Arseniy
Galashov
a,
Ekaterina
Kazakova
a,
Christian E.
Stieger
b,
Christian P. R.
Hackenberger
ab and
Oliver
Seitz
*a
aDepartment of Chemistry, Humboldt-Universität zu Berlin, Brook-Taylor-Straße 2, 12489, Berlin, Germany. E-mail: oliver.seitz@chemie.hu-berlin.de
bLeibniz-Forschungsinstitut für Molekulare Pharmakologie (FMP), Robert-Rössle-Strasse 10, 13125, Berlin, Germany
First published on 18th December 2023
The study of mucin function requires access to highly O-glycosylated peptides with multiple tandem repeats. Solid-phase synthesis would be a suitable method, however, the central problem in the synthesis of mucin glycopeptides is the need to use precious and potentially vulnerable glycoamino acid building blocks in excess. In this article, we report the development of a method based on SPPS and native chemical ligation/desulfurization chemistry that allows the rapid, reliable, and glyco-economical synthesis of long multi-O-GalNAcylated peptides. To facilitate access to the glycosyl donor required for the preparation of Fmoc-Ser/Thr(αAc3GalNAc)-OH we used an easily scalable azidophenylselenylation of galactal instead of azidonitration. The problem of low yield when coupling glycoamino acids in small excess was solved by carrying out the reactions in 2-MeTHF instead of DMF and using DIC/Oxyma. Remarkably, quantitative coupling was achieved within 10 minutes using only 1.5 equivalents of glycoamino acid. The method does not require (microwave) heating, thus avoiding side reactions such as acetyl transfer to the N-terminal amino acid. This method also improved the difficult coupling of glycoamino acid to the hydrazine-resin and furnished peptides carrying 10 GalNAc units in high purities (>95%) of crude products. Combined with a one-pot method involving native chemical ligation at a glycoamino acid junction and superfast desulfurization, the method yielded highly pure MUC5AC glycopeptides comprising 10 octapeptide tandem repeats with 20 α-O-linked GalNAc residues within a week.
Fig. 1 (A) Tandem repeat motif (TTSTTSXP) for MUC5AC. (B) Fmoc-Ser/Thr(αAc3GalNAc)-OH building blocks. |
Due to difficulties with the cloning of extended repeated sequences and the microheterogeneity of native glycans, conventional methods based on genetic engineering cannot provide access to mucin TR regions of defined length and glycan composition. Instead, chemical synthesis has been used to provide glyco-defined mucins, most frequently for applications in immunological research.8–10 Many research groups have contributed to the advancement of the synthesis methodology.11 While N-linked glycopeptides can be prepared by connecting elaborated carbohydrates with the full-length peptide, no such approach is feasible for the synthesis of long mucin peptides. Instead, the synthesis of O-glycopeptides relies on preformed glycoamino acids, with Fmoc-Ser/Thr(αAc3GalNAc) as the most frequently used building blocks (Fig. 1B).
Synthesis of O-glycopeptides is challenging. Couplings to and with glycoamino acids are more difficult than with conventional amino acid building blocks. For one thing, reactivity is low due to steric requirements. For another, glycoamino acids run the risk of racemization or β-elimination reactions under forced coupling conditions.12,13
Most of the previous work yielded O-glycopeptides that rarely contained more than 20 amino acids and only a few glycan units, some of which were of impressive complexity.14–25 However, the inherent properties of mucins, such as viscoelasticity, gelation, shielding, and multivalency, emerge from multirepeat TRs.26–28 We, therefore, embarked on a research program focusing on the development of methods for the synthesis of multi TR mucins. In the synthesis of long mucin peptides containing many GalNAc units, the difficulties of O-glycopeptide synthesis become even more apparent. A typical approach to get around difficult steps in solid-phase peptide synthesis is to use the building blocks in large excess and employ double couplings. Even with commercially available Fmoc-Ser/Thr(αAc3GalNAc), the easiest glycoamino acids to synthesize, such an approach is extremely costly if a large number of glycosylated building blocks must be incorporated. Notable contributions to the synthesis of highly GalNAcylated MUC1 and MUC2 TRs, with up to 141 amino acids in length, have been reported.29–32 Middle-sized segments (<40 amino acids) containing up to 7 GalNAc units were synthesized on solid phase either by single amino acid extension or fragment condensation. Purified fragments were subsequently stitched together by a series of silver-assisted thioester ligations29 or Cu-click reactions.30 Recently, Wang31 described three native chemical ligation (NCL) reactions used to convergently assemble a 240 amino acid antifreeze protein from two 30-mers containing none or 10 GalNAc residues. A common theme of the solid-phase synthesis tactics used to access fragments for ligation is that Fmoc-Thr(αAc3GalNAc) building blocks were coupled by using a twofold excess.30–32 Considering the reported difficulties,12 this number of excess equivalents is small. Possible truncations, side reactions, and purity of the products prior to purification have not been reported.
