Chemical Modifications in Aggregates of Recombinant Human Insulin Induced by Metal-Catalyzed Oxidation: Covalent Cross-Linking via Michael Addition to Tyrosine Oxidation Products
- First Online:
- Cite this article as:
- Torosantucci, R., Mozziconacci, O., Sharov, V. et al. Pharm Res (2012) 29: 2276. doi:10.1007/s11095-012-0755-z
- 935 Views
To elucidate the chemical modifications in covalent aggregates of recombinant human insulin induced by metal catalyzed oxidation (MCO).
Insulin was exposed for 3 h at room temperature to the oxidative system copper(II)/ascorbate. Chemical derivatization with 4-(aminomethyl) benzenesulfonic acid (ABS) was performed to detect 3,4-dihydroxyphenylalanine (DOPA) formation. Electrospray ionization-mass spectrometry (ESI-MS) was employed to localize the amino acids targeted by oxidation and the cross-links involved in insulin aggregation. Oxidation at different pH and temperature was monitored with size exclusion chromatography (SEC) and ESI-MS analysis to further investigate the chemical mechanism(s), to estimate the aggregates content and to quantify DOPA in aggregated insulin.
The results implicate the formation of DOPA and 2-amino-3-(3,4-dioxocyclohexa-1,5-dien-1-yl) propanoic acid (DOCH), followed by Michael addition, as responsible for new cross-links resulting in covalent aggregation of insulin during MCO. Michael addition products were detected between DOCH at positions B16, B26, A14, and A19, and free amino groups of the N-terminal amino acids Phe B1 and Gly A1, and side chains of Lys B29, His B5 and His B10. Fragments originating from peptide bond hydrolysis were also detected.
MCO of insulin leads to covalent aggregation through cross-linking via Michael addition to tyrosine oxidation products.
Key wordshuman insulinoxidationMichael additionaggregationfragmentation
4-(aminomethyl) benzenesulfonic acid
sodium citrate buffer
2-amino-3-(3,4-dioxocyclohexa-1,5-dien-1-yl) propanoic acid
electrospray ionization-mass spectrometry
- FT-ICR MS
Fourier transform ion cyclotron resonance mass spectrometry
metal catalyzed oxidation
sodium phosphate buffer
size exclusion chromatography
During pharmaceutical production and storage, therapeutic proteins can be exposed to components that are able to induce metal-catalyzed oxidation (MCO) (1), i.e. redox active transition metals, peroxides and reductants (2,3). MCO of therapeutic proteins (4) such as recombinant human interferon alfa and recombinant human interferon beta has been reported to form highly immunogenic aggregates (5–7), providing evidence that protein oxidation can potentially lead to severe side reactions and loss of therapeutic effect. Furthermore, it has been shown that in diabetic complications (8), copper ion concentrations are higher than in normal subjects (9). Therefore, MCO is a potential cause of insulin degradation in vivo as well, when the production of reactive oxygen species (ROS) exceeds the endogenous antioxidant defense. Montes-Cortes et al. (10) showed that the carbonyl content of human insulin and the hydroxylation of phenylalanine, based on the formazan assay (11), is increased after exposure of insulin to the plasma of diabetic patients, which can contain high concentrations of oxidants (12). Moreover, in a recent paper we showed that MCO induces significant covalent aggregation of insulin in vitro (13). However, the underlying mechanisms of insulin aggregation through MCO are still unknown. Previous reports on the MCO of human insulin and glycated insulin indicate that both histidine residues are easily oxidized to 2-oxo histidine (2,9).
Materials and Methods
Recombinant human insulin containing 0.4 % (w/w) zinc was provided by Merck (Oss, The Netherlands). L-ascorbic acid, ethylenediaminetetraacetic acid sodium salt (EDTA), copper dichloride, arginine, monobasic and dibasic sodium hydrogen phosphate, ammonium bicarbonate (ABI), dithiothreitol (DTT), iodoacetamide (IAM) and the solvents glacial acetic acid and acetonitrile were purchased from Sigma–Aldrich (St.Louis, MO, USA). Millipore Q water was used for the preparation of all the formulations and solutions. All chemicals were of analytical grade and used without further purification. ABS was synthesized according to a published procedure (21). Glu-C endoproteinase used in this study was purchased from Promega (Madison, WI, USA). Centrifugal filter units with a volume capacity of 4 mL and a molecular weight cut-off of 3 kDa were purchased from Millipore (Billerica, MA, USA). Cassette dialysis slides with a 2 kDa cut-off were purchased from Thermo Scientific (Asheville, NC, USA).
