Autoxidation of ascorbate mediates lysine N-pyrrolation

Abstract Protein N-pyrrolation, which converts lysine residues to N ε-pyrrole-l-lysine (pyrK), is a naturally occurring covalent modification. The pyrrolated proteins have a unique property of binding to DNA-staining agents, such as SYBR Green I (SG), and anti-DNA antibodies, suggesting a physiologically relevant modification that gives rise to DNA mimic protein. These properties of pyrrolated protein are suggested to be associated with innate and autoimmune responses. Short-chain aldehydes derived from lipid peroxidation are thought to be involved in the formation of pyrK. We now report that similar lysine N-pyrrolation also occurs during the metal-catalyzed oxidation of proteins with ascorbate. When human serum albumin (HSA) was incubated with Fe2+/ascorbate in the presence and absence of docosahexaenoic acid, the protein was converted to SG-binding proteins even without the polyunsaturated fatty acid. The formation of SG-binding proteins by Fe2+/ascorbate was accompanied by the formation of pyrK, which was also detected in ascorbate-treated hemoglobin. Moreover, the metal-catalyzed oxidation of ascorbate produced the pyrrolation factors, glycolaldehyde and glyoxal. These results and the observations that sera from autoimmune-prone MRL-lpr mice recognized modified proteins with Fe2+/ascorbate and with glycolaldehyde/glyoxal suggest that the autoxidation of ascorbate, as well as lipid peroxidation, can be a source of autoantigenic N-pyrrolated proteins. Our findings revealed a possible function of ascorbate as an endogenous source of pyrrolated proteins and suggested that the pyrK residues generated in proteins may play a role in the innate and autoimmune responses associated with the oxidative metabolism of ascorbate.