Herein, we describe the development of an efficient and economically viable method for the smooth synthesis of highly O-GalNAcylated mucin multi tandem repeats. We report an optimized multigram synthesis of both Fmoc-Thr(αAc3GalNAc) and Fmoc-Ser(αAc3GalNAc) based on azidophenylselenylation of galactal.33 During the optimization of glycoamino acid coupling, we observed acetyl transfer reactions resulting in the undesired capping of the growing peptide chain. Most surprising were the results from a thorough analysis of coupling conditions, which led to a method enabling quantitative couplings within 10 minutes by using only 1.5 equivalents of glycoamino acid. This method also facilitated the notoriously unreliable loading of the hydrazine resin required for the synthesis of peptide thioesters according to the methods of Liu and Dawson.34,35 Combined with an extremely efficient ligation–desulfurization sequence the new method enabled the smooth synthesis of 10 TR long MUC5AC peptides containing 20 GalNAc residues on Ser or Thr.
For the purpose of an application in the α-selective glycosylation of Fmoc-threonine tert-butyl ester 3t, 2-azido-2-deoxy-phenylselenogalactosyl donor 2 was prepared from triacetylgalactal 1 as previously described (Scheme 1).33,48 Next, this donor was allowed to react with the threonine aglycon at −10 °C under activation with NIS and TMSOTf in a mixture of diethyl ether and dichloromethane. The 2-azido-galactosyl-threonine conjugate 4t was formed as a 2.5:1 mixture of α- and β-isomers (Table 1). Chromatographic purification proved challenging. Therefore, the α/β-mixture was subjected to the zinc-mediated azide reduction in the presence of acetic anhydride furnishing the N-acetylgalactosamine-threonine conjugate 6t, which was easier to purify than 4t. After chromatography, the α-isomer 6t was isolated in 24% yield. The same synthetic route was applied to the glycosylation of Fmoc-serine tert-butyl ester 3s, providing the α-linked serine conjugate 6s in 26% yield.
The relatively moderate yields of the glycosylation, caused by difficult separation from by-products, prompted us to consider an alternative donor. The 2-azido-2-deoxygalactosyl trichloroacetimidate 5 was readily prepared from selenogalactoside 2 in 84% yield in two steps. First, 2 was quantitatively hydrolyzed to hemiacetal S1 (Scheme S1†), followed by the known reaction with trichloroacetonitrile.44,49 Glycosylation of amino acid 3t was induced upon the addition of trimethylsilyl triflate. With a 2.3:1 ratio of α/β-anomers, this reaction provided comparatively the same α-selectivity as the glycosylation with the selenophenyl donor 2 but almost twice as high yields (Table 1). Importantly, the outcome was independent of the reaction scale (100 mg to 9.3 g). To facilitate the purification of 6 after reductive acetylation, it proved beneficial to adjust the donor/acceptor ratios. A 1:1.1 ratio allowed the isolation of Fmoc-Thr(αAc3GalNAc)-OtBu in 47% yield over two steps. For the synthesis of Fmoc-Ser(αAc3GalNAc)-OtBu, the donor/acceptor ratio was adjusted to 1.2:1. The α-selectivity and the yields were slightly higher than for glycosylation of the threonine derivative. Cleavage of the tert-butyl ester was uneventful and furnished the glycoamino acid building blocks 7t and 7s on a multigram scale.