Metal-Catalyzed Oxidation of Insulin
Insulin oxidation was performed at three pH values: pH 7.4 in 50 mM sodium phosphate buffer (PB), pH 3.0 in 50 mM sodium citrate buffer (CB) and at pH 8.0 in 250 mM ammonium bicarbonate (ABI). For all the oxidation reactions, insulin was first dissolved in 0.1 M hydrochloric acid and then diluted into the corresponding buffer. Depending on the desired pH, 0.1 M sodium hydroxide was used. When CB was used, the final pH of 3 was achieved without further addition of base. Insulin concentration was measured by UV spectroscopy using a molecular weight of 5.8 kDa and an extinction coefficient of 6200 M−1 cm−1 at 276 nm (22). Further dilutions in PB, ABI or CB were performed to obtain a final insulin concentration of 1 mg/mL. Controls included 1 mg/mL of insulin in 50 mM PB, pH 7.4; in 50 mM CB, pH 3.0; and in 250 mM ABI, pH 8.0. MCO was performed by addition to 1 mL of 1 mg/mL insulin, 100 μL of 0.4 mM CuCl2 in 50 mM PB, pH 7.4, or in 50 mM CB, pH 3, or in 250 mM ABI, pH 8, depending on the desired pH, to a final concentration of 40 μM CuCl2. The reaction was performed in 2-mL Eppendorf tubes covered with aluminum foil to protect the reaction mixture from light. After 10 min of incubation of insulin with Cu2+, to allow copper to bind to insulin, the oxidation reaction was started by the addition of 110 μL of 40 mM L-ascorbic acid in 50 mM PB, pH 7.4, or in 50 mM CB, pH 3, or in 250 mM ABI, pH 8, depending on the desired pH, to a final concentration of 4 mM. The reaction was quenched after 3 h of incubation at room temperature by adding 12.1 μL of a 100 mM EDTA in 50 mM PB, pH 7.4, or in 50 mM CB, pH 3, or in 250 mM ABI, pH 8, depending on the desired pH, to a final concentration of 1 mM (15). To monitor the presence of protein fragments, the oxidation was also performed (in PB with the same amount of copper and L-ascorbic acid) at room temperature and at 37°C for 24 h before quenching. The oxidized samples were extensively dialyzed at +4°C against 50 mM ammonium bicarbonate (ABI) buffer, pH 8.0, for 24 h. In another experiment, the oxidized sample in 50 mM sodium citrate buffer pH 3, was extensively dialyzed at + 4°C in 50 mM sodium phosphate buffer, pH 7.4, using centrifugal filter units. After that the sample was left to equilibrate for 12 h at room temperature.
ABS-Derivatization of Undigested Samples
Solutions of native insulin and insulin oxidized at pH 3.0 (in CB) and pH 7.4 (in PB), dialyzed into 50 mM ABI, were treated with 0.1 M sodium hydroxide to a final pH of 9.0. Subsequently, to 100 μL of these solutions, 11 μL of a solution of 100 mM ABS dissolved in water were added to a final concentration of 10 mM. Next, 1.1 μL of 50 mM K3Fe(CN)6 dissolved in water was added to a final concentration of 0.5 mM. The reaction was conducted for 1 h at room temperature before performing SEC analysis. Controls for non-specific fluorescence included derivatization reagents (without protein) incubated under the same conditions and non-derivatized protein.
Reduction, Alkylation and ABS-Derivatization of Insulin
Solutions of native insulin and oxidized insulin, dialyzed into 50 mM ABI, pH 8.0, were reduced using 50 mM dithiothreitol (DTT), freshly prepared in 50 mM ABI, pH 8.0, added to a final concentration of 5 mM. The samples were incubated for 45 min at 45°C using a thermo heating bath (Thermo Scientific, Asheville, NC, USA). Subsequently, 200 mM iodoacetamide (IAM), freshly prepared in 50 mM ABI, pH 8.0, was added to a final concentration of 20 mM. The pH of 100 μL of dialyzed, reduced and alkylated insulin samples was adjusted to 9.0 with 0.1 M sodium hydroxide. Then, 11 μL of a solution of 100 mM ABS dissolved in water were added to a final concentration of 10 mM. Subsequently, 1.11 μL of 50 mM K3Fe(CN)6 dissolved in water was added to a final concentration of 0.5 mM. The reaction was conducted for 1 h at room temperature before performing digestion with Glu-C.
The proteolytic digestion was performed after reduction, alkylation and ABS-derivatization of insulin, by incubating the samples with Glu-C endoproteinase in a ratio insulin: Glu-C endoproteinase 10:1 (w/w), overnight at 37°C.
