facilitating protein oxidation, crosslink formation, and protein aggregation. Interestingly, possible or novel therapeutic approaches for AD (23, 325) and PD (169) rely on metal ion-targeting drugs. Indeed, of the 800 compounds used, tested, or proposed for PD treatment between 2014 and 2019, 250 present metal ion-chelating properties (646). Figure 6B summarizes the above information, displaying a hypothetical pathway for the role of metal ions in Tyr-Tyr crosslink formation and protein aggregation in the brains of AD or PD patients.

In conclusion, the characterization of many protein crosslinks formed by oxidative mechanisms occurred only recently, and many others remain uncharacterized (252). Advances in the strategies to detect and quantify protein crosslinks in biological samples are required, particularly in protein aggregates, which are possibly relevant to age-related diseases. With these and other advances, detailed investigations into the role of protein crosslinks in the pathogenic mechanism of diseases will be feasible.

1. Tyrosine nitration

The PTM of tyrosine by nitration consists of the addition of a nitro (−NO₂) group to the aromatic ring of a Tyr residue to form 3-nitrotyrosine (3-NT).

2. Cysteine nitrosation

The PTM of cysteine by nitrosation (also called S-nitrosation) consists of the covalent addition of a nitrosonium (−NO+) to the sulfur atom of a deprotonated Cys residue (P-CysS⁻), forming an S-nitrosothiol (P-S-NO).

1. Thiol/disulfide switches

Initially, we describe some thiol/disulfide switches (Fig. 8) in proteins as mechanisms underlying catalysis and signaling. The two-electron oxidation of thiols in proteins to disulfides is a common theme in biology initially associated with structural roles as permanent crosslinks that stabilize the protein structure (115). These structural disulfides adopt the most favorable conformations when the dihedral angle is 90°C, and the distance between the two sulfur atoms is 2.05 Å (6, 475, 550a).

OPEN IN VIEWER

FIG. 8.

Thiol switches. In this scheme, we describe how three thiol switches are related. The RSH/RSOH switch is the simplest one. The RSH/RSSR requires the formation of sulfenic acids (RSOH) as intermediates and two thiol/disulfide exchange reactions to return to the initial form (RSH). The RSH/RSO₂H also required the sulfenic acids (RSOH) as intermediates. The reduction of sulfinic acids back to the thiol form requires Srx in an ATP-dependent manner. Srx, sulfiredoxin.

Nonetheless, disulfides are also recognized to play functional roles through redox processes (115, 267). In this situation, disulfide bonds are continuously formed and disrupted. Therefore, a highly stable disulfide is not suitable, and this covalent bond then assumes less stable structural conformations that are more easily reduced back to dithiols. Indeed, these disulfides are strained in stereochemically disfavored conformations that present intrinsic torsional energy (267). These “functional” disulfides are sometimes referred to as “forbidden” (267).

A specific type of functional disulfide is the allosteric one. To understand this concept, we need to consider that an allosteric effector is a compound that changes the activity of a protein by binding in a region that is distant from the active site (440). Allostery is possible due to dynamics in the protein structure. Therefore, an allosteric disulfide represents a covalent bond that, once formed, triggers a conformational change that affects protein activity (115).

These dithiol/disulfide pairs may also function as redox switches during the catalysis of enzymes, such as ribonucleotide reductase, a protein involved in the synthesis of deoxyribonucleotides, the DNA building blocks. The dithiol of ribonucleotide reductase can reduce ribonucleotides into deoxyribonucleotides by a complex sequence of electron transfer reactions that result in the formation of a disulfide (185).

2-Cys Prxs are other Cys-based enzymes in this dithiol/disulfide category that play central roles in hydrogen peroxide metabolism. In this case, a highly reactive thiol present in the so-called peroxidatic Cys (C_P) reacts with H₂O₂ giving rise to sulfenic acid (R-Cys –SOH) intermediates that then undergo condensation with a resolving Cys (C_R) to form a disulfide (Fig. 9).

OPEN IN VIEWER

FIG. 9.

Mechanisms of H₂O₂ signaling by thiol switches. Signaling protein here refers to a protein, such as a phosphatase or transcription factor. (A) Signaling at low levels of H₂O₂. (I) Direct oxidation of a signaling protein (e.g., OxyR); (II) oxidation mediated by Prx (e.g., Yap1; STAT3); (III) Prx oxidizes Trx, consequently less reducing power is available to maintain the signaling protein in the reduced state (e.g., Ask-1). (B) At high levels of H₂O₂ or in the presence of CO₂, hyperoxidation of Prx is facilitated. In this case, direct oxidation of the target protein is feasible (e.g., PTP-1B) and increases the pool of reduced Trx. Ask-1, apoptosis signaling kinase 1; Prx, peroxiredoxin; Trx, thioredoxin.

The regeneration of the reduced forms of ribonucleotide reductase and 2-Cys Prx involves an S_N2 nucleophilic displacement reaction, called the thiol/disulfide exchange reaction. Uncatalyzed thiol/disulfide exchanges proceed at slow rates, but some oxidoreductases, such as Trx, glutaredoxin (Grx), and PDI, can accelerate these reactions by six to seven orders of magnitude (475). Since thiol/disulfide oxidoreductases display specificity to their substrates, they are involved in the redox regulation of signaling pathways (475).

Indeed, some transcription factors are redox-regulated by dithiol/disulfide switches that involve the participation of Trx or Grx enzymes. For instance, OxyR is a bacterial transcription factor that, upon oxidation of critical Cys residues by H₂O₂, forms a disulfide bond (Fig. 9A, arrow I), resulting in the induction of gene expression (760). Following the oxidative insult, Grx reduces the disulfide bond back to the dithiol state (760); consequently, the transcription of target genes ceases.

OhrR is another redox-regulated protein, but, in contrast to OxyR, it is a repressor rather than an activator of transcription (208). In the reduced state, OhrR binds to the promoters of target genes, inhibiting their expression. Upon oxidation of critical Cys residues by organic hydroperoxides, OhrR undergoes structural movements that provoke its release from the DNA, exposing the promoters of the target gene to the transcriptional machinery. Similar to OxyR, the oxidation of OhrR results in the formation of a disulfide that, in this case, is reduced explicitly by Trx instead of Grx (590).

Thiol/disulfide switches are also involved as a posttranscriptional mechanism that controls intracellular iron levels. Iron is required for several processes, such as DNA synthesis, respiration, and heme biosynthesis (123). However, high levels of iron may promote undesirable oxidation of vital cellular components (123). Therefore, the levels of this metal in cells are strictly regulated. Iron regulatory proteins 1 and 2 (IRP1 and IRP2) are the key iron sensors in mammalian cells that activate posttranscriptionally the synthesis of proteins involved in iron metabolism. IRP1 and IRP2 are cytosolic proteins that bind RNA stem-loops known as iron-responsive elements located in untranslated regions of mRNAs. IRP2 is the predominant RNA-binding protein in vivo (358), and unlike IRP1, IRP2 protein does not coordinate a [4Fe-4S] cluster and does not display aconitase activity (242). Under iron-replete conditions, IRP2 is degraded by a process involving polyubiquitylation and proteasomal degradation that requires a 73-amino acid sequence that is not present in IRP1 (298, 299). NO⁺ donors and exogenous heme increase IRP2 degradation in a process that depends on Cys residues present in the 73-amino acid segment (71). Notably, the redox state of two other Cys residues (Cys512 and Cys516) can regulate the mRNA-binding activity of IRP2 through a thiol/disulfide switch (767).

2. Thiol/sulfenic acid switches

Besides disulfides, redox regulation of thiol groups in proteins can also proceed through other intermediates such as nitrosothiols (Section II.D). We now focus our attention on sulfenic acid (R-Cys-SOH) derivatives that are the direct product of the two-electron oxidation of thiols (Fig. 9). The reaction between sulfenic acids and thiols gives rise to disulfides. Therefore, oxidants such as H₂O₂ generate disulfides in proteins with the intermediacy of sulfenic acids. In some cases, however, the sulfenic acids are protected from thiols by the protein environment.

