Chemical and Enzymatic Synthesis of Sialylated Glycoforms of Human Erythropoietin.

Recombinant human erythropoietin (EPO) is the main therapeutic glycoprotein for the treatment of anemia in cancer and kidney patients. The in-vivo activity of EPO is carbohydrate-dependent with the number of sialic acid residues regulating its circulatory half-life. EPO carries three N-glycans and thus obtaining pure glycoforms provides a major challenge. We have developed a robust and reproducible chemoenzymatic approach to glycoforms of EPO with and without sialic acids. EPO was assembled by sequential native chemical ligation of two peptide and three glycopeptide segments. The glycopeptides were obtained by pseudoproline-assisted Lansbury aspartylation. Enzymatic introduction of the sialic acids was readily accomplished at the level of the glycopeptide segments but even more efficiently on the refolded glycoprotein. Biological recognition of the synthetic EPOs was shown by formation of 1:1 complexes with recombinant EPO receptor.

Introduction

The hematopoietic cytokine erythropoietin (EPO) is a glycoprotein hormone with a central role in erythrocyte homeostasis. Under hypoxic conditions the expression of EPO in renal and hepatic cells is upregulated. Secreted EPO binds to the EPO receptor on progenitor cells in the bone marrow and stimulates their differentiation into erythrocytes. Seminal studies of the 166 amino acid glycoprotein EPO revealed that its in vivo bioactivity strongly depends on the presence of multiantennary sialylated N‐glycans on Asn residues 24, 38 and 83. The typical N‐glycans on EPO are tetraantennary and core fucosylated. The O‐glycan on Ser‐126 is not essential for activity but contributes to serum half‐life. Enzymatic desialylation of EPO resulted in increased binding to the EPO receptor, however, the in vivo activity was highly reduced.

Due to the heterogeneity of the three N‐glycans the separation of recombinant EPO into pure glycoforms is not feasible even with the most advanced methods. Thus, synthetic approaches for analogs of EPO and homogeneously glycosylated EPO were established. The use of sequential native chemical ligation (NCL) has provided pure glycoforms of EPO with excellent control of site specific glycosylation. Sialylated N‐glycans can be installed via glycosyl‐asparagine derivatives bearing protected sialic acids, by Lansbury aspartylation or by enzymatic modification. Here, we describe the chemical and enzymatic synthesis of glycoforms of EPO with terminal galactose or sialic acid using a set of five building blocks. Based on a generally applicable protecting group strategy the syntheses are robust and reproducible. A straight‐forward alternative to the multi‐step enzymatic synthesis and ligation of sialylated glycopeptides was the enzymatic sialylation of refolded EPO in a single step.

We intended to establish an approach to sialylated EPO by NCL using enzymatically sialylated glycopeptide segments. The introduction of the sialic acids at a late stage of the synthesis was expected to facilitate access to the biologically relevant sialylated glycoforms. A prerequisite for the enzymatic sialylations of EPO glycopeptides was a stable C‐terminal functionalization resistant to hydrolysis. Thus, the use of glycopeptide thioesters was not considered and highly stable hydrazides suitable for sequential NCL were chosen instead. Due to the lack of appropriately placed cysteines in EPO artificial ligation sites at alanine 68, 98 and 128 were introduced leading to five segments with a length of about forty amino acids (Scheme 1). This approach requires a desulfurization step converting the three non‐natural cysteines to native alanines after the ligations. A key to this endeavour was to develop a reliable protecting group strategy for the native cysteines compatible with the synthesis of the sialylated glycopeptides, the sequential ligations and the desulfurization step.

Scheme 1

Retrosynthesis of human EPO 1–166 with three asialo N‐glycans leading to fragments 1–5.

Results and Discussion

Our goal was to first synthesize an EPO 1–166 glycoprotein bearing three biantennary N‐glycans terminating with galactose via the fragments 1–5 and to subsequently extend the approach towards sialylated EPO glycoforms. This initial strategy was based on Acm protection of the native cysteines 33, 161, and temporary protection of the N‐terminal cysteines in fragments 2 and 4 by thiazolidines (Scheme 1). The synthesis of the three glycopeptide hydrazides 1–3 was envisioned by convergent pseudoproline‐assisted Lansbury aspartylation in solution using allyl esters at the glycosylation sites. An exploratory attempt to assemble the demanding EPO 29–97 sequence on the solid phase by fragment condensation and subsequently attach N‐glycans to Asp 38 and 83 was feasible for GlcNAcNH2 but not for larger N‐glycan amines (data not shown). Thus, the idea of a four‐segment approach to EPO was abandoned and the synthesis was continued with the more readily accessible glycopeptide segments 2 and 3.