An Insulin HMWP column (7.8 × 300 mm, Waters, Milford, MA, USA) was connected to a Shimadzu HPLC system (UFLC Shimadzu Instrument equipped with two LC-20AT pumps, Columbia, MD) coupled to a photo-diode array detector (Shimadzu, SPD-M20A) and a fluorescence detector (Shimadzu, RF-20A). The photo diode array detector allowed for the recording of the UV spectrum in the region between 200–800 nm. Undigested insulin samples, without ABS treatment, were injected to calculate the percentages of the aggregates and to measure tyrosine fluorescence. Moreover, undigested insulin samples that were treated with ABS were injected to measure the benzoxazole fluorescence, which can be formed only in presence of DOPA and/or DOCH. The percentage of aggregates was calculated based on peak areas of the UV peaks at 276 nm (chromatograms not shown), as reported by Hermeling et al. (5,6). The fluorescence detector was set at various different excitation (Ex) and emission (Em) wavelengths, depending on the type of analysis: Ex 275 nm/Em 302 nm was used for monitoring Tyr fluorescence and Ex 360/Em 490 nm was used for the detection of the benzoxazole fluorescence of the ABS-derivatized samples. The mobile phase was composed of a mixture of 1 g/L L-arginine in water/acetonitrile/glacial acetic acid 65:20:15 (v/v/v) as reported in the United States and European pharmacopeias (23,24). The elution buffer was freshly prepared, filtered using a regenerated cellulose filter (Sartorius Stedim Biotech, Arvada, CO) and degassed prior to use.
Steady-State Fluorescence Spectroscopy
The benzoxazole fluorescence of the ABS-derivatized samples was measured upon excitation at 360 nm. The emission spectra were recorded from 490 nm to 600 nm with a 5-nm bandwidth on a Shimadzu RF-5000U spectrofluorometer. To monitor the presence of dityrosine, the non-derivatized, oxidized samples were analyzed after dialysis in 50 mM ABI, pH 8.0, using an excitation wavelength of 315 nm and detection at 420 nm (25).
After ABS-derivatization and dialysis, sample volumes were diluted with 50 mM ABI to 0.5 mL. Spectra were recorded in 0.5 mL 1-cm light-pass fluorescence cuvettes (Hellma, Plainview, NY, USA). Controls for non-specific fluorescence included samples without a substrate or reagents incubated under the same conditions.
Digested and non digested peptides were analyzed by means of an LTQ-FT hybrid linear quadrupole ion trap Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer (Thermo-Finnigan, Bremen, Germany) (26) and a SYNAPT-G2 (Waters Corporation, Milford, MA), both located in the Mass Spectrometry Laboratory of the University of Kansas, under the conditions as described by Ikehata et al. (26). In short, peptides were separated on a reversed-phase LC Packings PepMap C18 column (0.300 × 150 mm) at a flow rate 10 μL/min with a linear gradient from 0 to 65 % acetonitrile in 0.06 % aqueous formic acid over a period of 55 min using LC Packing Ultimate Chromatograph (Dionex). LC-MS experiments were performed in a data-dependent acquisition manner using Xcalibur 2.0 software (Thermo Scientific). Five most intensive precursor ions in a survey MS1 mass spectrum acquired over a mass range of 300–2000 m/z were selected and fragmented in the linear ion trap by collision-induced dissociation. The ion selection threshold was 500 counts. The MS/MS spectra obtained were analyzed with the software MassMatrix (27–30). MassMatrix was used to generate the theoretical fragment tables of the b and y ions of the different oxidized and cross-linked products. The theoretical fragments were compared to the experimental MS/MS spectra to validate the structures, which were taken into consideration only if the difference between the theoretical and the experimental m/z of the parent ion (and the fragment ions) was strictly below 0.1 Da. The SYNAPT-G2 instrument was operated for maximum resolution with all lenses optimized on the [M + 2 H]2+ ion from the [Glu]1-fibrinopeptide B. The cone voltage was 30 V and Ar was admitted to the collision cell. The spectra were acquired using a mass range of 50–2000 m/z. The data were accumulated for 0.7 s per cycle. The CID data, at the MS2 level, acquired with the FT-ICR instrument were obtained after an attenuation of the parent ion of 35 %. The mass window to collect the parent ion was fixed to 0.1 Da. Deconvolution of the electrospray ionization data of the undigested insulin was obtained using the maximum entropy distribution algorithm implemented in the Masslynx MaxEnt software (Waters Corporation, Milford, MA) using an adduct of 1 proton. Assuming a normal statistical distribution of the noise, a uniform Gaussian with a width at half height of 0.5 Da was used. A number of fifty iterations and a range between 0–36000 Da were used to build the most probable mass spectrum of the parent ions.
To guide the reader through the results, these will be presented in three sections: 1) Aggregation Profile of Oxidized Insulin and Elucidation of Aggregation Mechanism, 2) MS Analysis of Undigested Insulin Samples, and 3) Identification of Chemical Modifications by MS/MS Analysis of Reduced, Alkylated, ABS-Derivatized and Digested Samples.