Sulfenic acids are unstable compounds that can react not only with other thiols but also with other nucleophiles. The half-lives of sulfenic acids can also be limited by oxidation to sulfinic (-SO₂H) and sulfonic (-SO₃H) acids or self-condensation to yield thiosulfinates (9). In addition, sulfenic acids can also reversibly react with backbone amides, generating sulfenamides (248). However, sulfenic acids can gain stability in specific environments, where steric hindrance effects due to bulky substituents can protect them from organic compounds (108) or protein microenvironments (745). In these proteins, thiol/sulfenic acid switches can function in catalytic or signaling processes.

Therefore, as sulfenic acids are present as a dynamic and transient oxidation product of thiols, these species are generally analyzed by dimedone (5,5-dimethyl-1,3-cyclohexanedione)-based labeling reagents that trap these intermediates before they undergo other reactions (14, 473). The reaction of dimedone with sulfenic acid is slightly selective but very slow (340, 583). As an alternative, strained alkynes and alkenes have been developed (523) that selectively trap sulfenic acids with rates more than 100 times faster than reaction with dimedone (524). In the same line, benzothiazine (205)- and trans-cycloocten-5-ol (574)-based probes were described. Noteworhty, electrostatic or steric effects can likely modulate the rate of any probe reactivity toward a protein sulfenic acid (583). Perhaps the description of a thiol/sulfenic acid switch in GAPDH was the first with functional implications. Notably, the conversion of the glyceraldehyde phosphate dehydrogenase activity into acylphosphatase activity of GAPDH was paralleled with the formation of a sulfenic acid on a critical Cys residue (Cys 149) (742).

Several other studies also described thiol/sulfenic acid switches in catalysis and redox signaling. Nevertheless, the development of chemoselective probes has allowed for the investigation of sulfenic acids in cellular contexts. For instance, when H₂O₂ oxidizes Cys-797 of epidermal growth factor receptor (EGFR) in human epidermoid carcinoma A41 cells, the kinase activity of this protein is enhanced (501). These processes are consistent with the mechanism described in Figure 9A, arrow I.

Although the formation of sulfenic acids by the oxidation of thiols is well studied, the reverse process (reduction of sulfenic acids to thiols) is still poorly characterized. Indeed, the identity of the reducing agent is controversial. As reviewed by Kettenhofen and Wood (328), trivalent arsenicals are considered specific reductants for sulfenic acids, but they are not physiological compounds (328). In biological systems, Trx and GSH are considered general reductants of sulfenic acids (132). In both cases, two sulfhydryl groups are required for the reduction of one sulfenic acid in a two-step process. In the first step, a mixed disulfide between the protein and the thiol is generated. When GSH is the thiol involved, the protein is glutathionylated. These reactions are generally fast, attaining values in the 10⁵ M ⁻¹·s⁻¹ range for nonprotein molecules (462).

On the contrary, the corresponding rate constant for sulfenic acid in human albumins is in the 3–100 M ⁻¹·s⁻¹ range (659). In the second step, a second sulfhydryl moiety is required to resolve the mixed disulfide. In the case of GSH, an oxidoreductase such as Grx is generally required to catalyze the reaction.

However, the identity of sulfenic acid reductants varies from protein to protein. Furthermore, these reductions of sulfenic acids in proteins display a high degree of specificity. For instance, mammalian Prdx6 (1-Cys Prx enzyme) is not reducible by GSH or Trx enzymes (199), although Trx and GSH are considered general reductants for sulfenic acids (132). Regeneration of the reduced form of Prdx6 can be achieved by its heterodimerization with πGST, which can then be reduced by GSH (199). In contrast, the homodimer of 1-Cys Prx1 from yeast (ScPrx1) can be reduced by Trx (504). Therefore, structural aspects such as protein/protein interactions should probably be considered when contemplating the specificity of the protein sulfenic acid reductions.

Ascorbate is another compound relevant to the reduction of protein sulfenic acids in biological systems but not frequently considered. Indeed, ascorbate is a known reductant for low-molecular-weight sulfenic acids (142). The direct reduction of protein sulfenic acids by ascorbate was first described in GAPDH (742) and later in 1-Cys Prx enzymes and papain (441, 526, 546, 766). Our results indicated that, in most cases, the rate constants for the reactions of diverse protein sulfenic acids lie in the 0.4–2.2 × 10³ M ⁻¹·s⁻¹ range, indicating that ascorbate is a broad-spectrum reductant (18). Possibly the relevance of ascorbate for the reduction of sulfenic acids might be higher in compartments, where the concentration of this reducing agent is elevated. Notably, the overall reduction of protein sulfenic acids by ascorbate was understood as an underlying mechanism in the noncanonical scurvy phenotype of triple mutant mice, lacking systems to reoxidize PDI (766).

3. Thiol-sulfinic/sulfonic acid switches

In some cases of higher oxidative insult, protein sulfenic acids can be hyperoxidized to sulfinic (R-Cys-SO₂H) (Fig. 9) or sulfonic (R-Cys-SO₃H) acids that were initially considered irreversible processes and, as a consequence, would provoke the inactivation of redox-active proteins harboring this PTM. In this way, the hyperoxidation of proteins would be toxic to cells due to the loss of function of the corresponding proteins. The finding that motifs that facilitate the hyperoxidation of 2-Cys Prx were positively selected for throughout evolution was surprising, indicating a physiological role for this PTM (720). This observation led to the coining of the so-called flood gate hypothesis (Fig. 9B), in which the inactivation of 2-Cys Prx would increase local H₂O₂ concentrations, allowing proteins that react slowly with this oxidant to be oxidized (720). Later on, other motifs that facilitate the hyperoxidation of 2-Cys Prx were identified (63, 627). Notably, not only hyperoxidation but also phosphorylation of Prdx1 provides another means to inactivate 2-Cys Prx, and this PTM was also proposed to have a role in increasing local H₂O₂ concentrations (719). Furthermore, acetylation of mammalian 2-Cys Prx (Prx1 and Prx2) increases their resistance to hyperoxidation (497). Hence, the thiol/sulfinic acid switch (Fig. 9) appears to fine-tune protein control.

Besides the local accumulation of H₂O₂, another mechanism was proposed to explain the positive selection of hyperoxidation motifs throughout evolution and involves the inactivation of 2-Cys Prx to preserve pools of reduced Trx (Fig. 9B) for other vital cellular processes (446). According to this model, the disulfide form of 2-Cys Prx actively consumes reduced Trx to turn over the peroxidatic cycle, thus reducing the availability of this oxidoreductase to reduce critical enzymes, such as ribonucleotide reductase and methionine sulfoxide reductase (156). When hyperoxidized, 2-Cys Prx enzymes do not react with Trx, increasing the reduced pool of this enzyme in cells and keeping the reducing equivalents required to sustain the DNA synthesis and repair oxidative lesions, among other essential processes.

A breakthrough in the field occurred with the discovery that sulfinic acids in 2-Cys Prx could be reduced back to sulfhydryl forms (717, 718) and the identification of an ATP-dependent system called Srx (58). At that moment, it became clear that the thiol/sulfinic acid pair is a reversible switch (Fig. 9), a convenient feature for a regulatory process. Accordingly, the thiol/sulfinic acid pair is involved in converting 2-Cys Prx enzymes from peroxidases into chaperones (holdase) upon hyperoxidation to sulfinic acids (305). Furthermore, hyperoxidized 2-Cys Prx can recruit chaperones to misfolded proteins in aggregates (255). Other lines of evidence indicate the biological relevance of Srx and sulfinic acids in 2-Cys Prx. For instance, the regulation of adrenal steroidogenesis involves the reduction of sulfinic acids in mitochondrial Prdx by Srx in a circadian pattern (332, 333).

