An Optimised Di-Boronate-ChemMatrix Affinity Chromatography to Trap Deoxyfructosylated Peptides as Biomarkers of Glycation

We report herein a novel ChemMatrix® Rink resin functionalised with two phenylboronate (PhB) moieties linked on the N-α and N-ε amino functions of a lysine residue to specifically capture deoxyfructosylated peptides, compared to differently glycosylated peptides in complex mixtures. The new PhB-Lys(PhB)-ChemMatrix® Rink resin allows for exploitation of the previously demonstrated ability of cis diols to form phenylboronic esters. The optimised capturing and cleavage procedure from the novel functionalised resin showed that only the peptides containing deoxyfructosyl-lysine moieties can be efficiently and specifically detected by HR-MS and MS/MS experiments. We also investigated the high-selective affinity to deoxyfructosylated peptides in an ad hoc mixture containing unique synthetic non-modified peptides and in the hydrolysates of human and bovine serum albumin as complex peptide mixtures. We demonstrated that the deoxyfructopyranosyl moiety on lysine residues is crucial in the capturing reaction. Therefore, the novel specifically-designed PhB-Lys(PhB)-ChemMatrix® Rink resin, which has the highest affinity to deoxyfructosylated peptides, is a candidate to quantitatively separate early glycation peptides from complex mixtures to investigate their role in diabetes complications in the clinics.


Introduction
In high-glucose concentration, typical of different diabetes conditions, the reducing sugar reacts with free amino functions, such as the N ε -side chain of lysine residues in proteins, In the current study, we selected the ChemMatrix ® Rink resin (CMRR) to increase selectivity of detection of deoxyfructosylated peptides as an alternative to the previously proposed m-aminophenylboronic acid (m-APBA) immobilised on the agarose gel [6]. In particular, the advantages of the ChemMatrix ® Rink resin are: compatibility with a large variety of solvents, including water; high chemical stability; optimal swelling properties; and the possibility to build on it any type of peptide sequence. Moreover, the ChemMatrix ® resin (initially designed for solid-phase synthesis of difficult peptide sequences, including long and hydrophobic ones) [24,25], was recently used for peptide-affinity chromatography based on combinatorial strategies for protein purification [26]. Last, but not least, derivatisation of the CMRR is relatively simple following the optimised protocols set up for peptide synthesis [26].
Therefore, CMRR has been selected to design a novel stationary phase for affinity chromatography of deoxyfructosylated peptides. In particular, we propose herein a simple synthetic protocol using the low-cost Fmoc-Lys(Fmoc)-OH as starting material to act as a linker between the CMRR and two phenylboronate moieties (Scheme 1). In particular, we developed the PhB-Lys(PhB)-CMRR as a capturing support for reversible binding of deoxyfructosylated peptides for further mass spectrometry identification. According to the literature, there is no evidence that the lysine moiety with two phenylboronate moieties linked to CMRR has been reported and applied for a specific affinity chromatography technique. Therefore, we examined the peptide-capturing properties of the novel functionalised resin PhB-Lys(PhB)-CMRR as a function of time to determine its efficiency. We confirmed its stability in different solvents and we developed analytical methods to perform its quality control, including loading and homogeneity of PhB-Lys(PhB)-NH 2 once cleaved from the resin. A synthetic protocol for the preparation of the novel functionalised resin PhB-Lys(PhB)-CMRR was optimised. We proved that the introduction of two phenylboronate moieties attached to the lysine linker has a significant impact on binding the deoxyfructosylated peptides. We showed that the PhB-Lys(PhB)-CMRR is recyclable having demonstrated the same ability to separate deoxyfructosylated peptides in solution after several cycles of capturing. Moreover, the resin was tested to demonstrate that deoxyfructosylated peptides could be captured specifically compared to other glycopeptides modified with different glycosyl moieties. In fact, its high selectivity and affinity for the deoxyfructosylated peptides were investigated using an ad hoc mixture containing unique synthetic glycopeptides and non-modified peptides, and complex peptide mixtures obtained from hydrolysis of human and bovine serum albumin (HSA and BSA).