To evaluate various coupling conditions, we prepared a Tentagel R Fmoc-NH-NH-TRT resin from Tentagel R Cl-TRT resin.56 This low loading resin was used because of its known advantages in the synthesis of long and difficult peptides. However, instead of suspending Fmoc-NH-NH2 in DMF/CH2Cl2, we used anhydrous DMSO as a solvent. The Fmoc-load was determined before Fmoc cleavage and after subsequent coupling of the starting amino acid. Conventional double coupling of Fmoc-Ser(OtBu)-OH (10 eq.) under activation with HATU in the presence of DiPEA (condition 1, Table 2) afforded only 49% loading yield. Microwave heating did not remedy the situation (condition 2). To avoid the possible formation of guanidino adducts, PyOxim was used as an activator with little success (condition 3). A higher loading yield was achieved when a coupling with DIC/Oxyma activation was performed overnight (condition 4). In our further attempts, we were inspired by the work of de la Torre/Albericcio57 and the Schönleber/Pedersen groups,58 who used 2-MeTHF as a solvent for coupling reactions promoted by DIC/Oxyma. Replacing DMF with 2-MeTHF dramatically improved reaction rates (conditions 5 and 6). Quantitative yields were obtained after 10 min reaction time.
Coupling conditionsa | Yieldb (%) |
---|---|
a 10 eq. of amino acid and activating reagents were used, 20 eq. of DiPEA was used. In condition 4 1 eq. of DiPEA was added. b Yields calculated according to Fmoc-test. In conditions with 2-MeTHF preactivation (2 min) was used. | |
(1) 2 × HATU/DiPEA, 45 min in DMF, rt | 49 |
(2) 2 × HATU/DiPEA, 30 min each in DMF, 50 °C | 26 |
(3) 3 × PyOxym/DiPEA, 5 min in DMF, 75 °C | 25 |
(4) 1 × DIC/Oxyma/DiPEA, 17 h in DMF, rt | 63 |
(5) 1 × DIC/Oxyma, 5 min in 2-MeTHF, rt | 92 |
(6) 1 × DIC/Oxyma, 10 min in 2-MeTHF, rt | 100 |
Next, we investigated the non-trivial loading of the H2N–NH–TRT resin with Fmoc-Ser(αAc3GalNAc)-OH (Table 3). Given its high cost, coupling with 10 equivalents is not an option. Gratifyingly, with the DIC/Oxyma system in 2-MeTHF, a 75% loading yield was obtained by using only 1 eq. glycoamino acid (condition 1, Table 3). Increasing the coupling time from 2 to 20 hours did not improve the yield (condition 2), but with only 0.5 excess equivalents, >90% yield was achieved (condition 3), which was deemed sufficient in terms of yield and building block economy.