Introduction
Ascorbate is a powerful antioxidant and can neutralize harmful free radicals. Although the direct antioxidant protection provided by ascorbate is limited to watersoluble environments, it plays an important antioxidant role in lipids through the regeneration of the fat-soluble antioxidant vitamin E [1,2]. When acting as an antioxidant, ascorbate is readily oxidized to dehydroascorbate. Due to its instability under physiological conditions, dehydroascorbate undergoes an irreversible ring-opening to produce 2,3-diketogulonic acid [3], which is further converted to degradation products, such as erythrulose, 3-deoxythreosone, and threosone [4] (Supplementary Scheme S1). This decomposition cascade of ascorbate is accelerated by the oxidized forms of the metals, such as Fe 3þ and Cu 2þ . On the other hand, many of these ascorbate-derived products are reactive aldehydes that can covalently modify lysine residues of proteins to form heterogeneous groups of molecules called advanced glycation end products (AGEs) [5][6][7]. High ascorbate levels in various tissues suggest that dehydroascorbate and its degradation products may be causally involved in the formation of AGEs in vivo. These processes are thought to be involved in the pathogenesis of many human diseases, such as diabetes and atherosclerosis, as the formation and accumulation of AGEs have a profound effect on the structure and function of proteins [8,9].
Protein N-pyrrolation, which converts lysine residues to N e -pyrrole-L-lysine (pyrK), was recently discovered by Miyashita et al. [10]. The uniqueness of this covalent protein modification is that the N-pyrrolation of lysine residues (Scheme 1(A)) are associated with an increase in the net negative charges of the proteins due to the modification of epsilon-amino group. More interestingly, proteins acquire an electrical conductivity through pyrrolation. It has been suggested that the underlying mechanisms of these electronic properties are due to stacking interactions between the pyrrole rings and/or between the pyrrole ring and aromatic amino acid residues [10]. Due to these characteristics, N-pyrrolated proteins have been claimed to be DNA mimetic molecules. In fact, DNA-binding molecules, such as the anti-DNA autoantibodies and DNA intercalators, recognize N-pyrrolated proteins, and the immunization of mice with N-pyrrolated proteins causes the enhanced production of anti-DNA autoantibodies [10]. Moreover, pyrK has been detected in mouse renal immune complex deposits using a sensitive and specific LC-MS/MS-based method [10], indicating that lysine N-pyrrolation is a physiologically relevant reaction. Recently, to understand the potential physiological and pathophysiological significances of the lysine N-pyrrolation, we searched for pyrrolation factor(s) generated during the peroxidation of polyunsaturated fatty acids (PUFAs) and identified glycolaldehyde as a source of pyrK [11]. We also showed that the glycolaldehyde-mediated N-pyrrolation of lysine could be promoted by the addition of glyoxal, an oxidized product of glycolaldehyde and proposed a mechanism by which the N-pyrrolation of lysine may proceed via two sets of nucleophilic addition reactions: (i) between a primary amino group of the lysine residue and glycolaldehyde and (ii) between a secondary amine derivative (Amadori product) and glyoxal (Scheme 1(B)) [11]. We now report the observation that proteins are converted to DNA mimetic molecules during ascorbate-mediated protein modification, independent of the presence of PUFAs. We also report that pyrK, previously described as a lipid peroxidation product, is formed during lysine modification by oxidized products derived from ascorbate.
Staining of nondenaturing polyacrylamide gel by SYBR Green I HSA was incubated with ascorbate or pyrrolation factors at 37 C for indicated times. The pyrrolated HSA (pyrHSA) was prepared by the incubation of HSA (1 mg/ml) with 1,4-butanedial (BDA, 1 mM) overnight in PBS at 37 C. The buffer was then exchanged with PBS by ultrafiltration (Amicon Ultra filter, 10 kDa MWCO, Merck Millipore, Burlington, MA). The protein samples were electrophoresed through a 10% nondenaturing polyacrylamide gel using Tris-glycine (pH 8.4) buffer as a running buffer. After electrophoresis, the gels were rinsed with ultrapure water and stained with SYBR Green I (SG) (Takara Bio Inc., Otsu, Japan) dissolved in TAE buffer (pH 8.3) (10,000 times dilution) for 30 min. Fluorescence scanning of the gel was carried out using a Safe Imager Transilluminator (Invitrogen, Carlsbad, CA). After the SG staining, the gel was fixed and stained with Coomassie Brilliant Blue (CBB). (100 mM) in 50 mM sodium phosphate buffer (pH 7.4) at 37 C for 12 h. After incubation, the oxidation products were reacted with NAK (1 mM) in 50 mM sodium phosphate buffer (pH 7.4) at 37 C for 18 h. The pyrNAK standard was prepared by the incubation of NAK with BDA overnight at 37 C. The formed pyrNAK was detected by LC-ESI-MS/MS.
Formation of pyrNAK in the reaction of NAK with ascorbate and Fe 21 NAK (1 mM) was reacted with ascorbate (10 mM) and FeCl 2 (100 mM) in 50 mM sodium phosphate buffer (pH 7.4) at 37 C for the indicated times. Ascorbate-and iron-dependent formation of pyrNAK was investigated by the reaction of NAK (1 mM) with ascorbate (0-1 mM) in the presence of FeCl 2 (100 mM) and with ascorbate (10 mM) in the presence of FeCl 2 (0-100 mM) at 37 C for 24 h. The reaction mixtures underwent a quantitative analysis of pyrNAK using LC-ESI-MS/MS.
Lysine N-pyrrolation of HSA and human Hb HSA (0.1 mg/ml) was incubated with different concentrations of ascorbate at 37 C in 50 mM sodium phosphate buffer (pH 7.4) containing FeCl 2 (100 lM). Similarly, the human Hb (0.1 mg/ml) was incubated with different concentrations of ascorbate at 37 C in 50 mM sodium phosphate buffer (pH 7.4). After 72 h, the reaction mixtures were extracted with an equal volume of chloroform:methanol (1:3 (v/v)) and centrifuged at 6000Âg for 10 min following agitation by a vortex mixer. An interphase was formed, and the fluid above was removed, then the remaining solution was evaporated. The protein was suspended in 4 M NaOH, then hydrolyzed by heating under argon for over 15 h at 110 C following the addition of [U-13 C 6 , 15 N 2 ] N e -pyrK as the internal standard. The resulting solution was neutralized with 6 N HCl, and centrifuged at 6000Âg for 10 min. The supernatant was loaded onto an Oasis HLB extraction cartridge (Waters, Milford, MA), then washed with 1 ml of pure water, followed by elution with 1 ml of acetonitrile (20%). The eluted fraction was evaporated and dissolved in pure water and was applied to the LC-ESI-MS/MS for pyrK quantification.