Development of a Specific ChemMatrix ® Rink Resin Functionalised to Selectively Capture Deoxyfructosylated Peptides
We report herein a novel stationary phase designed and optimised for affinity chromatography technology, to selectively capture deoxyfructosylated peptides, as we demonstrated on different samples. We selected the ChemMatrix ® Rink resin (CMRR) as an ideal solid support for its welldescribed intrinsic properties [24,25]. Fmoc-Lys(Fmoc)-OH was used as a low-cost, commerciallyavailable building block to introduce a linker between CMRR and two phenylboronate moieties (PhB). The proposed novel capturing support (e.g., PhB-Lys(PhB)-CMRR) was designed to carry two PhB (linked to the N-α and N-ε amino functions of the lysine residue respectively), to significantly improve trapping of deoxyfructosylated peptides. The CMRR is fully compatible with a solid-phase synthetic protocol to develop a collection of different phenylboronate derivatives. Therefore, the synthetic procedure (Scheme 1) initially binds Fmoc-Lys(Fmoc)-OH to CMRR using 2-(1H-Benzotriazole-1-yl)-1,1,3,3-tetramethylaminium tetrafluoroborate (TBTU) as a coupling reagent and, after removing the two Fmoc-protecting groups in a piperidine solution, the 4-carboxyphenylboronic acid (PhB-OH) was reacted with the two lysine-amino functions by PyBOP. We previously reported an efficient strategy to link PhB-OH to a lysine side chain without any protecting group on hydroxyl functions [23]. In contrast to the m-aminophenylboronate agarose support previously proposed [6], successful functionalisation and loading of the PhB-Lys(PhB)-ChemMatrix ® Rink resin can be easily and efficiently monitored by quantitative LC-ESI-MS determination of the acidic hydrolysis product PhB-Lys(PhB)-NH2. In fact, the HPLC ( Figure S1) and ESI-MS spectrum profiles ( Figure S2) of the PhB-Lys(PhB)-NH2 crude show an intensive signal corresponding to the m/z value and the characteristic isotopic pattern consistent with the hydrolysis product corresponding to the molecular formula C20H25N3O7B2Na ( Figure S3).

General Capturing Protocol of Model Deoxyfructosylated Peptide 1
PhB-Lys(PhB)-CMRR (5 mg) were swelled in a syringe for 30 min in ammonium bicarbonate buffer (50 mM) at pH 8 in H 2 O/MeCN 1:1 (v/v). Equimolar amount of deoxyfructosylated peptide H-K(Dabcyl)AK(1-DeoxyFru)AF-NH 2 (1) was dissolved in the same buffer (1 mL), and added to the resin and mixed for 1 h. Then the solution was filtered off and the resin was washed three times with the buffer solution to remove the possible unreacted deoxyfructosylated peptide. All the fractions were collected and lyophilised. Then, the cleavage mixture containing 0.1% HCOOH in H 2 O/MeCN 1:1 (v/v) was added and the syringe was stirred for 1h. The solution was filtered off and the resin was washed four times with the cleavage mixture. All the cleavage fractions were collected, lyophilised, and analysed by LC-MS, LC-MS/MS, and UV-VIS (Scheme S1, see SI).

Capturing Conditions of Peptides 2-8
The general procedure optimised for peptide 1 was applied separately to the synthetic peptides 2-8 using equimolar concentrations of all the peptides (1.4 µmol/mL). The capturing reaction time was extended to 2 h. The remaining steps of the procedure were applied without changes. All the cleavage fractions were collected, lyophilised, and analysed by LC-UV-MS. Three ad hoc aqueous solutions containing the equimolar concentrations of non-glycosylated peptide sequences (1.4 µmol/mL), including the deoxyfructosylated peptide 2, at the same concentration (1.4 µmol/mL), or 5% w/w (10.5 nmol/mL), or 2% w/w (4.2 nmol/mL), were prepared and used following the general capturing protocol described above, extending the time to 2 h. The remaining steps of the protocol were applied without changes. All the fractions (uncaught and released from PhB-Lys(PhB)-ChemMatrix @ Rink resin) were collected, lyophilised, and analysed by LC-UV-MS.

