Introduction

Protein pharmaceuticals may be exposed to ambient and some UV light during production, formulation and storage, which can lead to protein degradation (1,2,3,4), and potentially affect the potency of a formulation. For instance, the photo-irradiation of monoclonal antibodies can lead to aggregation (3,4,5,6), oxidation (4,7), fragmentation (8) and reduced bioactivity (3,5). To assess, predict and prevent such detrimental protein degradation a detailed understanding of product formation and the chemical mechanisms of photo-degradation is crucial. For mechanistic studies, we have subjected a small model protein, hGH, to photo-irradiation at λ = 254 nm and λ > 295 nm. These wavelengths were selected as an investigative tool (9) to generate a maximal amount and manifold of photo-degradation products within a reasonable experimental time frame. Specifically, wavelengths of λ > 295 nm overlap with the UV spectrum of Trp, which can extend to λ = 320 nm (10). Trp has been shown to be sensitive to photo-degradation not only by UV light (11) but also by ambient light (4), and photo-induced electron transfer from Trp can lead to disulfide degradation (12,13,14,15,16). The photo-degradation products identified in this study will provide protein and formulation scientists with a database for the characterization of proteins and/or protein pharmaceuticals exposed to milder stress conditions (e.g., ambient light) or simply long-term storage.

Aromatic amino acids and disulfides are the main chromophores in a protein and are, therefore, predisposed for photo-degradation (1,17,18). The direct UV-irradiation of disulfides at λ = 250 - 300 nm can generate a pair of Cys thiyl radicals (CysS●) (19,20,21,22) (reaction 1), which are reactive key intermediates in the photodegradation of peptides and proteins (23,24,25,26). Besides homolysis, CysS● radicals are generated indirectly by the one electron reduction of disulfide bonds (reaction 2, equilibrium 3) (27,28), for instance by photo-induced electron transfer from Trp or Tyr (12,13,14,15,16). Furthermore, Cys thiolate (CysS−), generated via equilibrium 3, can be photo-ionized to a CysS• thiyl radical (reaction 4). In addition, thiyl radicals may be generated via thiol/thiolate oxidation through various radicals such as phenoxyl, peroxyl and alkoxyl radicals, represented by XO● in reaction 5 (29,30,31). Moreover, the one electron reduction of thioaldehydes (32) may also contribute as a minor secondary pathway to the formation of thiyl radicals (reaction sequence 6) (33).

$$ \mathrm{R}-\mathrm{S}-\mathrm{S}-\mathrm{R}+\mathrm{h}\upupsilon \to 2\ \mathrm{R}-{\mathrm{S}}^{\bullet } $$
(1)
$$ \mathrm{R}-\mathrm{S}-\mathrm{S}-\mathrm{R}+\mathrm{e}\hbox{-} \to {\left(\mathrm{R}-\mathrm{S}\therefore \mathrm{S}-\mathrm{R}\right)}^{\bullet}\hbox{-} $$
(2)
$$ {\left(\mathrm{R}-\mathrm{S}\therefore \mathrm{S}-\mathrm{R}\right)}^{\bullet}\hbox{-} \rightleftharpoons \mathrm{R}-{\mathrm{S}}^{\bullet }+\mathrm{R}-\mathrm{S}\hbox{-} $$
(3)
$$ \mathrm{R}-\mathrm{S}\hbox{-} +\mathrm{h}\upupsilon \kern0.5em \to \kern0.5em \mathrm{R}-{\mathrm{S}}^{\bullet }+\mathrm{e}\hbox{-} $$
(4)
$$ \mathrm{R}-\mathrm{S}\hbox{-} +{\mathrm{XO}}^{\bullet}\kern0.5em +{\mathrm{H}}^{+}\to \kern0.5em \mathrm{R}-{\mathrm{S}}^{\bullet }+\mathrm{XO}-\mathrm{H} $$
(5)
$$ \mathrm{R}\hbox{-} \mathrm{CH}=\mathrm{S}+\mathrm{e}\hbox{-} +{\mathrm{H}}^{+}\to \mathrm{R}\hbox{-} {\mathrm{C}}^{\bullet}\mathrm{H}\hbox{-} \mathrm{SH}\rightleftharpoons \mathrm{R}\hbox{-} {\mathrm{C}\mathrm{H}}_2\hbox{-} {\mathrm{S}}^{\bullet } $$
(6)

CysS● radicals can initiate a variety of different degradation pathways through disproportionation (reaction 7) (23), hydrogen abstraction (34,35,36,37) (reaction 8) or combination with other peptide or protein radicals, representatively referred to as P● (reaction 9) (16,34).

$$ 2\mathrm{R}-{\mathrm{CH}}_2-{\mathrm{S}}^{\bullet}\to \mathrm{R}-\mathrm{CH}=\mathrm{S}+\mathrm{R}-{\mathrm{CH}}_2-\mathrm{SH} $$
(7)
$$ \mathrm{R}-{\mathrm{S}}^{\bullet }+\mathrm{R}-\mathrm{H}\rightleftharpoons \mathrm{R}-\mathrm{SH}+{\mathrm{R}}^{\bullet } $$
(8)
$$ \mathrm{R}-{\mathrm{S}}^{\bullet }+{\mathrm{P}}^{\bullet}\to \mathrm{R}-\mathrm{S}-\mathrm{P} $$
(9)

The disproportionation reaction 7 yields a free thiol and a thioaldehyde (23). The latter is an electrophile that can form dithiohemiacetals with thiols (23) or undergo hydrolysis to Ser semialdehyde (SerSA) (35). On the other hand, thiyl radicals can enter intramolecular 1,2- and 1,3-hydrogen transfer reactions and are capable of abstracting hydrogen atoms from amino acid residues as well as from the peptide backbone (31,33,34,35,36,37). Hydrogen abstraction from side chain residues may ultimately result in the formation of alcohols and carbonyls (38,39,40,41), while hydrogen abstraction from the Tyr hydroxyl group yields a phenoxyl radical (31), and, potentially, a Tyr-Cys crosslink (25). The 1,3-hydrogen transfer within CysS● yields a ●Cα radical from which the hydrogen sulfide radical HS● can be eliminated to form dehydro-Ala (Dha) (23,33). In the presence of oxygen, ●Cα radicals can convert into peroxyl radicals, which may initiate peptide bond cleavage by the diketo pathway (41). When converted to alkoxy radicals, peptide cleavage by the diamide pathway may occur (41).

We report here on the disulfide degradation pathways of hGH when photo-irradiated at λ = 254 nm or λ > 295 nm in aqueous solution, and provide mass spectrometric evidence for various new photo-degradation pathways, that include: the transformation of Cys to Gly, the formation of a vinyl ether between Ser-184 and a photoproduct of Cys-189, several novel cross-links, and two peptide cleavage reactions adjacent to Cys-53 leading to carbamate or an α-keto acid. The photo-irradiation experiments at λ = 254 nm were performed as proof-of-concept studies under conditions where the intensity of a single wavelength is well-defined. These conditions also mimic photolytic stress that protein pharmaceuticals may encounter during sterilization by UV-light, a proposed method for viral decontamination (6), or during chromatographic purification coupled to UV-detection (42,43). Photo-irradiation experiments at λ > 295 nm represent exposure to UV-B light that is not filtered by glass, i.e. conditions that may be encountered during transportation and storage, and exposure to fluorescent white light or sunlight. This article complements an earlier photodegradation study of hGH (44), in which we described an unusual cleavage between Glu-88 and Pro-89 in proximity to Trp-86 after photo-irradiation at λ = 254 nm, in addition to Met oxidation and new backbone cleavage reactions between Phe-1 and Pro-2.