Coupling conditions | Yielda (%) |
---|---|
a Yields calculated by integrating peaks in UPLC traces. Based on four independent replicates performed for condition 10, a 4% standard deviation is estimated. TMP, 2,4,6-trimethylpiperidine. | |
(1) 1.5 eq. 7t/DIC/Oxyma, 5 min, 75 °C | NA |
(2) 1.5 eq. 7t/HATU/HOAt, 3 eq. DiPEA, 30 min, rt | 75 |
(3) 1.5 eq. 7t/HATU/HOAt, 3 eq. DiPEA, 60 min, rt | 78 |
(4) 1.5 eq. 7t/HATU/HOAt, 3 eq. TMP, 30 min, rt | 70 |
(5) 1.5 eq. 7t/HATU/HOAt, 3 eq. TMP, 60 min, rt | 82 |
(6) 1.5 eq. 7t/HATU/HOAt, 3 eq. TMP, 90 min, rt | 82 |
(7) 1.5 eq. 7t/HATU/HOAt, 3 eq. TMP, 30 min, rt | 82 |
(8) 1.5 eq. 7t/HATU/HOAt, 3 eq. TMP, 60 min, rt | 90 |
(9) 1.5 eq. 7t/HATU/HOAt, 3 eq. TMP, 90 min, rt | 92 |
(10) 2.25 eq. 7t/HATU/HOAt, 4.5 eq. TMP, 90 min, rt | 94 |
(11) 2 × [0.95 eq. 7t/HATU/HOAt, 1.9 eq. TMP, 30 min, 40 °C] | 89 |
(12) 2 × [1.15 eq. 7t/HATU/HOAt, 2.3 eq. TMP, 30 min, 40 °C] | 93 |
(13) 2 × [1.3 eq. 7t/HATU/HOAt, 2.6 eq. TMP, 30 min, 40 °C] | 97 |
(14) 2 × [1.5 eq. 7t/HATU/HOAt, 3 eq. TMP, 18 min, rt] | 100 |
In search of (glyco) efficient coupling, we reviewed conditions based on the use of HATU as one of the strongest activation reagents.65 We used DMF as a solvent and varied the base, reaction time, and temperature (Table 4). Again, couplings were carried out at the resin-bound 30-mer peptide 8*, and again, rather than simply determining the Fmoc load, we monitored the reactions by HPLC analysis of crudes obtained after TFA cleavage (Fig. S5†). With only 1.5 equivalents of glycoamino acid/HATU/HOAt, the reaction required 90 minutes to afford a 92% yield (conditions 9). Based on a report from the Gildersleeve lab on the epimerization of glycoamino acids with DIPEA12 and slightly slower couplings (compare conditions 2 and 3 with 4 and 5), we focused on the use of TMP as a base. Increasing the excess from 1.5 to 2.25 equivalents at a 90 minute reaction time provided only a minor improvement in coupling yields (conditions 10). With an aim to reduce coupling time, we investigated mild microwave irradiation (25 W) and increased the temperature to 40 °C (conditions 11–13). At the same time, double coupling (18–30 min each) was performed with a total of 1.9–3 equivalents. At this temperature, we did not observe acetyl transfer. A double coupling for 18 min with 1.5 equivalents each provided a quantitative yield (condition 14). However, the need for 3 equivalents in each glycoamino acid coupling is costly.
Motivated by success with difficult couplings to the hydrazine resin, we explored reactions in 2-MeTHF with activation by the DIC/Oxyma system (Table 5). In our first attempt, a double coupling with substoichiometric amounts of glycoamino acid was carried out with little success. Remarkably, a single coupling with 1.5 equivalents of glycoamino acid in 2-MeTHF for 10 min proved sufficient to give a quantitative yield. No heating was required, and there was no evidence of acetyl transfer (Fig. S6†).
Coupling conditions | Yielda (%) |
---|---|
a Yields calculated by integrating peaks in UPLC chromatograms. For condition 1 double coupling was performed. | |
(1) 2 × [0.78 eq. 7t/DIC, 10 eq. Oxyma in 2-Me-THF, 18 min, rt] | 56 |
(2) 1.09 eq. 7t/DIC/Oxyma in 2-MeTHF, 10 min, rt | 94 |
(3) 1.5 eq. 7t/DIC/Oxyma in 2-MeTHF, 10 min, rt | 100 |
The highly efficient coupling of Fmoc-Ser/Thr(αAc3GalNAc) in 2-Me-THF is noteworthy. Research by de la Torre and Albericcio has shown that coupling reactions with DIC/Oxyma activation proceed remarkably swiftly in 2-MeTHF, whereas DMF seemed to be a better solvent for activation with HATU or HBTU.57 There is an increasing amount of evidence to suggest that the rate of coupling reactions increases with decreased polarity of the solvent.58,66 We assume that the high reactivity in apolar solvents can counterbalance the reduction of reaction rate caused by low concentration of the building block. Hence, apolar solvents such as 2-MeTHF could prove particularly advantageous when expenses for building blocks are high.
