Solid-phase synthesis of d-fructose-derived Heyns peptides utilizing Nα-Fmoc-Lysin[Nε-(2-deoxy-d-glucos-2-yl),Nε-Boc]-OH as building block

Aldoses and ketoses can glycate proteins yielding isomeric Amadori and Heyns products, respectively. Evidently, d-fructose is more involved in glycoxidation than d-glucose favoring the formation of advanced glycation endproducts (AGEs). While Amadori products and glucation have been studied extensively, the in vivo effects of fructation are largely unknown. The characterization of isomeric Amadori and Heyns peptides requires sufficient quantities of pure peptides. Thus, the glycated building block Nα-Fmoc-Lys[Nε-(2-deoxy-d-glucos-2-yl),Nε-Boc]-OH (Fmoc-Lys(Glc,Boc)-OH), which was synthesized in two steps starting from unprotected d-fructose and Fmoc-l-lysine hydrochloride, was site-specifically incorporated during solid-phase peptide synthesis. The building block allowed the synthesis of a peptide identified in tryptic digests of human serum albumin containing the reported glycation site at Lys233. The structure of the glycated amino acid derivatives and the peptide was confirmed by mass spectrometry and NMR spectroscopy. Importantly, the unprotected sugar moiety showed neither notable epimerization nor undesired side reactions during peptide elongation, allowing the incorporation of epimerically pure glucosyllysine. Upon acidic treatment, the building block as well as the resin-bound peptide formed one major byproduct due to incomplete Boc-deprotection, which was well separated by reversed-phase chromatography. Expectedly, the tandem mass spectra of the fructated amino acid and peptide were dominated by signals indicating neutral losses of 18, 36, 54, 84 and 96 m/z-units generating pyrylium and furylium ions. Supplementary Information The online version contains supplementary material available at 10.1007/s00726-021-02989-7.


Introduction
Protein glycation is a non-enzymatic posttranslational modification that was linked to diabetes, rheumatoid arthritis, Alzheimer's disease, and inflammatory diseases. Glycation is the initial stage of the Maillard reaction characterized by the reaction of electrophilic carbonyl groups, e.g., reducing sugars, with nucleophilic amino groups to form a Schiff base, which can rearrange to glycation products. Aldoses, such as d-glucose as dominant sugar in blood, can react with the ε-amino groups of lysine residues yielding 1-deoxy-1-ketosyllysine, i.e., fructosyllysine (glucated lysine), also referred to as Amadori product. In contrast, the rearrangement induced after the initial reaction of a ketose, e.g., d-fructose, yields 2-deoxy-2-aldosyllysine in two epimeric forms, i.e., glucosyl-and mannosyllysine (fructated lysine), which are both termed Heyns products. Both Amadori and Heyns products are prone to further reactions producing a class of heterogeneous compounds called advanced glycation endproducts (AGEs), which have been linked to the pathogenesis of diabetic complications, such as retinopathy, nephropathy, neuropathy, and accelerated atherosclerosis (Goh and Cooper 2008;Schalkwijk and Miyata 2012). Thus, Amadori and Heyns products appear to be promising precursors to monitor the formation of AGEs early. While Communicated by P. Meffre. a single glycation site can be quantified by ELISA, if a specific antibody is available, the large number of known glycation sites in serum or plasma proteins demands quantitative proteomics techniques, which typically rely on appropriate isotope-labeled standards. While synthetic Amadori peptides have been extensively used to study AGE formation (Greifenhagen et al. 2015) or to quantify prospective markers of type 2 diabetes (Spiller et al. 2017), only a few reports deal with d-fructose-modified (fructated) peptides. This was recently attributed to the challenging analysis of Heyns products and the Maillard reaction in general (Troise 2018). The lack of appropriate synthetic and analytical techniques is surprising, as fructation is more extensively involved in glycoxidation than glucation and, thus, represents an important research object (Hinton and Ames 2006;Semchyshyn 2013). As a consequence of increased sucrose or high-fructose corn syrup (HFCS) intake, preferentially consumed in the form of sugar-sweetened beverages, the fructose consumption raised markedly over the last 50 years (Bray et al. 2004). This excessive intake is associated with fatty liver (Jensen et al. 2018), metabolic syndrome, and type 2 diabetes (Malik and Hu 2015;Rodríguez et al. 2016). Fructose glycation may also contribute as a source of dietary AGEs, which can form in situ in the gastrointestinal tract (GI) lumen (DeChristopher 2017).