In contrast to the pathways in aqueous solution reported here, a solid state photo-degradation study of the structurally related bovine growth hormone revealed a Trp-dependent degradation mechanism, where an initial electron transfer from Trp to the disulfide bond, leads to the generation of CysS● radicals. After reaction with oxygen, the CysS● radicals eventually recombined to yield α-disulfoxide, thiosulfinate, and/or thiosulfonate (12).

Experimental Section

Materials

Human growth hormone (somatotropin, hGH) was provided as Nutropin AQ® formulations by Genentech, Inc. (South San Francisco, CA). The protein was separated from excipients by dialysis against water overnight with Slide-A-Lyzer Dialysis Cassettes of 3.5 K molecular weight cut off (MWCO) from Thermo Fisher Scientific Inc. (Rockford, Il). Alternatively, formulations were filtered and washed three times with water with Amicon Ultra-15 Ultracel - 10 k (10 kDa MWCO) centrifugal filters from EMD Millipore Corporation (Billerica, MA). Aliquots of ca. 0.1 ml were stored at −20° C and concentrations were measured by means of UV spectroscopy at 277 nm, where A277 = 0.82 cm−1 (45). TPCK-treated sequence-grade trypsin was obtained from Promega (Madison, WI) (some digestions were performed with trypsin from Sigma-Aldrich (St. Louis, MO)), and ThioGlo® 1 was obtained from Millipore. All other chemicals were of the highest commercial grade available and obtained from Sigma (Saint Louis, MO) or Fisher (Pittsburgh, PA). Chemicals were used as received and all solutions were prepared with MilliporeQ-water.

Photo-Irradiation

Photo-irradiations were carried out in a Rayonet photo-reactor RPR-200 equipped with four 35 W low pressure mercury lamps (RPR-2537 Å) or four phosphor-coated low pressure mercury lamps (RPR-3000 Å). The RPR-2537 Å lamp has an emission >80% at λ = 254 nm and the RPR-3000 Å lamp emits 95% of light between λ = 265-340 nm (λmax = 300 nm) (46). A rotating Rayonet RMA-500 Merry-Go-Around assured uniform light exposure of the samples that had a distance of 3.8 cm (1.5 in.) from the lamps. All equipment was from the Southern New England Ultra Violet Company (Branford, CT). Samples were placed in 13 × 100 mm quartz glass tubes when irradiated at λ = 254 nm or in equally sized borosilicate glass (akin to Pyrex) tubes to cut off light below 295 nm when irradiated with RPR-3000 Å lamps. The light intensity reading orthogonal to the lamps was 210 W/m2 when 16 RPR-2537 Å lamps were employed at the distance of 1.5 in. (46). Hence, we calculate a light intensity of 52.5 W/m2 for 4 lamps, and photo-irradiations of 5 min correspond to ca. 4.4 Wh/m2. In contrast, the current ICH guidelines for photostability testing require 200 Wh/m2 (47). However, we note that the ICH guideline makes use of a broad irradiation spectrum between 320 and 400 nm (47), whereas our experiments at 254 nm used light at a specific wavelength. Based on previous actinometry measurements, we calculate that samples were subjected to a total of 4.3 × 10−6 einstein (mol photons) when irradiated for 5 min with 4 RPR-2537 Å lamps. Due to the broad emission spectrum of the RPR-3000 Å lamps actinometry was not performed; the spectral distribution of irradiance density of the lamps is published by the manufacturer (46).

Preparation of Samples

Samples for irradiation had a volume of 400 μl, contained 10 μM (0.22 mg/ml) hGH and were buffered with 0.1 M phosphate at pH 7.4. After irradiation, samples were transferred to 2 ml centrifuge vials. Crystalline guanidine hydrochloride was added for a final concentration of 5 M. A 100-fold excess of N-ethylmaleimide (NEM) was added and the pH was adjusted to ca. 7.4 with 0.1 M tribasic sodium phosphate. After 10 min incubation at 37°C, hGH was precipitated with 1.4 ml 0.5 M HClO4 and centrifuged at 14,000 g for 10 min. The resulting pellets were washed twice with water. Thereafter, tryptic digestion (trypsin/protein = 1:50) was performed in 100 μl aqueous solution, buffered at pH 7.8 with 0.1 M NH4HCO3. After incubation at 37° C for 8 h, the digests were stored at −20° C prior to analysis.

Preparation of the Samples for Time-Dependent Analysis of the Photoproducts

HGH samples were reconstituted in 100 μL of NH4HCO3 buffer (50 mM, pH 7.5). The disulfide bonds were reduced by the addition of 50 μL of 5 mM stock solution of bis(2-mercaptoethyl)sulfone (BMS) in acetonitrile (ACN), followed by incubation at 45°C for 30 min. The reduced cysteine residues were alkylated by the addition of 50 μL of 25 mM iodoacetamide (IAA) in NH4HCO3 buffer (50 mM, pH 7.4), followed by incubation for 2 h at 37°C. After 2 h incubation at 37°C, hGH was precipitated with 700 μL of 0.5 M HClO4 and centrifuged at 14,000 g for 10 min. The resulting pellet was washed twice with milliQ water. Thereafter, the pellet was reconstituted in 200 μL of NH4HCO3 buffer (50 mM, pH 8). 5 μg of trypsin (protein/trypsin = 40) were added to each sample (we noted, that digestions performed with porcine pancreas trypsin obtained from Sigma-Aldrich revealed some chymotryptic cleavage sites). After overnight incubation at 37°C, the proteolytic digests were purified using Amicon ultra-0.5 centrifugal devices equipped with 10 kDa NMWL membranes to separate undigested protein and trypsin from the proteolytic peptides.

Analysis on a SYNAPT-G2 Mass Spectrometer

For HPLC-MS1 analysis samples were analyzed on an Acquity UPLC® system (Waters Corp., Milford, MA), equipped with a Vydac 218MS™ capillary column (25 cm × 500 μm C18, 5 μm) from Grace (Deerfield, IL) and connected to a Synapt G2 mass spectrometer (Waters Corp., Milford, MA). 10 μl of tryptic digests were eluted with a 30 min linear gradient of 10-60% (v/v) ACN in aqueous formic acid (0.06%, v/v), followed by a second linear gradient of ACN in aqueous formic acid (0.06%, v/v) to 80% (v/v) within 10 min. The flow rate was set to 20 μl/min. All lenses in the G2 mass spectrometer were optimized on the [M + 2H]2+ ion from the [Glu]1-fibrinopeptide for maximum resolution. The cone voltage was 45 V and Ar was admitted to the collision cell. The spectra were acquired between 50-2000 amu (amu = atomic mass unit) and the data were accumulated for 2 s per cycle. Analysis of raw data files was carried out with the MassLynx™ Software (Waters Corp., Milford, MA).

Analysis on an LTQ-FT-ICR Mass Spectrometer

The hybrid linear quadrupole ion trap Fourier transform-ion cyclotron resonance (LTQ-FT-ICR) mass spectrometer from ThermoFinnigan, Bremen, Germany was operated as described. (48) Liquid chromatography was carried out with an LC Packings Ultimate Chromatograph (Dionex, Sunnyvale, CA) system coupled to the same Vydac column (25 cm × 1 mm C18, 3.5 μm). MS2 spectra were recorded based on MS survey scans; MS2 and MS3 spectra of selected ions were recorded exclusively in separate runs for optimal spectrum quality. Raw data files were analyzed with the Xcalibur™ software package (ThermoFinnigan, Bremen, Germany). The program ReAdW version 4.3.1 was used to generate mzXML files for analysis with the MassMatrix program (49).

