HPLC–UV–MS Analysis: A Source for Severe Oxidation Artifacts

Letter pubs.acs.org/ac

Cite This: Anal. Chem. XXXX, XXX, XXX−XXX

HPLC−UV−MS Analysis: A Source for Severe Oxidation Artifacts Fritz Schweikart*,† and Gustaf Hulthe‡ †

Advanced Drug Delivery, Pharmaceutical Sciences, IMED Biotech Unit, AstraZeneca, Gothenburg 43183, Sweden Pharmaceutical Technology & Development, AstraZeneca R&D, Gothenburg 43183, Sweden

Downloaded via UNIV OF WINNIPEG on January 25, 2019 at 21:53:58 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

‡

ABSTRACT: HPLC coupled to both UV and MS is an established setup for purity assessments in many areas. With evolving technology, instrument sensitivity increases, calling for lower sample concentration, while light flux in a commercial UV detector cell is considerably higher than earlier. This evolution has now reached the point where radicals formed by UV light are abundant enough, compared to the analyte levels, to generate unwanted artifact signals in the MS spectrum. In this work we show several examples from pharmaceutical development where UV degradation in the UV detector leads to severely misleading mass spectra in typical day to day samples.

H

compounds are standards, commercially distributed by SigmaAldrich or in case of certain peptides by Bachem. For H2O2 measurements a Fluorometric Hydrogen Peroxide assay kit, Sigma-Aldrich, P/N MAK166, was used.

PLC−UV analysis coupled to online MS instrumentation is widely used in research and product development both in academia and pharmaceutical industry. During purity assessment of active drug compounds, including proteins and oligonucleotides, the quantitative aspect is often based on the UV signal, while identity and structural characterization are preferably assessed by the online coupled MS instrument. In modern low dead volume Diode Array Detectors (DAD) like the Agilent 1290 or Waters Acquity, the detector cell is designed in form of a fused silica capillary, transmitting UV light through the sample by total reflection. These features improve the sensitivity and dynamic range for all kinds of analytes in a wide range of application areas. Nevertheless, this high-performance UV detection may have severe impact on analytes where the mass is measured by MS in an online HPLC-UV-MS setup. This type of flow cells acts as a photochemical reactor. We show here how online UV detection leads to severe oxidative modifications of the analyte and subsequently may lead to incorrect structural characterization results.

■

RESULTS AND DISCUSSION Effect on Peptide and Concentration Dependency. Figure 1 shows, how the peptide Bombesin, containing one

■

Figure 1. Mass spectra of Bombesin (Pyr-QRLG-NQWAVGHLMNH2) in the (1) early, (2) main, and (3) late eluting fraction of the peak (Agilent/Bruker system). For comparison, the MS spectrum with the UV lamp switched off is shown as “no UV.”.

MATERIAL AND METHODS Instrumentation: Two different LC-UV-MS instruments have been used. 1. Agilent Infinity II HPLC equipped with an Agilent 1290 G7117A DAD detector, 10 mm G4212 flow cell coupled to a Bruker Maxis ETD II Q-Tof. 2. Waters Acquity IClass UPLC with 10 mm flow cell coupled to a Xevo G2 XS Q-Tof. Mobile phases for chromatography experiments have been composed of MeOH/H2O/0.1% FA or MeCN/H2O/ 0.1% FA respectively. Samples for infusion were dissolved in 50% MeCN/0.1%FA at 1 μM and were infused with 50 μL/ min passing the DAD detector into the MS instrument. In the presented experiments it is indicated which of the above systems was used. We verified that the herein described effects occur in both systems at a similar magnitude. The used © XXXX American Chemical Society

Trp and one Met residue (tryptophan and methionine) adds up to five additional oxygens. The measured relative portion of oxidation products is highest in the beginning and end of the peak where the peptide concentration is lowest. As low analyte concentration gives higher relative degradation we suggest that the light absorbing reaction is situated in the solvent and not in Received: December 19, 2018 Accepted: January 23, 2019 Published: January 23, 2019 A

DOI: 10.1021/acs.analchem.8b05845 Anal. Chem. XXXX, XXX, XXX−XXX

Letter

Analytical Chemistry sample molecule chromophores itself. Observations clearly show that at high analyte concentration, oxidation effects will be small on a relative scale, while at low sample concentrations severe degradation of analyte is observed. In Figure 1 also signals with +4 Da and +20 Da are detected, which correspond to the formation of two oxidized tryptophan metabolites Kynurenine and Hydroxykynurenine, respectively.1 It is also evident from the low UV absorption readings that the flux of available light in the flow cell is not the limiting factor for degradation. In the case of Ketoprofen (Figure 5) a known light sensitive compound, the formation of radicals seems to take place within the molecule, as the highest degradation is observed at peak maximum (data not shown). Effect on Protein and Amino Acids. Since time-of-flightMS (TOF-MS) and chromatography separation behave poorly when overloaded, typically sample loads of 1−20 picomoles are injected, at which levels severe degradation artifacts can be anticipated as discussed above. We infused α-lactalbumin into a DAD detector (Agilent 1290 or 1200) and subsequently into the MS (Bruker Maxis) with UV light “on” or “off”. The results are shown in Figure 2. In the absence of UV light (Agilent