Aggregation Profile of Oxidized Insulin and Elucidation of Aggregation Mechanism
Percentages of Monomer, Dimer and High-Molecular-Weight Oligomers (HMWO) for the Insulin Oxidized Under Different Conditions and Native Insulin
Oxidized insulin in PB, pH 7.4
74.4 ± 0.9
18.8 ± 1.4
6.8 ± 1.2
Oxidized insulin in ABI, pH 8.0
96.1 ± 0.7
3.9 ± 0.7
Oxidized insulin in sodium citrate, pH 3
99.2 ± 0.3
0.8 ± 0.3
Oxidized insulin in sodium citrate, pH 3.0, spiked into PB, pH 7.4
88.8 ± 1.3
9.3 ± 1.1
1.9 ± 0.1
Native insulin in PB, pH 7.4
99.4 ± 0.2
0.6 ± 0.2
To evaluate the potential of MCO to induce fragmentation, already reported by other authors for the MCO of glycated insulin (9,33), and by our group for the MCO of PEGylated insulin (13), human insulin was oxidized in 50 mM PB, pH 7.4, for 24 h at room temperature and 37°C. Figure 2f clearly shows the elution of species after the monomer peak, which indicates the formation of lower molecular weight species.
MS Analysis of Undigested Insulin Samples
The FT-ICR MS analysis presented in this section was performed on oxidized insulin and native insulin (our control), prior to reduction, alkylation, ABS-derivatization and digestion with Glu-C. This analysis was executed to monitor the formation of oxidized monomers and to confirm the MCO-induced fragmentation, as well as to quantify these species relative to native insulin.
The MS spectra of the oxidized and native insulin that were used for this calculation are displayed in Fig. 4c and d, respectively. Figure 5 represents the histograms plotted using the intensity values of the same spectra. The total intensity of the control (i.e. native insulin), was calculated between masses of 0 Da and 36,000 Da, i.e. until just beyond the theoretical molecular weight of the insulin hexamer. Based on this calculation, the total yield of fragments in the non-oxidized control was 0.3 % compared to 3.1 % in insulin exposed to Cu2+ and L-ascorbate (percentages have been calculated dividing the total intensity of the fragments over the total intensity of the native insulin used as a control). For a better view of the fragments, we provide a zoom of the range between 100–5,805 Da of native insulin and oxidized insulin (Fig. 4c, d). Altogether these results suggest that MCO induces peptidic fragmentation. We did not explore in more detail the mechanism(s) by which such fragmentation occurs, nor did we search for all the possible cross-link in which the new N-termini could be involved. However, we did note the generation of new cross-links between insulin and such fragments (see below), which contain new N-termini and C-terminal amino acids not generated by Glu-C digestion.
Identification of Chemical Modifications by MS/MS Analysis of Reduced, Alkylated, ABS-Derivatized and Digested Samples
Fragments Identified in Reduced, Alkylated, ABS-Derivatized and Digested (Glu-C) Insulin and Oxidized Insulin, as Determined by Mass Spectrometry
Glu-C fragments oxidized
Glu-C fragments ABS-derivatized
RB22GFFaYaTPKTB30 Fig. S1 A, B, C
RB22GFFYcTPKTB30 Fig. S5
FaB1VNQHLCGB8 Fig. S7
LA13YQLEA17 Fig. S13
A B14LYaLVCGEB21 Fig. S2
RB22GFFYdTPKTB30 Fig. S6
EB21RGFbFYTPKTB30 Fig. S8
SA12LYcQLEA17 Fig. S11
IA10CSLYA14 Fig. S14
FaB1VNQHLCGSHLVEB13 Fig. S3
FaB1VNQHLCGSHLVEALB15 Fig. S9
SA12LYdQLEA17 Fig. S12
SA12LYQLEA17 Fig. S15
FbB1VNQHLCGSHLVEALYLVCGEB21 Fig. S4
HB5LCGSHaLVEALB15 MS/MS not shown
SA12LYQLA16 Fig. S16
FB1VNQHaLCGSB9 MS/MS not shown
YA14QLEA17 Fig. S17
FB1VNQHaLCGSHLB11 MS/MS not shown
GB8SHLVEB13 Fig. S18
HB10LVEALB15 Fig. S19
FB1VNQHLCGSB9 Fig. S20
YB16LVCB19 Fig. S21
GB20ERGFFYB26 Fig. S22
GA1IVEQCA6 Fig. S23
AB14LYLVCGB20 Fig. S24
QA15LENYCNA21 Fig. S25
LB6CGSHLVEB13 Fig. S26
RB22GFFaYaTPKTB30 Fig. S1 A, B, C
GB23FFYTPKTB30 Fig. S27
LA13YQLEA17 Fig. S13
Oxidized Glu-C Fragments
The C-terminal region of chain B contains two Phe and one Tyr in positions B24, B25 and B26, respectively. DOPA and DOCH can arise from the oxidation of all these residues. Thus it becomes of primary importance to identify the ions, which allows us to discriminate between different oxidation products. The MS/MS spectra reported in supplementary Fig. S1A provides evidence for the oxidation of Phe B25 to Tyr where * indicates the incorporation of one oxygen atom. The ion b3+, despite its low intensity, indicates that Phe B24 is present in the native, non-oxidized state, where m/z 361.2 corresponds to the singly charged ion of the sequence RGF. Instead, the ion b4+ provides evidence for oxidation of Phe B25. A careful analysis of the spectra (Figures S1B) reveals the contemporary presence of two different y5+ ions, indicated as (F4)y5+ with m/z 625.5 (which corresponds to a structure containing Tyr B26 oxidized) and, (HO-F4)y5+ with m/z 609.4 (which corresponds to a structure containing Phe B25 oxidized). Panel C shows the b4+ ions for these two coexisting structures, again representing oxidation of Phe B25 or Tyr B26, respectively. Hence, both Phe B25 and Tyr B26 are targets for the incorporation of one oxygen atom.
