Degradation and de novo formation of nine major glucose degradation products during storage of peritoneal dialysis fluids

Reactive glucose degradation products (GDPs) are formed during heat sterilization of glucose-containing peritoneal dialysis fluids (PDFs) and may induce adverse clinical effects. Long periods of storage and/or transport of PDFs before use may lead to de novo formation or degradation of GDPs. Therefore, the present study quantified the GDP profiles of single- and double-chamber PDFs during storage. Glucosone, 3-deoxyglucosone (3-DG), 3-deoxygalactosone (3-DGal), 3,4-dideoxyglucosone-3-ene (3,4-DGE), glyoxal, methylglyoxal (MGO), acetaldehyde, formaldehyde, and 5-hydroxymethylfurfural (5-HMF) were quantified by two validated UHPLC-DAD methods after derivatization with o-phenylenediamine (dicarbonyls) or 2,4-dinitrophenylhydrazine (monocarbonyls). The PDFs were stored at 50 °C for 0, 1, 2, 4, 13, and 26 weeks. The total GDP concentration of single-chamber PDFs did not change considerably during storage (496.6 ± 16.0 µM, 0 weeks; 519.1 ± 13.1 µM, 26 weeks), but individual GDPs were affected differently. 3-DG (− 82.6 µM) and 3-DGal (− 71.3 µM) were degraded, whereas 5-HMF (+ 161.7 µM), glyoxal (+ 32.2 µM), and formaldehyde (+ 12.4 µM) accumulated between 0 and 26 weeks. Acetaldehyde, glucosone, MGO, and 3,4-DGE showed time-dependent formation and degradation. The GDP concentrations in double-chamber fluids were generally lower and differently affected by storage. In conclusion, the changes of GDP concentrations during storage should be considered for the evaluation of clinical effects of PDFs.

. UHPLC-DAD analysis of α-dicarbonyls. (a) Derivatizing reaction with OPD to yield corresponding quinoxaline derivatives and (b) chromatogram of a typical PDF after derivatization with OPD, recorded at 316 nm. The indices "qx" refer to the quinoxaline derivatives of the α-dicarbonyl GDPs. The index "bfm" refers to (5-(1H-benzo[d]imidazol-2-yl)furan-2-yl) methanol, which is the benzimidazole derivative of 5-HMF. www.nature.com/scientificreports/ For the analysis of the hydrazone derivatives, a UHPLC-DAD method was established and validated (Fig. 2b). To determine the appropriate time for complete and stable derivatization, the derivatization of the four monocarbonyls 5-HMF, acetaldehyde, furfural, and formaldehyde was monitored between one and twelve and a half hours. The signal for formaldehyde and furfural remained stable over the whole period, whereas the signal of acetaldehyde decreased slightly during prolonged derivatization, but deviated less than 5% within a period of up to eight hours (see Supplementary information, Fig. S-1a). A slight increase of 5-HMF was observed, which was also less than 5% within a period of up to eight hours (Supplementary information, Fig. S-1b).
Thus, a derivatization time between one and eight hours was set. Chromatographic signals for the E-and Z-isomers of the 5-HMF and furfural derivatives could be detected, but only at high concentration levels. At low concentrations, only the signal of the predominant isomer was detectable or quantifiable, respectively. Because the ratio of both isomers was constant, the more intense signal was used for quantification. The calibration curves met the prerequisite of coefficients of determination R 2 > 0.999 for all calibration models. The relative errors, however, ranged from 74 to 133% and the homogeneity of variances across the concentration range was violated (F-test, P < 0.05). Therefore, we applied weighted linear regression models (Table 1) resulting in relative errors below 5%.