Quantification of pyrK and pyrNAK by mass spectrometric analysis
The pyrNAK and pyrK were detected by a Xevo TQD triple stage quadrupole mass spectrometer (Waters, Milford, MA) equipped with an ACQUITY ultra-performance LC system (Waters, Milford, MA). The reaction mixture (injection volume of 10 ll) was separated using an ACQUITY UPLC BEH C18 1.7 lm column (2.1 mm Â 100 mm, Waters, Milford, MA) at the flow rate of 0.3 ml/min. The elution was performed by a discontinuous gradient using mobile A (0.1% formic acid) and B (0.1% formic acid in acetonitrile) under the following elution conditions: 1% B from 0 to 1 min, 30% B from 1 to 8 min, 99% B from 8 to 9 min. A mass spectrometric analysis in the positive ion mode was performed using the selected reaction monitoring (

Measurement of antibody titer in the sera from systemic lumps erythematosus (SLE)-prone mice
Female MRL-MpJ mice and female MRL-lpr mice were purchased from Japan SLC (Hamamatsu, Japan). The care and use of all the mice were performed in accordance with prescribed national guidelines, and the Animal Care and Use Committee of Nagoya University granted ethical approval for the study. The antibody titer against the Fe 2þ /ascorbate-treated HSA and double strand DNA (dsDNA) in the sera from the MRL-MpJ and MRL-lpr mice (n ¼ 5) was analyzed by ELISA as previously described [10]. Each antigen (50 lg/ml, 100 ll/well) was added to each well of a 96-well ELISA plate (Nunc, Roskilde, Denmark). A 100 ml aliquot of a 100-times dilution of serum was used as the first antibody.

Statistical analysis
All experiments for quantification were repeated for at least three different preparations of one experiment. All data are expressed as means ± S.D. Statistical significance was evaluated using the unpaired Student's t-test or, when appropriate, Dunnett's test or Tukey's test using GraphPad Prism 7 (La Jolla, CA).

Results
Ascorbate-mediated conversion of HSA to SGbinding proteins in the presence and absence of PUFAs The autoxidation of ascorbate by metal ions is often used to initiate lipid peroxidation. The reaction is associated with the production of various autoxidation products originating from ascorbate in addition to the lipid peroxidation products. Taking advantage of the fact that the N-pyrrolated proteins can be recognized by DNA-staining agents, such as SG, we evaluated the possibility of ascorbate being involved in the N-pyrrolation of lysine. Ascorbate and/or Fe 2þ were pre-incubated in the presence and absence of DHA for 12 h, then the reaction mixture was further incubated with HSA for 6 h. As shown in Figure 1, the protein was converted to the SG-binding proteins by ascorbate or Fe 2þ /ascorbate, regardless of the presence of DHA. The presence of DHA moderately enhanced the formation of the SG-binding proteins. These data suggest that lipid peroxidation is not necessarily essential for the lysine N-pyrrolation and that ascorbate, probably in the presence of metal ions, can mediate the conversion of HSA to the SG-binding proteins.
Lysine N-pyrrolation during metal-catalyzed autoxidation of ascorbate Next, we evaluated whether ascorbate could mediate the N-pyrrolation of lysine in the presence and absence of DHA. To this end, DHA was incubated with either or both Fe 2þ and ascorbate at 37 C for 12 h, then the reaction mixtures were incubated with N a -acetyl-Llysine (NAK) for an additional 18 h. Conversion of NAK to N e -pyrrolated NAK (pyrNAK) was monitored by LC-ESI-MS/MS (Supplementary Fig. S1). As shown in Figure  2(A), pyrNAK was formed when NAK was incubated with Fe 2þ , ascorbate, or Fe 2þ /ascorbate in the presence of DHA (Figure 2(A), right). pyrNAK was also formed when NAK was incubated with ascorbate or Fe 2þ /ascorbate in the absence of DHA (Figure 2(A), left). In addition, in agreement with the results of Figure 1, the effect of the DHA addition on the formation of pyrNAK was moderate, indicating that ascorbate present in the reaction mixtures is sufficient to mediate the N-pyrrolation of lysine. Figure 2(B) describes the time-dependent conversion of NAK to pyrNAK by Fe 2þ /ascorbate. The yields of pyrNAK also depended on the concentrations of ascorbate (Figure 2(C)) and Fe 2þ (Figure 2(D)). When NAK was incubated with ascorbate or Fe 2þ alone, only trace amounts of pyrNAK were formed.
To establish the formation of pyrK in proteins during the ascorbate autoxidation reactions, HSA was incubated with either or both Fe 2þ and ascorbate, and conversion of the lysine residues to the pyrK residues was monitored by LC-ESI-MS/MS. As shown in Figure 3(A), the incubation of HSA with ascorbate alone or Fe 2þ /ascorbate yielded the pyrK residues in the protein.
Ascorbate mediated the N-pyrrolation of the lysine residues in a concentration-dependent manner ( Figure  3(B)). The ascorbate-mediated lysine N-pyrrolation was also observed without the addition of Fe 2þ in Hb, an iron-containing protein (Figure 3(C)). These results indicated that the iron-catalyzed autoxidation of ascorbate and subsequent formation of oxidized products (pyrrolation factors) may be involved in the lysine N-pyrrolation (Figure 3(D)).