Analysis on an In-house Packed 75 cm C18 Nano-column Coupled to a Xevo-G2 Mass Spectrometer

We have constructed an ultra-high pressure chromatography system based on slight modifications of a published design (50). This chromatography system was equipped with an in-house prepared C18 nano-column (75 cm x 0.75 μm), packed with particles supplied by Waters Corporation (Charge Surface Hybrid, CSH™). The analytes were eluted through the column at a flow rate of 0.25 μL/min at a pressure of 34,000 psi. The mobile phases consisted of water/acetonitrile/formic acid at a ratio of 99%, 1%, 0.08% (v:v:v) for solvent A and a ratio of 1%, 99%, 0.06% (v:v:v) for solvent B. A linear gradient, starting with 7% A and changing A by 0.15%/min over 250 min was prepared using a NanoAcquity ultra-high performance liquid chromatography (nanoAcquity-UPLC) system (Waters Corporation, Milford, MA), and stored in a 200 μL loop (50) prior to be pushed through the analytical column. More details about the chromatography system will be reported elsewhere. The chromatographic system was coupled to a Waters Xevo-G2 (Waters Corp., Milford, MA) mass spectrometer operating in the positive mode. A nanoflow gas pressure was set to 0.2 bar, with a cone gas flow set to 4 L/h and a source temperature of 100°C. The capillary voltage and cone voltage were set to 2800 V and 35 V, respectively. The Xevo-G2 acquisition rate was set to 0.5 s with a 0.0 s inter-scan delay. Ar was employed as the collision gas. The instrument was operated in the MSE mode. The instrument was operated with the first resolving quadrupole in a wide pass mode with the collision cell operating with different alternating energies. To acquire the non-fragmented MS1 spectrum the collision cell was operated at 5 eV. The fragmented MS1 ion spectra were acquired by ramping the collision cell energies between 15 eV and 45 eV. The data were collected into separate data channels. All analyses were acquired using the lockspray to ensure mass accuracy and reproducibility; [Glu]1-fibrinopeptide B was used as the lock mass (m/z 785.8426, doubly charged) at a concentration of 2 pmol/μL and flow rate of 1 μL/min. Data were collected in centroid mode, the lockspray frequency was set to 5 s, and data were averaged over 10 scans.

Estimation of the Yields of Degradation Products

We estimated the relative yields of the detected degradation products through the peak areas in the extracted ion chromatograms of the most abundant ions of the respective tryptic peptides. Since signal intensity strongly depends on the molecular weight and chemical structure of an ion, we only compared the signal intensities of products to signal intensities of the native, unmodified tryptic peptides. These are presented in Tables I and III. The relative abundances of the cross-links reported in Table II were estimated by comparing signal intensities of all detected cross-links, which, however, resulted in potentially large errors. Due to the inherent uncertainty of uncalibrated relative quantitation, relative abundances are only estimated and products are described as major (++), minor (+; one order of magnitude lower signal intensity compared to ++) or as trace (t), when signals were below the quantitation limit. The loss of disulfide-containing tryptic fragments was calculated based on peak areas, and referenced to the more photolysis-resistant fragments T12 and T13 (LEDGSPR and TGQIFK) as internal standards. Yields of cross-links formed between T6 - T16 and T20 - T21 were estimated by dividing the peak area of the new cross-link by the peak area of the respective native disulfide in the control sample, corrected for concentration differences with T12 and T13 as internal standards.

Table I Amino Acid Modifications Observed After Irradiation at λ = 254 or λ > 295 nm and Subsequent Derivatization with NEM, Tentatively Assigned by MS and MS2 Analysis
Full size table
Table II Crosslinks Observed After Irradiation at λ = 254 or λ > 295 nm and Derivatization with NEM
Full size table

The yields of free thiols were measured by derivatization with ThioGlo® 1 (51). Briefly, to irradiated samples 400 μl of 6 M guanidine hydrochloride were added and the pH was adjusted to 7 with 0.1 M Na3PO4. ThioGlo® 1 reagent (2 mM in ACN) was added for a final concentration of 20 μM and the mixture was incubated for 30 min at room temperature. For calibration, non-irradiated samples were reduced with 0.1 M NaBH4 in 400 μl guanidine chloride (6 M) at pH 9 for 15 min at 37°C. The remaining NaBH4 was quenched by lowering the pH to 6.0 with 0.2 M NaH2PO3. Thereafter, the pH was adjusted to 7 with 0.1 M Na3PO4 and the volume was adjusted to 800 μl with water. ThioGlo® 1 was added for a final concentration of 20 μM. After incubation for 30 min at room temperature, the fluorescence was measured (Ex: λ = 379 nm, Em: λ = 513 nm). Fluorescence readings were corrected for background fluorescence by subtracting the fluorescence of non-irradiated, non-reduced samples that were treated with ThioGlo® 1.

Data Analysis

Targeted data analysis for known disulfide degradation products was performed by the MassMatrix program (49). Non-targeted data analysis was performed by analyzing strong newly formed peaks manually.

Results

We photo-irradiated hGH at λ = 254 nm and λ > 295 nm for typically 5 min and 1 h, respectively, and thereafter free thiols were alkylated with NEM. The mass spectrometric analysis of tryptic digests (Tables I, II, and III, Figs. 1, 2, 3, 4, 5, 6, 7, 8, and 9 and Supplementary Information, Figs. S11-S63) showed that such irradiation conditions degraded the disulfide-containing tryptic peptides by ca. 80% (λ = 254 nm) and 20% (λ > 295 nm), respectively. Chromatograms of tryptic maps comparing control and irradiated hGH under both photo-irradiation conditions are displayed in Figs. S64 and S65. After the irradiation at both wavelengths we observed an increase of background signals in the total ion count chromatograms (TIC) during HPLC-MS analysis, very similar to that observed when hGH was subjected to oxygen centered radicals derived from AAPH (52), likely indicating the formation of a variety of products at relatively low yields.

Table III Peptide Cleavages Observed After Irradiation at λ = 254 or λ > 295 nm and Subsequent Derivatization with NEM
Full size table
Fig. 1
figure 1

MS2 spectrum and tentative structure of the doubly charged precursor ion with m/z = 551.79 assigned to NYGLLYGFR and derived from T16 (NYGLLYCFR). The spectrum is in excellent agreement with the Cys-165 to Gly-165 transformation. The sample contained 10 μM hGH in 0.4 ml phosphate (0.1 M, pH 7.4) and was photo-irradiated at λ = 254 nm for 10 min with four RPR-2537 Å lamps. After irradiation a 100-fold excess of NEM and crystalline guanidine for final concentration of 5 M was added; the pH was adjusted to ca. pH 7.4 with tribasic phosphate 0.1 M. After precipitation with 0.5 M HClO4, and two washing steps with water, the protein pellets were resolubilized and digested. Tryptic peptides were analyzed by an LTQ-FTICR mass spectrometer.

Full size image
Fig. 2
figure 2

MS2 spectrum and tentative structure of the doubly charged precursor ion with m/z = 636.82 derived from T16-T17 (NYGLLYCFRK). The spectrum is in agreement with a dehydrogenation of Cys-165 to a thioaldehyde and a dehydrogenation of Phe-196, Arg-167 or Lys-168. Peaks with m/z = 448.18 and 549.29 are assigned to the y3 and y4 fragments which correspond to cleavage before and after Cys-165 thioaldehyde. Conditions as in Fig. 1.

Full size image
Fig. 3
figure 3

MS2 spectrum and tentative structure of the doubly charged precursor ion with m/z = 781.28 derived from T21 (SVEGSCGF). The spectrum is in agreement with a dehydrogenation of Cys-189 to a thioaldehyde and a dehydrogenation of Phe-191. Peaks with m/z = 430.22 and 412.14 could originate from a coeluting isobaric species that is characterized by a mass shift of −4 Da shared by Ser-188 and Cys-189. Conditions as in Fig. 1.

Full size image
Fig. 4
figure 4

MS2 spectrum and tentative structure of the singly charged precursor ion with m/z = 749.32 derived from T21 (SVEGSCGF). The cyclic peptide structure exclusively leads to strong b7 and b6 fragments. Conditions as in Fig. 1.

