2.1. Relative Response of Methylglyoxal Derivatives in LC-MS Analysis
Although theoretically indicative, the ESI-MS responsiveness of carbonyl derivatives can be quite different from that of the original amine reagent; therefore, the responsiveness of the reagent is only a rough indicator for predicting the relative response of the corresponding derivative. For example, Figure 3
a compares the change in response pattern of the protonated molecular ion of the phenylenediamines (left) and either their methylglyoxal (middle) or glyoxal (GO, right) quinoxaline derivatives depending on the phenylenediamine substituent R3
(see Figure 2
a). The corresponding m/z
values of the MGO derivatization products that were used for quantification are given in Supplementary Table S1
Among the phenylenediamines, signal intensity of the 4-methoxyphenylenediamine (4-PDA) product was consistently higher compared with those of other phenylenediamines, which was confirmed for different molar ratios of reagent and aldehyde (data not shown). As expected, the derivatives of reagents with strong mesomeric effect exhibited the highest signals with +M (e.g., methoxy) being better than -M (e.g., NO, nitro), and para
-substituted (4-position) better than meta
-substituted ones (3-position). According to these results, 4-PDA would be the best phenylenediamine reagent for detection of low amounts of methylglyoxal. Apart from ionization efficiency, the relative response of the reaction products may be caused by different reaction yields, e.g., due to insufficient incubation time. Thus, after 3 h incubation time, only 50% reaction yield was reported when using 2,4-dinitrophenylhydrazine (DNPH) for MGO derivatization [30
], while reactions with 4-PDA and 4,5-methylenedioxyphenylenediamine (4,5-PDA) have been reported to be nearly complete after just 1 h [26
]. These findings were in agreement with our results. Consequently, the characteristics supporting ionization efficiency, e.g., electron-donating substituents enhancing the basicity of compounds as the most important criterion for ESI-MS responsiveness of nitrogen bases [28
], often foster the reactivity of the compound as well, resulting in faster reaction times. (Note that DNPH, which is the most commonly employed phenylhydrazine, was not included because with mass spectrometry the derivatives are usually detected in ESI negative ion mode, meaning the response cannot be easily compared to the other reagents. Moreover, the slow reaction times contradict the purpose of our study, which was to find a reagent candidate with a fast reaction time). However, we did not notice a significant time-dependent response effect between 2 and 4 h incubation time for all of the commercially available reagents and no response enhancement for the derivatives after that time, which also generally resulted in a satisfactory standard deviation of the replicates. As a general rule, signal response of the derivatives from reagents with electron-donating substituents slowly decreased up to 50% during the course of a day, so we prepared all samples for response comparison in situ.
4-PDA has only recently been introduced as a derivatization agent for methylglyoxal [26
]. Among the tested, commercially available phenylenediamines, this reagent seems to produce the most sensitive signal response. Thus, we compared this particular reagent to phenylhydrazines, another group of commonly employed reagents for analysis of carbonyls that also promises to provide excellent performance for derivatization. The results are illustrated in Figure 3
It is clear that 3-methoxyphenylhydrazine (3-MPH)—which to our knowledge has not been suggested yet for derivatization of methylglyoxal—outperformed all other commercially available reagents we tested in our experiments. Only our synthesized phenylhydrazine carrying a permanent charge in para
position (4-AEH) provided a higher intensity, although it mostly reacted with just the aldehyde function and we observed mainly monomers. For comparison, Table 1
lists the corresponding estimated limits of detection (LOD) for selected protocols, i.e., using the reagents 4-PDA, 3-MPH, 3-ammonioethoxyphenylhydrazine (3-AEH), 3-ammonioethoxyaniline (3-AEA), coumarin carbohydrazide (DCCH), Amplifex™ Keto Reagent (AKR, ABSciex, Framingham, MA, USA), and 4-MPH.