Formation of pyrrolation factors from Fe 21 /ascorbate
It was previously suggested that glycolaldehyde, an aldehyde derived from lipid peroxidation, and its oxidation product, glyoxal (Figure 4(A)) might be involved in the formation of pyrK [11]. To establish the mechanism of the ascorbate-mediated lysine N-pyrrolation, we attempted to detect these short-chain aldehydes during the iron-catalyzed autoxidation of ascorbate. Ascorbate was incubated with either Fe 2þ or DTPA (diethylenetriamine-N,N,N 0 ,N 00 ,N 00 -pentaacetic acid), a metal chelator, in sodium phosphate buffer (pH 7.4) at 37 C for 24 h. The aldehyde products generated during ascorbate autoxidation were derivatized with ABA followed by reduction with NaBH 3 CN. The ABA derivatives of glycolaldehyde (GA-ABA) and glyoxal    Figure 4(B,C) shows that, after 24 h of incubation, the yield of glycolaldehyde from Fe 2þ /ascorbate was significantly greater than that from ascorbate alone, whereas the addition of DTPA almost completely inhibited the glycolaldehyde formation. Similar results were also observed for the formation of glyoxal (Figure 4(D,E)). These results suggest that ascorbate may contribute to the formation of pyrK through the production of glycolaldehyde and its oxidation product, glyoxal, during the autoxidation reactions in physiological buffer (Scheme 2).

Lysine N-pyrrolation during metal-catalyzed oxidation of glycolaldehyde
Based on the detection of both glycolaldehyde and glyoxal from ascorbate, we investigated whether the iron-catalyzed glycolaldehyde oxidation is involved in the N-pyrrolation of lysine. As shown in Figure 5(A), the addition of Fe 2þ to the reaction of NAK and glycolaldehyde showed a marked increase in the pyrNAK formation. The formation of pyrNAK was steadily enhanced by Fe 2þ in a concentration-dependent manner ( Figure 5(B)). The glycolaldehyde-mediated pyrrolation of NAK was partially inhibited by DTPA ( Figure  5(C)). In addition, the formation of glyoxal from glycolaldehyde was examined. Glycolaldehyde was incubated with Fe 2þ in the sodium phosphate buffer (pH 7.4) at 37 C for 8 h, then glyoxal was analyzed by LC-ESI-MS/MS following reductive amination with ABA. As shown in Figure 5(D), the formation of glyoxal from glycolaldehyde was significantly enhanced by the addition of Fe 2þ and inhibited by the addition of DTPA.

Combined effect of glycolaldehyde and glyoxal on lysine N-pyrrolation
Finally, we investigated the N-pyrrolation of NAK and conversion of HSA to the DNA mimetic proteins by the combined reaction of glycolaldehyde and glyoxal. As shown in Figure 6(A), the time-dependent formation of pyrNAK in the reaction of NAK with glycolaldehyde was remarkably accelerated by the addition of glyoxal. In addition, it was also observed that the formation of pyrNAK was dependent on the concentrations of glycolaldehyde and glyoxal, and the maximum yield of pyrNAK was obtained when equal concentrations of glycolaldehyde and glyoxal were incubated with NAK ( Figure 6(B)). Glycolaldehyde slightly mediated the formation of pyrNAK, probably due to the formation of glyoxal by the autoxidation of glycolaldehyde, whereas glyoxal alone hardly mediated the lysine N-pyrrolation. We also observed the combined effect of glycolaldehyde and glyoxal on the conversion of HSA to the SGbinding proteins (Figure 6(C)). These results support the hypothesis that the autoxidation of glycolaldehyde followed by the formation of glyoxal is involved, at least in part, in the lysine N-pyrrolation.