Initially, 2-Cys Prxs were the only known substrates of Srx, an enzyme that reduces sulfinic acid. However, proteomic studies using an electrophilic diazene probe revealed that 55 other proteins have the potential to be also turned over by Srx (8). The reduction of sulfinic acids in protein tyrosine phosphatase (PTP), nonreceptor type 12 (PTPN12/PTP-PEST), and DJ-1 was confirmed by an enzymatic assay using recombinant Srx (8). Despite the notable progress, several aspects of the thiol/sulfinic acid switch likely remain to be elucidated.

Therefore, distinct thiol switches can mediate redox signaling (Fig. 9). Regarding H₂O₂ signaling, many mechanisms are possible (474), some of which were already mentioned above. The simplest of them is the direct oxidation of signaling proteins by H₂O₂ (Fig. 9A, arrow I). Perhaps the best example of such a pathway is the activation of the bacterial transcription factor OxyR. OxyR possesses two distinct globular domains, the DNA binding domain (N-terminal) and the regulatory domain (C-terminal domain). A highly reactive Cys in the regulatory domain (Cys-199 in OxyR from E. coli) is rapidly oxidized (∼10⁵ M ⁻¹·s⁻¹) to a sulfenic acid intermediate, which then forms an intramolecular disulfide bond with another Cys residue of the same domain (Cys-208 residue in OxyR from E. coli) (22, 310, 359). The formation of disulfide bonds in each of the four subunits of the OxyR tetramer triggers substantial structural changes in this transcription factor, resulting in the induction of the expression of its target genes (116, 310, 359). After oxidative stress subsides, OxyR is deactivated upon the reduction of the intramolecular disulfide bond by the Grx system for OxyR from E. coli (760) or the Trx system for OxyR from Pseudomonas aeruginosa (699).

The formation of sulfenic acids in the EGFR (501) and Src kinase (274) are two other examples consistent with the direct oxidation of protein thiols by H₂O₂. This proposal is because it is difficult to propose an alternative mechanism for the protein sulfenic acid formation other than the direct oxidation of thiols by H₂O₂.

However, most thiols react only slowly with H₂O₂ (10–50 M ⁻¹·s⁻¹) (711). In contrast, the peroxidatic Cys (C_p) of Prxs can react with H₂O₂ 1 to 10 million times faster (482, 498, 512, 642, 654). Although most of the protein thiols react slowly with H₂O₂, some of them are detected in oxidized forms in biological systems (710). One model that proposed to account for this apparent contradiction between biology and chemistry was the so-called relay mechanism (Fig. 9, arrow II), in which a sensor protein (Prx or Gpx) mediates the oxidation of slow reacting thiols by H₂O₂. These sensor proteins can react very rapidly with H₂O₂ and, through physical protein/protein interactions, transmit the signal as oxidizing equivalents to a slow-reacting thiol that can be a signaling protein, such as a transcription factor or a phosphatase.

The first experimental support for this model came from studies on the mechanism by which Yap1 is activated. Yap1 is the transcription factor that mediates the responses of the yeast Saccharomyces cerevisiae to H₂O₂ (235, 301). The finding that Gpx (also called Orp1) mediates the oxidation of Yap1 by H₂O₂ through thiol/disulfide exchange reactions provided the first experimental evidence for this redox relay mechanism (158).

There are also examples of the redox relay mechanism in mammalian cells. For instance, Prx1 from mammalian cells can physically interact in a redox-dependent pattern with various signaling proteins (94, 307, 661). Interestingly, Prx2-mediated activation of the transcription factor, STAT, displays features similar to the Gpx/Orp1–Yap1 pathway in yeast (596). In all cases, a fast-reacting thiol (Gpx/Orp1 or Prx2) senses H₂O₂ and then relays the oxidizing equivalents through physical interactions and thiol/disulfide exchange reactions to the transcription factors (Yap1 or STAT). Genetic approaches using deletion or depletion of cytosolic Prxs broadly inhibited protein thiol oxidation by H₂O₂, which contrasts with what was predicted by a model that assumes direct oxidation of proteins by H₂O₂ (617). Furthermore, a proteomic approach identified several partners of 2-Cys Prx (617).

The relay mechanism can be even more complicated, with the intermediacy of Trx and Prx (Fig. 9A, arrow III), which we have previously named the Prx-Trx model (474). Accordingly, several signaling proteins are redox-regulated by Trx enzymes, as reviewed (50). For example, only reduced Trx1 binds apoptosis signaling kinase 1 (Ask-1) and, as a consequence, inhibits its enzymatic activity (561). NF-κB is another signaling protein regulated by Trx1, as the binding of this transcription factor to its target DNA sequence requires the reduction of a single Cys residue (268, 416). In this sense, 2-Cys Prxs can participate in redox signaling not only by regulating H₂O₂ levels but also by modulating the redox status of Trx (474).

4. Relationships between H₂O₂ signaling and tyrosine phosphorylation

H₂O₂ signaling is closely associated with phosphorylation-dependent pathways, especially those related to Tyr residues (549). Indeed, the catalytic Cys of PTP enzymes directly involved in removing a phosphate group of its substrate also undergoes reversible oxidation/inactivation by H₂O₂. This dephosphorylation catalyzed by PTPs is a kind of nucleophilic displacement reaction that proceeds via an essential catalytic Cys residue conserved within the PTP signature motif, Cys(X)5Arg (163, 164, 492). The sulfhydryl group of the catalytic Cys displays an acidic pKa (within the 4.0–6.0 range), indicating that a major fraction of the sulfhydryl group is present as thiolate (RS⁻) ion at neutral pH, which is a better nucleophile than the corresponding thiol (RSH). This low pKa of the catalytic cysteine is necessary for its dephosphorylation activity; but, at the same time, renders PTPs more susceptible to oxidative inactivation, which appears to be a reversible and regulatory process (485, 644).

As mentioned above, the primary product of the two-electron oxidation of the nucleophilic cysteine of PTPs is sulfenic acid (RSOH). The outcome of sulfenic acids in PTPs varies depending on the enzyme (393). Nevertheless, most of the oxidative intermediates generated in PTPs are reversible, a convenient feature for a regulatory process. For instance, in the cases of Cdc25 (23 –25, 82, 193, 598) and phosphatase and tensin homologue (PTEN) (371), an intramolecular disulfide bond linking the catalytic Cys residue to a proximal Cys (named backdoor Cys) is generated, which can also be an intermolecular disulfide bond in other phosphatases (HePTP) (393). Furthermore, a stable sulfenic acid species is formed as the oxidation-dependent inactive state of other PTPs (164, 309). Finally, cyclic and reversible sulfenilamide is generated in other PTPs, which is described next.

It is important to mention that the PTP superfamily can be divided into “classical PTPs” (38 members) and “dual-specificity phosphatases” (61 members). The classical PTPs are further subdivided into receptor-like transmembrane PTPs (rPTPs) and nontransmembrane PTPs (nPTPs or cytosolic PTPs) (485, 644). While all nPTPs comprise a single catalytic domain linked to diverse regulatory or targeting domains, most rPTPs have two PTP domains arranged in tandem (485, 644).

The dual-specificity phosphatases can often also dephosphorylate phosphoserine/threonine residues, in addition to phosphotyrosine. This group comprises the MAP kinase phosphatases and the tumor suppressor PTEN that dephosphorylates lipids (phosphatidylinositol (3,4,5)-triphosphate [PIP3]). As mentioned above, sulfenilamides, intra- and intermolecular disulfides, and sulfinic/sulfonic acids were all detected as oxidized intermediates in PTPs belonging to these distinct subfamilies using crystallography and MS, among other techniques (485).

In the case of PTP1B, which is perhaps the best-studied among all classical PTPs, oxidation is reversible due to the rapid conversion of the sulfenic acid form of the catalytic Cys to a 5-atom-ring structure, a cyclic sulfenamide. In PTP1B, the juxtaposition of His214 with Cys215 polarizes the amide bond, promoting nucleophilic attack by the amide nitrogen of Ser216 on the sulfur atom of the sulfenic acid in Cys215, leading to the formation of a covalent bond between the sulfur and nitrogen atoms of these neighbor residues. Of note, the cyclic sulfenamide can be readily reduced back to the thiolate form (443, 562). Possibly this sulfenamide might have two roles: (i) prevent irreversible hyperoxidation of catalytic Cys and (ii) facilitate the reduction of the sulfenamide to restore the active form of the PTPs (644).