The pattern of LODs did not exactly follow the relative response as obtained from signal intensities at 50 µM MGO (0.5 nmol injected). Thus, for 3-AEA, a surprisingly high value was found, while 4-PDA had a surprisingly low LOD in contrast. Our comparison confirms the notion that assessment of relative sensitivities based on peak areas far above the detection limit does not necessarily help to find the most sensitive protocol. According to these results, 4-PDA might be the preferred reagent by far for analysis of α-oxo-aldehydes, which quinoxaline derivative(s) of methylglyoxal (and glyoxal, not shown) were detected particularly sensitive. However, at higher concentrations of monocarbonyl compounds, multiple derivatives could be produced from phenylenediamines, which can be avoided when using the second-best candidate reagent, i.e., 3-MPH. Furthermore, the production of a diagnostic daughter ion from 3-MPH derivatives, m/z 122, provides a great advantage for identification of unknown carbonyl compounds by MS/MS analyses.
A drawback of using phenylhydrazines for quantification of nonsymmetric dicarbonyl compounds, such as methylglyoxal, compared to phenylenediamines is the formation of two mono- and two main bis-derivatives (for the latter, four isomers would be possible). For 3-MPH, the less abundant bis-derivative was present as a shoulder peak under our conditions, so we quantified the signal as the more robust sum of both peak areas. The same applied to the coumarin carbohydrazide DCCH, which was also the only reagent considerably retained beyond the dead volume of the LC column. Here, the two MGO monoderivatives eluting within the broad tailing of the reagent peak were nearly twice as abundant as the late eluting bis-derivatives, indicating insufficient reactivity against ketones. (Note that another reason for the observed low abundance of the bis-derivative might be the expected very poor solubility, with a logS of 6.78 in the ChemAxon’s solubility predictor [34
] corresponding to maximal 3 µM dissolved substance vs. 50 µM starting material used in this experiment.). Considering the use of the reagent for a generally applicable multimethod to analyze aldehydes and ketones beyond the analysis of α-oxo-aldehydes, 3-MPH would still be favorable due to its nonselective high reactivity. In return, the presence of two isomers also enables an easier identification of corresponding derivatives in an unknown mixture or for derivatization of unknown carbonyl compounds.
The most curious result, however, is the bad performance of 4-MPH as derivatizing reagent. With this reagent, the base peak chromatogram of the reaction mixture was dominated by multiple degradation products rather than by the expected MGO derivatives. Thus, the only intermediate response of the main derivative (m/z
296) and the corresponding high variance of this species may be the consequence of ion suppression by a very abundant coeluting reagent dimer (m/z
243, see Section 2.2.1
or Figure 4
). Stability tests suggested that extensive oxidation processes within the reaction mixture hampered reproducibility of the results and led not only to the degradation of derivatives and reagent but also to precipitations in reaction mixtures with high concentration of GO.
2.3. Application of 3-MPH as Derivatizing Agent to Explore the Contamination of Laboratory Water with Methylglyoxal
Another well known, serious problem in the quantitative analysis of methylglyoxal is the fact that it is hard to produce a clean solvent blank as required, for instance, for LOD determination [41
]. As illustrated below using 3-MPH as reagent, even high-quality equipment and several tested cleaning procedures were not able to completely remove methylglyoxal from the blank (Figure 6
MGO response in methanol equaled the one from distilled water and water purification systems 3, 4, and 5, suggesting that these purification protocols indeed performed the best. However, successful purification by SPE procedures appeared to be much more cumbersome, expensive, and laborious and required a very careful optimization in advance. Thus, we believe the favorable value achieved with the PFBHA (III) protocol could at least partially be a consequence of signal suppression by bulk unreacted PFBHA, which was not satisfactorily removed by SPE. Excess PFBHA is usually removed by l/l extraction after selective protonation of the reagent [16
], after which a higher response of blank contamination can be obtained (protocol IV). Moreover, complete elimination of the reagent employed for MGO removal from the solvent used for analysis is indispensable to avoid subsequent competitive reaction with the reagent of choice for determination of MGO in the sample. Using the same reagent for purification of water and analytical derivatization itself [26
], on the other hand, may still lead to an overestimation of MGO in the sample in case the residue is not completely removed.