Full size image
Fig. 5
figure 5

MS3 spectrum of the singly charged ion with m/z 509.16 which is the b6 - H2O fragment of the precursor ion with m/z = 749.32 (Fig. 4). The fragmentation pattern used to assign peaks is displayed in Scheme 1. The sample was prepared as described in Fig. 1.

Full size image
Fig. 6
figure 6

MS2 spectrum and tentative structure of the doubly charged precursor ion with m/z = 556.78 derived from T16 (NYGLLYCFR). The spectrum is in agreement with a linear peptide of the sequence NY(+10 Da)GLLGFR which is an unlike structure from a chemical viewpoint. Figure 7 shows the MS3 spectrum of y8 2+ that suggest a crosslinked Tyr-Cys derived structure modified by a mass shift of −36 Da.

Full size image
Fig. 7
figure 7

MS3 spectrum of the y8 2+ fragment (m/z = 499.92) derived from the doubly charged the precursor ion with m/z = 556.78 (Fig. 5) which is a −36 Da modification of T16 (NYGLLYCFR). The spectrum would be in agreement with a structure that features the shown cross-link (referred to as structure cY in the text), characterized by a mass shift of −36 Da. Conditions as in Fig. 1.

Full size image
Fig. 8
figure 8

Magnified and full (insert) MS2 spectrum as well as tentative structure of the doubly charged precursor ion with m/z = 1260.59 derived from T6 (YSFLQNPQTSLCFSESIPTPSNR). The loss of CO2 from MH+ is in agreement with a carbamate structure. The lack of the y10 fragment of T6 but presence of the y9 - H2O fragment suggest the proposed cyclic structure. Conditions as in Fig. 1.

Full size image
Fig. 9
figure 9

Magnified and full (insert) MS2 spectrum as well as the tentative structure of the doubly charged precursor ion with m/z = 1306.58 derived from T6 (YSFLQNPQTSLCFSESIPTPSNR). The y11 fragment gives evidence of a Cys53-derived modification of m/z = 71.97 on the N-terminus of the cleavage fragment. Conditions as in Fig. 1.

Full size image

In Tables I, II, and III, which summarize the photochemical products, the majority of structures can be assigned by MS1, MS2 and/or MS3 analysis and by comparison to analogous products, which we characterized for the photo-degradation of a series of model peptides (23). Nevertheless, we must consider the structural assignments tentative as we could not fractionate sufficient quantities for complementary NMR analysis. Tryptic digests were screened for known disulfide degradation products by HPLC-MS analysis and non-targeted screening was performed on prominent newly formed peaks after photo-irradiation. The detected modifications are summarized in Tables I, II, and III as single amino acid modifications, cross-links and cleavage products, respectively. In Tables I, II, and III we have grouped all the different modifications of an individual amino acid in clusters, according to the position of the original amino acid in the hGH sequence and its respective tryptic fragment (referred to as T). The corresponding mass spectra in support of the product structure(s) are referenced in the last columns of the Tables. Figures 1, 2, 3, 4, 5, 6, 7, 8, 9, and 10 are presented in the manuscript while Figs. S11-S63 are located in the Supplementary Information.

Fig. 10
figure 10

HPLC-MS analyses on a 75 cm C18 column of the peptide digests of hGH, obtained after photoirradiation of hGH at λ = 254 nm and λmax > 295 nm. (a) Resolution of dithiohemiacetal 34 and thioether 36 from three isomers of vinyl thioethers, 35a, 35b, and 35c, obtained after photo-irradiation at λ = 254 nm. (b) Analyses of products 34, 35a, b, c, and 36, obtained after various times of photo-irradiation of hGH at λ = 254 nm. (b) Analyses of products 34, 35a, b, c, and 36, obtained after various times of photo-irradiation of hGH at λ > 295 nm.

Full size image

Single Amino Acid Modifications on Individual Tryptic Peptides

The photo-degradation products with highest signal intensities observed for all originally Cys-containing peptides, except Cys53 in T6, irradiated at λ = 254 nm, were the reduced Cys residues, subsequently alkylated with NEM (+125 Da) (Table I, products 1, 6, 17, and 24 containing Cys-NEM). The hydrolysis products of NEM (+143 Da), which can form during tryptic digestion (53), were detected only in low yields and are, therefore, not included in Table I. Analysis of irradiated protein samples with ThioGlo® 1 revealed that the ratio [Cys]:[0.5 cystine], i.e. a measure for the formation of reduced Cys per Cys residues present in the original disulfide bond, was ca. 0.3 after 5 min of irradiation at λ = 254 nm and ca. 0.08 after 1 h of irradiation at λ > 295 nm. Consistent with a disproportionation mechanism of CysS• radical pairs, (23) we also detected the formation of Cys thioaldehyde (2-amino-3-thioxopropanoic acid; MWCys − 2 Da) for two Cys residues (products 7 and 28), albeit at lower yields as compared to reduced Cys residues. Such lower yields of Cys thioaldehydes are expected based on their potential for hydrolysis to serine aldehyde (SerSA), or subsequent reactions of Cys thioaldehyde, and its tautomeric form, 2-amino-3-mercaptoacrylic acid, to form dithiohemiacetals, vinyl thioethers and other products (see below). In fact, the derivatization of photo-irradiated hGH with NEM yields NEM-adducts of 2-amino-3-mercaptoacrylic acid in tryptic fragments T6, T20 and T21 (products 3, 23 and 32).

Through 1,3-hydrogen transfer, CysS• radicals exist in equilibrium with ●Cα radicals, which can subsequently eliminate HS• to yield Dha with a rate constant on the order of 5x103 s−1 (33). Consistent with such mechanism, we detected Dha formation for three Cys residues present in hGH (products 10, 20, and 25). We note, however, that some conversion of Cys165 (T16) and Cys189 (T21) to Dha is also detected in control samples, which may be the result of β-elimination during sample preparation for analysis.

We observed trace amounts of S-sulfocysteine (CysSSO3H; product 5) in T6, likely originating from minor yields of Cys perthiyl radicals (CysSS•) generated through reaction 10 (reaction 10 is representatively shown for isolated cysteine). (54) Perthiyl radicals add oxygen, and a plausible (54) route to S-sulfocysteine is provided by reactions 11-15, where D− represents a suitable electron donor, e.g. CysSH or CysS− (in competition to reaction 15, the radical CysSS(O)2O• can release SO3 to yield CysS•) (54).

$$ \mathrm{CysSSCys}+\mathrm{hv}\to {\mathrm{CysSS}}^{\cdot }+\left({\mathrm{H}}_3{\mathrm{N}}^{+}\right)\left({{\mathrm{CO}}_2}^{-}\right)\mathrm{CH}-{{\mathrm{CH}}_2}^{\bullet } $$
(10)
$$ {\mathrm{CysSS}}^{\bullet }+{\mathrm{O}}_2\to {\mathrm{CysSS}\mathrm{OO}}^{\bullet } $$
(11)
$$ \mathrm{CysSS}{\mathrm{OO}}^{\bullet }+\mathrm{hv}\to {\mathrm{CysSS}}^{\bullet }{\left(\mathrm{O}\right)}_2 $$
(12)
$$ {\mathrm{CysSS}}^{\bullet }{\left(\mathrm{O}\right)}_2+{\mathrm{O}}_2\to \mathrm{CysSS}\ {\left(\mathrm{O}\right)}_2{\mathrm{O}\mathrm{O}}^{\bullet } $$
(13)
$$ \mathrm{CysSS}\ {\left(\mathrm{O}\right)}_2{\mathrm{O}\mathrm{O}}^{\bullet }+{\mathrm{RO}\mathrm{O}}^{\bullet}\to \mathrm{CysSS}\ {\left(\mathrm{O}\right)}_2{\mathrm{O}}^{\bullet }+{\mathrm{RO}}^{\bullet }+{\mathrm{O}}_2 $$
(14)
$$ \mathrm{CysSS}{\left(\mathrm{O}\right)}_2{\mathrm{O}}^{\bullet }+{\mathrm{D}}^{-}/{\mathrm{H}}^{+}\to {\mathrm{CysSSO}}_3\mathrm{H}+{\mathrm{D}}^{\bullet } $$
(15)

The addition of oxygen to CysS• (55,56) will ultimately lead to oxy acids. Consistent with the predominant formation of CysS• via reaction 1, we detect the formation of Cys sulfinic acid in T16 and T21 (products 8 and 27), and the formation of Cys sulfonic acid in T16 and T20 (products 9 and 19).

