Unveiling the Accurate Site-Specific N- and O‑Glycosylation of Hyperglycosylated Erythropoietin Drugs by an Integrated Approach

Abstract

Hyperglycosylated proteins with a high sialic acid content show great promise in the development of long-acting biotherapeutics. However, their structural complexity and heterogeneity pose significant challenges to traditional analytical methods, which often fail to provide comprehensive glycan information across all glycosylation sites, leading to ambiguities in characterization. Despite the fact that long-acting hyperglycosylated erythropoietin (hyperEPO) has been available on the market for over two decades, its site-specific glycosylation profile remains ambiguous due to closely spaced glycosylation sites and large glycans that contain labile sialic acids substituents. Here, using hyperEPO as an example, we developed an integrated workflow that incorporates an experimentally cross-validated glycan database, optimized nonspecific digestion, and a streamlined glycopeptide derivatization method to enhance sialylated peptide detection, enabling comprehensive site-specific characterization of both N- and O-glycosylation. We applied this strategy to compare the glycosylation profiles of the commercial hyperEPO drug darbepoetin alfa and a novel high-potency analogue, EPO-XL. We found that both proteins exhibited high site occupancy, large tetra-sialylated glycans, and similar types of sialic acid linkages but differed markedly in site-specific glycoform distributions. Notably, EPO-XL contained extensive LacNAc structures at all five N-glycosylation sites, and a previously unreported O-glycosylation site at S120 was identified alongside the canonical S126 site. Collectively, this study presents the first site-specific N- and O- glycosylation profiles of hyperEPO proteins, offering valuable guidance for the quality control and rational design of these therapeutics. The state-of-the-art analytical strategy introduced here holds great potential to advance the site-specific glycosylation characterization of proteins with complex glycosylation.

Graphical Abstract

INTRODUCTION

Glycosylation, one of the most common post-translational modifications (PTMs), plays a critical role in regulating protein structure and stability while mediating essential processes such as signal transduction, antigen recognition, and immune responses.^1–3 In protein-based biopharmaceuticals, both glycosylation sites and glycoforms profoundly impact drug efficacy and safety, highlighting the importance of site-specific glycosylation profiling in the quality control of glycoprotein therapeutics.^4,5 Hyperglycosylation is a compelling strategy for the development of long-acting and efficient biotherapeutics by increasing the glycosylation sites and/or glycan content, thereby modulating the absorption, distribution, metabolism, and excretion (ADME) processes.^6,7

Hyperglycosylated proteins often feature multiple adjacent glycosylation sites, large complex glycans, and high levels of sialic acid. Despite their therapeutic potential, achieving accurate site-specific glycosylation analysis remains a significant challenge. First, the compact arrangement of glycosylation sites leads to numerous missed cleavages and the generation of glycopeptides with multiple glycosylation sites during traditional enzymatic digestion, such as tryptic digestion, which complicates data interpretation.⁸ Second, the presence of larger glycans complicates glycopeptide identification that limit mass spectrometry (MS) ionization and data analysis.^9–11 Third, sialic acids are particularly labile and prone to loss during sample preparation and in MS analysis.^12–14 While sialic acid derivatization can improve the detection of sialylated glycans and glycopeptides, few applications on hyperglycosylated proteins have been reported.^15–17 These challenges often exceed the capabilities of conventional analytical methods, creating significant obstacles to the characterization of hyperglycosylated proteins.

Recombinant human erythropoietin (rhEPO) was the first approved glycoprotein drug and is widely used in the treatment of anemia associated with chronic kidney disease and cancer.^18,19 The human EPO protein contains three N-glycosylation sites (Asn24, Asn38, and Asn83) and one O-glycosylation site (Ser126), with the glycans accounting for approximately 40% of its molecular weight.²⁰ However, rhEPO drug has a short half-life in vivo, necessitating frequent dosing and resulting in poor patient compliance.^21,22 One of the major advancements in long-acting EPO therapy has been achieved through glycosylation engineering.²³ The second-generation EPO drug, darbepoetin alfa (trade name Aranesp or NESP), introduced two additional N-glycosylation sites at Asn30 and Asn88, resulting in a total of six potential glycosylation sites (Asn24, Asn30, Asn38, Asn83, Asn88, and Ser126).^24,25 Compared to rhEPO, darbepoetin alfa exhibits a much larger molecular weight, a higher sialic acid content, and, particularly, a 3-fold longer half-life.^24,26,27 Extensive studies have shown that glycosylation sites and glycan composition are critical factors influencing the efficacy and pharmacokinetics of EPO.^28,29 Recent research has demonstrated the development of novel erythropoietin-based therapeutic candidates with additional N-glycan sites that inhibit hematopoiesis while preserving neuroplasticity.³⁰ Notably, a rational computational strategy identified compatible regions of human EPO for the insertion of N-linked glycosylation motifs (N-X-S/T), resulting in Analog-71.1, which exhibited a high glycosylation probability and strong receptor binding stability, as confirmed by molecular docking and molecular dynamics simulations.³¹