Supplementary Figure S2 displays the MS/MS data for the peptide ALY*LVCGE, which contains an expected Glu-C cleavage site, where Y* represents the incorporation of one oxygen into Tyr, i.e. the formation of DOPA (the Cys residue is alkylated with IAM). This product is confirmed through the presence of the ions b3+-b5+ (although b3+ and b4+ show low intensities) and y5+. The singly charged ions b3+, b4+ and b5+ indicate that the oxygen is in one of the following sequences: ALY, ALYL, or ALYLV. The singly charged ion y5+ indicates that the sequence LVCGE is not oxidized. Therefore, oxygen incorporation must have occurred on one of the first three amino acids, ALY, where Tyr represents the most oxidation-sensitive target amino acid. Further evidence for Tyr B16 oxidation through ABS derivatization is given below.
The N-terminus of chain B, Phe B1, displays a mass increase consistent with the oxidation of Phe B1 to cyclohexadienone. The MS/MS analysis (Figure S3) of the sequence F*VNQHLCGSHLVE indicates the formal addition of one oxygen atom and loss of 2 hydrogens from Phe B1, i.e. a mass increase of 14 Da through the appearance of the following ions: y9+, which suggests that the sequence HLCGSHLVE is not oxidized, and b6+, indicating the oxidation of the sequence FVNQHL. In this sequence, HL can be excluded as an oxidation target because of the nature of y9+. Since in the sequence FVNQ, F is most sensitive to oxidation we conclude that oxidation targets Phe B1.
Phe B1 can be further oxidized to DOCH (Figure S4). The observed masses of the ions y17++-H2O and y17+ exclude any oxidation of other amino acids sensitive to oxidation in F**B1VNQHLCGSHLVEALYLVCGEB21. The ion y17+ with m/z 1899.82 shows an intensity which is about 80 % of that of b16+ with m/z 1898.75; thus, it should not be considered the first isotope peak of b16+, which would be expected at m/z 1899.75.
ABS-Derivatized Glu-C Fragments
RB22GFFY#TPKTB30 and RB22GFFY##TPKTB30
Supplementary Figure S5 shows that Tyr B26 is converted to DOPA and/or DOCH, indicated by derivatization with one molecule of ABS (indicated with the symbol #). The ions with m/z 361.28 and m/z 508.26, corresponding to b3+ and b4+, respectively, suggest that both Phe B24 and Phe B25 are not oxidized in this sequence. If for instance, Phe B24 (in the sequence RGFB24FYTPKT) had been oxidized, the b3+ ion would have been expected with m/z 377.28 (i.e. 361.28 + 16 Da). Instead, the ions b5+ and b6+ provide evidence that Tyr B26 is the target of oxidation. For additional evidence, supplementary Figure S6 displays the MS/MS spectrum of RGFFY##TPKT, derivatized with two molecules of ABS, as shown in the displayed structure. Here, the ions b3+ and b4+ are identical to the ones reported in Figure S5, although more intense. The ion with m/z 1037.29 corresponding to the ions b5+ and y5+ reported in the inset of Figure S6, confirm that the original Tyr contains two ABS molecules. We had realized in the past (19,20), that derivatization with ABS can lead to the incorporation of one or two molecules of ABS into the final benzoxazole product (see scheme 1), depending on the availability of ammonia, resulting in competition of ammonia and ABS for the 6-position in benzoxazole. In addition to that, it seems that the competition between ammonia and ABS depends on the steric hindrance of the peptide which is derivatized: i.e., derivatization with two ABS molecules of sterically less accessible DOPA and DOCH can be kinetically less favorable than that of more exposed residues. It must be noticed that the sequence RB22GFFYTPKTB30, depicted in Figures S5 and S6, does not necessarily belong to the same B chain: in other words, the sequence derivatized with one ABS molecule could be involved in the formation of HMWO, while the doubly ABS-derivatized sequence can actually be present in the monomer as well, although oxidized, and subsequently be derivatized with two ABS molecules.