Cross-reactivity of autoantibodies to Fe 21 /ascorbate-treated proteins
Systemic lumps erythematosus mouse and patient sera have been reported to cross-react with SG-binding Npyrrolated proteins [10]. Hence, to further characterize the formation of DNA mimetic proteins during ascorbate autoxidation reactions, we evaluated the crossreactivity of the SLE-prone MRL-lpr mouse sera to Fe 2þ /ascorbate-treated proteins. An ELISA analysis showed that the MRL-lpr mouse sera more significantly cross-reacted with ascorbate-and Fe 2þ /ascorbatetreated HSA than the control MRL-MpJ mouse sera (Figure 7(A)). The level of the anti-dsDNA titers in the MRL-lpr mice was significantly higher than that in the MRL-MpJ mice (Figure 7(B)), and the cross-reactivity of the MRL-MpJ and MRL-lpr mouse sera to Fe 2þ /ascorbate-treated HSA showed a linear correlation with the cross-reactivity to the dsDNA (Figure 7(C)), indicating that ascorbate might contribute to the production of antigenic structures against anti-DNA autoantibodies. Moreover, based on the combined effect of glycolaldehyde and glyoxal on the conversion of HSA to SG-binding proteins (Figure 6(C)), we evaluated the cross-reactivity of the SLE-prone MRL-lpr mouse sera to proteins treated with glycolaldehyde in the presence and absence of glyoxal. Consistent with the formation of the SG-binding proteins, the maximum cross-reactivity was observed when equal concentrations of glycolaldehyde and glyoxal were incubated with the protein (Figure 7(D)).