Capturing Conditions of Peptide 2 and 18 in HSA or BSA Hydrolysate
Human Serum Albumin (HSA, BioLabs) was hydrolysed according to the procedure reported in SI. The deoxyfructosylated peptide 2 was added to the HSA hydrolysate at the same concentration used for the non-glycosylated peptides (1.4 µmol/mL), or 5% w/w (10.5 nmol/mL), or 2% w/w (4.2 nmol/mL), and applied in the general capturing protocol, also extending the time reaction to 2 h. The remaining steps were followed without any changes. All the cleavage fractions were collected, lyophilised, and analysed by LC-MS, LC-UV-MS, or LC-MS/MS. The LC-MS/MS data were analysed using PEAKS search engine.
The general protocol described above was applied to a mixture of the commercially available Bovine Serum Albumin hydrolysate (500 pmol, BSA, BioLabs) including peptide 18 (300 pmol), extending the time reaction to 2 h. The remaining steps were followed without any changes. All the cleavage fractions were collected, lyophilised, and analysed by LC-MS/MS. Obtained LC-MS/MS data were analysed using PEAKS search engine.

Bioinformatic Analysis with PEAKS
LC-MS/MS analyses of the samples were performed on a SHIMADZU IT-TOF instrument using automatic fragmentation mode. The spectra were analysed by a bioinformatic approach using the LabSolution software to convert LC data file to mzXML format. Data were analysed using PEAKS search engine. For this purpose, we downloaded a FASTA file from the UniProt database. The following search parameters were set: precursor mass error tolerance ±0.1 Da; fragment ion tolerance ±0.2 Da; CID fragmentation, semi-specific digestion mode by trypsin, assuming up to three missed cleavages per peptide. Moreover, carbamidomethylation, as a fixed post-translational modification (PTM), was chosen. Deamidation, dehydration, oxidation of methionine, and hexose modifications were set as variable PTMs and three maximum variable PTMs per peptide were allowed. A false discovery rate (FDR) < 2% at the spectra level was used to filter search results. Additionally, the presence of at least two unique peptides was set for the identification of the corresponding protein. The obtained results, especially the statistical data, were reported into the Supporting Information.

Development of a Specific ChemMatrix ® Rink Resin Functionalised to Selectively Capture Deoxyfructosylated Peptides
We report herein a novel stationary phase designed and optimised for affinity chromatography technology, to selectively capture deoxyfructosylated peptides, as we demonstrated on different samples. We selected the ChemMatrix ® Rink resin (CMRR) as an ideal solid support for its well-described intrinsic properties [24,25]. Fmoc-Lys(Fmoc)-OH was used as a low-cost, commercially-available building block to introduce a linker between CMRR and two phenylboronate moieties (PhB). The proposed novel capturing support (e.g., PhB-Lys(PhB)-CMRR) was designed to carry two PhB (linked to the N-α and N-ε amino functions of the lysine residue respectively), to significantly improve trapping of deoxyfructosylated peptides. The CMRR is fully compatible with a solid-phase synthetic protocol to develop a collection of different phenylboronate derivatives. Therefore, the synthetic procedure (Scheme 1) initially binds Fmoc-Lys(Fmoc)-OH to CMRR using 2-(1H-Benzotriazole-1-yl)-1,1,3,3-tetramethylaminium tetrafluoroborate (TBTU) as a coupling reagent and, after removing the two Fmoc-protecting groups in a piperidine solution, the 4-carboxyphenylboronic acid (PhB-OH) was reacted with the two lysine-amino functions by PyBOP. We previously reported an efficient strategy to link PhB-OH to a lysine side chain without any protecting group on hydroxyl functions [23].