Initially, the site-specific modification of lysine residues by sugars used a global postsynthetic approach on solid phase (Frolov et al. 2006b). Thermal glucation of partially protected peptides provides good conversion rates, but requires long reaction times and high temperatures (70 °C or 110 °C) and purification is often challenging (Frolov and Hoffmann 2008). This approach was also applied to synthesize d-ribose and d-fructose-modified peptides to study their fragmentation behavior (Frolov et al. 2006a). However, poor yields were reported for fructation, which was also true for other sequences (Jakas et al. 2008). Similarly, the protocol reported by Heyns et al. (Heyns and Noack 1962) provided also low yields of fructated hippuryl-lysine using anhydrous conditions in DMSO at 75 °C (Krause et al. 2008). All three approaches yield a mixture of two epimers, i.e., 2-deoxy-2-glucosyl-and 2-deoxy-2-mannosyllysine, already reported in 1962 (Heyns and Noack 1962).
Here, we describe a new solid-phase approach using a N α -Fmoc-and N ε -Boc-protected and N ε -fructated lysine for the site-specific synthesis of Heyns peptides containing epimerically pure 2-deoxy-2-glucosyllysine. The N ε -Boc-protection in combination with acidic purification conditions enabled a stereoselective conversion of the glucosyl-epimer. Finally, a fructated 15mer peptide corresponding to HSA Lys233 was synthesized on solid phase in high yield and purity. The stereoisomer configuration of glucosyllysine was confirmed by NMR spectroscopy.
An aliquot of the crude product (133 mg) was reconstituted in 30% (v/v) aqueous acetonitrile containing 0.1% (v/v) formic acid and centrifuged (3030×g, 5 min, RT, Centrifuge 5702; Eppendorf AG, Hamburg, Germany). The supernatant was purified by reversed-phase high-performance liquid chromatography (RP-HPLC), as described below, and fractions corresponding to the expected molecular mass were lyophilized. The purified product (36.9 mg, 8% yield) was obtained as a mixture of epimers Fmocl-Lys(Glc)-OH 1a and Fmoc-l-Lys(Man)-OH 1b as white solid, which will be referred to as Fmoc-l-Lys(Glc/Man)-OH 1. A second purification of the crude product (125 mg) by RP-HPLC collecting only the early eluting fraction yielded the pure glucosyl-epimer Fmoc-Lys(Glc)-OH 1a as a brown, crystalline solid after lyophilization (12.0 mg, 98% purity according to UV-HPLC, 3% yield).
The presence of residual Fmoc-Lys(Glc)-OH leads to additional signals at 5.30 (d, J = 3.7 Hz).
Samples were analyzed by RP-HPLC on an 1100 LC system (Agilent Technologies, Santa Clara, CA, USA) equipped with an UV detector (absorbance recorded at 214 nm and 301 nm) coupled online to an ion trap mass spectrometer equipped with an electrospray ionization source (ESI-IT-MS, Esquire HCT, Bruker Daltonics) operated in positive or negative ion mode. Separations relied on a Jupiter C 18 -column (ID: 2 mm, length: 150 mm, particle size: 5 µm, pore size: 300 Å, Phenomenex) using a column temperature of 60 °C and a flow rate of 0.2 mL/min. Eluents were 0.1% (v/v) aqueous formic acid (eluent C) and 60% (v/v) aqueous acetonitrile containing 0.1% (v/v) formic acid (eluent D). Gradients used a linear slope from 5 to 95% eluent D in 30 min. The ESI source was operated at a source temperature of 365 °C using nitrogen as curtain gas (40 psi) and dry gas (9 L/min).
Peptides modified by glucose or fructose were analyzed on a nanoAcquity UPLC (Waters GmbH) coupled online to a Synapt G2-Si mass spectrometer (Waters GmbH) equipped with a nanoESI source (Waters GmbH) operated in positive ion mode. Eluents were eluent C and acetonitrile containing 0.1% (v/v) formic acid (eluent E). Peptides were trapped on a nanoAcquity Symmetry C18-column (ID: 180 μm, length: 2 cm, particle diameter: 5 μm) at a flow rate of 5 μL/ min (3% eluent E). Separation was achieved on a BEH 130 column (C18-phase, ID: 75 μm, length: 10 cm, particle diameter: 1.7 μm) using a flow rate of 0.35 μL/min and a column temperature of 35 °C. Peptides were eluted with linear gradients from 3% E to 40% eluent E in 19 min and to 85% eluent E in 5 min. Tandem mass spectra were recorded in data independent acquisition (DIA, MS E ) mode using the following settings: capillary voltage 3 kV, sampling cone of 30 kV, source offset of 60 kV, source temperature of 80 °C, cone gas flow of 20 L/h, nano-flow gas pressure of 0.20 bar, and a scan time of 0.5 s. Ions were fragmented in the trap cell using a collision energy ramp from 18 to 50 V.