Discussion
We have previously shown that lipid peroxidation is a source of pyrK formed during the metal-catalyzed oxidation of LDL in vitro [11]. Following this discovery, glycolaldehyde was identified as an intermediate in the formation of pyrK from PUFAs. In addition, we showed that the glycolaldehyde-mediated N-pyrrolation of lysine could be promoted by the addition of glyoxal. On the other hand, since glycolaldehyde and glyoxal are known to be produced by multiple pathways, including the glucose metabolism, it was predicted that lysine N-pyrrolation may occur in biological systems. In the present study, we eventually showed that pyrK is formed during lysine modification by ascorbate-derived glycolaldehyde and glyoxal. These findings revealed a possible function of ascorbate as an endogenous source of DNA mimetic pyrrolated proteins. They also suggested that the ascorbate-derived pyrK residues generated in proteins may play a role in the innate and autoimmune responses associated with the glucose metabolism and oxidative stress.
The present study unequivocally established that both glycolaldehyde and glyoxal are formed from ascorbate, and glyoxal is formed from glycolaldehyde ( Figure 4, Scheme 2). The data suggest that pyrK, previously described as a product of the lipid peroxidation modification of lysine, could also be generally formed from reducing sugars during autoxidation reactions. In addition, glycolaldehyde is an enzymatic metabolite formed from fructose 1,6-bisphosphate as catalyzed by ketolase in an alternate glycolysis pathway. Therefore, beside the non-enzymatic oxidation of reducing sugars and PUFAs, the glucose metabolism may also be involved in the lysine N-pyrrolation in biological systems. The presence of multiple pathways to form these key intermediates in the formation of pyrK suggest that the lysine N-pyrrolation may generally occur in vivo.
Ascorbate-derived pyrK was detected even without the addition of Fe 2þ , although to a much lesser extent than with the supplementation of Fe 2þ (Figure 2). Similar results were also observed in the formation of glycolaldehyde and glyoxal (Figure 4). These findings may be due to that trace levels of iron and other transition metal ions are present in the buffer solution. It was previously estimated that tens of millimolar concentrations of phosphate buffer contain a submicromolar concentration of iron [12,13]. As ascorbate is oxidized to dehydroascorbate by hydroxyl radical, superoxide, hydroperoxyl radical, and Fe 3þ , but not Fe 2þ itself [14], the oxidation mechanism of ascorbate to form glycolaldehyde was speculated that (i) Fe 2þ directly reacts with oxygen to generated Fe 3þ and reactive oxygen species such as superoxide anion radical [15] and (ii) the generated Fe 3þ or reactive oxygen species oxidize ascorbate to dehydroascorbate followed by further degradation of dehydroascorbate to glycolaldehyde/glyoxal.
The autoxidation of reducing sugars, as well as lipid peroxidation, produces many other short-chain aldehydes, which can frequently cause complex effects of various aldehydes on lysine residues [16,17]. Glycolaldehyde can readily react with biological molecules, including proteins, and has been implicated in the pathogenesis of vascular diseases, such as atherosclerosis and renal disorders [18,19]. Upon the reaction with proteins, glycolaldehyde primarily reacts with lysine amino groups to form a variety of products, including brown-colored cross-linked structures [20] and a pyridinium-containing lysine adduct [19]. Argirov et al. [21] have also reported that glycolaldehyde generated during the decomposition of dehydroascorbate forms a hydroxymethyl-oxidopiridinium derivative of lysine (HOP-lysine). On the other hand, glyoxal reacts with lysine residues to form glyoxal-lysine adducts, including N e -(carboxymethyl)lysine (CML). CML is a product of both the carbohydrate and lipid peroxidation reactions [22] and has been identified as a major AGE antigen in tissue proteins [23]. Chikazawa et al. [11] proposed that pyrK would be composed of both glycolaldehyde and glyoxal. This proposed mechanism was established in this study by detecting glycolaldehyde and its oxidized product (glyoxal) during the autoxidation of ascorbate. Given the presence of multiple pathways of the lysine N-pyrrolation, it may be difficult to identify the major pathways for the formation of pyrK in complex biological matrices; however, pyrK is expected to be a potential marker of oxidative stress and long-term damage to proteins. Furthermore, since iron-catalyzed reactions are involved in the lysine N-pyrrolation, pyrK is likely to be formed during ferroptosis, a type of programmed cell death dependent on iron [24]. Further research is needed to determine whether the lysine N-pyrrolation is involved in this new type of cell death mechanism.
In the present study, we evaluated the cross-reactivity of the SLE-prone MRL-lpr mouse sera to Fe 2þ /ascorbate-treated proteins and observed that the MRL-lpr mouse sera cross-reacted more significantly with ascorbate-and Fe 2þ /ascorbate-treated HSA than the control MRL-MpJ mouse sera (Figure 7(A)). Moreover, the cross-reactivity of the MRL-MpJ and MRLlpr mouse sera to the Fe 2þ /ascorbate-treated HSA showed a linear correlation with the cross-reactivity to dsDNA (Figure 7(B,C)). The MRL-lpr mouse sera also showed the highest cross-reactivity to the protein modified with equal concentrations of glycolaldehyde and glyoxal (Figure 7(D)). These data suggest that ascorbate autoxidation followed by the formation of short-chain aldehydes, such as glycolaldehyde, may contribute to the production of antigenic structures against anti-DNA autoantibodies. They also emphasize the association of the autoxidation of reducing sugars with the anti-DNA response in autoimmune diseases. Although the potential role of reducing sugars and their oxidized products in pathogenesis needs to be further investigated, the lysine N-pyrrolation may not be a related side effect of disease progression but may be a possible cause of SLE and other autoimmune diseases.
In conclusion, pyrK, first identified as one of the products originating from the lipid peroxidationderived protein modification, was established to be the product of the lysine modification by ascorbate. It was shown that the same aldehydes, glycolaldehyde and glyoxal, were involved as common sources of pyrK in both reactions. The autoxidation of reducing sugars, as well as lipid peroxidation, produces many other shortchain aldehydes, which can frequently cause complex effects of various aldehydes on lysine residues [16,17]. However, only a few such adducts have been so far identified. Therefore, the present data suggest the need for further research into the involvement of short-chain aldehydes, especially the C2 and C3 aldehydes, derived from reducing sugars and PUFAs during the chemical modification of proteins in the mechanisms of oxidative stress and pathways of oxidative damage to protein and other biomolecules in aging, atherosclerosis, and diabetes.