Sulfenamides were also detected in the crystal structures of the D2 domain of rPTPα, which is structurally very similar to and with an identical chemical composition of sulfenamides generated in PTP1B (730). It has been shown that rPTPs have a tandem arrangement of two PTP domains, with the catalytic residues in the membrane proximal domain (D1). In contrast, the membrane distal domain (D2) predominantly functions as a regulatory domain, displaying little to no catalytic activity (511, 730). Intriguingly, although virtually devoid of phosphatase activity, most rPTP D2 domains share the conserved catalytic Cys and Arg residues of the PTP signature motif, suggesting a redox sensor role (730). As demonstrated by an antibody-based assay for detecting oxidation-inactivated PTPs, the second regulatory domain of rPTPα is more susceptible to oxidation (511).

Remarkably, using this antibody, it was possible to detect platelet-derived growth factor (PDGF)-induced oxidation of endogenous Src homology 2 domain-containing protein tyrosine phosphatase 2 (SHP-2; PTPN11), an nPTP (511), which is a very physiological condition. SHP2 is a ubiquitous multidomain nonreceptor phosphatase implicated in diseases such as cancer, diabetes, and Noonan syndrome. Noonan syndrome is one of the most common genetic disorders associated with congenital heart disease, and approximately half of the patients with Noonan syndrome present mutations in SHP2, such as N308D. Remarkably, SHP2^N308D is more prone to oxidation than wild-type SHP2, but an association of this kinetic feature with Noonan syndrome is premature (392).

Methods to detect oxidation of phosphatases in cells are a challenge. Traditionally, most of the methods involved thiol alkylation of the reduced form of PTPs (485). However, there are some drawbacks, as many of these alkylating agents also modify sulfenic acids (523). Therefore, assays based on dimedone, which specifically alkylate and therefore trap Cys sulfenic acids, were developed (473), as described above. Another immunochemical approach to directly profile oxidized PTPs involved the generation of an antibody that recognizes the conserved sequence of the PTP signature (VHC_DMDSAG) harboring the catalytic cysteine modified with dimedone (C_DMD) that chemoselectively reacts with Cys sulfenic acids to form a stable thioether adduct (217).

Employing these and other methods, PTPs of all subfamilies were found to be oxidized in cells upon physiological stimuli such as growth factor (EGF and PDGF), insulin, and cell receptor (B cell and T cell) stimulation (362, 644). A remarkable feature is the pronounced selectivity for the oxidation of each specific pathway by a different stimulus. For instance, while exogenous H₂O₂ treatment of fibroblasts caused oxidation of multiple PTPs, stimulation of the same cells with PDGF selectively oxidizes SHP2 phosphatase (421). In contrast, insulin stimulation preferentially oxidized PTP1B and TC-PTP, two PTPs that are relevant for negative regulation of insulin receptors (421).

Although the thiolates of catalytic Cys of PTPs display acidic pKas, their reactivity toward H₂O₂ is low (k_inact  = 10–20 M ⁻¹·s⁻¹) (164). In contrast, Prx and Gpx react with H₂O₂ 1 to 10 million times faster and are abundant protein thiols. Furthermore, GSH is present in cells in millimolar concentrations. Therefore, it is reasonable to consider that on chemical grounds, Prxs, Gpxs, and GSH outcompete PTPs (710). However, as mentioned above, PTPs are frequently found oxidized in cellular systems, showing an apparent contradiction (710). The relay mechanism (described above) is frequently considered to account for this apparent contradiction, but at least so far, no physical interaction between Prx and PTPs has been described (617).

Although the Prx redox relay model has gained increased appreciation as described above, the other mechanisms of H₂O₂ signaling cannot be ruled out. Indeed, because the H₂O₂ reactivity can be increased two to three orders of magnitude by the CO₂/bicarbonate pair (656), the mediation of Prx or Gpx enzymes may not be required in some circumstances. Notably, bicarbonate concentrations in mammalian tissues are generally very high (in the millimolar range), as CO₂ is a product of energetic metabolism (19). This stimulating effect of the CO₂/bicarbonate pair on H₂O₂ reactivity is likely related to the formation of peroxymonocarbonate (HCO₄ ⁻). In equilibrated H₂O₂ and CO₂/bicarbonate solutions and at neutral pH, HCO₄ ⁻ generation depends on the slow perhydration of CO₂ (31). Indeed, at physiological conditions, the CO₂/bicarbonate pair can accelerate the oxidation of signaling protein such as PTP1B (139, 761) and other slow reacting thiols, such as bovine serum albumin (652) and AhpE (548). Furthermore, the hyperoxidation of 2-Cys Prx can also be accelerated by the CO₂/bicarbonate pair (513, 656), allowing the flood gate process to occur, preserving the reduced Trx pools.

The activity of Prx enzymes can be inhibited not only by hyperoxidation but also by phosphorylation. For instance, Prx isoforms can be phosphorylated at distinct residues, affecting (increase or inhibit) their peroxidase activities (592). Therefore, redox and phosphorylation pathways appear to be a two-way road. Oxidation of critical Cys residues affects the activity of kinases and phosphatases, whereas phosphorylation of each one of the six mammalian Prx changes their ability to remove H₂O₂.

Of note, the pool of Prx1 located at the plasma membrane can be phosphorylated at Tyr194 by Src kinase in response to stimuli such as EGF and PDGF, leading to its inactivation (719). Consequently, H₂O₂ locally rises, allowing the oxidation of nearby less reactive thiols, such as the catalytic Cys of PTPs. Therefore, in response to growth factor stimulation, H₂O₂ generation by NADPH oxidases (NOX1) in lipid rafts is necessary to simultaneously activate Src family kinases and inactivate PTPs sustaining further phosphorylation and inactivation of Prx1 (592). Therefore, NADPH oxidases, Prx1, Src kinases, and PTPs take part in a retroalimentation cycle in lipid rafts that operate in response to growth factor stimulation (592). In contrast, Prx2 is inactivated preferentially by hyperoxidation (719).

In summary, distinct thiol switches (RSH/RSSR; RSH/RSOH; RSH/RS₂OH) and various mechanisms of H₂O₂ signaling can co-occur in cells, possibly in a coordinated manner that can also interact with other pathways, such as those dependent on tyrosine phosphorylation. Understanding how these processes occur from a global perspective may represent a challenge in the redox field that could be circumvented using systems biology (SB) approaches.

1. Reversal mechanism

The most studied enzymes that carry out protein S-deglutathionylation are Grxs (213). As already described, Srx can deglutathionylate specific Prxs at specific Cys-SG sites (92, 496). GSTO1-1 was reported as a deglutathionylating enzyme in epithelial breast cancer cells with specificity toward actin (422). Amazingly, when the same cells were exposed to S-nitrosoglutathione (GSNO), GSTO1-1 was associated with the S-glutathionylation of several proteins because, as mentioned above, some S-nitroso proteins are labile to subsequent S-glutathionylation (233). Importantly, a GSTO1-1 polymorphism has been associated with several diseases (422). Evidence of enzymatic specificity for S-deglutathionylation comes from human Prx1, in which three out of four Cys residues are found to be S-glutathionylated. Two Cys residues are preferentially deglutathionylated by Srx and the last one by Grx (496). Despite these examples, more direct evidence for the specificity of enzymatic deglutathionylation is missing in the literature.