Phe B1 is oxidized to hydroxylated Phe, indicated through the MS/MS data displayed in Figure S7: the ion y7+ shows that none of the amino acids in the sequence C-terminal to Phe B1, VNQHLCG, is oxidized. Therefore, the ion b5+ confirms oxidation of Phe B1. In fact, Phe hydroxylation can occur in positions ortho, meta and para; however, only the latter would lead to Tyr.
The MS/MS data displayed in Figure S8 are consistent with the oxidation of Phe B24 to DOCH. The ions b4+ and y6+ are fundamental to confirm the oxidation of Phe B24 since the first indicates the presence of the sequence ERGF**, and the second, suggests that the sequence FYTPKT is not oxidized. Further confirmation of the chemical structure in which Phe B24 is doubly oxidized arises from the detection of the internal fragment GF**FY with m/z 545.39, which can be produced during the analysis via specific pathways consistent with the mobile proton model (35).
Phe B1 oxidation is also evident in the sequence F*VNQHLCGSHLVEAL, indicated by the MS/MS data displayed in Figure S9. Here it is sufficient to consider the ions y13+ and b14++. The ion y13+ indicates that the sequence NQHLCGS is not modified. On the other hand b14++ indicates oxidation in the sequence FVNQHLCGSHLVEA. By exclusion, this limits oxidation to the N-terminal subsequence FV, where F represents the most oxidation-sensitive amino acid.
We further corroborated the formation of DOPA through derivatization of the oxidized peptide ALY*LVC with ABS (Figure S10). The only possible site for derivatization in this peptide is an oxidized Tyr residue, since ABS derivatization requires the presence of DOPA (which during the derivatization is oxidized with K3Fe(CN)6 to DOCH).
SA12LY#QLEA17 and SA12LY##QLE A17
The MS spectra presented in Figures S11 and S12 provide evidence for the ABS derivatization of Tyr A14 within the sequence SLYQLE. Successful derivatization with one and two molecules of ABS indicates oxidation of Tyr A14 to either DOPA or DOCH. In both figures the ions b3+ with m/z 560.11 and 730.07, respectively and, the ion y4+ with m/z 748.39 and 918.15, respectively, indicate derivatization of the original Tyr residue with one and two molecules of ABS.
Two types of non-oxidized fragments were detected in the control (Table II). They are likely generated during production or storage as a consequence of low traces of transition metals. Instead, 14 non-oxidized fragments were detected as a result of MCO, summarized in the last column of Table II. The MS/MS data of all these fragments are reported in the Supplementary Material (Figures S13-S27).
Gly A1 –Tyr A14 Cross-Link
Gly A1—Tyr B 26 Cross-Link
Figure 7c represents the cross-link of Gly A1 of the sequence GIVE to Tyr B26. The mass spectra displayed in Fig. 7c represents sodium adducts of the respective peptide fragment(s), likely due to incomplete removal of buffer during dialysis. Therefore, all the m/z of the ions presented in Fig. 7c have an increased mass of 23 Da. The ion [By2 + Na]+, representing the singly charged sequence VE, suggests that the amino acids Val and Glu are not involved in the new covalent bond. If this sequence had been involved in the new covalent bond with Tyr, it would not have been possible to detect the ion [By2 + Na]+, since only dissociation of peptide bonds occurs during the low energy collision with the inert gas. [Bb2 + Na]+ indicates that the sequence GI is covalently bound to the sequence FFY. We excluded the possibility of DOPA as the product of one of the Phe residues, since in such case the ion [Bb2 + Na]+ would have a mass increase (relative to the expected mass of the fragment of unmodified insulin) larger than 16 Da (the oxidation of Phe to DOPA and DOCH requires two oxygen atoms); if one of the Phe residues were exclusively oxidized to Tyr, the ion [Bb2 + Na]+ would have the same molecular weight as the one detected in this MS/MS spectrum. In such case, however, it would not be possible to observe Michael addition, since the new Tyr residue (generated from the mono-oxidation of Phe) is not a Michael acceptor unless it is further oxidized to quinone.