In contrast to the m-aminophenylboronate agarose support previously proposed [6], successful functionalisation and loading of the PhB-Lys(PhB)-ChemMatrix ® Rink resin can be easily and efficiently monitored by quantitative LC-ESI-MS determination of the acidic hydrolysis product PhB-Lys(PhB)-NH 2 . In fact, the HPLC ( Figure S1) and ESI-MS spectrum profiles ( Figure S2) of the PhB-Lys(PhB)-NH 2 crude show an intensive signal corresponding to the m/z value and the characteristic isotopic pattern consistent with the hydrolysis product corresponding to the molecular formula C 20 H 25 N 3 O 7 B 2 Na ( Figure S3).
A calibration curve was set up with the commercially available 4-carboxyphenylboronic acid (purity > 99%) as a standard ( Figure S4). In particular the crude PhB-Lys(PhB)-NH 2 ( Figure S5) at 236 nm (not interfering with amide bond absorptions) [27] allows calculation of the loading of PhB-Lys(PhB)-ChemMatrix ® Rink resin, after cleavage of PhB-Lys(PhB)-NH 2 from an exactly weighted quantity of resin, lyophilisation, and final UV analysis (Table 1). Table 1. Loading of PhB-Lys(PhB)-ChemMatrix ® Rink resin (M = 0.28 mmol/g, SD = 0.007). We demonstrated that the novel PhB-Lys(PhB)-ChemMatrix ® Rink resin is stable to capture reaction conditions and the resin loading is constant. In fact, after incubation of the functionalised resin in different solvents, accurate ESI-MS analysis of the lyophilised solutions did not display any signal corresponding to the possible cleaved linker PhB-Lys(PhB)-NH 2 .

PhB-Lys(PhB)-CMRR (mg) Sample Dilution Abs (OD) n [mmol] PhB-Lys(PhB)-CMRR
We used, as first model the simple synthetic pentapeptide visible sequence, H-Lys(Dabcyl)Ala-Lys(1-DeoxyFru)Ala-Phe-NH 2 (1) to demonstrate the capturing properties of the novel PhB-Lys(PhB)-ChemMatrix ® Rink resin. Therefore, we built a calibration curve to discriminate, in mixtures, deoxyfructosylated peptides in presence of differently glycosylated peptides. In particular, peptide 1 contains the deoxyfructopyranosyl moiety linked to Lys3 side chain (already known to form esters with phenylboronic acid) [12], but also the 4-((4-(dimethylamino)phenyl)azo)benzoyl moiety (Dabcyl) linked to Lys1 side chain, exploiting its broad and intense visible absorption, to determine the minimal detectable concentration of the deoxyfructosylated model peptide 1 (Figures S6-S10). The synthetic building block Fmoc-Lys(Dabcyl)-OH (I) was used to prepare the calibration curves (Figures S11 and S12) and determine the concentration of the model peptide 1 before and after the capturing reaction. The maximum absorption of Fmoc-Lys(Dabcyl)-OH (I) in the capturing solution was ca. 445 nm, but in the cleavage solution the absorption was shifted to 455 nm ( Figure S13).
To determine the maximum capacity of the PhB-Lys(PhB)-ChemMatrix ® Rink resin (substitution grade 0.28 mmol/g) in the capturing reaction conditions, we performed an experiment in saturating conditions of model deoxyfructosylated peptide 1 (Scheme S1). PhB-Lys(PhB)-CMRR (5 mg) was able to capture the model peptide 1 (0.0014 mmol), considering that only one active phenylboronic group attached to the lysine moiety could bind cis diol groups.
According to the obtained data, 5 mg of resin is capable to capture 1.11 mg of pure peptide. Therefore, the capturing efficiency of the proposed procedure is 81%. We proved that both the new resin and the regenerated resin batches have the same ability to capture the deoxyfructosylated peptide model 1 from the solution ( Table 2). Table 2. Comparison of capturing efficiency of new and reused (once or twice) batches of the PhB-Lys(PhB)-ChemMatrix ® Rink resin. Moreover, we determined the minimum time to dry the functionalised resin after synthesis obtaining a constant mass ( Figure S14), showing that resin storage under atmospheric pressure causes insignificant changes in its weight ( Figure S15).