2. Methods to detect protein S-glutathionylation

Several techniques are used to detect S-glutathionylated proteins. The first methods described were based on radio- and immune-labeling (227) and the utilization of biotinylated GSH (77). The most promising approaches are based on trapping the reduced protein Cys residues with alkylating agents, followed by treatment with Grx or reducing agents (e.g., dithiothreitol [DTT]) to remove the –SSG moiety and subsequent exposure of the free protein sulfhydryls to fluorescent probes or biotinylated thiols for detection. MS techniques allow the identification of the protein and the Cys-containing sequences. These methodologies are also able to map linear protein motifs prone to S-glutathionylation (729). However, the few MS studies so far do not permit mapping the motifs, if any, prone to S-glutathionylation.

3. S-glutathionylation as a modulatory mechanism of protein function

The number of proteins found to be S-glutathionylated is not considered significant when considering the total proteome of mammalian cells (240). Among them, there are well-documented examples where the S-glutathionylation is associated with the modulation of protein function coupled to general intracellular or compartmental redox shifts. In these situations, protein S-glutathionylation functions as a regulatory cell response mechanism. As established elsewhere (424), some of the criteria necessary to fulfill this condition would include (i) specificity of S-glutathionylation toward the protein Cys residue; (ii) occurrence of S-glutathionylation at high ratios of GSH/GSSG; (iii) S-glutathionylation occurrence at physiological conditions upon a stimulus, altering a functional or physiological endpoint; (iv) existence of a rapid and efficient mechanism for reversing the reaction thereby re-establishing the initial physiological condition when the stimulus ceases. Among the proteins that fulfill these criteria, examples include Prx, as mentioned above and the yeast catalytic unit of the 20S catalytic unit of the proteasome where 2 out of 32 Cys residues are S-glutathionylated in vivo, promoting the opening of the 20S catalytic chamber and increasing the degradation of oxidized proteins. This process was shown to be in vitro reversed by Grx2 and Trx1 or Trx2 and dependent on the intracellular redox status (160, 587, 588). In addition, many other proteins play a regulatory role when S-glutathionylated. Examples include several mitochondrial proteins that adjust the energy metabolism (398), ER proteins to cope with Ca²⁺ metabolism (754), and Fas or CD95, members of the TNF family of death receptors in the amplification of cell death pathways (15).

4. Physiopathological processes associated with protein S-glutathionylation

The S-glutathionylation of actin and hemoglobin has been proposed to be biomarkers in patients with Friedreich ataxia, diabetes, hyperlipidemia, and uremia (480, 629). A GSTP polymorphism found in European ancestors (259) alters the substrate-binding site of the enzyme and is considered a risk factor for a response to redox imbalance in at least one example provided by the recycling of Prx6 [(403); see above, the role of GSTP in Prx6 recycling]. The studies reported in cellular models (403) suggest that the GSTP polymorphism implies that individuals will have potential differences in the antioxidant response based on Prx6 activity, impacting the protection of cell membranes against lipid peroxidation, as Prx6 acts toward lipid peroxides. As GSTP interacts with several proteins, there are some speculative interpretations about the consequences the GSTP polymorphism has on protein S-glutathionylation (240). Therefore, more investigation is necessary to clarify that point. The S-glutathionylation of specific substrates is believed to involve the immune response and allergic inflammation (754). The DNA binding of the S-glutathionylated tumor suppressor p53 has been found inhibited in human cancers, a fact that is expected to determine a relationship between redox control of p53 in cancers and healthy tissues (674).

As pointed out, protein S-glutathionylation is promoted by a redox buffer (GSH), interacting with an important redox-sensitive protein residue (Cys). Moreover, as it is reversible, it represents a recycling mechanism. Hence, protein S-glutathionylation might be considered a prominent redox mechanism of protein modulation and signal transduction. However, impaired protein function is also caused by oxidative stress due to ROS and RNS, as is the case with eNOS activity (111, 404). In this example, there is a disruption in the coupling of NADPH oxidase to eNOS, resulting in decreased NO synthesis and superoxide generation, thus suggesting an angiotensin II-induced impairment of endothelium-dependent vasorelaxation (214).

1. Mechanisms involved in protein aggregation induced by protein oxidation

Tanase et al. (632) analyzed aggregates from bone marrow and splenic cells from 3-, 12-, and 22-month-old mice. They found a significant association between carbonylation of the proteome and the accumulation of high-molecular-weight protein aggregates with age. More than that, the authors found that over 90% of the carbonylated proteome was aggregated, suggesting that protein carbonylation is intimately linked to age-induced protein aggregation (632). Aggregates from aged mice had different carbonyl modifications, especially aminoadipic acid and glutamic semialdehyde, generated from the oxidation of lysine and arginine, respectively. The aggregates were also enriched in kynurenine and oxolactone, which are tryptophan oxidation products, and pyrrolidone and pyrrolidinone, which are generated from proline oxidation (632). The cellular turnover of aggregates is slower than of nonaggregated proteins. Thus, it would be interesting to know whether there is a causal relationship between protein oxidation and aggregation or whether increased oxidation is simply due to slower aggregate turnover.

Experimental data clearly show that Met oxidation is intimately linked to human prion protein (PrP) aggregation (540, 714, 744). As shown, sodium periodate-induced Met oxidation increased the aggregation of recombinant human prion protein (rhPrP) proportionally to the number of oxidized Met residues (human PrP has nine of them) (714). Interestingly, when PrP Met was substituted by methoxinine (a hydrophilic chemically stable Met) to mimic oxidized Met residues, a stronger aggregative behavior was observed. The opposite behavior was obtained when norleucine (a hydrophobic and chemically stable analogue that mimics reduced Met) was used (714), indicating that Met oxidation is the main trigger of PrP aggregation. Similar results were obtained in a yeast model of prion disease. Oxidative stress conditions promoted by incubation with hydrogen peroxide, menadione, or silencing of antioxidant enzymes increased the amount of aggregation of the yeast PrP. The presence of aggregates was accompanied by Met oxidation, and both were significantly reduced by overexpression of Met sulfoxide reductase (179).

Met oxidation was also described in amyloid aggregates composed of apoA-I (106, 716) or α-synuclein (282, 521). In the latter case, Met oxidation was shown to enhance the formation of soluble oligomers (757), arguably the main toxic form of aggregate in PD (300). The oxidation of each one of the four Met present in α-synuclein was shown to increase the polarity of the protein, elevating the degree of unfolding and preventing its organization into mature fibrils, locking aggregated α-synuclein in the soluble oligomeric form. Of note, the degree of fibrillation inhibition (i.e., the assembly of α-synuclein into amyloid fibrils) was found to be proportional to the number of oxidized Met residues (282, 521). Interestingly, dopamine-quinone, a product of dopamine oxidation, was shown to form adducts with α-synuclein and was also suggested to inhibit fibrillation (757). Synuclein oligomerization was also shown to be induced by peroxynitrite in the presence of CO₂ or MPO in the presence of H₂O₂ (601), which are possible in vivo aggregation-inducing agents. Finally, the formation of toxic oligomers was also found to occur after the covalent modification of α-synuclein by 4-hydroxy-2-nonenal (HNE) (527).

Oxidation of SOD1 Trp and His residues was shown to occur and be involved in the aggregation of the wild-type human enzyme. While Trp oxidation was the result of the enzyme bicarbonate-dependent peroxidase activity (125, 749), His oxidation to 2-oxohistidine was obtained by MCO (CuCl₂ in the presence of ascorbic acid) (536). More recently, Dantas et al. (143) reported that one of the products of cholesterol oxidation, cholesterol 5,6-secosterol aldehyde B, was able to covalently modify SOD1 Lys leading to significant aggregation of the enzyme. Of note, secosterol aldehydes were also shown to promote β-amyloid peptide (670) and α-synuclein (66) aggregation.

Different states of Cys oxidation were described as being involved in protein aggregation. For instance, the folding and stability of the FF domain of the yeast URN1, a model protein, were significantly impacted upon irreversible oxidation of a free Cys residue to sulfonic acid. Indeed, the oxidative modification of this single residue was sufficient to induce the protein aggregation into amyloid structures (407). The establishment of intermolecular adventitious disulfide bonds between two RRM2 domains of the transactive response DNA-binding protein 43 (TDP-43) was described as a determinant for dimerization, unfolding, and aggregation of the TDP-43 protein in vitro (127, 531). Interestingly, it was proposed that those adventitious bonds impair the ability of the RRM2 domain to refold, thus facilitating misfolding and ensuing aggregation (531).