NMR
All NMR spectra were recorded on a 700 MHz Bruker Avance III spectrometer equipped with a cryocooled TCI probe at 298 K. The 1 H-NMR spectra were recorded over a spectral width of 25 ppm. The free induction decay (FID) was digitized with 64 k data points and multiplied by an exponential window function (lb = 0.3) prior to Fourier transformation. All 1 H chemical shifts are referenced to DMSO-d 5 (δ = 2.50 ppm). The HSQC spectra (pulse program: hsqcedetgpsisp2.4) were recorded with a (F2, F1) resolution of (1024, 256) data points over a spectral width of (12 ppm, 165 ppm). The data points were zero-filled to (1024, 1024) points and multiplied by a qsine (SSB 2, SSB 2) window function prior to Fourier transformation. The HMBC spectra (pulse program: hmbcgpl2ndqf) were recorded with a (F2, F1) resolution of (1024, 512) data points over a spectral width of (12 ppm, 230 ppm). The data points were zero-filled to (1024, 1024) points and multiplied by a qsine (SSB 2, SSB 0) window function prior to Fourier transformation. The COSY spectra (pulse program: cosygpqf) were recorded with a (F2, F1) resolution of (1024, 512) data points over a spectral width of (12 ppm, 12 ppm). The data points were zero-filled to (1024, 1024) points and multiplied by a qsine (SSB 0, SSB 0) window function prior to Fourier transformation.

Synthesis of the d-fructose-modified lysine building block Fmoc-Lys(Glc,Boc)-OH
The synthesis of glycated peptides in high yields and purities requires building blocks suitable for peptide synthesis. While glucated peptides can be routinely synthesized by the Fmoc/ t Bu-strategy using protected building blocks (Carganico et al. 2009;Stefanowicz et al. 2010), a suitable strategy for the synthesis of fructated peptides containing N ε -glucosyl-or N ε -mannosyllysine is still missing. Therefore, we synthesized a building block based on a N ε -fructose-modified Fmoc-Lys(Boc)-OH (Fig. 1). Fmocl-Lys-OH was glycated adopting a method reported for fructation of lysine (Srinivas and Harohally 2012), which yields a mixture of l-Lys(Glc) and l-Lys(Man). Fmoc-Lys-OH, which was well soluble and stable in the aprotic solvent system, remained in the crude product (35%) besides two partially separated substances eluting on RP-HPLC at 20.3 min (39%) and 20.7 min (19%). Both products displayed the same monoisotopic mass of 531.1 Da (calculated 531.2 Da) indicating glucosyl-and mannosyl-epimers (Supplement, Fig. S1). Presumably, the earlier eluting species represented the glucosyl-epimer, as its formation is thermodynamically favored due to the equatorial position of the bulky lysine residue (Heyns et al. 1967). Tandem mass spectra acquired on a MALDI-TOF/TOF-MS displayed the typical fragmentation pattern reported for d-fructose-modified peptides (Supplement, Fig. S2) (Frolov et al. 2006a).
As the glycated ε-amino group was partially acylated during SPPS (data not shown), a Boc-protecting group was introduced using the purified mixture Fmoc-Lys(Glc/ Man)-OH (1). The reaction monitoring by RP-HPLC in the presence of 0.1% (v/v) TFA as ion pair reagent indicated that the first peak corresponding to Fmoc-Lys(Glc)-OH disappeared after one hour, while a new signal appeared around 6 min later (21% of all peak areas, Supplement, Fig.   Fig. 2 Zoomed proton NMR spectra (4.5-5.5 ppm) of purified N α -Fmoc-Lys[N ε -(2-deoxy-d-glucos-2-yl)]-OH (1a), N α -Fmoc-Lys[N ε -(2-deoxy-d-mannos-2-yl)]-OH (1b), AEFAEVSK Glc LVTDLTK (3a) and AEFAEVSKLVTDLTK (3b) from top to bottom. Chemical shifts and coupling constants of the proton at C 1 ′ are labeled, allowing conclusions about the sugar modification and its anomeric equilibrium. Peptide-derived signals (4.60-4.51 ppm, m) are marked with spades (♣) S7). Surprisingly, the area of the second peak remained stable even after two hours. When the same reaction mixture was separated with formic acid as ion pair reagent by LC-MS, the peak area of Fmoc-Lys(Man)-OH peak was reduced after 2 h (Supplement, Fig. S8), while a second Boc-protected product peak appeared. Due to the protection of both amino groups, the mass spectra were recorded in negative ion mode. They confirmed isobaric products at three different retention times. After purification by RP-HPLC (TFA as ion pair reagent), the presumed Fmoc-Lys(Glc,Boc)-OH (2) was obtained with a yield of 