Curiously, we still obtained a response for MGO when injecting 3-MPH in methanol only. This finding prompted us to further test the potential origins of the contamination. Considering that excess reagent is used for derivatization, MGO response in the water blank was expected to stay constant with decreasing reagent concentration until a certain threshold. Instead, we observed an immediate decrease in response with decreasing concentrations of the reagent at all levels (Figure 6
b), indicating reagent contamination with the derivative. Reagent contamination was further confirmed by dissolving one reagent in a solution of another after 2 h incubation, i.e., 3-MPH in 4-PDA and 4-PDA in 3-MPH solutions without adding MGO, where derivatives of MGO with both reagents were found (Figure 6
c). We conclude that the reagents themselves are a source of blank contamination and require thorough purification and subsequent storage under appropriate conditions, particularly if quantification near or below the concentration of the corresponding contamination is anticipated (5 nM for 3-MPH in our case). In addition, appropriate blank replication is required for quantification above the blank level for which, theoretically, the reagents can be used without prior purification employing blank subtraction. Thus, 3-PDA, 4-PDA, and 3-MPH were all suitable to determine concentrations down to 10 nM MGO (0.1 pmol injected) before blank contamination prevented quantification of lower concentrations. For comparison, typical concentrations of MGO in biological samples are, for example, ~250 nM in whole blood and 170 nM in human plasma [17
], 1 nmol/g in U87 cancer cells [22
], ~15 µM in wine [42
], and >200 nM in urine [26
2.4. Analysis of Carbonyl Reaction Products of Linoleic and Linolenic Acids Oxidation by Cu(II) and Hydrogen Peroxide After Derivatization with 3-MPH
As a proof of concept and to show the applicability of our derivatization agent, we analyzed reaction mixtures after oxidation of linoleic and linolenic acids—two very important native fatty acids that are present, for instance, in human epidermis. Oxidation of these acids is known to produce mainly hydroxylated acids, such as 8,13-dihydroxy-9,11-octadecadienoic and 9,14-dihydroxy-10,12-octadecadienoic acids [43
], but oxo-octadecadienoic acids were also observed [44
]. Recently, the formation of several carbonyl compounds, such as acrolein and crotonaldehyde as well as glyoxal and methylglyoxal after oxidation of linoleic and linolenic acids, has been reported [45
] but no quantitative information has been added. Because lipoxidation mechanisms are a highly interesting topic in biomedical research and because there is already information available to rate the validity of our own results, we selected this model system to show the performance of our method. Firstly, we detected signals that could be produced by approximately 60 possible RCC candidates (although this tentative identification would have to be subjected to future confirmation). Moreover, we were able to quantify methylglyoxal in the reaction mixture (4.6 µM in the linoleic acid and 1.1 µM in the linolenic acid mixtures) and used the average comparison for other test candidates. As an example, we present a small table with ratios and average comparisons of selected potential target compounds (Table 2
Although an exhaustive evaluation of such an experiment would require prior establishment of a proper multimethod, in particular the assessment of appropriate retention times and MS-MS data for differentiation of derivatives with the same mass but different structure (e.g., methylglyoxal and malondialdehyde), this experiment already shows the capability of the reagent for application in such multitargeted methods to assess the presence and concentration of many carbonyl compounds in a sample in one analytical run. However, compared to our standard samples, we observed not only a significantly enhanced background in the hydrogen peroxide and metal ion (Cu II)-containing solutions of this experiment but also an enhanced formation of the reagent dimer (m/z 243) as indicator of oxidative degradation of the reagent. (Note that similar effects were observed for 4-PDA as derivatization agent in a glycation mixture from a glucose solution with Fe II/III; data is not shown). Copper and iron salts are additives with the particular purpose of enhancing the reactivity of lipids and carbohydrates toward oxidative degradation; therefore, not surprisingly, we found that the presence of these Lewis acids may become very problematic in investigations such as ours. Their interaction with the lone electron pairs of the heteroatoms could enhance the reactivity of the reagents and the carbonyls. Thus, in such experiments, the immediate analysis of the derivatized sample would be of particular importance. Our advanced protocol, which changed the solvent directly after derivatization, is expected to decrease the concentration of these metal ions in the reaction mixture and would therefore provide a significant advantage to improve experimental variance.