The EPO 29–67 peptide was obtained by Fmoc‐SPPS on a trityl‐ChemMatrix‐resin (Trt‐CM) using pseudoproline dipeptides and DMB‐glycine to reduce aggregation during synthesis. To avoid sulfoxidation Met‐54 was replaced by norleucine. The peptide was cleaved from the resin and converted in situ to hydrazide 6 (Scheme 2 a). In the course of Pd0‐catalyzed deallylation of 6 a considerable loss of the Acm group was observed (7:8=3:1). Thus, a Phacm group was installed at Cys‐33, which remained stable during deallylation and provided the desired 29–67 glycopeptide hydrazide 9 (Scheme 2 b). However, in the following conversion of hydrazide 9 to thioester 10 the N‐terminal thiazolidine gave rise to the stable side product 11 with a higher mass (M+29, presumably N‐nitroso) and resistance to ring opening with methoxyamine[ , , ] (data not shown). Hence, Cys‐29 and Cys‐33 in EPO 29–67 were protected with a Phacm group and after three sequential ligations using fragments 3, 4 and 5 the desulfurized EPO 29–166 glycopeptide 12 could be obtained. At this stage the removal of the Phacm and Acm groups using AgI‐ or HgII‐salts, 2,2′‐dithiobis(5‐nitropyridine) or penicillin G acylase was very troublesome and suffered from incomplete conversion and low recovery (data not shown).

Scheme 2

Drawbacks of the initial protecting group strategies for native cysteines: a) loss of Acm during Pd0‐catalyzed deallylation; b) N‐terminal Thz forming a stable nitroso derivative under diazotization conditions; c) difficult removal of Phacm at a late stage using various deprotection methods; d) low overall yield for S‐tritylation and low solubility in following transformations.

To overcome the difficulties associated with Acm‐type protecting groups in the 29–67 glycopeptide S‐trityl protecting groups were selectively reinstalled at both cysteines after global deprotection of the glycopeptide 13. However, the two proximal trityl groups of segment 14 led to low solubility and difficult purifications even after the following ligation (data not shown).

Since all the initially tested protecting group combinations exhibited serious drawbacks, we finally returned to using the Acm protecting group in the EPO glycopeptide 29–67 but replaced the allylester in Asp‐38 by an acid‐labile phenylisopropyl ester (PhiPr), thus avoiding the use of Pd0 in the presence of Acm groups. The EPO 29–67 peptide was synthesized on Trt‐CM resin, cleaved with 20 % HFIP in CH2Cl2 and equipped with a C‐terminal hydrazide (Scheme 3). After purification by flash chromatography the PhiPr ester was removed with 1 % TFA/CH2Cl2. The deprotection was monitored by RP‐HPLC to minimize cleavage of other acid‐sensitive protecting groups. After pseudoproline‐assisted aspartylation of the resulting EPO 29–67 aspartyl peptide (S4) with N‐glycan 16 the desired glycopeptide hydrazide 20 was obtained in 43 % yield.

Scheme 3

Solid phase synthesis and functionalization of the five EPO segments. The glycopeptide hydrazides 17, 20 and 23 were obtained using pseudoproline‐assisted Lansbury aspartylation.

Glycopeptide hydrazide 17 (EPO 1–28) was synthesized very efficiently from the 1–28 peptide hydrazide 15 bearing a phenylisopropyl ester at the glycosylation site (Scheme 3). The key to the high yield was the amenability of the selectively deprotected aspartyl peptide to purification by non‐aqueous HPLC.