36% and a purity of 96% according to UV-HPLC (eluents containing formic acid); the overall yield was 3%. The monoisotopic mass of 630.28 Da was confirmed by ESI-MS in negative ion mode (629.95 Da, Fig. 4). The peak areas indicated that the fraction at 26.8 min (76%) contained most likely the α-glucopyranose form and the fraction at 27.7 min probably the β-glucopyranose form (20%) considering the ratios of the tautomeric forms of Fmoc-Lys(Glc)-OH (1a) (Fig. 2,  first panel). The recorded NMR showed multiple signal sets due to rotatory isomers from 1.34 to 1.55 ppm representing the aliphatic protons of a tert-butyl group (Supplement, Fig. S9). The anomeric proton region (4.5-6.1 ppm) was very crowded making it difficult to annotate the sugar signals. Most likely the lyophilized fraction contained trace amounts of earlier eluting glucosyl-(20.5 min, 2%) and mannosyllysine (20.7 min, 1%) as well as an unknown byproduct (23.8 min, 1%) detected with a 26 Da higher mass than Fmoc-Lys(Glc/Man)-OH. This may indicate the hydrolysis of the Boc-group, which appears highly pH-dependent, during RP-HPLC, as different conversion rates of the product were determined for different ion pair reagents (Supplement, Fig. S7/S8). Fractions eluting late from preparative column implied that preferentially Fmoc-Lys(Man,Boc)-OH was prone to hydrolysis in TFA-containing eluents (Supplement, Fig. S10). The reaction of the manno-epimer might be favored by electrostatic interactions between preferentially the 3-OH or 6-OH groups and the neighbored Boccarbonyl group. This may enhance its electrophilic character promoting Boc-hydrolysis by a nucleophilic attack of water. The use of amino sugar-derived organocatalysts increasing the enantioselectivity of a reaction and to activate carbonyls due to hydrogen bonding with hydroxyl groups was reported for asymmetric organic syntheses (Singh et al. 2008;Agarwal and Peddinti 2011). Importantly, purification of Fmoc-Lys(Man,Boc)-OH appears possible, but low pH conditions may promote its decomposition. The NMR of Fmoc-Lys(Glc)-OH presented above supports a separation of tautomers, as only the α-and β-pyranose forms were present, explaining the signals observed in RP-HPLC. Similarly, Carganico et al. observed four peaks for the Amadori analog N α -Fmoc-Lys[N ε -1-deoxy-d-fructos-1-yl,N ε -Boc)]-OH, which were explained by four tautomeric forms resulting from mutarotation of the sugar moiety (Carganico 2010).

Peptide synthesis
Fmoc-Lys(Glc,Boc)-OH (2) was used to synthesize the human serum albumin sequence AEFAEVSK Glc LVTDLTK (3a) ( Table 1), which is a glucation site proposed as biomarker of type 2 diabetes (Frolov et al. 2014). A quantitative coupling of the glycated amino acid (2) was achieved in SPPS, when Fmoc-Lys(Glc,Boc)-OH was added in twofold molar excess, as indicated by the absence of the acetylated truncated peptide (Fig. 5). The fructated (3a) and the unmodified peptides (3b) synthesized in parallel were detected in the crude products with peak areas of 55% (Fig. 5) and 65% (data not shown), respectively, based on RP-HPLC (Table 1). The sequences were confirmed by the b-and y-ion series displayed in the tandem mass spectra acquired on a MALDI-TOF/TOF-MS (data not shown). Importantly, the unmodified sequence was not detected in the crude glycated product. However, a major byproduct (18%) was detected at 22.5 min with a mass shift of 26 m/z-units at Lys8 compared to 3a or 188 m/z-units compared to 3b (Fig. 5). The tandem mass spectra of this byproduct did not display the typical neutral loss pattern of glycated peptides (data not shown) indicating a modification on a hydroxyl group. Most likely the anomeric sugar hydroxyl group reacted with the Bocgroup during acidic cleavage triggering an intramolecular cyclization yielding a 2-oxazolidinon derivative not prone  to water losses in MALDI-MS/MS due to the stabilized ring structure (Supplement, Fig. S14). A similar byproduct was obtained at the amino acid level, when acidic treatment of Fmoc-Lys(Glc,Boc)-OH yielded a minor byproduct (Fig. 4). This proposed mechanism would be similar to the reaction of triphosgene or carbon dioxide (Niemi et al. 2018), when a cyclic carbamate modification is purposely introduced on Amadori products (Gallas et al. 2012) or oligosaccharides (Whitehead et al. 2017) enabling