The PhB-Lys(PhB)-ChemMatrix ® Rink Resin Is Specific for Deoxyfructosylated Peptides and Not for Unrelated Non-Glycosylated Peptides
We tested the binding efficiency of the PhB-Lys(PhB)-ChemMatrix ® Rink resin for deoxyfructosylated peptides as early glycation biomarkers of diabetes, in a peptide mixture mimetic of a digestion procedure of biological fluids. Therefore, we prepared three ad hoc aqueous solutions containing equimolar concentrations of all the synthetic peptides reported in Table 4 (1.4 µmol/mL), including the deoxyfructosylated peptide 2: (i) at the same concentration (1.4 µmol/mL); (ii) 5% w/w (10.5 nmol/mL); (iii) 2% w/w (4.2 nmol/mL), to estimate the limit of detection of the deoxyfructosylated model peptide 2 by LC-MS. The ten synthetic peptides 7 and 9-17 (Table 4 and Table S2) were dissolved in an ammonium bicarbonate buffer solution at pH 8 and then subjected (as model matrix) to the capturing reaction conditions described above. These peptides were selected for the following characteristics: amino acid sequence, hydrophobicity, different length (from 4 to 35 amino acids) and post-translational modification (DeoxyFru, Ac, PO 3 H 2 , and Pam). The hexapeptide LSETTI (11) was selected as a previously developed peptide mimicking the HNK1 tri-saccharide epitope [37]. It is interesting to notice that Frolov et al. [19] previously reported that the efficiency of the m-aminophenylboronic acid (m-APBA) immobilised on agarose resin to capture Amadori peptides, after digestion of plasma protein from type 2 diabetes, was affected by unspecific recognition of peptide sequences containing several, often neighbouring, hydroxyamino acid residues. Therefore, in order to demonstrate the ideal and optimised capacity of the newly-functionalised PhB-Lys(PhB)-ChemMatrix ® Rink resin (herein proposed), only for deoxyfructosylated peptides, we selected the series of synthetic peptides containing those amino acid residues (Table 4). In addition, the tetrapeptide EKEK (13) was designed as a control sequence, because it contains only lysine residues (but no Tyr and Thr) that in diabetes are putative glycation motifs.
Our experiments demonstrated that the only peptide captured from the reaction mixture was the peptide [(1-DeoxyFru)Lys 7 ]CSF114 (2). None of the non-deoxyfructosylated peptides were captured by the novel CMRR resin, demonstrating the high selectivity of our procedure for deoxyfructosylated peptides (Figures S22-S25). The lowest detectable concentration with proper signal-to-noise ratio is 5% (Figures S26 and S27).

The PhB-Lys(PhB)-ChemMatrix ® Rink Resin Is Also Specific for Deoxyfructosylated Peptides in Complex Peptide Mixtures (i.e., Human Serum Albumin and Bovine Serum Albumin Hydrolysates)
The affinity of the PhB-Lys(PhB)-ChemMatrix ® Rink resin for the deoxyfructosylated peptides 2 and 18 was tested in complex hydrolysate mixtures of human and bovine serum albumin (HSA and BSA, respectively). Therefore, the above described procedure was applied to a mixture of the commercially available hydrolysate of HSA (97-99%, Sigma-Aldrich) in the presence of the deoxyfructosylated peptide 2 that was added: (i) at the same concentration (1.4 µmol/mL); (ii) 5% w/w (10.5 nmol/mL); or (iii) 2% w/w (4.2 nmol/mL) to estimate the limit of detection of the deoxyfructosylated peptide model 2 by LC-UV-MS (Figures S28 and S29). Interestingly, the most intensive signal detected corresponds to the deoxyfructosylated peptide model 2. Detailed LC-MS/MS of the HSA hydrolysate containing peptide 2 at 5% w/w concentration, followed by bioinformatic analysis using the PEAKS software (see Materials and Methods), showed that only a negligible amount of two unrelated peptide sequences from HSA hydrolysate (ALVLIAFAQYLQQCPFEDHVK and DVFLGMFLYEYAR) were captured by CMRR ( Figure S30). Interestingly, one of the sequences (including two tyrosine residues) was also reported by Frolov et al. [19] Moreover, the deoxyfructosylated peptide 2 at 2% w/w concentration was detectable only by LC-MS, but not by UV (Figure 1). At this concentration, no peptide sequence derived from the HSA hydrolysate could be detected (PEAKS software).