We detected the conversion of Cys to Ala for Cys165 and Cys182 (products 11 and 18), consistent with earlier reports on the photolysis of isolated cystine (57,58,59). Mechanistically, Ala can be obtained through reduction of Dha by electrons generated via reaction 4 (35), consistent with the observation of Dha (products 10, 20 and 25; see above). However, an alternative route for Ala formation would be reduction of alanyl-Cβ radicals generated via reaction 10.

The conversion of Cys to Gly, detected for Cys165, Cys182 and Cys189 (products 12, 21 and 29) represents an important novel modification. Figure 1 displays representative data for the MS/MS characterization of Gly originating from Cys165 in the tryptic peptide T16. Mechanistically, the formation of Gly can be rationalized through 1,2-hydrogen transfer of an initial CysS• radical (33) (see Discussion). The latter mechanism also provides an entry for the formation of serine aldehyde (SerSA), detected in peptides T16 and T21 (products 13 and 30). However, SerSA may also be formed through the hydrolysis of thioaldehydes, such as observed for peptides T16 and T21 in products 7 and 28 (see above), and through oxygen addition to alanyl-Cβ radicals, followed by the Russell mechanism (60). Importantly, the Russell mechanism would also yield Ser, a product which was, in fact, observed in peptides T16 and T21 (products 14 and 31).

An important pathway for peptide sequence modification is also the conversion of Trp to Gly (product 33), detected for Trp86 in peptide T9. Originally, we had reported such a modification for several Trp residues in monoclonal antibodies and model peptides, initiated through photo-induced electron transfer from Trp to a nearby disulfide bond (8).

Multiple Amino Acid Modifications on Individual Tryptic Peptides

We detected several products which contained multiple modified amino acids on individual tryptic peptides. These products will be described in the following sections; however, for most of these products an unambiguous structural assignment was not possible as the mass spectrometry data would be consistent with several isobaric structures.

One modification of T6 displayed a mass shift of -4 Da (product 2). This product showed the highest signal intensity of all T6-derived photo-degradation products. The MS/MS data are consistent with a loss of 2 Da from each Cys53 and Phe54, possibly indicating the formation of thioaldehyde (or its tautomer, 2-amino-3-mercaptoacrylic acid) and dehydrophenylalanine. However, we were unable to detect abundant b12 and y11 fragments, indicating that fragmentation between Cys53 and Phe54 may not be possible. This feature may indicate a cross-link generated from the original Cys53 and Phe54 residues, such as a vinyl thioether cross-link formed by reaction of 2-amino-3-mercaptoacrylic acid with dehydrophenylalanine (because of the cross-link nature of the latter, this potential structure is also listed as product 48 in Table II; see below). An additional modification of T6 displayed a mass shift of -36 Da (product 4). The MS/MS data are consistent with the conversion of Cys53 to Dha and either Leu52 to dehydroleucine or Phe54 to dehydrophenylalanine.

We detected a modification of T16 with a mass shift of -4 Da (product 15). The MS/MS spectrum suggests that the modification is located on the subsequence Cys165-Phe166-Arg167. However, the spectrum cannot discriminate between two separate modifications of -2 Da on two amino acids or a vinyl thioether cross-link between 2-amino-3-mercaptoacrylic acid derived from Cys165 and either dehydrophenylalanine or dehydroarginine (hence, this product is also listed as product 47 in Table II).

An additional modification, on the tryptic peptide T16-T17 (containing one missed tryptic cleavage site), displayed a mass shift of -4 Da (product 16). Here, the MS/MS data clearly suggest the conversion of Cys165 into thioaldehyde (or its tautomer, 2-amino-3-mercaptoacrylic acid) and an additional dehydroamino acid, either dehydrophenylalanine166, dehydroarginine167 or dehydrolysine168 (Fig. 2).

We detected a modification of T20 that displayed a mass shift of −4 Da (product 22). The MS2 spectrum suggests two modifications, the conversion of Cys182 into the respective thioaldehyde and of Arg183 into dehydroarginine.

A product of the tryptic peptide T21 displayed a mass shift of −4 Da (product 26). The spectrum displayed in Fig. 3 indicates that two isobaric species could be coeluting: i) a product where Cys189 is converted into thioaldehyde (or its tautomer, 2-amino-3-mercaptoacrylic acid) and Phe191 is converted into dehydrophenylalanine, and ii) a product with a vinyl thiohemiacetal crosslink between 2-amino-3-mercaptoacrylic acid generated from Cys189 and SerSA generated from Ser188.

Crosslinks

The photo-degradation of hGH led to the formation of various covalent cross-links, which are summarized in Table II and described in detail below. We detected a dithiohemiacetal crosslink (product 34) between Cys182 and Cys189 within tryptic peptides T20 andT21, alkylated with NEM (+125 Da mass shift relative to the disulfide). This crosslink was generated with an estimated yield of 2.5% when hGH was irradiated for 5 min at λ = 254 nm and an estimated yield of 1.5% when hGH was irradiated for 1 h at λ > 295 nm. Such dithiohemiacetal crosslinks have been characterized in detail by mass spectrometry and NMR for the photo-degradation of cystine-containing model peptides (23) but also observed during the photo-irradiation of insulin (25) and a monoclonal antibody (61). Mechanistically, they are likely generated through the reaction of a Cys thiol with a thioaldehyde (though, technically, the combination of a Cys thiyl radical with a Cys-derived α-mercaptoalkyl radical would provide an alternative route) (23). Another cross-link generated with significant yields was a vinyl thioether between photo-products generated from Cys182 and Cys189 (product 35) and between photo-products generated from Cys53 and Cys165 (product 39). Product 35 was generated with an estimated yield of 3.9% when hGH was irradiated for 5 min at λ = 254 nm and a yield of 0.4% when hGH was irradiated for 1 h at λ > 295 nm while product 39 was only generated in trace amounts. Mechanistically, such vinyl thioether can form via at least two pathways: (i) the reaction of 2-amino-3-mercaptoacrylic acid with Dha, both of which have been detected for Cys182 and Cys189 (Table I) (these vinyl thioethers can be resolved into various isomers on a 75 cm nano-column; see below), and (ii) the photochemical decomposition of dithiohemiacetal (23) (such as product 34). The photochemical decomposition of dithiohemiacetal generates H2S/HS−, which, during light exposure, can reduce the vinyl thioether further to the thioether (23). Not surprisingly, thioether crosslinks between Cys182 and Cys189 were detected as major products (product 36), with an estimated yield of 2.2% when hGH was irradiated for 5 min at λ = 254 nm and a yield of 1.2% when hGH was irradiated for 1 h at λ > 295 nm. The formation of such thioether can have a significant impact on the activity of hGH (62). Interestingly, a vinyl disulfide (of minor yield) was detected in place of the original disulfides in the tryptic peptide T6-T16 (product 40; originating from Cys53 and Cys165). In addition, traces of a vinyl disulfide were generated from Cys165 and Cys189 (product 41), and traces of a divinyl disulfide were generated between Cys182 and Cys189 (product 37).