Recovery and precision were determined at three different concentration levels. In all cases, the recovery did not deviate more than 8.4% from the actual concentration and the variation coefficient was less than 6.3% ( Table 2). The results verify that the described procedure is a reliable and precise method to quantify formaldehyde, acetaldehyde, and 5-HMF in glucose-based PDFs.  www.nature.com/scientificreports/ During analysis, a hypsochromic shift at the tail (full width at half maximum) and the apex of the furfural signal was observed in heated PDFs, but neither in standard solutions nor in spiked unheated PDFs (see Supplementary information, Fig. S-2). Therefore, an unknown compound was assumed to coelute with furfural in heat-sterilized glucose-containing PDFs, which may lead to an overestimation of the furfural content. Furfural was the least quantifiable GDP in the PDFs (double-chamber PDFs: ≤ 0.3 µM, single-chamber PDFs: ≤ 1.4 µM; apparent uncorrected concentrations; Table 3). Although furfural shows satisfying validation parameters, we therefore excluded furfural from further analysis. No chromatographic interferences were observed for any of the other GDPs.
Quantitative profiling of GDPs during storage of single-chamber PDFs. Changes of the GDP profile in conventional single-chamber PDFs were investigated during storage for up to 26 weeks at 50 °C and 45% relative humidity. The initial total GDP load of 496.6 µM increased slightly to 519.1 µM after 26 weeks of storage (Fig. 3a). The initial contents of 171.1 µM 3-DG and 127.1 µM 3-DGal decreased over time to 88.5 µM 3-DG and 55.9 µM 3-DGal after 26 weeks (Fig. 3b). The concentration of glucosone was stable within four weeks (25.0 µM) and decreased during ongoing storage to 12.0 µM (Fig. 4b).
The MGO content also remained stable within two weeks (8.0 µM) and dropped to 3.1 µM during further incubation (Fig. 3c). The 3,4-DGE concentration increased during the first week from 10.6 to 30.6 µM and decreased afterwards to 16.1 µM within 26 weeks of storage (Fig. 3c). Acetaldehyde was initially present at a concentration of 129.5 µM. Its concentration peaked after 13 weeks (147.4 µM) and decreased later to 111.6 µM (Fig. 3d). The concentration levels of glyoxal, formaldehyde, and 5-HMF rose during the incubation period. At the beginning, the PDFs contained 9.2 µM glyoxal and 5.0 µM formaldehyde. After 26 weeks, 41.4 µM glyoxal and  Table 3. Concentration of four monocarbonyl and six α-dicarbonyl GDPs as well as total GDPs in PDFs in µM after 0 and 26 weeks of storage at 50 °C and 45% relative humidity. a n.q., not quantifiable; b n.a., not assessed because of a coeluting compound; c n.d., not detectable, d excluding furfural. Mean values ± standard deviation of three different PDF bags are displayed. www.nature.com/scientificreports/ 17.7 µM formaldehyde were present (Fig. 3c,e). The largest increase in concentration was observed for 5-HMF ( Fig. 3f). At the beginning of the experiment, 11.1 µM 5-HMF were measured, but the content increased up to 172.8 µM after 26 weeks (Fig. 3d).

Quantitative profiling of GDPs during storage of double-chamber PDFs. Double-chamber PDFs
were stored up to 26 weeks at 50 °C and 45% relative humidity to monitor the GDP contents. Prior to storage, the products contained 77.6 µM GDPs in total. The most abundant GDP was 3-DG (28.6 µM), followed by 5-HMF (22.5 µM), and 3-DGal (19.7 µM). The compounds 3,4-DGE (2.8 µM), acetaldehyde (2.2 µM), and glucosone (1.6 µM) were present at much lower concentrations. Formaldehyde was below the limit of quantification (LOQ; 0.6 µM) and glyoxal below the limit of detection (LOD; 0.2 µM) at week 0. MGO was not detectable in any double-chamber PDF (LOD: 0.2 µM, Table 3). Up to four weeks of storage, the total GDP load remained constant, followed by an increase to 99.0 µM after 13 weeks and 142.4 µM after 26 weeks (Fig. 4a). As storage progressed, the concentrations of 3-DG, 3-DGal, 3,4-DGE, and acetaldehyde decreased so that neither 3-DGal nor 3,4-DGE was quantifiable at the end of the storage period ( Fig. 4b-d). After 26 weeks, only 9.4 µM 3-DG and 1.5 µM acetaldehyde were measured (Fig. 4b,d). During the first four weeks, the glucosone concentration remained almost stable and increased slightly with ongoing storage so that 6.1 µM glucosone was present after 26 weeks (Fig. 4b). Glyoxal, which was not detectable at the www.nature.com/scientificreports/ beginning, was formed during prolonged storage resulting in 0.5 µM glyoxal after one week and 1.0 µM after 26 weeks (Fig. 4c). Formaldehyde was only quantifiable after 26 weeks (2.5 µM) indicating formation during storage (Fig. 4d, Table 3). The content of 5-HMF increased remarkably during storage (Fig. 4e,f). While 22.5 µM 5-HMF was present at the beginning, its contents increased up to 121.6 µM. At the end of the study, 5-HMF was the major GDP in the double-chamber PDFs (Fig. 4f).