Molecules 2020, 25, x FOR PEER REVIEW 9 of 14 bioinformatic analysis using the PEAKS software (see Materials and Methods), showed that only a negligible amount of two unrelated peptide sequences from HSA hydrolysate (ALVLIAFAQYLQQCPFEDHVK and DVFLGMFLYEYAR) were captured by CMRR ( Figure S30). Interestingly, one of the sequences (including two tyrosine residues) was also reported by Frolov et al. [19] Moreover, the deoxyfructosylated peptide 2 at 2% w/w concentration was detectable only by LC-MS, but not by UV (Figure 1). At this concentration, no peptide sequence derived from the HSA hydrolysate could be detected (PEAKS software). Unspecific interactions between the new functionalised resin and peptides derived from HSA hydrolysate were carefully investigated, applying the capturing procedure described above without addition of any deoxyfructosylated peptide. Detailed LC-MS analyses are presented in Figure S31. In particular, PEAKS bioinformatic analysis revealed that protein sequence coverage was only 16% and the seven low-intensity non-modified peptides identified ( Figure S32) could be detected only if the deoxyfructosylated peptide 2 (besides having the highest affinity for the CMRR) is not present in the mixture and therefore, it cannot compete in binding. Unspecific interactions between the new functionalised resin and peptides derived from HSA hydrolysate were carefully investigated, applying the capturing procedure described above without addition of any deoxyfructosylated peptide. Detailed LC-MS analyses are presented in Figure S31. In particular, PEAKS bioinformatic analysis revealed that protein sequence coverage was only 16% and the seven low-intensity non-modified peptides identified ( Figure S32) could be detected only if the deoxyfructosylated peptide 2 (besides having the highest affinity for the CMRR) is not present in the mixture and therefore, it cannot compete in binding.
An additional experiment was performed using the complex mixture of the commercially available BSA hydrolysate (500 pmol, BioLabs) in the presence of the analytically pure deoxyfructosylated BSA fragment DTEK(1-DeoxyFru)QIKKQT (18) (300 pmol). Before treating the complex mixture in the capturing reaction conditions, a detailed LC-MS/MS analysis was performed (Figure 2). Sequence coverage was 61% ( Figure S33) and identity of the peptide DTEK(1-DeoxyFru)QIKKQT (18) was confirmed by XIC (Extracted Ion Chromatography), ESI-MS, and ESI-MS/MS (Figure 2). The concentration of the deoxyfructosylated peptide 18 was chosen unquestionably lower than the other peptides to display a lower mass intensity. Nevertheless, LC-MS/MS analysis revealed that the most intense signal (after reaction with the functionalised resin) corresponded to the deoxyfructosylated peptide 18 (Figure 3 and Figure S34). Moreover, bioinformatic analysis showed in BSA hydrolysate nine low intensity, but negligible peptide sequences, compared to peptide 18 ( Figures S34 and S35). In conclusion, we demonstrated that decrease in deoxyfructosylated peptide concentration in the hydrolysate mixture might cause unspecific but insignificant interactions. Therefore, these unspecific but negligible interactions do not hamper the quality of the analysis, because the deoxyfructosyl-lysine containing the peptide 18 continues maintaining the highest affinity for the novel PhB-Lys(PhB)-ChemMatrix ® Rink resin presented herein and specifically-designed to trap deoxyfructosylated peptides to be efficiently detected by an enhanced mass spectrometry technique. unquestionably lower than the other peptides to display a lower mass intensity. Nevertheless, LC-MS/MS analysis revealed that the most intense signal (after reaction with the functionalised resin) corresponded to the deoxyfructosylated peptide 18 (Figures 3 and S34). Moreover, bioinformatic analysis showed in BSA hydrolysate nine low intensity, but negligible peptide sequences, compared to peptide 18 ( Figures S34 and S35). In conclusion, we demonstrated that decrease in deoxyfructosylated peptide concentration in the hydrolysate mixture might cause unspecific but insignificant interactions. Therefore, these unspecific but negligible interactions do not hamper the quality of the analysis, because the deoxyfructosyl-lysine containing the peptide 18 continues maintaining the highest affinity for the novel PhB-Lys(PhB)-ChemMatrix ® Rink resin presented herein and specifically-designed to trap deoxyfructosylated peptides to be efficiently detected by an enhanced mass spectrometry technique.