MS2 and MS3 spectra (Figs. 4 and 5) are consistent with a vinyl ether crosslink between Ser-184 and a reaction product from Cys-189 in T21 (product 38). A tentative structure of product 38 is displayed in the insert in Fig. 4. The MS2 spectrum of the precursor ion with m/z 749.31 (Fig. 4) displays dominant b6 and b7 fragments, which indicate selective fragmentation exclusively N-terminal of Gly-190 as well as between Gly-190 and Phe-191. This strongly suggests a cyclic structure connecting reaction products of Ser-184 and Cys-189. In comparison, the native peptide, SVEGSCGF, alkylated with NEM, shows highly abundant y5 and y6 fragments (cleavages between Val-185, Glu-186 and Gly-187). Further evidence for a crosslink between products of Ser-184 and Cys-189 is obtained by examining the MS3 spectrum of the dehydrated b6 ion (b6-H2O), a putative cyclic peptide (Fig. 5). A tentative mechanism for the fragmentation of product 38 during MS2 and MS3 analysis (of the b6-H2O fragment) is displayed in Scheme 1. The prime peaks in the MS3 spectrum can be well assigned by the following fragmentation sequence: First, small neutral molecules such as H2O and CO are lost, or alternatively, the cyclic structure is ruptured followed by the elimination of Gly187 to give the product ion with m/z = 450.2. Second, either Ser188, or Glu186 are lost. The sites for dehydration of the precursor ion b6 − H2O are either Ser188 or Glu186 and, hence, fragmentation patterns of both the native and dehydrated amino acids Ser and Glu are observed. The loss of Glu86 is followed by that of Val185 which leads to the products ions with m/z = 224.02 and m/z = 206.06.

Scheme 1
scheme 1

Tentative fragmentation of the cyclic precursor ion with m/z = 749.31.

Full size image

We observed a total of three disulfide scrambling products (products 42-44), rationalized by the photolytic formation of Cys (products 1, 6, 17 and 24; see above) followed by disulfide exchange processes during photo-irradiation (note, that NEM was added after photo-irradiation minimizing disulfide exchange during sample preparation for analysis). Importantly, one of the new disulfides observed was a symmetric disulfide (product 42), i.e. formed between identical peptides on different hGH molecules.

We detected a modification of T16, generated through the loss of 36 Da (original sequence: NYGLLYCFR) (product 45). Product 45 displays complex MS2 and MS3 spectra (Figs. 6 and 7). The MS2 spectrum (Fig. 6) of the doubly charged molecular ion (m/z 556.78) shows a series of distinct fragmentation patterns. We observe the ions y1 + (m/z 175.17) and y2 + (m/z 322.21), indicating the cleavage of unmodified Arg and Arg-Phe from the C-terminus. We also observe the ions b7 + (m/z 791.21), b8 + (m/z 938.34), y8 + (m/z 998.54) and y8 ++ (m/z 499.92), indicating that the loss of 36 Da from the original sequence must have occurred on the subsequence YGLLYC. A new ion with m/z 768.61 is generated through the loss of 230 Da from y8 + (Fig. 6), confirmed by MS3 experiments on the y8 ++ fragment (Fig. 7).

The loss of 230 Da from y8 + and y8 ++ would be consistent with the formation of a vinyl ether between Tyr164 and a photoproduct of Cys165 (tentatively assigned to structure cY, shown below), and gas phase cleavage of an internal fragment from y8 +. The formation of cY is rationalized by the conversion of Cys165 into thioaldehyde (23,37) (see product 7), followed by thiohemiacetal formation between Tyr164 and the thioaldehyde, and photo-induced elimination of H2S (23). This mechanism will be discussed in more detail in the Discussion section).

MS3 experiments on y8 ++ do not only confirm a neutral loss of 230 Da, but also the formation of the corresponding protonated species, with m/z 231.15 (Fig. 7). In the following, the individual signals in the MS3 spectrum of y8 ++ will be rationalized with the product ions given in parenthesis: y8 ++ can cleave Tyr (m/z 835.64) or Phe-Arg (m/z 677.41), where the latter can be followed by cleavage of Tyr-Gly (m/z 457.36), Tyr-Gly-Leu (m/z 344.19) or Tyr-Gly-Leu-Leu (m/z 231.15). The side chain of Arg can cleave (63) the peptide bond C-terminal of Gly, followed by the cleavage of Leu-Leu-X (m/z 542.43) and, possibly, Tyr (m/z 379.39). Alternatively, the side chain of Arg can induce cleavage (63) C-terminal of Leu163, followed by the loss of X (m/z 768.51), or C-terminal of Leu162, followed by the loss of Leu-X (m/z 677.41). Most of these fragments are also present in the MS2 spectrum derived from the ion with m/z 556.78 (Fig. 6). In addition, the MS2 spectrum reveals the loss of 287 Da (m/z 825.55) and the presence of an ion with m/z 288.14. This ion would be consistent with an isobaric structure of X, generated through a vinyl ether crosslink between Tyr160 and an oxidation product of Cys165 (product 46), connected to Gly161, i.e. Z-G. However, as the formation of Z-G would require multiple gas phase fragmentation processes, and constitutes the only indication for the formation of such isobaric cross-link, additional experimental evidence for such structure would be necessary. MS3 experiments on the ion with m/z 288.14 (Fig. S62), and of the ion with m/z 260 (Fig. S62) were inconclusive.

Additional evidence for the formation of product 45 was obtained in the tryptic fragment T16 + T17, containing one missed cleavage (Fig. S37).

Product 47 represents a putative vinyl thioether cross-link between a thioaldehyde (or its tautomer, 2-amino-3-mercaptoacrylic acid) generated from Cys165 and dehydrophenylalanine generated from Phe166. This cross-link would essentially derive from product 15, which was discussed above. The absence of b2 and y7 ions in the MS/MS spectrum (Fig. S40) of products 15 and 47 would be consistent with such a cross-link.

Product 48 represents a putative cross-link on T6 between a thioaldehyde (or its tautomer, 2-amino-3-mercaptoacrylic acid) generated from Cys53 and dehydrophenylalanine generated from Phe54. This cross-link would derive from product 2, discussed above. The absence of b12 and y11 fragments in the MS/MS spectrum of products 2 and 48 would be consistent with such a cross-link.

We detected two putative Cys-Trp crosslinks between T9 and T16. One of these cross-links (product 49) displays the mass of the two tryptic fragments T9 and T16 minus 2 Da. Based on the MS/MS data (Fig. S41), this cross-link is tentatively assigned to a thioether cross-link of Cys165 with the indole of Trp86. Notably, thioether cross-links between Cys and an aromatic amino acid residue, i.e. Tyr, have been detected during the photo-irradiation of insulin (25). The other cross-link (product 50) displays the masses of the two tryptic fragments T9 and T16 minus 34 Da. The MS/MS data (Fig. S42) of product 50 are consistent with a cross-link between Cys165 and Trp86, generated after the loss of H2S (34 Da). A tentative structure of this cross-link would be a Michael addition product of the indole ring to DHA, generated from Cys165. The Michael addition of indole to α,β-unsaturated ketones has been reported (64), though unsubstituted indoles tend to add via the C-3 position. In the case of Trp, the indole ring would likely add via the indole nitrogen.

We detected a putative Trp-Tyr crosslink (product 51) between T9 and T14 which displays the mass of the two tryptic fragments minus 2 Da. The MS/MS data (Fig. S36) of this cross-link are consistent with a covalent bond between Tyr143 and Trp86, analogous to covalent bonds formed between two Tyr residues in dityrosine cross-links. In fact, trace amounts of such a dityrosine cross-link were detected between tryptic peptides T10 and T14 (product 52). The MS/MS data of product 52 (Fig. S35) are consistent with a covalent bond between Tyr143 and Tyr111.

