Strategies for the highly efficient synthesis of erythropoietin N‐glycopeptide hydrazides

A convergent synthesis for erythropoietin (EPO) 1‐28 N‐glycopeptide hydrazides was developed. In this approach, EPO 1‐28 peptides were synthesized on the solid phase and converted to C‐terminal hydrazides after cleavage from the resin. After selective deprotection of the Asp24 side chain, the desired glycosylamine was coupled by pseudoproline‐assisted Lansbury aspartylation. Although the initial yields of the EPO 1‐28 glycopeptides were satisfactory, they could be markedly improved by increasing the purity of the peptide using a reversed‐phase high‐performance liquid chromatography (RP‐HPLC) purification of the protected peptide.


| INTRODUCTION
Erythropoietin (EPO) is a cytokine needed for the homeostasis of erythrocytes. Therapeutic EPO is expressed recombinantly in Chinese hamster ovary (CHO) cells and used mainly to treat anemic patients suffering from renal failure or cancer. 1 Owing to its high therapeutic relevance, human EPO is one of the best studied glycoproteins. The biological activity of EPO is modified by the sugar part 2 ; however, the inherent microheterogeneity of glycoproteins at each glycosylation site precludes the availability of pure glycoforms for detailed structure-activity studies. The presence of three N-glycans is crucial for the biological activity of recombinant EPO, because nonglycosylated EPO expressed in Escherichia coli is suffering from low stability and a short serum half-life. [3][4][5][6] Thus, a number of solubilityenhancing mutations need to be introduced when expressing nonglycosylated EPO. 7 Additionally, the presence of a single glycan can significantly stabilize EPO against aggregation. 8 Single pure glycoforms of glycoproteins are currently available only by synthetic methodology. 9 Despite its complexity (three Nglycans and one O-glycan), a number of EPO glycoforms and variants were successfully synthesized, providing the protein with natural or surrogate linkages for the glycans. 8,[10][11][12][13][14][15][16][17][18][19] Synthetic approaches for EPO based on native chemical ligation require the synthesis of several glycopeptide building blocks with a length of up to 40 amino acids.
The convergent synthesis of N-glycopeptides following the aspartylation method developed by Lansbury 20 involves the coupling of a glycosylamine to an aspartyl side chain of a protected peptide.
Activation of the aspartyl side chain carboxylate by coupling reagents can lead to an extensive formation of aspartimide and other byproducts. These side reactions can be significantly reduced by using a pseudoproline-assisted Lansbury aspartylation either in solution 21 or on the solid phase. 22 This methodology takes advantage of the Asn-X-Ser/Thr sequon for N-glycosylation. It was found that within this sequon, the incorporation of an n+2 Ser/Thr pseudoproline efficiently reduces the formation of aspartimide both during the synthesis of the peptide and in the subsequent coupling to the glycan. 22 Even though pseudoprolines are widely used in peptide synthesis, a mechanistic explanation of how pseudoprolines efficiently reduce the formation of aspartimides remains to be established.  Towards this endeavor, EPO 1-28 glycopeptides were found to be versatile building blocks serving for the synthesis of EPO variants with one N-glycan (N24, see Figure 1). Herein, we present a systematic investigation of the large-scale convergent synthesis of EPO 1-28 glycopeptide hydrazides 25 suitable for native chemical ligation.
We found that the efficient synthesis of EPO 1-28 glycopeptides requires a high purity of both the glycosylamine and the selectively deprotected peptide. A key feature of this approach was the establishment of conditions allowing the prepurification of the protected EPO 1-28 hydrazide building blocks using nonaqueous reversedphase high-performance liquid chromatography (RP-HPLC). 26