Figure 3. Traces show the base peak intensity of bovine α-lactalbumin under infusion with 50 μL/min in MeCN/H2O/0.1% FA or MeOH/ H2O/0.1% FA respectively. The signal increases if UV light is switched off (factor given in legend for each trace). It is also indicated whether an Agilent 1290 G7117A or 1200 G1315B DAD was used. BHT was added as radical quencher with 250 ppm.

switched off. When water is illuminated with UV light, it has been shown that radicals like OH· and O2·are formed.2 They may rapidly react with double bonds or sulfides, resulting in e.g. new alcohol or carbonyl functionality or sulfur oxides. The radical electron is carried on, potentially attacking new molecules in a chain reaction. It has been suggested that a radical moiety can travel along a protein backbone3,4 until it finds a suitable reaction site or is trapped by a disulfide bridge and finally lost. When analyzing proteins, peptides and amino acids we detect a substantial proportion of oxidized analyte after it passed the UV flow cell. For some unknown reasons, often the third, sixth and even ninth oxidation states (for certain proteins) are more abundant compared to the addition of 1, 2; 4, 5 and 7, 8 oxygens, respectively. Tryptic digests were carried out to find the oxidized amino acids that may explain this pattern, but mainly singly oxidized Met and doubly oxidized Trp were identified. But still we see, that in two of our tested peptides (LRRWSLG and GHWSYLLRP), with Trp as the only susceptible amino acid, oxidation occurs both singly, doubly but predominantly 3fold (+48da). If this reaction is catalyzed by only one radical associated with one molecule this would explain the preferred 3, 6 and even 9fold oxidation pattern. Some Met/Trp in proteins were hardly affected suggesting they were difficult to reach from the solvent or red/ ox shielded by the protein environment.5 To investigate this further, several peptides were investigated for UV mediated oxidation. Tryptophan gave rise to the most severe photooxidations, but when situated in either of the terminal positions, only a minor effect was detected. In Figure 4 several examples are presented, including the extent to which the first, second and third oxidation states are detected by MS after the analyte has passed the DAD. Experiments with free amino acids were conducted, and only oxidation of Met, Trp, Cys (cysteine) and Tyr (tyrosine) were seen (Figure 4). Sulfur and aromatic systems have low energy barriers to form radicals,6 which explain their susceptibility for radical attack. Another explanation is that they carry the needed chromophore to capture light directly, reach triplet state and through electron transfer produce a radical.7,8 Effects on Other Compound Classes. So far we have discussed analytes composed of amino acids, but we have also seen that other classes of compounds degrade. In Figure 5 two examples are shown, an antisense oligonucleotide and Ketoprofen, a nonsteroidal anti-inflammatory drug. In this example the oligonucleotide is stabilized by substituting all phosphate to phosphorothioate groups. When the oligonucleotide is exposed to light in the UV detector, it shows two types of degradation: 1) The thioate groups are targets for free

Figure 2. Deconvoluted mass spectra of infused α-lactalbumin under different conditions. Labeling of the different obtained spectra show the used DAD detector, UV/noUV, organic phase MeCN or MeOH, and applied BHT (2,6-di-tert-butyl-4-methylphenol) radical quencher (250 ppm).

1290) virtually no oxidation is detected. Also, by using the Agilent 1200 DAD (UV light “on”) no additional oxidations are visible. When using acetonitrile in the mobile phase the oxidative degradation is significantly higher compared to using methanol. Upon adding a radical quencher (BHT, 250 ppm) in the mobile phase, the oxidations are suppressed by >90%. It is also noteworthy, that a strong decrease in the ion signal intensity of the main signal is observed for all investigated compounds. In Figure 3 this is shown for α-lactalbumin. When measuring the protein in MeCN/H2O/0.1% FA the base peak signal increased with >500% and by >250% when UV light is B