In Fig. 7d the original sequence FFY shows additional oxidation of both Phe residues (in contrast to Fig. 7c, which shows the native Phe residues in FFY): one of them to DOPA and the second one to a 6-amino substituted DOPA, likely originating from Michael addition of ammonia to a DOPA oxidation product. The ions By1+, with m/z 612.22, displayed in the inset, and Ab1+, with m/z 626.29, unequivocally indicate that the covalent bond is between Gly and Tyr: By1+ represents the singly charged sequence Y-GIVE and Ab1+ corresponds to the singly charged sequence FFY-G. The combination of both demonstrates that the new covalent bond, not dissociable during MS/MS analysis, is located between Tyr and Gly. Importantly, our results show that predominantly Tyr oxidation products serve as Michael acceptors for cross-link formation during the MCO of insulin (see also below). There are several possible rationales for this behavior, including protein conformation and the respective yields of Tyr and Phe oxidation products: the oxidation of Phe to DOCH requires one additional oxidation step compared with that of Tyr to DOCH, suggesting that DOCH formation from Tyr is kinetically favorable.
Gly A1—Tyr B16 Cross-Link
Figure 7e displays MS/MS data consistent with a covalent cross-link between Gly A1 and Tyr B16. The most relevant ion in this figure is Ab2+ with m/z 479.27, which demonstrates that the sequence GI is covalently bound to LY. Since Ile and Leu do not contain functional groups amenable to Michael addition, we conclude that the covalent bond is formed between Gly and Tyr.
Phe B1—Tyr A14 Cross-Link
Phe B1 is not only a possible target for oxidation; the free N-terminal amino group can react via Michael addition with DOCH. The spectrum depicted in Fig. 7f shows the relevant ions consistent with a cross-link between Phe B1 and Tyr A14: Ab1+ with m/z 697.36 shows the existence of a peptide containing the following amino acids: Phe, Tyr, Gln and two Leu, plus the molecular weight of one oxygen atom. However considering only the ion Ab1+, it is not possible to localize the oxygen addition and to characterize the nature of the cross-link. The ion Bb2+, with m/z 797.36, indicates that the sequence FVNQ is connected to the sequence LY. The ion Ay3+ -NH3, with m/z 343.16, excludes any covalent cross-link of the sequence VNQ. In the sequence LY only Tyr oxidation can lead to a Michael acceptor for the amino terminal of Phe B1, which suggests a cross-link such as depicted in Fig. 7f.
His B10—Tyr B16 Cross-Link
Figure 7g displays a cross-link between His B10 of the sequence SH and Tyr B16 of the sequence YLVCG. The relevant ions in this spectrum are Ab1+ and Ay4+. Ion Ab1+, with m/z 420.14, suggests that the sequence SH is covalently bound to the original Tyr residue (after oxidation of Tyr to DOCH). The deprotonated imidazole nitrogen of His is an appropriate nucleophile for Michael addition, demonstrated for example for His addition to dehydroalanine (36).
His B5—Tyr A19 Cross-Link
Figure 7h displays a cross-link between His B5 of the sequence HLCGS and Tyr B16 of the sequence YCN and the MS/MS data supporting this assignment. We note that the ion Ab1+-NH3 shows an intensity which is about 80 % of that of Ab1-H2O. Thus, the ion with m/z 533.14 should not be considered the second isotope of Ab1+-H2O, suggesting that the N-terminal amino group of the His is free and not involved in the new covalent bond.
Lys B29—Tyr A14 Cross-Link
Figure 7i represents the cross-link between Lys B29 and Tyr A14. The relevant ion in the spectra is Ab1+, since it localizes the cross-link to Tyr and the amino acids Pro and Lys in the sequence PK. Because the MS/MS data in Fig. 7i shows no B ions, we cannot conclude whether the cross-link involves the side chain amino group of Lys or the amino group of Pro. The latter would require that non-oxidative fragmentation between Thr B27 and Pro B28 occurs prior to cross-linking; the side chain amino group of Lys B29 represents a good nucleophile and it can therefore be assumed that at least some of the cross-links are formed through the reaction of Tyr A14 with Lys B29.
We performed a detailed mass spectrometric analysis of insulin in order to elucidate the mechanisms and amino acid residues involved in MCO and MCO-mediated aggregation. The oxidation of Phe and Tyr leads to the formation of catechol structures that can be further oxidized to DOCH. These oxidation products serve as Michael acceptors for cross-links with several nucleophiles present in the insulin sequence. Formation of dityrosine, as a potential structure involved in the cross-linking of insulin, was not detected, neither with fluorescence measurement (see material and methods), nor with the help of MassMatrix software. This is not surprising since it has been reported that only the hydrogen peroxide/copper system is capable of inducing dityrosine formation (37). Schiff base formation could potentially be in competition with Michael addition, nonetheless such cross-links can be reversed relatively easily and based on our results does not appear to be the main mechanism of insulin aggregation. Changes in the primary structure, as reported in this paper, may alter the biological activity of insulin with severe consequences on the glycemic control. Furthermore, changes in the secondary, tertiary and quaternary structure as a consequence of oxidation, as reported by us before (13), may generate new repetitive epitopes or open the access to hidden epitopes that could contribute to the formation of immunogenic products.