2. Cellular consequences of aggregate accumulation

Highly oxidized and crosslinked proteins were shown to be a significant part of the brownish age pigment lipofuscin. However, the amino acids involved in such crosslinks were not described (280, 487). The mechanisms through which lipofuscin arises are still debated (445), but a failure in the cellular proteostasis system certainly contributes to its accrual (543). Importantly, lipofuscin was shown to incorporate different metals (up to 2%), which afford lipofuscin the ability to produce ROS, further oxidizing proteins and lipids and contributing to accumulation in cells (543). The ability of lipofuscin to inhibit the proteasome (281) exacerbates the accumulation of oxidized proteins and leads to the accumulation of other proteasome substrates such as the proapoptotic protein Bax, consequently facilitating apoptosis (525).

Protein aggregates may disrupt cell homeostasis by promoting the loss of function of the aggregated proteins and/or acquiring a toxic gain of function. The mechanisms through which aggregates exert their toxic effects are far from being completely understood. However, they seem to be dependent on the supramolecular organization of the aggregate (81, 95), the cellular environment, and the cell type. However, some features have been reported in a number of situations in which toxic aggregates are present, such as the failure of the ubiquitin/proteasome system (UPS) (43, 48, 280). Although the mechanisms through which protein aggregates inhibit the UPS are still elusive (281, 381), part of the inhibition may be due to the proteasome sequestration into aggregates, as observed after brain ischemia (223). UPS inhibition is adeterminant for the toxic effects of aggregates. Indeed, aggregate dissolution brought about by HSP 104 overexpression in yeast increased the activity of the proteasome without affecting its levels and normalized the degradation of proteasome substrates in aged yeast cells. Importantly, protein disaggregation allowed an increase in the replicative life span of yeast cells, indicating that the accumulation of aggregates shortens the longevity of this organism (16).

Damage to membranes was also reported as part of the toxicity mechanism of α-synuclein aggregates, although the molecular details are still unclear (300). Importantly, aggregation-prone proteins, which play essential roles in cellular functions, were shown to be sequestered by amyloid-like aggregates in HEK293T cells, which undoubtedly contribute to tissue aging as described in C. elegans (289).

Most of the data gathered so far on protein aggregates used in vitro systems. While valuable to uncover possible formation mechanisms and other crucial molecular details of aggregates, in vitro data may not reflect in vivo situations as aggregate accumulation in living organisms is the result of an interplay between the proaggregative stimuli and the activity of the proteostasis system; thus, it is essential to study protein aggregates in vivo. Those studies will undoubtedly contribute to the understanding of the relative contribution of aggregates to pathologic states and organismal aging, a topic that is actively being investigated.

Despite the vast array of evidence relating oxidation-induced protein aggregation to pathological states, it is important to emphasize that oxidation-induced protein aggregation may have protective or functional purposes. As an example, elegant works by Benjamin Tu's group showed that the yeast ortholog of ataxin-2 (Pbp1) coordinates the target of rapamycin complex 1 (TORC1) signaling and autophagy in response to mitochondrial activity (734). In conditions of intense mitochondrial respiration, Pbp-1 protein organizes into puncta composed of labile β-polymers. Upon addition of small concentrations of hydrogen peroxide, specific Met residues present on a low complexity domain of Pbp1 are oxidized, leading to β-polymer disassembly and subsequent TORC1 activation and autophagy inhibition. The reduction of the oxidized Met residues reverses the process (322). Interestingly, the organization of Pbp1 into labile β-polymers has a functional purpose and depends on the redox state of Met residues located at the low complexity domain, as shown by site-directed mutagenesis (322). For more examples of functional or protective roles of redox-dependent aggregation, the reader is referred to the review by van Dam and Dansen (141).

3. Human pathologies associated with oxidized protein aggregation

Age-associated cataract is a prototypical example of a disease caused by high-molecular-weight protein aggregates, which lead to lens opacity. Oxidative modification of amino acid residues of crystallin proteins was reported to play a significant role in cataract etiology (444). Indeed, ∼90% of protein sulfhydryl groups are lost in lenses with advanced cataracts (655). Moreover, oxidation of Trp 9 and Met 1 are among the major PTMs detected in aggregated α-crystallin, a protein present in all insoluble aggregates in lenses (606). Finally, Trp-Trp and Trp-Tyr crosslinks were recently characterized in human nuclear cataracts (Fig. 6A) (498).

The presence of protein aggregates in brain tissue is a hallmark of many NDs, such as Alzheimer's, Parkinson's, and Huntington's, among others (see for instance Fig. 6B). In the same way, high levels of oxidized biomolecules were reported in postmortem brains of patients with NDs (113). Considering the data obtained in vitro showing that oxidation is a destabilizing factor for protein folding, it is likely that protein oxidation contributes to the abnormal protein deposits found in the brains of patients with NDs. However, the etiology of NDs is still elusive, and no work to date has demonstrated a direct relationship between the oxidation of proteins and their abnormal deposit in the brain or the development of disease in humans. More studies are required to define the participation (if any) of oxidized proteins and their aggregation in the setting of NDs.

1. Proteasome

The description of the polyubiquitination of proteins as a PTM that directs them to degradation in the early 1980s was a significant achievement to the understanding of proteostasis (121, 122, 275). Afterward, the discovery of the proteolytic complex responsible for the recognition and degradation of polyubiquitinated proteins, namely the 26S proteasome (26SPT), was reported (697). The 26SPT is composed of a central cylindrical-shaped catalytic unit called the 20SPT, coupled on one or both sides to the most abundant regulatory unit, the so-called 19S unit (316, 681). The 19S regulatory unit is responsible for recognizing polyubiquitinated proteins and their subsequent deubiquitination, unfolding, and translocation to the catalytic 20S unit. The polyubiquitination machinery, together with the 26S protease, is presently referred to as the UPS.

Initially, it was assumed that the 20SPT was permanently coupled to the 19S regulatory unit, degrading exclusively polyubiquitinated proteins. Some investigators challenged this proposition about the degradation of oxidized proteins. The 20SPT was mentioned as an independent particle called the “multicatalytic proteinase complex” (707) that was associated with the degradation of oxidized and nonubiquitinated proteins (150, 187, 316). Throughout the last decades, free 20SPT is increasingly accepted as an essential player in the degradation of oxidized proteins through a ubiquitin- and ATP-independent pathway. However, its role in vivo is not fully appreciated (159). A debate is still in place on whether oxidized proteins are polyubiquitinated or not before degradation. Indeed, there is no described biochemical feature that precludes oxidized proteins from being polyubiquitinated. Therefore, the question is whether the prompt degradation of oxidized protein substrates by the 20SPT could bypass their polyubiquitination and subsequent degradation by the 26SPT. In other words, there are no data to consider which mechanism would prevail on kinetic grounds. The mechanism proposed and explored in vitro for the recognition of oxidized proteins by the 20SPT relies on the exposure of hydrophobic patches due to structural rearrangement upon oxidation (149). A recent study with yeast cells exposed to oxidative stress revealed that 50% of highly abundant oxidized proteins were polyubiquitinated (405). However, this study does not discuss the fate of the remaining 50% of the oxidized proteins, suggesting that distinct mechanisms of proteolytic degradation of oxidized proteins can coexist.