Peptide-Bond-Cleavages

Products derived from peptide bond cleavages are summarized in Table III. We detected an N-terminal cleavage fragment of T6, the peptide Y41SFLQNPTQTS51 (product 53), where the C-terminal Ser51 displays a carboxylic acid, such as would be expected for peptide bond hydrolysis. This product is absent in non-irradiated controls, indicating that it is not generated through enzymatic processes (e.g., it is not generated during proteolytic digestion). In addition, we detected the same N-terminal cleavage fragment of T6, where the C-terminal Ser51 is present as an amide, i.e. Y41SFLQNPTQTS51-amide (product 54). An analogous cleavage product of T6 was detected C-terminal of Leu52, i.e., the peptide Y41SFLQNPTQTSL52-amide (product 55). Consistent with the formation of these N-terminal cleavage fragments of T6, we also detected a series of C-terminal cleavage products of T6, i.e. a peptide corresponding to C 53(−31 Da)FSESIPTPSNR64 (product 56), a peptide corresponding to F 54(+26 Da) SESIPTPSNR64 (product 57), and a peptide corresponding to C 53(−19 Da)FSESIPTPSNR64 (product 58). In these products, italic one letter codes followed by a mass loss in parenthesis signal a modification to the original amino acid residue. These modifications will be rationalized in the Discussion section.

Two cleavage fragments were also detected for tryptic peptide T16, i.e. N159YGLLY164 (product 59) and N159YGLLY164-amide (product 60).

Time-Dependent Analysis of the Dithiohemiacetal, Thioether, and Vinyl Thioethers on a 75 cm Nano-column

A time-dependent analysis of product formation was carried out representatively for the dithiohemiacetal, thioethers and vinyl thioethers. To better separate potential isomeric products, the photo-irradiated samples were injected onto our in-house packed 75 cm C18 column. Photo-irradiation at λ = 254 nm led to three isomers of product 35, referred to as 35a, 35b, and 35c (Fig. 10a). The collision-induced dissociation spectra of 35a, 35b, and 35c were similar to that presented in Fig. S41, consistent with a vinyl thioether structure. Quantitatively, thioether 36 was generated in significantly higher yields compared to products 34 and 35a, b, c, when hGH was photo-irradiated at λ = 254 nm, shown in Fig. 10b, where the respective yields are plotted as a function of irradiation time. In contrast, the dithiohemiacetal 34 was generated at higher yields compared to thioethers and vinyl thioethers, when photo-irradiation was carried out at λ > 295 nm (Fig. 10c). This is consistent with the light-dependent conversion of dithiohemiacetal into thioether at λ = 254 nm. (23) Our mechanistic analysis with model peptides (23) revealed that the light-dependent conversion of dithiohemiacetal into thioether proceeds via the intermediary generation of vinyl thioether. Consistent with the intermediacy of vinyl thioethers in the analogous processes in hGH, vinyl thioethers represented the photoproducts with the lowest yields in Fig. 10b and c.

Discussion

The objective of the present work was to provide a comprehensive characterization of products originating from the photo-degradation of protein disulfides. Hence, this is not a classical pharmaceutical stability study but a forced degradation study applying selected stress conditions (2) for generation of a maximal number of photoproducts.

Mechanistic Considerations

The photo-degradation patterns initiated by UV- irradiation at λ = 254 nm and λ > 295 nm are very similar, which suggests common degradation pathways. Indeed, at both wavelengths disulfide bridges of Cys can be cleaved to form CysS● radicals (19,20,21,22), which are the key intermediates leading to the various products observed (e.g. reactions 7-9). In addition, the aromatic residues of Trp and Tyr can be photo-ionized at both wavelengths, which leads to CysS● radicals via the secondary reactions 2, 3, 5 and 6. Once formed, thiolates can be further photo-ionized to CysS● radicals (reaction 4) at λ = 230 - 270 nm (65,66). This likely contributes to the considerably faster photo-degradation at λ = 254 nm compared to irradiation at λ > 295 nm. The broad increase in background signals in MS-experiments after irradiation is probably due to the formation of a large number of degradation products of low yields, which could not be identified at present.

Amino Acid Modifications

The main degradation products detected were Cys residues alkylated with NEM (products 1, 6, 17 and 24). Cys can be generated by at least three distinct mechanisms: i) disproportionation of thiyl radicals to a thiol and thioaldehyde (reaction 7) (23); ii) hydrogen transfer (reaction 8) (35,36,37); and iii) equilibrium 3. On the other hand, sulfinic and sulfonic acids (products 8, 9, 19, and 27) were detected only in minor or trace amounts. They are formed via the reaction of thiyl radicals with oxygen (55,56) or the oxidation of thiols by a variety of oxidizing species such as dioxygen and peroxides (67,68). Furthermore, we detected traces of S-sulfocysteine (product 5), a product also reported for the photo-irradiation of cysteine (69).

Dha (products 10, 20, and 25) is formed either via the elimination of a sulhydryl radical (HS•) from a Cys •Cα-radical (23,33) or via a non-radical dehydrosulfuration reaction during protein digestion (70). Since we detected Dha on T16 and T21 in non-irradiated controls both mechanisms likely contribute to the formation of Dha in our experiments. Ala is the reduction product of Dha (35), which was detected in products 11 and 18. The reduction of Dha to Ala likely involves hydrated electrons (35) generated during light exposure from thiolate and HS− (35) (formed during dehydrosulfuration of Cys; see above). Such mechanism has been confirmed for the reduction of Dha in small model peptides (35).

Thioaldehydes (products 7, and 28) generated by reaction 7 were detected as minor product from all Cys-containing tryptic fragments. Hydrolysis of the thioaldehyde (35), or elimination of HS• from an intermediary alkoxyl radical (reaction 14 in Scheme 2; see below) leads to the respective aldehyde (SerSA), which was detected in trace levels (products 13 and 30). Subsequent reduction of the SerSA (likely analogous to the reduction of DHA) can generate Ser, which was also detected in trace amounts (products 14 and 31). Alternatively, the 2-amino-3-mercaptoacrylic acid tautomers of thioaldehydes can be derivatized with NEM. This process leads to products 3, 23, and 32, where mass spectrometric data confirm the formation of NEM-alkylated Cys derivatives with a mass difference of −2 Da relative to Cys.

Scheme 2
scheme 2

Tentative mechanism for the conversion of Cys to Gly.

Full size image

To the best of our knowledge, the conversion of Cys to Gly (products 12, 21 and 29) represents a novel modification, which requires a multistep mechanism. We propose the mechanism displayed in Scheme 2, where each individual step has been documented for either Cys or comparable radicals in the literature: the CysS● radical is in equilibrium 16 with the α-mercaptoalkyl radical (33) which, analogous to α-hydroxyalkyl radicals (71,72,73), reacts with dioxygen to yield a peroxyl radical (reaction 11). Elimination of superoxide (71,72,73) would yield a thioaldehyde (reaction not shown); however, either the bimolecular reaction with another peroxyl radical (74), or a formal oxygen transfer to a thiol or thioether (52), such as present in Met, could yield an alkoxyl radical (reaction 18) (75). The alkoxy radical could eliminate either HS• to generate SerSA (reaction 19) or thioformic acid (O = CH-SH) to yield a Cα radical of Gly (●Cα-Gly) (reaction 20). The final product Gly is then generated by H-atom transfer from any thiol to ●Cα-Gly (reaction 21) (8). The intermediary formation of ●Cα-Gly is also key to the conversion of Trp86 to Gly (product 33). We have recently documented an analogous transformation in a monoclonal antibody (8) and octreotide (76). The underlying chemistry involves the photolytic formation of a Trp+• radical cation, followed by elimination of 3-methyleneindoleine (3-MEI) (8,76) to yield ●Cα-Gly, followed by H-atom transfer to ●Cα-Gly from any thiol (8).