The synthesis of glycopeptide hydrazide EPO 68–97 was also examined using a PhiPr ester at Asp 83. However, cleavage of the PhiPr moiety of the 68–97 hydrazide using 1 % TFA in CH2Cl2 also opened one of the two pseudoprolines (71 or 85) according to RP‐HPLC‐MS. The extensive aspartimide formation during the subsequent aspartylation with 16 suggested that the acetonide at Ser 85 was mainly affected. Since less acidic conditions did not provide fully orthogonal cleavage of the PhiPr ester in this segment, the EPO 68–97 hydrazide 22 was assembled with an allylester at Asp 83. A modified deallylation of 22 was carried out using immobilized PdII acetate in combination with the water‐soluble phosphorus ligand DCHT, which improved the workup and the purity of the aspartyl peptide S8. Subsequent attachment of the glycan 16 provided the glycopeptide hydrazide 23 in 37 % yield. The glycopeptide hydrazides 17 and 20 were readily converted to the corresponding thioesters 18 and 21 by first transforming the hydrazide to an acylazide (NaNO2, pH 3, −15 °C) followed by thiolysis with MESNa at near‐neutral pH. The C‐terminal segments EPO 98–127 thioester 4 (42 % yield) and the Cys‐peptide EPO 128–166 5 (58 % yield) were obtained by Fmoc‐SPPS.

With all five EPO segments in hand the ligations and the following transformations were investigated (Scheme 4). The NCL of thioester 4 and Cys‐peptide 5 catalyzed by MPAA and the subsequent deprotection of thiazolidine‐98 were performed in a one‐pot approach furnishing the EPO 98–166 peptide 26 in 89 % yield. Glycopeptide hydrazide 23 was ligated with equimolar amounts of EPO 29–67 glycopeptide thioester 21 giving EPO 29–97 glycopeptide hydrazide 25 in 72 % yield after purification. Conversion of the glycopeptide hydrazide 25 to the corresponding thioester S14 required strict pH‐control, to avoid the formation of an unreactive lactam at the C‐terminal lysine. The lactamization can be avoided efficiently by use of side chain protecting groups or by lowering the pH during thiolysis of the acyl azide. Cys‐peptide 26 was ligated with the glycopeptide thioester S14 furnishing the EPO glycopeptide 29–166 27 in high yield (73 %). A simultaneous radical desulfurization of the three thiol groups using VA‐044/TCEP proceeded best with glutathione as a thiol additive and gave quantitative conversion to 28 within 3 h. The reaction progress was monitored by RP‐HPLC‐MS since the partially desulfurized species showed no difference in retention time on HPLC.

Scheme 4

a) Synthesis of the asialoglycoform EPO A bearing three N‐glycans by sequential native chemical ligation, desulfurization, Pd‐mediated cleavage of the Acm groups and oxidative refolding; b) RP‐HPLC‐MS of the final ligation of glycopeptide thioester 18 and glycopeptide 29 to full‐length EPO 1–166 30 after 6 d, insert shows SDS‐PAGE of crude ligation mixture; c) SEC after oxidative refolding of ligation mixture; d) RP‐HPLC‐MS of EPO A after RP‐HPLC purification; e) HR‐MS of purified EPO A (H2O, direct injection); f) simulated and measured isotope pattern of the [M+15 H]15+ HR‐MS peak; g) SDS‐PAGE (reduced and oxidized) of EPO A.

At this point the cleavage of the Acm groups of 28 was investigated. With respect to the experiences obtained from the inital syntheses we considered the use of aqueous PdII[35] to remove the three Acm groups as an alternative to commonly applied AgI. To our delight the Acm groups were rapidly cleaved by PdCl2 in water. However, due to the poor solubility of PdCl2 in water the results were not consistent and the deprotected glycopeptide retained a brownish color after purification by HPLC. By adding NaCl to the suspension of PdCl2 in water the well‐soluble tetrachloropalladate complex formed leading to quantitative cleavage of the Acm groups with stoichiometric amounts of PdII. Subsequently, the addition of a large excess of DTT was required to fully remove bound Pd from the thiol groups of the deprotected glycopeptide as evidenced by the formation of colored, soluble Pd‐DTT‐chelates. The use of other scavengers for PdII (dimethylglyoxime or 3‐mercaptopropionic acid ) was less efficient. Although the relative amount of PdII had to be increased to 30 equivalents when carrying out the deprotection of 28 in buffered guanidinium chloride, the subsequent removal of bound Pd with excess DTT proceeded smoothly under these conditions. Residual palladium led to poor ionization of the deprotected glycopeptide 29 under LC‐MS conditions (Figure S31‐Pd). After purification by RP‐HPLC the desired EPO 29–166 glycopeptide 29 was obtained as a white solid in 83 % yield. The buffered deprotection conditions avoid the acidic pH of aqueous solutions of palladium chloride and should thus also be compatible with acid‐sensitive sialylated EPO glycopeptides.