Discussion
The comprehensive profiling of nine major GDPs in single-and double-chamber PDFs during 26 weeks of storage confirmed the assumption that the GDP concentration and composition in PDFs can differ between fresh products and fluids that patients actually use for dialysis. PDFs are subject to various storage conditions depending, e.g., on the type of transport (overseas or land transport), the transport routes (long/short), or the ambient temperatures (summer/winter). In particular, temperatures can fluctuate remarkably and may reach almost 60 °C 17 . Since elevated temperatures are expected to have the most severe effects on the GDP composition, the present study assumed worst-case conditions. However, other typical storage conditions, e.g. in dialysis clinics or at home, should be investigated in a next step. Besides, www.nature.com/scientificreports/ samples drawn from PDFs directly before administration could provide additional information on the actual exposure of patients to GDPs. The current analysis investigated products ready for dispatch with the consequence that the PDFs had been pre-stored at room temperature for two and five months, respectively. Even though this random selection represented realistic conditions, the pre-storage may be a limitation of the study, because changes of the GDP profile between production and dispatch cannot be excluded.
In the present study, the GDP contents in single-chamber PDFs and double-chamber solutions with lower GDP load correspond to the concentration levels previously reported in the literature 2 . In both types of PDF, the concentrations of 3-DG and 3-DGal decreased during prolonged storage. This result is in line with previous studies that also observed a decrease of 3-DG during storage over 21 days at 40 and 60°C 14 , or six months at 25, 30, and 40 °C, respectively 13 . 3-DG and 3-DGal undergo reactions leading, for example, to the formation of 3,4-DGE 9,18,19 and further to 5-HMF, which seems to be a stable end-product in PDFs 2 (Fig. 5).
At week 26 of the present storage experiments, we measured additional 161.7 µM 5-HMF and 5.5 µM 3,4-DGE, which corresponds to the degradation of 3-DG (− 82.6 µM) and 3-DGal (− 71.2 µM). Interestingly, the concentration of 3,4-DGE peaked after one week of incubation. Erixon et al. also reported increased 3,4-DGE concentrations, especially at incubation temperatures of 40 °C and 60 °C, during one week of incubation and observed that the 3,4-DGE content decreased again until the end of the 21-day storage period at 60°C 14 . The previous study and our data indicate that the formation of 3,4-DGE from 3-DG and 3-DGal and its degradation to 5-HMF compete with each other at elevated temperatures. Whereas the relatively fast formation dominates at the beginning of storage, it is overcompensated by 3,4-DGE degradation at the end of the storage period (Fig. 5).
The concentrations of glucosone and MGO decreased in single-chamber PDFs over time, but the glucosone content increased slightly in double-chamber solutions. MGO was below LOD in the double-chamber-bag fluids at any time point. Glucosone is formed by the oxidation of glucose 20,21 . Recently, it was shown that traces of redox-active metal ions, such as iron(II), can promote the formation of the oxidized GDP glucosone and, to a lesser extent, also glyoxal and MGO 22 . The concentrations of 3-DG, 3-DGal, and 3,4-DGE, which are formed by www.nature.com/scientificreports/ non-oxidative mechanisms 2 were not affected by metal ions 22 . Thus, traces of metal ions from raw materials or production equipment may be responsible for the formation of glucosone during the storage of double-chamber PDFs.