| RESULTS AND DISCUSSION
As shown before, 22 EPO 1-28 glycopeptides can be synthesized on the solid phase in a straightforward manner. The aspartate of the Nglycosylation site 24 was protected with an allyl ester. 22 After mild cleavage from the resin, the 1-28 glycopeptides were converted to a thioester. 22,27 The glycopeptide thioester containing GlcNAc gave an overall yield of 37%. In contrast, the yields of the corresponding glycopeptide thioester containing a biantennary N-glycan nonasaccharide reached only 24% owing to retention of the glycopeptide on the solid phase. 22 This led to a redesigned strategy where the peptide carrying a C-terminal hydrazide serving as a latent thioester is synthesized first, and after Asp side chain deprotection, the desired glycan can be coupled in solution. 28 To avoid residual palladium species in the peptide after Pd-catalyzed deallylation 22,29 in solution Asp24 was protected with a phenylisopropyl ester. 30 Cysteine7 was protected as a mixed disulfide, reducing the risk of oxidation after global deprotection of the peptide.
The EPO 1-28 peptide was assembled on Fmoc-Gly-trityl-ChemMatrix resin (1) 31 incorporating two pseudoproline dipeptides (see Figure 2). After mild cleavage from the resin (2) barely exceeded the overall yield obtained after the aspartylation of EPO 1-28 on the solid phase (24% for the biantennary N-glycan nonasaccharide vide supra). We suspected that impurities were responsible for the moderate yield of glycopeptide 8. Because glycosylamine 7 was prepared from the corresponding glycosyl azide 6, which was purified by HPLC, 28 we focused on increasing the purity of the selectively deprotected peptide 5.
However, the solubility of the selectively deprotected aspartyl peptide 5 in MeOH was low, whereas the fully protected precursor 4 showed satisfactory solubility. In CH 3 CN, both peptides 4 and 5 were poorly soluble. Thus, preparative purification conditions for 4 were initially investigated on a C8 column using MeOH/water as F I G U R E 2 Synthesis of erythropoietin (EPO) 1-28 glycopeptide hydrazide 8. The 1-28 peptide was assembled by Fmoc-solid-phase synthesis on a trityl-ChemMatrix resin, cleaved, and converted to the hydrazide 4. Selective cleavage of the phenylisopropyl ester rendered the aspartyl peptide 5, which was coupled with the complex-type glycosyl amine 7. After global deprotection, the glycopeptide hydrazide 8 was obtained. On the right side, the high-performance liquid chromatography (HPLC) traces from the HPLC-MS analysis of the peptides are shown an eluent. With the use of an isocratic elution (95% MeOH/water), peptide 4 could be separated from a number of overlapping peaks with shorter retention time.
Owing to the broad shape of most of the peaks and low reproducibility, a nonaqueous 26 mixture of organic solvents (70% CH 3 CN/MeOH isocratic) was tested for the purification, which separated 4 reliably and additionally provided 11 minor fractions ( Figure 3C).
Because the masses of the protected peptides in the minor fractions could not be determined by LC-MS directly, the fractions were separately deprotected and then analyzed. HPLC-MS revealed that all the peaks eluting prior to peptide 4 corresponded to truncated peptides. Only the two fractions with longer retention times gave a higher mass than the target peptide 4d (Data S10b). The peptide 4 purified by HPLC resulted in a significantly increased purity of 4d where only traces of the previously observed side products were visible ( Figure 3E). Despite a good separation of the side products on a 10-mg scale using a 2 × 25-cm C8 column, the purification of amounts over 15 mg decreased the resolution significantly. Thus, another stationary phase was tested. Gratifyingly, a polystyrene-based HPLC column using a nonaqueous gradient system (0-40% CH 3 CN/MeOH) also provided 4 in similar purity as before ( Figure 3E). In a single run, over 30 mg of 4 could be purified over a 2.5 × 30 cm Nucleogel column, removing many side products and contaminants mainly with shorter retention times.
Subsequently, the phenylisopropyl ester of peptide 4 purified by HPLC was selectively removed by 1% TFA/DCM, and the resulting peptide 5 was aspartylated using glycosyl amine 7. To our delight, the final yield of glycopeptide 8 was raised to 60%, thus virtually doubling the yield in this step. It can be assumed that the numerous side products previously present in crude 5 (no HPLC) were also converted to glycopeptides and other products, which needed to be removed in the final HPLC step, thus lowering the coupling yield considerably.
We next attempted to also increase the purity of 5 by HPLC and was subsequently deprotected with a TFA cocktail. After workup and purification, nearly 5 mg of the EPO 1-28 glycopeptide 11 containing a galactosylated tetraantennary N-glycan was isolated, corresponding to a yield of 65%.

| CONCLUSION
In summary, the yields of EPO 1-28 glycopeptide hydrazides by convergent synthesis were found to depend strongly on the purity of the selectively deprotected aspartyl peptide. Conditions were established to monitor the synthesis and also purify the peptide hydrazides by nonstandard conditions (nonaqueous reversed-phase chromatography). With the protected peptides purified by HPLC, the pseudoproline-assisted Lansbury aspartylations proceeded in high yields and thus permitted the rapid and efficient derivatization of a single peptide with the desired glycosyl amines. Because the solubility of protected peptides is affected by the sequence and the protecting groups, purification schemes need to be established for each peptide individually.