Reported alterations of the UPS upon oxidative stress, such as the uncoupling of the 20S catalytic unit from the 19S regulatory unit, the increased expression of the regulatory unit 11S, and decreased polyubiquitination, are interpreted as adaptive responses that UPS uses to cope with oxidative shifts (344) and highlight the role of the 20SPT catalytic unit as the main player in the degradation of oxidized proteins. Nonetheless, all the investigations in the field use oxidative challenges to cells that do not necessarily resemble aging or degenerative processes, in which protein oxidation proceeds at a slow pace. It has been very challenging to design experimental conditions related to these human pathologies. Indeed, a large amount of data are available, showing that proteasomal activation is associated with extended life span (117). On the other hand, proteasomal activity declines with age and in age-related pathological conditions, with consequent accumulation of oxidized proteins that correlate with a decrease in the expression of antioxidant enzymes in aged cells (271, 493, 520). The decline of proteasomal activity in those conditions is widely reported. However, it is unclear whether it is because of its decreased expression or increased PTMs, including oxidative damage. Major oxidative modifications reported are modifications by HNE, the formation of carbonyl moieties, and glycation of the 20SPT catalytic unit, implying inhibition of protein degradation (35). The proteasome likely has a high recycling rate that is decreased during aging and in age-related pathologies. The most significant alteration to compromise protein degradation under these conditions would be the decreased proteasome expression rather than inhibition through oxidative modifications.

In conclusion, regardless of the underlying mechanism, the proteasome is a critical component of the quality control systems involved in the degradation of oxidized and misfolded proteins.

2. Lon protease

Mitochondria are the primary intracellular source of O₂ ^•− and H₂O₂ (197, 455). As a consequence, the mitochondrial proteome is continuously exposed to oxidative damage. Oxidatively damaged mitochondrial proteins represent a risk to the organelle and need to be repaired or degraded. In human mitochondria, the LONP protease (also known as LomP1 or simply LON) is considered the main component of the mitochondrial PQC system. It plays a central role in the degradation of oxidatively modified proteins (683, 684).

LON is a highly conserved ATP-dependent protease localized in the mitochondrial matrix compartment. While specific membrane-integrated proteases (namely, m-AAA protease and i-AAA protease) cleave damaged mitochondrial proteins of the inner membrane, the LON protease preferentially degrades soluble proteins in the matrix (166). The LON substrates include nonassembled polypeptides that arise from incorrect folding during mitochondrial protein biogenesis and proteins damaged by oxidants (633). The yeast ortholog of the mammalian LON, Pim1, has been widely studied as a model to uncover LON protease activities. Pim1 is a large homo-oligomeric protein complex composed of seven 100 kDa subunits organized in a ring-shaped heptameric structure (611). The proteolytic domain responsible for the hydrolysis of peptide bonds is located inside the internal proteolytic cavity or chamber, shielded from the aqueous environment. Therefore, similar to other proteases such as the proteasome complex, protein degradation by LON requires prior unfolding and translocation of the client protein into the proteolytic cavity. This task is performed by the ATPase domain situated at the edges of the complex (516, 684).

Bender et al. (44) investigated the role of Pim1 in the maintenance of mitochondrial proteostasis under oxidative stress conditions. By using a proteomic approach, the authors identified a set of mitochondrial proteins that are mainly degraded by Pim1 in response to the exposure of isolated mitochondria to three distinct oxidative conditions: H₂O₂, menadione, and succinate combined with antimycin A (an inhibitor of electron transfer from ubiquinone to complex III). Although some proteins were found to be degraded only under a single type of oxidative treatment, Fe-S cluster-containing enzymes were the primary targets of degradation. It was proposed that after the oxidative modification of their Fe-S clusters, the corresponding polypeptides become destabilized and thus more prone to degradation by Pim1 (44). This observation is consistent with previous studies showing that Fe-S clusters containing enzymes are major targets of oxidative modifications (42, 68, 89, 400).

However, protein oxidative damage is not restricted to Fe-S-containing enzymes. The mechanism proposed is based on the generation of partially unfolded protein regions upon oxidation, which are recognized by LON (41, 306, 400, 484). Hence, the conformational state of a polypeptide chain serves as the primary criterion for its selection as a protease substrate. This idea is further supported by a study that revealed that Pim1 requires an unstructured segment longer than 50–60 amino acids at the N-terminal to initiate protein degradation (306). These unstructured segments can access the interior of the Pim1 proteolytic chamber, allowing degradation to occur. In this scenario, tightly folded substrates representing native enzymes remain mostly resistant to proteolysis. As a result, the proteolysis reaction is selective to unfolded or damaged polypeptide chains.

As mentioned above, LON plays a crucial role in the control of mitochondrial proteostasis. This statement is highlighted by the fact that homozygous Lomp1 knockout mice exhibit embryonic lethality (528). It has been known for a long time that LON activity declines in aged animal models, which is accompanied by the accumulation of oxidatively modified proteins in the mitochondria (70, 360). This decline may contribute to the progress of various age-associated diseases, such as NDs (69, 668).

3. Autophagy

Although autophagy is mainly studied as a catabolic process involved in the degradation of intracellular components during starvation conditions, it is now clear that this process participates in many other pathways necessary for cellular homeostasis (191, 349). In addition to a nonselective process of self-consumption (nonselective autophagy), it has been demonstrated that some cargoes can be degraded with high selectivity in response to diverse conditions (335). Numerous studies have reported the selective autophagic degradation of damaged or superfluous organelles, including mitochondria (321, 468, 469, 743), peroxisomes (165, 192, 653, 753), and ER (330, 430, 467).

Although the degradation of organelles is the best-described type of selective autophagy to date, some evidence indicates that intracellular protein aggregates can also be selectively degraded in a process known as aggrephagy (59, 293, 387, 388, 491, 569). The selective incorporation of protein aggregates into the autophagosome requires selective autophagy receptors (SARs) that bind simultaneously to the protein aggregate cargo and the core components of the autophagic machinery. Examples of these proteins include the p62/SQSTM1 (sequestosome 1), the next to BRCA1 gene 1 (NBR1), and optineurin (OPTN) (222, 311, 618).

The most studied ubiquitin-dependent SAR in the context of aggrephagy is p62. For many years, p62 was considered a common component of ubiquitin-positive protein aggregates, such as Mallory and Lewy bodies, neurofibrillary tangles (β-amyloid), and Huntingtin aggregates (335, 353, 354, 460, 748). Subsequently, it was demonstrated that the recruitment and oligomerization of p62 into ubiquitinated aggregates are required for efficient cargo degradation by the autophagic machinery (59). The trafficking of ubiquitinated protein aggregates into autophagosomes depends on specific motifs present in the p62 structure. This protein simultaneously interacts with the MAP1LC3/GABARAP (ATG8 in yeast cells) family of proteins (a core component of autophagosome membrane) and ubiquitin via the LC interacting region (LIR) motif and ubiquitin-binding (UBA) domain, respectively (491). As a consequence of these interactions, the ubiquitinated protein aggregates are selectively sequestered and delivered to the autophagosome.

In parallel to the description of p62, numerous other SARs that recognize ubiquitinated cellular cargos and LC/GABARAP (Atg8) were described, including the Cue5 from yeast cells, which is required for clearance of aggregation-prone proteins (388). The fact that these SARs recognize ubiquitin molecules attached to cargos led researchers to propose that ubiquitin is the primary signal that cells use for discriminating which cargo is degraded by autophagy (329, 336, 345, 556). Considering that the UPS also uses ubiquitin as a degradation signal for substrates, it was initially thought that both pathways could degrade the same type of substrates. However, it is now acknowledged that they exhibit a high degree of specificity involving selective enzymatic reactions and discriminatory receptors, ensuring delivery of the correct substrate to either the proteasome or lysosome. Substrate size plays an important role in pathway choice. Soluble monomeric proteins are mostly conducted to the UPS. In contrast, larger protein complexes such as protein aggregates, which are unable to enter into the narrow channel of the proteasome and thus resistant to proteasomal degradation, are degraded by the autophagic machinery (518). Moreover, the arrangement of the ubiquitin chains attached to the cargo might play an important role in substrate fate. While Lys 48-linked ubiquitination preferentially targets substrates to proteasomal degradation, Lys 6-linked chains serve as a recognition signal for multiple autophagy receptors (175).