In addition to the characterization and localization of oxidation products, we provided quantitative data on the amount of fragments and oxidized monomers. An important feature of insulin exposed to MCO is the appearance of fragments, which increases the number of polypeptide nucleophiles available for Michael addition. At this point, we do not know whether the observed insulin fragmentation precedes or follows cross-link formation. Furthermore, it is unknown whether Glu-C might have a different specificity towards oxidized protein. The actual mechanism of cleavage is currently unknown, but we can exclude the classical oxidative cleavage (38) since we did not detect the α,β-dicarbonyl products expected for such a mechanism. Instead, all non-specific cleavages appear to involve a hydrolytic cleavage of the respective peptide bonds. It may be possible that the addition of Cu2+ and L-ascorbate promote the formation of peptide-metal complexes with hydrolytic activity, such as described for the cleavage of amides through Cu2+ (39). Surprisingly, most of the non-specific peptide fragments we detected were already reported by other authors, e.g. after the oxidation of glycated insulin by Fenton chemistry (9,33); however, Guedes et al.(33) interpreted the cleavage mechanism as oxidative cleavage, which is inconsistent with the intact C- and N-termini of the detected peptide fragments.
Through MCO experiments at various pH, and in the presence of ammonia, we are able to confirm that Michael addition between insulin chains causes covalent aggregation. In the presence of ammonia, which competes for Michael addition to DOCH, aggregation was almost completely inhibited. Although such experimental conditions allow to prevent aggregation, we noted that the introduction of an amino group on oxidized Tyr or Phe might still have consequences for the immunogenicity of this small polypeptide hormone.
Several proteins have been shown to form aggregates after exposure to Cu2+/L-ascorbic acid, such as recombinant SHa(29–231) prion protein (40), superoxide dismutase (41), monoclonal IgG2 (42), interferon alpha 2a (5,6) and interferon beta 1a (7). The results presented in this paper may serve as a model to rationalize the chemical mechanisms of aggregation during MCO of proteins in general. Although insulin lacks Trp and Met, which are also prone to oxidation (43,44) and are present in most other proteins, DOPA and DOCH should not be formed from oxidation of Trp or Met, thus they cannot act as electron acceptor for Michael addition. Since MCO was performed at pH 7.4, where insulin is present mainly in the hexameric state (45), the cross-links we measured are likely to be intermolecular cross-links, i.e. formed between different insulin molecules. As an example we calculated the distance of Gly A1 and Tyr A14 and Gly A1 and Tyr A19 within the same A-chain (using Swiss pdb viewer (46,47) and T6 human insulin at 1.0 Å resolution, 1MSO pdb file). An average of 18.5 Å was measured for a theoretical covalent bond between the amino group of Gly A1 and the carbon in ortho of the aromatic ring of Tyr A14, and an average of 8.3 Å for the amino group of Gly A1 and the carbon in ortho of the aromatic ring of Tyr A19, which seems to be too far for an intramolecular reaction (e.g., the length of the peptidic bond between Gly A1 and Leu A2 is 1.3 Å). We also noticed that Phe residues, oxidized to DOCH, were not involved in cross-links, nor were they derivatized with ABS. In both cases this could be due to low accessibility of Phe residues which might belong to chains B involved in the formation of HMWO, in which only DOCH originated from Tyr is accessible for ABS derivatization. Moreover, in the case of cross-links it could be speculated that the oxidation of Tyr to DOCH (which requires the addition of only one oxygen atom), is kinetically favorable over the oxidation of Phe to DOCH (which requires two oxidation steps). This can potentially lead to a depletion of the nucleophiles available in the insulin molecule.
Oxidation is one important degradation process that proteins can undergo. In this work we highlighted the potential consequences of DOPA and DOCH formation and also illustrated how the knowledge of oxidation mechanisms can be utilized to investigate the mechanism of covalent aggregation. To this end, recombinant human insulin was used as a model: oxidation of aromatic amino acids residues, besides others, leads to α,β unsaturated carbonyl compounds, which are electron acceptors for Michael addition. These reactive groups, resulting from oxidation, not only can lead to covalent protein aggregation, as shown for insulin in this paper, but also may lead to cross-links between protein molecules and amino acids, which are typically present during cell culture and are often used as formulation excipients. This must be taken into consideration during production and formulation development of therapeutic proteins.
Acknowledgments & DISCLOSURES
The authors thank Dr. Nadya Galeva for performing MS measurements, Merck (Oss, The Netherlands) and The University of Kansas and the National Institutes of Health (PO1AG12993) for financial support, and Merck for providing insulin for this project, Leiden University Fund/Slingelands (LUF, 1111/19-4-11\O, Sl) and Nederlandse Stichting voor Farmacologische Wetenschappen (NSFW) for the travel grants to Riccardo Torosantucci.
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.