the synthesis of complex glycoconjugates. Although the presumed carbamate byproduct was well separated on RP-HPLC, its formation could be prevented for example by peracetylation of the sugar moiety. Nevertheless, such an approach would require an additional deprotection step reducing also the peptide yield (Tavernaro et al. 2015). The need for a separate carbohydrate deprotection can be eliminated by protecting hydroxyl groups as acid-labile acetals or 4-methoxybenzyl ethers, which are cleaved simultaneously with TFA-catalyzed release from solid support. The glycated (3a, Fig. 5) and unmodified peptides (3b, Supplement Fig. S16) were obtained in similar yields of 35%-40% and high purities (~ 96%). The structures were verified by NMR confirming the successful incorporation of glucosyllysine while excluding the mannosyllysine form (Figs. 4, S15, S17). Chemical shifts of the anomeric proton signals of the glucosyllysine derivative (1a) are in agreement with those detected for the glycated peptide (3a) indicating that the α-glucopyranosyl-and the β-glucopyranosl-forms are present at a 4:1 ratio. The tandem mass spectra of glucated (synthesized as described by (Spiller et al. 2017)) and fructated AEFAEVSK Fru/Glc LVT-DLTK (3a) were dominated by characteristic neutral loss patterns (Fig. 6), i.e., consecutive losses of up to three water molecules (18 Da, 36 Da, and 54 Da) and a loss of three water and one formaldehyde molecules (84 Da). The Heyns product 3a displayed an additional loss corresponding to 96 Da (m/z 858.99), which we reported earlier as specific for ketohexose-derived compounds (Frolov et al. 2006a;Corzo-Martínez et al. 2009).

Conclusions
The synthesis of a fructose-modified lysine derivative in two steps and its application as a building block for solidphase peptide synthesis allowed the fast production of a Heyns peptide in high purity and good yield. This protocol overcomes the currently applied post-synthetic onresin glycation of partially deprotected peptides. Despite using an unprotected sugar moiety, no epimerization was observed in the purified peptide. Only one major byproduct Compounds were analyzed by RP-HPLC using a linear 30-min gradient from 3 to 57% and a linear 20-min gradient from 24 to 36% aqueous acetonitrile containing 0.1% formic acid, respectively (absorbance recorded at 214 nm). The mass spectrum from 20.0 to 22.0 min was recorded online on an ESI-ion trap-MS in positive ion mode from m/z 500 to 1000. The small insert shows the isotope pattern of the doubly protonated quasimolecular ion at m/z 906.72. The observed byproduct (22.5 min) is described by its proposed structure, assuming an irreversible rearrangement in the context of Boc-group cleavage (see Fig. S14). Further details are provided in methods and the full mass spectra of crude 3a are shown in Supplement, Fig. S13 was formed, which was well separated by RP-HPLC, presumably due to incomplete cleavage of the Boc-group. The low yield of the synthesis of Fmoc-Lys(Glc,Boc)-OH was partially attributed to the purification by RP-HPLC in the presence of 0.1% TFA. Better yields might be observed by ethyl acetate extraction to remove DMF and excess Boc 2 O (Stefanowicz et al. 2010). The extraction may even prevent the hydrolysis of the presumed acid-labile Fmoc-Lys(Man,Boc)-OH. However, the demonstrated chemoselective hydrolysis of epimeric Fmoc-Lys(Glc/Man, Boc) with subsequent purification by RP-HPLC might provide pure mannosyllysine in good yields. The presented synthetic strategy gives access to welldefined Heyns compounds of high purity, which will allow studying fructation and evaluating their potential to form toxic AGEs in more detail. Fig. 6 Tandem mass spectra of the doubly protonated quasimolecular ions detected at m/z 907.02 corresponding to fructosamine-modified (AEFAEVSK Fru LVTDLTK, Amadori product, top) and to glucosamine-modified peptides (3a, AEFAEVSK Glc LVTDLTK, Heyns product, bottom). The neutral losses characteristic for the sugar moieties are indicated above the peaks. Heynsproduct fragment ions, which are exclusive (m/z 858.99) or significantly increased in their intensities (m/z 898.00) are highlighted with a frame. Data were acquired on an ESI-QTOF instrument using collision induced dissociation (CID) with an energy ramp from 18 to 50 V (32.1 V for m/z 907.02)