The EPO 29–166 glycopeptide 29 was subsequently ligated with the EPO 1–28 glycopeptide thioester 18. After 6 d of reaction time high conversion to the full‐length glycopeptide 30 was observed, which was directly subjected to refolding by dialysis of the ligation mixture.[ , ] Arginine was added to suppress the aggregation of folding intermediates. Monomeric folded EPO A was separated from misfolded and oligomeric EPO species by size exclusion chromatography (SEC) and purified further by RP‐HPLC. The purity of the refolded EPO A was confirmed by SDS‐PAGE, HPLC‐MS and high resolution ESI‐MS. The gaussian charge state distribution in the mass spectrum indicates conformational homogeneity of the sample. Moreover, the CD‐spectrum of EPO A was nearly identical to that reported for recombinant EPO. The residual Pd‐content of the synthetic EPO A was determined by inductively coupled plasma optical emission spectrometry (ICP‐OES) and remained below the detection limit of 0.01 ppm.

We next investigated the enzymatic sialylation of the various glycopeptide hydrazides used in the synthesis of EPO A (Scheme 5). The α‐2,6‐sialyltransferase from Photobacterium damsela (ST6) readily catalyzed the transfer of sialic acid to the terminal galactose residues of the biantennary N‐glycopeptide hydrazides 17, 20 and 23 in concentrations ranging from 0.7–5 mM. Alkaline phosphatase (CIAP) was added to reduce product inhibition and sialidase activity of the bacterial enzyme. After repeated addition of CMP‐Neu5Ac the conversions to the disialylated derivatives typically exceeded 90 %. To prevent loss of the acid‐sensitive 2,6‐sialosides in the course of preparative HPLC, TFA was replaced by formic acid as an additive and the eluates were immediately neutralized with dilute Na2CO3. After a final desalting step, the sialoglycopeptide hydrazides were obtained in yields of 67 % (31), 45 % (32) and 82 % (33).

Scheme 5

The efficiency of the enzymatic 2,6‐sialylation of EPO glycopeptides decreases with increasing chain length: a) sialylation of 30–40mer glycopeptide hydrazides 17, 20, 23 and RP‐HPLC‐MS of crude mixtures; b) sialylation of EPO 29–97 glycopeptide 25 and RP‐HPLC‐MS of crude mixtures; c) sialylation of EPO 29–166 glycopeptide 28; Total Ion Chromatogram (TIC) of reaction mixture RP‐HPLC‐HRMS after 31 h; RP‐HPLC‐HRMS of reaction mixtures after different reaction times (only 11+ charged peaks of MS are shown).

Despite the presence of the same biantennary N‐glycan the reactivity (rate of conversion) and the overall reaction times of the three glycopeptide acceptors 17, 20 and 23 (28–39 amino acids) varied strongly. All glycopeptides were well‐soluble in water even at a concentration of 5 mM. EPO 1–28 glycopeptide 17 already reached nearly full conversion after 3 h whereas EPO 68–97 (23) required 16 h of reaction time. Remarkably, EPO 29–67 (20, 5 mM) gave only 52 % conversion to 32 after 16 h. By varying the reaction parameters (see Table S7) it was found that at lower concentrations (0.5–1 mM 20) the sialylation proceeded significantly better leaving less than 10 % of incompletely sialylated compounds.