Conclusions
Deoxyfructosylation of peptides in human proteins is the early event of the complex glycation pathway typical of diabetes. Up to now, specific, selective, and high affinity early diagnostics of deoxyfructosylated peptides is always an unmet need for clinicians, to prevent further and, in most cases, fatal complications. In the present paper, we report the preparation, chemical characterisation, and capturing properties specific for deoxyfructosylated peptides, of a novel ChemMatrix ® Rink resin functionalised with two phenylboronate moieties (PhB) linked to the two amino functions of a lysine residue. The new PhB-Lys(PhB)-ChemMatrix ® Rink resin allows exploitation of the previously demonstrated ability of cis diols (present in the deoxyfructosyllysine moiety) to form phenylboronic esters. We demonstrated, by accurate ESI-MS analysis, that the novel PhB-Lys(PhB)-ChemMatrix ® Rink resin is stable to capturing reaction conditions. In fact, no signal corresponding to the possibly cleaved linker PhB-Lys(PhB)-NH2 was detected. Moreover, the resin loading is constant. Resincapturing and -cleavage procedures were optimised to ensure that only the peptides containing deoxyfructosyllysine moieties can be efficiently and specifically detected by HR-MS and MS/MS experiments. The selectivity and affinity of the PhB-Lys(PhB)-ChemMatrix ® Rink resin was demonstrated in the case of three different peptides used in complex mixtures. First, we used, as a model, the simple synthetic pentapeptide H-Lys(Dabcyl)Ala-Lys(1-DeoxyFru)Ala-Phe-NH2 (1) containing Lys(Dabcyl), which, with its characteristic broad and intense visible absorption, allowed for the determination of the minimal detectable concentration. Using peptide 1, the capturing efficiency of PhB-Lys(PhB)-ChemMatrix ® Rink resin was approximately 80%. It is worth highlighting that CMRR allows both the performance of orthogonal multidirectional synthesis, as well as accurate control by the specific PhB-Lys(PhB)-linker on the resin. The product can be removed from the resin at any time and can be characterised by analytical methods, which is not possible in the case of the agarose resin.