In addition to its function as an SAR, p62 can also escort ubiquitinated substrates to the proteasome by acting as a proteasomal shuttle factor via its Phox and Bem1p (PB1) domain (556). The structural architecture of p62 determines which pathway is used by p62-delivered substrates. While p62 dimerization leads substrates to the proteasome, oligomerization of p62 into ubiquitinated cargos promotes substrate channeling to autophagy (722). The underlying mechanisms that control the switch of this receptor between the dimeric and oligomeric states have been recently unraveled. Under a condition termed ubiquitin stress (a condition characterized by an accumulation of free ubiquitin, as observed upon heat shock or prolonged proteasome inhibition), p62 is ubiquitinated, which disrupts the dimerization of its UBA domain and facilitates its ability to recognize polyubiquitinated cargoes for selective autophagy (506). This model is also supported by a recent work showing that oxidation of p62 redox-sensitive Cys residues in response to oxidative stress induces the formation of disulfide-linked conjugates, which facilitates p62 oligomerization with consequent induction of the autophagy pathway (98). Importantly, oxidized p62-dependent autophagy activation seems to be essential for cell survival in the context of oxidative stress associated with age-related diseases.

In conclusion, the data cited above suggest a functional connection between UPS and autophagy pathways toward ubiquitinated protein degradation. Autophagy would be a compensatory mechanism when the proteasome is overwhelmed or inhibited (175, 415, 518).

Disturbances in PQC pathways are clinical hallmarks of aging and many age-related diseases, including NDs. There is a decline in these pathways during aging with direct consequences, including damaged protein accumulation. On the contrary, accumulating evidence indicates that interventions that stimulate the UPS and autophagy can prolong the maintenance of a healthy proteome and consequently increase longevity (368). Recent findings in the same direction are discussed in Section V.B.2.

Figure 11 illustrates the fate of oxidized proteins, including aggregation and proteolysis.

OPEN IN VIEWER

FIG. 11.

Fate of oxidized proteins. [1] Oxidized proteins present increased surface hydrophobicity because of the loss of tertiary and secondary structures that may determine their degradation [2, 3], rescue by chaperones [4], and possible refolding [5] or aggregation [6]. Aggregates may burden the proteostasis system leading to further protein aggregation, increasing their molecular mass and further modifications, including metal binding. The accumulation of aggregated proteins [7] is a hallmark of many neuropathological conditions and cataracts. Autophagy [8] is another important fate of aggregated proteins. Into the mitochondrial matrix [9], LON protease accomplishes the role of degrading oxidized mitochondrial proteins. Section IV describes and discusses each of the processes illustrated above. Color images are available online.

1. Antioxidants

Prevention of macromolecule oxidation relies on the maintenance of the redox homeostasis, whose physiological regulation is not yet completely understood. One of the most explored tools to counteract redox imbalance is the utilization of antioxidants, particularly natural sources. A search in the PubMed platform utilizing the term antioxidant therapy returned more than 11,800 reviews among 84,000 publications in the last decade. By taking only publications based on Human Randomized Controlled Trials presenting antioxidant therapy in the title in the last decade, about 4500 results emerged. In these publications, a set of nutritional proposals and specific compounds or natural products were tested and proposed as therapies against several pathologies. There is no consensus established by the scientific community on the real benefits of such therapies. The idea of the importance of healthier nutritional intake is gaining more and more traction, including advisements by worldwide public health services to the general and specific populations (e.g., scholars) to increase the intake of natural products (fruits and vegetables), even though the main goal is the control of obesity, diabetes, hypertension, and others. It is important to emphasize that such recommendations are not yet based on systematic scrutiny on their benefits. It is out of the scope of this review to explore the subject.

Noteworthy is the rereading of natural “antioxidants” based on the hormesis phenomenon. Instead of antioxidant protection, many phytochemicals induce cellular stress response (454). This is based on the fact that humans consume many toxic phytochemicals from natural sources at very small doses within the hormetic range. For many years it was assumed that the free radical scavenging property of phytochemicals would be responsible for their benefit as antioxidants. However, many epidemiological studies have failed to demonstrate the benefits of dietary supplementation. Instead, much evidence suggests that hormesis underlies the health benefits of phytochemicals (91). A general hypothesis is that these phytochemicals would induce adaptive cellular stress response pathways that induce the expression of genes encoding antioxidant enzymes, chaperones, detoxifying enzymes, and other protective proteins. Nevertheless, more experimental data are needed for a better understanding of the role natural products play in redox biology.

2. Elimination of oxidized proteins and aggregation prevention

Besides the loss of function, one of the most critical outcomes of protein oxidation is the proteotoxic event of aggregation, as discussed above (Section IV.A). Aggregation takes place if oxidized proteins are not promptly eliminated. Three intracellular systems accomplish the degradation of oxidized proteins: proteasome based or not on the labeling by polyubiquitin chains, autophagy, and the mitochondrial LON protease. However, those systems are usually compromised under the same pathological conditions; thus, the pool of oxidized proteins increases. There is an emerging field of investigation for proteasome activators as pharmacological tools (128, 313, 676, 677), but no drug has been clinically tested so far.

In the last years, a powerful technique has been developed based on the induction of protein degradation by proteolysis-targeting chimeras, namely PROTAC (84, 120, 464). The methodology is based on the utilization of small bifunctional molecules capable of interacting with the protein substrate and ligases to induce the polyubiquitination of the protein substrate, directing it for proteasomal degradation. Polyubiquitination is a three-step process where the final one is the PTM of the protein substrate by ubiquitin through E3 ligases that transfer ubiquitin to an Lys residue of the substrate either directly or indirectly. The PROTAC-mediated degradation is based on targeting the E3 ligase-specific substrate by degraders that combine affinity for both substrate and E3 ligase. This approach promotes the degradation of selected proteins as E3 ligases are substrate specific. Now, it has become an important tool in both experimental biology, for the understanding of cell regulation and signaling, and in drug discovery. Several PROTAC studies on the degradation of aggregate-prone proteins have been conducted (118, 215, 294, 589). Moreover, the same technique has also been used to direct proteins to the lysosome (628). This approach seems robust in the way it targets proteins considered undruggable so far and introduces a new perspective on pharmacological interventions (120). In principle, this approach might be extended to the degradation of any protein. In the case of oxidized proteins, as important as the identification of the oxidative process these proteins undergo is identifying the proteins prone to oxidation in all pathologies where they accumulate. That might be a possibility to intensify their degradation and establish either their pathological role or avoid proteotoxicity.

Chaperone-like molecules might be an important tool for avoiding the aggregation of oxidized proteins. Although numerous proteins were described to have moonlighting chaperone-like activity (315, 593, 624), some were described to prevent protein aggregation in an oxidative scenario. One example is the 19-amino acid sequence derived from α-crystallin, which prevents the aggregation of UV-irradiated or H₂O₂-treated γ-crystallin proteins in vitro (351). The same group of researchers modified the original 19-amino acid sequence giving rise to a set of short peptides with varying degrees of chaperone-like activity, some of which have improved activity compared with the original sequence (535). Peptides derived from α-crystallin are taken up by primary human fetal retinal pigment epithelium cells in culture and were able to prevent apoptosis induced by H₂O₂ (605). In vivo assays also demonstrated that the peptides prevented selenite-induced cataracts when administered intraperitoneally to rats (463), evidencing the therapeutic potential of these molecules.

Another compound with interesting chaperone-like properties is polyphosphate. Polyphosphate is made up of a thousand phosphate monomers linked through phosphoanhydride bonds. Numerous functions were reported for polyphosphate in a wide range of organisms, including humans (723), and some authors argue that polyphosphate is a primordial chaperone (238). Both in vitro and in E. coli exposed to conditions that lead to oxidative protein unfolding, polyphosphate was able to stabilize those proteins, significantly decreasing the amount of protein aggregation (238). Moreover, polyphosphate significantly accelerates the formation of mature fibrils from amyloid proteins in vitro, effectively inhibiting amyloid cytotoxicity (135). While polyphosphate has significant effects in vitro, its therapeutic potential in vivo remains unexplored.