It can be assumed that the reactivity of the N‐glycans in the three glycopeptide segments 17, 20 and 23 is affected by several factors. The solubility in neat water and buffers is good for all three fragments. Additionally, their hydrophilicity is quite similar indicated by a close range of retention times on an RP‐18 HPLC‐column (see Scheme S39a). For more insight into the secondary structure of the glycopeptides in solution we recorded CD spectra in dilute sialylation buffer (50 μM glycopeptide in 5 mM Tris, pH 9) (Scheme S39a). The two segments 17 and 23 displaying the highest reactivity of the N‐glycan showed CD spectra with defined peaks suggesting a random coil. Glycopeptide 20 gave a similar CD spectrum, albeit less defined and with multiple small peaks, which might reflect conformational deviations from the typical random coil of segments 17 and 23. Thus, CD spectra can provide a starting point to investigate the different reactivities of the three glycopeptides with respect to the peptide part. Most likely the local conformation of the peptide regions flanking the N‐glycans affect the accessibility by the enzyme. We assume that the secondary structures found in EPO should also be present in the corresponding segments to a certain extent. In the most reactive segment 17 the N‐glycan (Asn 24) is located at the end of the complete N‐terminal helix of the protein. In the less reactive segment 23 the N‐glycan (Asn 83) is placed between two incomplete α‐helices connected by a short loop (7 aa). In the least reactive segment 20 the N‐glycan (Asn 38) is within an extended loop region (19 aa) and directly flanked by a short β‐sheet. We assume that an unfavorable combination of local conformation and aggregation in the reaction buffer lowers the reactivity of 20. Revealing the explicit mechanisms, which govern the different reactivities of the EPO glycopeptides 17, 20 and 23 during enzymatic sialylation requires an in‐depth study and is beyond the scope of this paper.

The enzymatic sialylation was also tested with longer glycopeptides bearing two N‐glycans. Multiple sialic acid residues were transferred onto the well‐soluble glycopeptides EPO 29–97 25 and EPO 29–166 28, but despite repeated additions of excess CMP‐Neu5Ac the conversions were generally sluggish and incomplete. In the case of hydrazide 25 the large excess of CMP‐Neu5Ac led to an oversialylation of the glycopeptide (36+Sia, Scheme 5 b), which presumably occurred at the hydrazide function. Despite the length of glycopeptide 25 the different sialylation products were chromatographically resolved (RP‐HPLC‐MS). We assumed that the free thiol group of 25 might interfere with the reaction progress and thus tested the desulfurized 29–166 glycopeptide 28. However, the starting material 28 and the various sialylation products eluted in a single peak during LC‐MS analysis. Thus, the reaction progress was estimated from the relative intensity of the MS peaks of the various sialylation products after HPLC‐HRMS analysis (Scheme 5 c). The reactivity of 28 was marginally lower compared to 25.

Since the complete enzymatic sialylation of longer EPO glycopeptides with multiple glycans was not promising, the more readily accessible 2,6‐sialylated glycopeptides 31, 32 and 33 were used for assembling a sialylated EPO glycoform by sequential ligations following the procedures established for EPO A (Scheme 6). The glycopeptide hydrazides 31 and 32 were converted to the corresponding thioesters 34 and 35. Diazotization of the glycopeptides at pH 3 (−15 °C) did not affect the sialosides. The ligation of the sialoglycopeptide thioester 35 with glycopeptide 33 proceeded smoothly (65 % yield) and the resulting hydrazide 36 was efficiently converted to the corresponding thioester S15 maintaining pH 5.9 in the thiolysis step to avoid lactamization of the C‐terminal lysine. Despite a good conversion in the following ligation with 26 the 29–166 glycopeptide 38 was difficult to purify owing to a similar retention of 38 and the hydrolysis product of thioester 35. The subsequent desulfurization step and the removal of the Acm groups with PdII gave lower yields compared to the non‐sialylated compounds. This trend continued in the final ligation where the conversion to the sialylated full‐length EPO 1–166 glycopeptide 39 was lower than for 30. As a consequence, the oxidative refolding of the ligation mixture resulted in a higher proportion of oligomeric EPO 29–166 relative to oligomeric EPO 1–166 in the purification of EPO S by size exclusion chromatography (Scheme 6 b, c). After preparative RP‐HPLC the desired sialylated glycoform EPO S was obtained in 37 % yield. EPO S was characterized by SDS‐PAGE, HPLC‐MS, HR‐ESI‐MS and CD‐spectroscopy (Scheme 6 d–h). The purity and the CD spectrum of EPO S were nearly identical to that of EPO A indicating that the synthetic approach to EPO is robust and applicable to non‐sialylated as well as sialylated glycoforms.