DOI: 10.1021/acs.analchem.8b05845 Anal. Chem. XXXX, XXX, XXX−XXX

Letter

Analytical Chemistry

Figure 4. Comparison of various proteins, peptides, and amino acids. The relative intensity of the first, second, and third oxidized compound compared to the nonoxidized molecule. For example: α-lactalbumin oxidation figures relate to Figure 2, panel “1290, UV, MeCN” and Bombesin relate to Figure 1.

radical attack due to their low ionization energy barrier, resulting in sulfur being exchanged to oxygen (loss of 16 mass units per exchanged sulfur). 2.) We also notice additions of oxygens, most likely as a result from radical attack on pyrimidine or purine bases.9 The aromatic structure of Ketoprofen shows three oxidative degradation products. Residence Time, Radical Quenching, and Solvent Effects. The extent of photooxidation in the UV detector is proportional to the residence time in the flow cell. In the case of α-lactalbumin around 20% was oxidized at 500 μL/min, while 90% was oxidized at 50 μL/min (data not shown). This is an irreversible reaction that becomes more obvious at lower flow rates. The flow cells used in this study were the standard 0.5 μL (Waters) and 1 μL (Agilent) cells. Similar, “high sensitivity” cells with even longer light paths are expected to have a proportional higher photodegradation effect. The extensive degradation in an Agilent 1290 DAD is striking when compared to an Agilent 1200 DAD, where light is not reflected on the walls of the detector cell, and no photodegradation was therefore seen. This type of flow cell was designed fundamentally differently, while the D2 lamps remained the same. It has been shown previously that electrospray ionization itself can be a source of artifact oxidation of peptides through electrochemical or corona induced radical formation.10,11 To separate this effect from what we observed, a fused silica ESI interface was tested. Here potential electrochemical oxidation of the sample at the electrode surface is omitted. A minor spectral difference was noticed, while a 100% change in oxidation pattern was seen when the UV lamp was switched off (data not shown). It can be stated that the reaction originates from radicals produced from light since the addition of the radical inhibitor BHT reduced the oxidation effect by >90% (see Figure 2). The

Figure 5. Degradation of a 18mer oligonucleotide (50 μg/mL in 16 mM TEA, 100 mM hexafluoro-2-propanol, infusion 50 μL/min, neg mode), with phosphorothioate backbone (Agilent 1290DAD) and Ketoprofen with proposed structure of degradation products (500 μL/min, MeCN/H2O/0.1% FA, Waters Acquity DAD).

C

DOI: 10.1021/acs.analchem.8b05845 Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry

Letter

■

CONCLUSION The most pronounced risk we urge the readers to consider is to avoid misinterpretations when oxidation impurities are quantified with LC−DAD−MS. If spectra are acquired to assess purity or modifications of proteins and oligonucleotides on a DAD−MS system, photooxidation products may severely affect the conclusions. The solution is simple, turn off the UV lamp or connect the two detectors in parallel instead of in series. However, with knowledge of this behavior, a free online UV reactor is readily available in the system to be used for structure elucidation by photochemistry or for radical reactivity elucidations.

effect of an organic modifier was investigated by replacing the 50% acetonitrile by 50% MeOH. MeOH reduces the photooxidation by roughly 50% (see Figure 2). This effect is to some extent explained by methanol’s higher UV cut off wavelength,12 by its reactivity with radicals,13 and by its higher rate constant for electron transfer quenching.14 O2 Level in Mobile Phase and H2O2 Formation. We have not fully investigated how the radicals are formed in the UV cell, but several observations (response to level of analyte, quencher, O2 level) point toward their being formed in the solvent and diffusing to the analyte where the oxidation occurs. The literature9,15 often explains the radical oxidation chain by a primary attachment of a radical to a resonance stabilized tertiary carbon, followed by an addition of O2, giving a R−O− OH. The final loss of a hydroxy radical leaves one oxygen atom on the molecule. As this chain reaction would need O2, an experiment was done with somatostatin 28 (3.1 kDa) with three levels of O2 in the solvent: first, with air saturation; second, at only the partial pressure of O2 left by the Acquity/ 1290 degassers; and third, after He-purging (very low pO2 level). No major difference in the photooxidation was noted, suggesting that OH • radicals being formed by water dissociation might be the major source of radicals, and the chain reaction that needs O2 may be of limited importance in this case. Many water purification systems have implemented UV radiation, though we could not measure significant H2O2 concentrations in the used Milli-Q water. Therefore, we measured the H2O2 concentration in water before it entered and after it left the DAD flow cell. To assess the radical concentration in the UV detector cell, we measured the H2O2 formation in the Agilent 1290, 10 mm flow cell when pumping purified water for injection (not UV radiated) with 50 and 100 μL/min. The idea is that ample OH• radicals would enable 2OH• → H2O2. H2O2 has also been shown to dissociate back to two OH• with further UV radiation. At 50 μL/min ca. 0.3 μM and at 100 μL/min ca. 0.18 μM H2O2 were formed (data not shown). This corresponds to ca. 3 pmol of OH• (same order of magnitude as analyte amount) that comes in contact with an analyte during a 6 s peak at 50 μL/min. To understand the reversible effect of H2O2 itself, a peptide oxidation experiment (Bombesin) was conducted with addition of 0.5 and 2 mM H2O2 in the mobile phase, with and without the UV lamp on. No oxidation was seen from H2O2 in the dark, while the oxidation rate was roughly ×1.5 by adding 0.5 mM H2O2 and exposure to light and ×3.2 with 2 mM. This shows the contribution from extra OH• radicals released from dissociated H2O2 under UV light. In the Waters Acquity DAD, a “UV210 nm blocking filter” is implemented. With this filter active, the photooxidation is reduced by ca. 50% (data not shown). Opportunities. Stability studies in pharmaceutical drug development including light stress studies today still are a lengthy process. Information gained in seconds by observing the radical reactions in LC−UV−MS may rapidly evaluate the stability of drugs to radical attacks. Not only reactivity is gained from that experiment, but also information on reaction products. Correctly used, this could be a great tool for fast recognition of sensitive substances and pointing out reaction pathways. If only traces of oxidation products are seen, the flow rate may be lowered to increase the reaction time in the UV cell. If a higher concentration of radicals is needed, an addition of 1 mM H2O2 to the mobile phase will boost this by photodissociation but may simultaneously initiate two-electron oxidation.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected] ORCID