Several tryptic peptides display changes of masses of −4 Da, consistent with the conversion of Cys into thioaldehyde (or its tautomer, 2-amino-3-mercaptoacrylic acid) and an additional dehydrogenation at a proximate amino acid such as Phe (products 2, 15 and 26), Leu or Phe (product 4), Arg (product 22) and/or Lys (product 16). The lack of y/b fragments between these modified amino acids in products 2, 4, 15 and 26 may hint to covalent cross-link formation, e.g. through Michael addition of 2-amino-3-mercaptoacrylic acid to the other dehydroamino acid. That is why several of these products have also been listed under “cross-links” in Table II (i.e., products 47 and 48). Overall, this finding indicates that proximate amino acids are prime targets for dehydrogenation and/or crosslinking. Furthermore, such modifications can be formed in high yields, where product 2 on T6 represents a major photoproduct. The dehydrogenation of two adjacent amino acids suggests that these processes are mechanistically coupled. A tentative mechanism is displayed in Scheme 3. The photolysis of a disulfide generates a pair of thiyl radicals (reaction 1) (23), each of which can formally abstract an H-atom from an adjacent amino acid residue, generating a pair of Cys and a dehydroamino acid (reaction 22) (23). The two H-atom transfers do not need to proceed simultaneously (23); an initial H-atom transfer to one thiyl radical may be followed by H-atom transfer from the product amino acid radical to the second thiyl radical. Subsequently, deprotonation of a thiol, followed by photo-induced oxidation, regenerates a thiyl radical (reaction 23), which exists in equilibrium 24 with an α-mercaptoalkyl radical (33). The latter can react with oxygen to yield thioaldehyde (and its tautomer, 2-amino-3-mercaptoacrylic acid).

Scheme 3
scheme 3

Tentative mechanism for the dehydrogenation of two adjacent amino acids.

Full size image

Crosslinks

Disulfide scrambling products (products 42-44) were among the most abundant new crosslinks formed, which is not surprising since a major disulfide photo-degradation product is free Cys. We also observed some disulfide scrambling (product 44) in the non-irradiated control samples, which likely took place during proteolytic digestion. To a much smaller extent, vinyl and divinyl disulfide products were observed (products 37, 40, and 41), that can be generated by reaction of the thioaldehyde tautomer, 2-amino-3-mercaptoacrylic acid, with original disulfides and vinyl disulfide products, respectively.

A novel vinyl ether crosslink was detected on T21 (SVEGSCGF), formed between Ser184 and a reaction product of Cys189 (product 38). The proposed mechanism is shown in Scheme 4 (part a), and resembles closely our mechanism published for the formation of a vinyl thioether crosslink (23): the homolytic cleavage of the disulfide bond (reaction 1) yields a pair of thiyl radicals, followed by disproportionation (reaction 7), and cyclization to yield a thiohemiacetal (reaction 25). The continuous light exposure of the thiohemiacetal leads to elimination of HS● (reaction 26), which abstracts an H-atom to generate H2S and the vinyl thioether (reaction 27) (23). Importantly, additional mechanistic support for this pathway has been derived from studies with Ser and Cys disulfide-containing model peptides, where the intermediary thiohemiacetal was trapped by reaction with NEM (unpublished results). In addition to the mechanism proposed in Scheme 4 (part a), an alternative mechanism can lead to vinyl ether formation through reactions 28-30 (Scheme 4; part b): any radical able to abstract an H-atom from the sulfhydryl group of the thiohemiacetal can generate a thiyl radical (reaction 28), amenable to 1,3-H-transfer (reaction 29) (33), followed by the elimination of HS• (reaction 30) (33).

Scheme 4
scheme 4

Tentative reaction mechanism for the formation of a vinyl ether crosslink.

Full size image

The vinyl ethers formed between Tyr164 and a photoproduct of Cys165 (product 45), or Tyr160 and a photoproduct of Cys165 (product 46), are likely generated via an analogous reaction, schematically displayed in Scheme 5 (reactions 31-33). Here, the nucleophilic addition of a phenolic hydroxyl group to a thioaldehyde generates a hemithioacetal, followed by photo-induced release of H2S.

Scheme 5
scheme 5

Formation of a vinyl ether cross-link between Tyr and a thioaldehyde.

Full size image

The disulfide bond between Cys182 and Cys189 undergoes photo-induced rearrangement into a dithiohemiacetal crosslink (product 34), generally formed by nucleophilic attack of a Cys thiolate on a Cys thioaldehyde (analogous to the reactions presented for Ser in Scheme 4, part a) (23) Under continuous exposure to light, dithiohemiacetals convert into vinyl thioethers (23), which are detected for hGH in products 35 and 39. The reduction of vinyl thioethers via photolytically generated (e.g., from HS−) hydrated electrons leads to thioethers (23). For hGH, such thioether product is detected on T20-T21 (product 36).

Mechanistically, product 49, a crosslink between Trp-86 and Cys-165 is likely formed via radical combination of a Trp● radical and a Cys● radical, as proposed for a Trp-Cys crosslink that was observed during photo-degradation of α-lactalbumin (16). However, also product 50 represents a cross-link between Trp86 and a product of Cys165, which is reduced in mass by 34 Da. This cross-link is likely the product of a Michael-addition (64) via the indole nitrogen of Trp86 to DHA (product 10) formed from Cys165.

Radical-radical combination reactions between aromatic radicals of Trp and Tyr are most likely responsible for covalent cross-links formed between Trp86 and Tyr143 (product 51) and between Tyr111 and Tyr143 (product 52).

Cleavage Products

Fragmentation was detected for sequences originally located in tryptic peptides T6 and T16, but not in T20 and T21 (probably due to the small peptide size). Backbone fragmentation via the diamide and α-amidation pathways is well-established in free radical chemistry, and involves the intermediary generation of alkoxy radicals at the Cα atom of the respective amino acid residues (41,77). The formation of alkoxy radicals is initiated by H-atom abstraction from Cα-H bonds by thiyl radicals, generating •Cα radicals (78,79). Scheme 6, reaction 34, representatively shows an intramolecular 1,3-H-transfer at the Cys residue, yielding radical III (33), but H-atom abstraction from neighboring amino acid residues has been observed as well (35,37,79). The Cα radical adds oxygen, yielding peroxyl radical IV (Scheme 6, reaction 35), which subsequently converts into alkoxyl radical V through oxygen transfer to the thiol (Scheme 6, reaction 36). Alternative pathways for alkoxy radical formation would include bimolecular reactions of peroxyl radicals (41). An α-β-cleavage of the side chain of structure V (reaction 37) leads to the oxamide derivative VI, which can further hydrolyze (reaction 38) to product 55 (structure VII) and product 56 (structure VIII). An alternative pathway to product 56 would be the α-amidation pathway (41).

Scheme 6
scheme 6

Tentative mechanism leading to peptide cleavage in T6.

Full size image

Mechanistically, the formation of products 53 and 54 is related to alkoxy radical formation on the Cα atom of Leu52. Here, diamide formation and hydrolysis (analogous to reactions 37 and 38) yields product 53 whereas product 54 may be the result of oxamide formation followed by hydrolysis (analogous to reaction 38). An alternative pathway to product 54 would be the α-amidation pathway (41). Analogous mechanistic considerations apply to the formation of products 59 and 60, respectively.

Significance

The photo-degradation of hGH at λ = 254 nm and λ > 295 nm under forced degradation conditions leads to a large number of disulfide degradation products. A total of 60 products have been identified, and structures tentatively assigned based on MS2 and MS3 data. We were able to identify single amino acid modifications, covalent cross-links, and backbone cleavage products. Mechanistically, the formation of several of the observed cross-links required multiple reactions steps such as, e.g., the formation of a dehydroamino acid, followed by the addition of a thioaldehyde tautomer, 2-amino-3-mercaptoacrylic acid, and photo-induced elimination of H2S. A photo-degradation study such as presented in this paper serves to build a data base of degradation products for the detailed analysis of all types of proteins under various experimental conditions. Future studies need to show whether, and to what extent, some of these products form during manufacturing and storage of protein pharmaceuticals.