Scheme 6

a) Synthesis of glycoform EPO S from sialylated glycopeptides; b) RP‐HPLC‐MS of the ligation of sialylated glycopeptide thioester 34 and Cys‐glycopeptide 29–166 S 16 to full‐length EPO 1–166 39 after 6 d, c) SEC after oxidative refolding; d) RP‐HPLC‐MS of EPO S after RP‐HPLC purification; e) HR‐MS of purified EPO S (H2O, direct injection); f) simulated and measured isotope pattern of [M+15 H]15+ HR‐MS peak; g) SDS‐PAGE (reduced and oxidized) of EPO S. h) CD‐spectra of EPO A and EPO S.

Finally, we also attempted an enzymatic 2,6‐sialylation of the synthetic glycoform EPO A (Scheme 7). In contrast to the unfavorable acceptor properties of the longer glycopeptides 25 and 28, the folded glycoprotein EPO A showed fast conversion. At an acceptor concentration of 0.5 mM corresponding to 11.5 mg of EPO A mL−1 the transfer of six sialic acid units could be driven to completion by multiple additions of CMP‐Neu5Ac. The conversion could only be followed by MS (Scheme 7 b) as all sialylation intermediates of EPO coeluted on RP‐HPLC (Scheme 7 b, TIC of LC‐MS). After purification by RP‐HPLC the glycoform EPO S was obtained in good yield and high purity. The glycoprotein was virtually identical to EPO S synthesized independently from the sialylated glycopeptides. We observed the same low degree of desialylation and other fragmentations of the N‐glycans during HRMS analysis of EPO S made from sialylated glycopeptides and EPO S obtained by enzymatic sialylation of EPO A (see Schemes 6 e, 7 d). This indicates that the minor MS‐peaks corresponding to EPO S with loss of Neu5Ac, GalSia and LacNAcSia mainly originate from fragmentation during HRMS analysis rather than incomplete enzymatic sialylation of EPO A.

Scheme 7

a) The enzymatic sialylation of EPO A yields the sialylated glycoform EPO S directly; b) RP‐HPLC‐HRMS of the enzymatic sialylation of EPO A after 22 h, sialylation intermediates coelute with product EPO S: TIC of LC‐MS (1.5 h) and MS after different reaction times are shown; c) RP‐HPLC‐MS of EPO S after RP‐HPLC purification; d) HR‐MS of purified EPO S (H2O, direct injection); e) simulated and measured isotope pattern of [M+15 H]15+ HR‐MS peak; f) SDS‐PAGE of EPO S (reduced and oxidized).

The unexpectedly high reactivity of the bacterial 2,6‐sialyltransferase towards the folded EPO A glycoprotein provides rapid access to the sialylated glycoform EPO S in a single step. This straight‐forward transformation bypasses the need to synthesize the three sialoglycopeptide building blocks 33–35 and carry them through a time‐consuming process involving multiple ligations, peptide modifications and protein refolding.

EPO A and the unfolded glycopeptides 25 and 28 are well‐soluble in aqueous buffers but show different reactivities with the bacterial sialyltransferase. We assume that the faster enzymatic sialylation of properly folded EPO A should mainly be attributed to a generally good accessibility of the three N‐glycans on the surface of the globular glycoprotein. In the case of the structurally less reinforced EPO fragments 25 and 28 unfavorable local conformations around each glycosylation site most likely affect the sialylation reactions and hamper full conversion.

The observed high efficiency of the late‐stage enzymatic sialylation of EPO A is very encouraging. Most likely other homogeneous glycoproteins can also be modified analogously, however, the feasibility and efficiency of this approach need to be evaluated on a case‐to‐case basis. The efficient enzymatic α‐1,4‐galactosylation of a folded N‐glycoprotein was recently shown using a synthetic glycoform of Saposin D containing a biantennary N‐glycan. From our experiences well‐behaved proteins are preferred substrates since they readily withstand the conditions of the enzymatic reactions and the following purifications leading to good overall yields. It is desirable that the enzymatic conversion of the homogeneous glycoprotein reaches completion unless reliable chromatographic means for the efficient removal of products resulting from incomplete reactions are available.