Scheme 3 Solid-phase synthesis of MUC5AC glycopeptides and their yields after isolation in parentheses. gS, gT = Ser/Thr(αGalNAc). Conditions: Fmoc-cleavage: 20% piperidine in DMF; coupling: HATU/Oxyma/DiPEA (5 eq., 5 eq., 15 eq.); coupling of 7t or 7s: see condition 3, Table 5; capping: Ac2O/DiPEA/DMF (70:20:10, v/v/v); TFA cleavage: for 11 and 12 TFA/TIS/H2O (96/2/2). For 13 TFA/TIS/EDT/H2O (94/2/2/2). |
MS analysis of the glycopeptides showed peaks with m/z values (−200 Da and −400 Da) corresponding to a loss of one or two GalNAc residues (Fig. 3D–F). This loss did not occur when glycopeptides were detached in O-acetylated form (Fig. S8†). Hydrolytic cleavage of sugar residues could occur during detachment with TFA. We therefore prepared glycopeptide 12 with O-acetyl groups on the GalNAc residues (which is known to increase their hydrolytic stability) and performed the O-deacetylation in solution. The mass spectra did not change (Fig. S9†). Therefore, we attributed the loss of GalNAc-residues to the conditions of the ESI-MS measurement. In the absence of readily ionizable side chains, protonation may occur also at the sugar, followed by possible cleavage of the glycosidic bond. Most importantly, this fragmentation reaction can be prevented by altering the conditions of ESI-MS (vide infra).
The mixtures of the NCL reactions were forwarded for desulfurization. We applied a very recent method from the Li group, which has been reported to enable almost instantaneous desulfurization upon reaction with tetraethylborate.67 A slight shift of retention to shorter times (compare Fig. 4A with 4B), consistent with the loss of a hydrophobic mercapto group, indicated that the reaction was completed after 1 minute. No side reactions were observed in the process of desulfurization. Of note, the apparent purity of crude products 14 and 15, both of which carry 20 GalNAc residues on 10 MUC5AC tandem repeats, was higher than 95%. After preparative HPLC and lyophilization, the glycopeptides 14 and 15 were obtained in 55% and 36% yield, respectively. To prevent fragmentation during high-resolution ESI-MS, it proved necessary to reduce the concentration of the collision gas (Fig. 4D).
(1) We introduced a facilitated, more reliable, and easily scalable synthesis of Fmoc-Thr/Ser(αAc3GalNAc)-OH, in which the potentially hazardous azidonitration of tri-O-acetylgalactal was replaced with azidophenylselenylation.
(2) For the challenging loading of hydrazine resin with a glycoamino acid coupling was performed in 2-MeTHF with DIC/Oxyma activation, which reliably provided higher yields than coupling in DMF.
(3) This approach also eliminated the third and probably most limiting bottleneck, namely the need for large amounts of expensive glycoamino acids in coupling reactions. By using 2-MeTHF as a solvent with DIC/Oxyma activation agents, only 1.5 equivalents of Fmoc-Thr/Ser(αAc3GalNAc)-OH were sufficient to achieve quantitative coupling yields within 10 minutes of reaction time. There was no need in time-consuming double coupling procedures or potentially harmful microwave heating. Without microwave heating users of “traditional” synthesizers can perform glyco-economic yet speedy syntheses of O-glycopeptides. In addition, the absence of microwave heating avoids side reactions, such as the acetyl transfer reaction we observed. Remarkably high purities of crude products of 90–95% show that the new method allows very smooth solid-phase synthesis of mucin peptides encompassing up to 5 MUC5AC tandem repeats and carrying 10 GalNAc units.
(4) Native chemical ligation between two highly glycosylated peptides proceeded rapidly and cleanly also at the glycoamino acid junction.
With the presented method, solid-phase synthesis and purification of a peptide 40-mer with α-O-linked GalNAc residues on 25% of all amino acids takes 4 days. Within 2 days, subsequent solution steps, including thioester preparation, ligation, and desulfurization, are accomplished. We believe that these advances will further mucin research by facilitating access to a greater variety of GalNAcylated peptides in a short time.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3sc05006h |
This journal is © The Royal Society of Chemistry 2024 |