The acetaldehyde content peaked in single-chamber PDFs at week 13, while it decreased in double-chamber bags from the beginning of storage. Acetaldehyde is formed from lactate but is considered as GDP, because its formation is glucose-dependent 1,13 . Single-chamber bags contain lactate and glucose, which mediate the formation of acetaldehyde 13 . In double-chamber bags, glucose and lactate are stored separately so that the reaction is inhibited. Thus, double-chamber bags contain considerably less acetaldehyde compared to single-chamber solutions. Previous studies reported concentrations below LOD or LOQ in double-chamber PDFs, which ranged from 2 to 18 µM, respectively 1,23,24 . The present method is more sensitive, so that acetaldehyde contents between 1.5 and 2.2 µM could be quantified in the PDFs. These results indicate that very minor concentrations of acetaldehyde may be formed from glucose either independently from lactate, during/after mixing the solutions of both compartments, or directly from lactate.
The concentrations of glyoxal and formaldehyde increased over the storage period in both types of PDFs, which may result from the degradation of long-chain GDPs such as glucosone. Glyoxal can be formed from glucosone via retro-aldol cleavage 21 . Alternatively, Yaylayan and Keyhani proposed a mechanism based on the loss of two water molecules from glucose and subsequent retro-aldol reaction 25 . In addition, glyoxal can be formed via oxidation of glycolaldehyde after retro-aldol cleavage of glucose 20 . The reaction mechanism leading to the formation of formaldehyde has yet to be elucidated.
The contents of 3-DG and 3-DGal decreased in both PDF types during 26 weeks of storage. However, 5-HMF, glyoxal, and formaldehyde were formed during the incubation time in both PDF systems. The acetaldehyde content decreased in double-chamber PDFs, while the concentration peaked in single-chamber fluids after 13 weeks. The concentration of 3,4-DGE reached its maximum in single-chamber PDFs after one week of incubation and decreased afterwards. After 26 weeks of storage, however, the 3,4-DGE concentration was still slightly higher than in week 0. In double-chamber bags, a continuous decline of 3,4-DGE was observed. The GDP profiles of single-and double-chamber bags underwent different changes. The PD solutions in both systems vary in pH as well as in glucose, buffer, and electrolyte concentrations. These factors influence the formation and degradation of the GDPs during heat sterilization and subsequent storage. At the beginning, and even at the end of the study, double-chamber PDFs contained less GDPs than single-chamber solutions (Figs. 3f, 4f).
Since it is well established that the individual GDPs feature distinct bioactivities, it is important to consider not only the total GDP content, but also the concentrations of the single GDPs in PDFs. For instance, 3,4-DGE has been reported to have the highest cytotoxicity among the GDPs and to reduce enzyme activity in vitro 9,10 . In single-chamber bags, which contain about threefold higher concentrations of 3,4-DGE compared to doublechamber bags, storage and, in particular, short-term storage may increase its concentration, whereas the storage of double-chamber bag fluids has a beneficial influence on 3,4-DGE concentrations. The biological relevance of 5-HMF is still not clear 26 and depends highly on the applied concentration. Adverse effects have been described after high doses of 5-HMF in the millimolar range, which reflect or even exceed dietary exposure [27][28][29] . These effects are mainly linked to the 5-HMF metabolite 5-sulfoxymethylfurfural 29,30 . In contrast, Zhao et al. observed antioxidative properties and antiproliferative activity of 5-HMF against cancer cell lines in vitro 31 . In 2011, the German Federal Institute for Risk Assessment classified 5-HMF as not harmful to health 32 . In PDFs, 5-HMF is present in much lower concentrations and administered peritoneally. In the context of peritoneal dialysis, 5-HMF had no effect on the viability of human peritoneal mesothelial cells 33 or L-929 fibroblasts 11,33 in vitro. Morgan et al. reported that 5-HMF did not attenuate the re-mesothelialization of human peritoneal mesothelial cells at concentrations present in PDFs 34 . Although 5-HMF is the most abundant GDP at the endpoint of the present study, it causes less adverse effects in peritoneal dialysis than other compounds such as 3,4-DGE, which are present in lower concentrations.
In conclusion, the described storage effects should be taken into consideration when the clinical relevance of GDPs in PDFs is evaluated.