The significant binding of deoxyfructosylated peptides by functionalised CMRR, in comparison to other glycosylated peptides, may be explained considering the equilibrium of the different cyclic

Conclusions
Deoxyfructosylation of peptides in human proteins is the early event of the complex glycation pathway typical of diabetes. Up to now, specific, selective, and high affinity early diagnostics of deoxyfructosylated peptides is always an unmet need for clinicians, to prevent further and, in most cases, fatal complications. In the present paper, we report the preparation, chemical characterisation, and capturing properties specific for deoxyfructosylated peptides, of a novel ChemMatrix ® Rink resin functionalised with two phenylboronate moieties (PhB) linked to the two amino functions of a lysine residue. The new PhB-Lys(PhB)-ChemMatrix ® Rink resin allows exploitation of the previously demonstrated ability of cis diols (present in the deoxyfructosyllysine moiety) to form phenylboronic esters. We demonstrated, by accurate ESI-MS analysis, that the novel PhB-Lys(PhB)-ChemMatrix ® Rink resin is stable to capturing reaction conditions. In fact, no signal corresponding to the possibly cleaved linker PhB-Lys(PhB)-NH 2 was detected. Moreover, the resin loading is constant. Resin-capturing and -cleavage procedures were optimised to ensure that only the peptides containing deoxyfructosyllysine moieties can be efficiently and specifically detected by HR-MS and MS/MS experiments. The selectivity and affinity of the PhB-Lys(PhB)-ChemMatrix ® Rink resin was demonstrated in the case of three different peptides used in complex mixtures. First, we used, as a model, the simple synthetic pentapeptide H-Lys(Dabcyl)Ala-Lys(1-DeoxyFru)Ala-Phe-NH 2 (1) containing Lys(Dabcyl), which, with its characteristic broad and intense visible absorption, allowed for the determination of the minimal detectable concentration. Using peptide 1, the capturing efficiency of PhB-Lys(PhB)-ChemMatrix ® Rink resin was approximately 80%. It is worth highlighting that CMRR allows both the performance of orthogonal multidirectional synthesis, as well as accurate control by the specific PhB-Lys(PhB)-linker on the resin. The product can be removed from the resin at any time and can be characterised by analytical methods, which is not possible in the case of the agarose resin. The significant binding of deoxyfructosylated peptides by functionalised CMRR, in comparison to other glycosylated peptides, may be explained considering the equilibrium of the different cyclic anomeric forms in solution, as reported in the literature (i.e., β-pyranose: ca. 75%; β-furanose: ca. 13%; and α-furanose: ca. 12%) [38,39]. Boronic acid is known to bind glucose preferentially in the α-furanose form and not in the more abundant α-pyranose form. β-D-Fructofuranose binds favourably to boronic acid because of the high presence of syn-periplanar hydroxyl groups [40]. This can explain the different binding constants to other sugars, such as galactose, mannose, or glucose. The binding of differently glycosylated and deoxyfructosylated peptides are strongly in agreement with data reported with free sugars, because only the deoxyfructosylated peptides bear the sugar in the furanose form. In the case of glycosylated peptides (e.g., galactosylated or glucosylated), only pyranose forms are present.
Moreover, we demonstrated that the regenerated resin has the same ability to capture the deoxyfructosylated peptide model 1 from the solution. Afterwards we investigated the high-selective affinity to deoxyfructosylated peptides 2 and 18 in an ad hoc mixture, containing also unique synthetic non-modified peptides, and in the hydrolysate of human and bovine serum albumin as an example of complex peptide mixture. We demonstrated that only peptide [(1-DeoxyFru)Lys 7 ]CSF114 (2) was captured from the reaction mixture of synthetic non-modified peptides by the novel CMRR resin, showing unambiguously that the deoxyfructopyranosyl moieties on lysine residues are crucial in the capturing reaction. Testing the affinity of the PhB-Lys(PhB)-ChemMatrix ® Rink resin for different concentrations of deoxyfructosylated peptides 2 and 18 in complex hydrolysate mixtures of HSA and BSA, LC-MS/MS analysis showed that the most intense signal (after reaction with the functionalised resin), corresponded to deoxyfructosylated peptide 18 (or 2). PEAKS bioinformatics analysis revealed that decreasing the concentration of deoxyfructosylated peptides in the hydrolysate mixture might cause unspecific, but insignificant interactions, because intensity of the identified peptides originating from hydrolysate was low compared with the deoxyfructosyl-lysine containing peptides (18 or 2). Therefore, the novel specifically-designed PhB-Lys(PhB)-ChemMatrix ® Rink resin, displaying highest affinity to deoxyfructosylated peptides, is a candidate to quantitatively separate early glycation peptides from complex mixtures to investigate their role in diabetes complications.