Fritz Schweikart: 0000-0002-2201-7896 Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS Dr. Sara Richardson is acknowledged for thorough proofreading and revision of the manuscript. REFERENCES

(1) Perdivara, I.; Deterding, L. J.; Przybylski, M.; Tomer, K. B. J. Am. Soc. Mass Spectrom. 2010, 21 (7), 1114−1117. (2) Attri, P.; Kim, Y. H.; Park, D. H.; Park, J. H.; Hong, Y. J.; Uhm, H. S.; Kim, K. N.; Fridman, A.; Choi, E. H. Sci. Rep. 2015, 5, 9332. (3) Bent, D. V.; Hayon, E. J. Am. Chem. Soc. 1975, 97 (10), 2599− 2606. (4) Teresa, M.; Petersen, S.; Prakash, G. UV Light Effects on Proteins: From Photochemistry to Nanomedicine. In Molecular Photochemistry - Various Aspects; Neves Petersen, M. T.; Teresa, M.; Petersen, S.; Prakash, G., Eds.; InTech: 2012; pp 125−158. (5) Sharp, J. S.; Becker, J. M.; Hettich, R. L. Anal. Chem. 2004, 76 (3), 672−683. (6) Pattison, D. I.; Rahmanto, A. S.; Davies, M. J. Photochemical & Photobiological Sciences 2012, 11 (1), 38−53. (7) Davies, M. J.; Truscott, R. J. W. J. Photochem. Photobiol., B 2001, 63 (1), 114−125. (8) Neves-Petersen, M. T.; Klitgaard, S.; Pascher, T.; Skovsen, E.; Polivka, T.; Yartsev, A.; Sundstrom, V.; Petersen, S. B. Biophys. J. 2009, 97 (1), 211−26. (9) Asmus, K.-D.; Bonifacic, M. Handbook of oxidants and antioxidants in Exercise; Elsevier: 2000; pp 3−54. (10) Morand, K.; et al. Rapid Commun. Mass Spectrom. 1993, 7 (8), 738−743. (11) Chen, M.; Cook, K. D. Anal. Chem. 2007, 79 (5), 2031−2036. (12) LoBrutto, R.; Kazakevich, Y. Reversed-Phase HPLC. HPLC for Pharmaceutical Scientists: Kazakevich/HPLC for Pharmaceutical Scientists; Wiley Blackwell (John Wiley & Sons): 2007; pp 139−239. (13) Monod, A.; Chebbi, A.; Durand-Jolibois, R.; Carlier, P. Atmos. Environ. 2000, 34 (29), 5283−5294. (14) Zanini, G. P.; Montejano, H. A.; Carlos, J. J.; Previtali, C. M. J. Photochem. Photobiol., A 1997, 109 (1), 9−14. (15) Denisov, E. T.; Afanas’ev, I. B. Oxidation and Antioxidants in Organic Chemistry and Biology; CRC Press, 2005; pp 23−76.

D

DOI: 10.1021/acs.analchem.8b05845 Anal. Chem. XXXX, XXX, XXX−XXX