Chemicals and reagents. All chemicals and reagents were of at least analytical grade and purchased from
Sigma-Aldrich (Steinheim, Germany) unless indicated otherwise. Liquid chromatography/mass spectrometry-grade solvents (Carl Roth, Karlsruhe, Germany) and formic acid (Acros, Geel, Belgium) were used in all experiments. 3-DG (purity > 95%) was obtained from Chemos (Regenstauf, Germany). Glucosone 35 , 3-DGal 36 , and 3,4-DGE 19 were synthesized as reported previously and stock solutions were prepared: glucosone 10 mM, 3-DGal 2.2 mM, and 3,4-DGE 4.9 mM. The concentrations of the stock solutions were determined as described by Mittelmaier et al. 19,35 . PDFs and storage conditions. The commercial single-chamber PDFs contained 4.25% (w/w) glucose, sodium d-lactate (3.9 g/L), calcium chloride dihydrate (0.26 g/L), sodium chloride (5.8 g/L), and magnesium chloride hexahydrate (0.10 g/L). The ready-to-use double-chamber PDFs contained 4.25% glucose (w/w), sodium d-lactate (3.9 g/L), calcium chloride dihydrate (0.26 g/L), sodium chloride (5.6 g/L), and magnesium chloride hexahydrate (0.10 g/L). Because the glucose solution and the buffer compartment of double-chamber systems must be mixed prior to sampling, the samples were always drawn from originally sealed bags. Directly after production, the PDFs were kept at room temperature before the present storage experiments started five months (single-chamber PDFs) or, respectively, two months (double-chamber PDFs) after production. For the storage experiments, the PDFs were kept at 50 °C and 45% relative humidity. Samples were drawn after 0, 1, 2, 4, 13, and 26 weeks from three new bags each, except for the single-chamber PDF at week 13, which was only UHPLC-DAD instrument. An Ultimate 3000RS system (degasser, binary pump, autosampler, column oven, and DAD; Thermo Fisher Scientific, Dreieich, Germany) was used with a Waters ACQUITY UPLC® Phenyl column (100 × 2.1 mm, 1.7 μm particle size; Waters, Eschborn, Germany) equipped with a corresponding guard column. System control, data acquisition, and processing were performed by Chromeleon 6.8 software.
Quantitative profiling of α-dicarbonyl GDPs. α-Dicarbonyl GDPs were quantified as previously reported with minor modifications 3 . Briefly, α-dicarbonyl compounds were converted into their corresponding quinoxaline derivatives by derivatization with OPD. For this purpose, 80 µL of the sample was mixed with 10 µL of derivatizing reagent (4% OPD in 1 M 2-[4-(2-hydroxyethyl)piperazine-1-yl]ethanesulfonic (HEPES) buffer, pH 7) and 10 µL of internal standard (2,3-dimethylquinoxaline, 50 µg/mL in water). The samples were incubated in the dark between 2 and 16 h. Afterwards, the solutions were analyzed by UHPLC-DAD using ammonium formate buffer ( To test whether the derivatization procedure yielded stable hydrazone derivatives in lactate-buffered PDFs, a commercial single-chamber PDF was treated as described above. UHPLC-DAD analysis was performed after incubation at room temperature for different periods of time up to 12.5 h. Each sample was analyzed in triplicate and the mean values of the peak areas were plotted against the derivatization time. Formaldehyde and acetaldehyde were quantified at 356 nm and furfural and 5-HMF at 390 nm by external calibration. Since the derivatization reagent contained traces of formaldehyde and acetaldehyde, solvent blanks were analyzed in the same way and the peak areas of formaldehyde and acetaldehyde in the samples were corrected for these background signals. A ten-point calibration curve from 0.6 to 320.0 µM was recorded for 5-HMF and a nine-point calibration curve from 1.3 to 320.0 µM for acetaldehyde. Seven-point calibration curves were obtained for formaldehyde and furfural ranging from 0.8 to 64.0 µM (formaldehyde) and 0.2 to 16.0 µM (furfural). Each calibration level was analyzed in duplicate. The linearity of the calibration curves was evaluated by linear regression analysis with a minimally acceptable coefficient of determination (R 2 ) of 0.990, a relative error of < 5%, and homogeneity of variances across the concentration range (F-test, P < 0.05).
The LOD and LOQ were determined using the calibration method according to German technical standard DIN 32645:2008-11 38

Data availability
Data is available upon request.