To ensure that synthetic EPO A and EPO S are biologically recognized we investigated binding to the human EPO receptor EPOR (Scheme 8). The extracellular domain (25–250) of the receptor was expressed as a His6‐SUMO fusion protein in E. coli and gave inclusion bodies (see supporting information). After purification of the fusion protein by Ni‐IMAC the SUMO domain was cleaved using the specific protease SENP2. Subsequently, EPOR was refolded from urea by dialysis and purified via gel filtration. The EPOR was characterized by SDS‐PAGE, HPLC‐MS, HR‐MS, and CD spectroscopy. The CD data are consistent with a previous report suggesting a proper fold of EPOR.

Scheme 8

a) Cartoon of the high affinity (1:1) complexes of EPO A and EPO S with EPOR based on a crystal structure (PDB code: 1EER); b) complex formation of EPO A, EPO S and EPO with EPOR monitored by SEC (apparent molecular weights of the complexes were calculated by linear approximation between the reference points of 75 and 44 kDa given by the manufacturer); c) SDS‐PAGE (non‐reducing) of the complexes of EPO A, EPO S and EPO with EPOR isolated by SEC; d) the isolated complexes of EPO A, EPO S and EPO with EPOR were first dissociated with 0.2 % TFA and subsequently analyzed by RP‐HPLC‐MS.

EPO can bind to its receptor via a high affinity and a low affinity site. In solution the high affinity complex between EPO and the soluble EPO receptor (1:1 ratio) is formed first. Only at relatively high receptor concentrations (≈8 mg of EPOR mL−1) binding of an additional EPOR occurs, leading to the 2:1 EPOR/EPO complex, which can be detected by gel filtration.

We first probed the binding of the refolded EPOR (24.8 kDa) with commercially available EPO recombinantly expressed in CHO cells (EPO≈37 kDa). Both proteins were mixed under dilute conditions (≈0.3 mg EPOR mL−1) and submitted to gel filtration over an analytical superdex 75 column. A newly formed peak (apparent MW≈62 kDa) eluting earlier than the individual proteins indicated the formation of an EPO+EPOR complex. Analysis of the peak by SDS‐PAGE clearly showed EPOR but EPO was visible only as a broad band due to N‐glycan microheterogeneity.[ , ] Thus, the EPO+EPOR complex was dissociated using dilute TFA (0.2 %, 1 h) followed by an RP‐HPLC‐MS analysis.[ , ] The relative intensity of the two protein peaks corresponded to the formation of the 1:1 complex. Subsequently, the complexation of the synthetic glycoforms EPO A and EPO S with EPOR was carried out and confirmed by gel filtration (complexes with an apparent MW of ≈51 or ≈55 kDa) and SDS‐PAGE. In both cases the formation of 1:1 EPO/EPOR complexes was revealed by RP‐HPLC‐MS after dissociation of the complexes with dilute TFA. These findings altogether indicate that the EPO receptor EPOR binds to the synthetic glycoforms EPO A, EPO S and to recombinant EPO in a similar manner. Thus, synthetic EPO A and EPO S should display the biologically relevant native fold.

Conclusion

In summary, we developed a robust and flexible chemical synthesis for EPO bearing three biantennary N‐glycans based on sequential native chemical ligation. Out of several protecting group schemes only one strategy rendered EPO reproducibly and in high purity. For glycopeptide segments containing Acm‐protected cysteines the use of a PhiPr ester instead of an allyl ester at the glycosylation site was instrumental. The final cleavage of Acm groups with PdII was found to be compatible with glycopeptides bearing multiple N‐glycans. Despite the use of excess PdII the residual content of Pd in the final glycoprotein EPO A was below the detection limit of 0.01 ppm. Sialic acids were conveniently introduced at the level of the glycopeptide hydrazide building blocks, which could be elaborated to the sialylated glycoform EPO S. A surprisingly efficient alternative approach to sialylated EPO was found in the enzymatic sialylation of EPO A furnishing the sialylated glycoform EPO S in a single step. The enzymatic modification of the N‐glycans of folded EPO should provide rapid access to multiple glycoforms from the same precursor. Biological recognition of synthetic EPO A and EPO S was shown to be similar to commercially available EPO by complexation with recombinant human EPO receptor EPOR. The chemoenzymatic routes presented herein should facilitate the synthetic access to even more complex glycoforms of EPO, thus providing the well‐defined tools needed to evaluate how carbohydrates affect the critical plasma lifetime of this circulatory cytokine therapeutic.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Click here for additional data file.