While the site-specific glycan profiles of rhEPO have been extensively characterized using a range of analytical techniques, the glycosylation analysis of the second-generation hyperglycosylated EPO (hyperEPO) has been largely limited to assessment of overall glycan composition.^32–37 To date, the site-specific glycan distribution of hyperEPO remains unexplored, primarily due to the close proximity of glycosylation sites and high tetra-antennary sialylated glycan composition, leaving a critical gap in our understanding of its structural and functional attributes.^38–43

A novel hyperEPO drug candidate, EPO-XL, which shares a protein sequence identical with that of darbepoetin alfa but differs in host cell lines, culture conditions, and downstream purification processes, was recently developed and exhibits superior in vivo activities and an extended half-life (data not shown). It is well-established that variations in expression systems and bioprocessing parameters can strongly impact the glycosylation of recombinant proteins, thereby modulating their biological activity, stability, and pharmacokinetics.^44,45 Hence, characterizing the site-specific glycosylation differences between these two hyperEPO proteins is critical for optimizing therapeutic performance and ensuring product consistency. In this study, we established a comprehensive workflow that incorporates an experimentally cross-validated glycan database for accurate glycan complexity characterization, a reliable nonspecific digestion method to overcome challenges associated with dense site cleavage, and a streamlined sialoglycopeptide derivatization strategy for sialic acid protection, enabling site-specific N- and O-glycosylation characterization of hyperEPO proteins. A systematic comparison between EPO-XL and darbepoetin alfa was conducted at the glycan, peptide, and protein levels, with rigorous site occupancy analysis to quantitatively evaluate glycosylation at each site. Our study provides valuable insights for the development of EPO therapeutics and the investigation of other hyperglycosylated proteins.

RESULTS AND DISCUSSION

Workflow for Site-Specific Glycosylation Analysis of Hyperglycosylated Proteins.

To address the challenges associated with site-specific glycosylation characterization of hyperglycosylated glycoproteins, we developed an integrated workflow incorporating both comprehensive released glycan analysis and detailed glycopeptide-level investigations (Figure 1). First, we constructed a high-accuracy, experimentally cross-validated glycan database derived from dedicated released glycan analyses. Hyperglycosylated proteins are often characterized by structurally diverse, highly branched, and large glycan moieties, which significantly complicate glycopeptide identification. As the molecular weight increases, the number of potential glycoforms with the same mass also rises, adding ambiguities to glycopeptide assignment. Additionally, the instability of sialic acids presents further challenges for released glycan characterization. Therefore, in our study, we integrated complementary glycan analysis methods to obtain accurate glycan compositions, sialic acid linkage information, and other glycan modification, while employing cross-validation to ensure reliability. Second, to maximize the exposure of all the glycosylation sites, we optimized a nonspecific enzymatic digestion protocol to produce peptides of relatively uniform length (5−10 amino acids) with high reproducibility. This method not only enhances digestion efficiency for hyperglycosylated proteins to reduce generation of peptides with multiple glycosylation sites, but also facilitates the enrichment of O-glycopeptides using HILIC SPE cartridges, enabling the simultaneously detection of N- and O-linked glycopeptides. Third, we implemented a novel one-tube dual-derivatization strategy to enhance glycopeptide detection, specifically to stabilize sialic acids and improve sialoglycopeptide identification. Moreover, site occupancy assays were conducted to quantitatively evaluate glycosylation levels at specific sites, while intact protein mass analysis corroborated the overall glycosylation abundance. Using hyperEPO as an example, these integrated methodologies enabled detailed site-specific glycosylation characterization of hyperEPO proteins, providing a robust framework for comparing the novel glycoprotein EPO-XL with the commercially available darbepoetin alfa.

Figure 1.

Schematic workflow for site-specific glycosylation analysis of hyperglycosylated proteins. This workflow centers on glycopeptide-level characterization, supported by released glycan profiling and intact protein mass analysis. Released glycans were independently derivatized via amidation and permethylation to cross-validate glycan identification and enhance confidence in the glycan database construction, thereby facilitating glycopeptide assignment. At the glycopeptide level, nonspecific enzymatic digestion was used to produce short and uniform glycopeptides, enabling comprehensive coverage of densely glycosylated regions. A one-tube dual derivatization strategy, involving dimethylation followed by amidation, was applied to stabilize sialic acids and improve the detection of sialylated glycopeptides. Additionally, site occupancy analysis and intact protein mass profiling were conducted to provide an overview of the glycosylation levels.

Experimentally Cross-Validated Glycan Database Construction.

Previous studies have demonstrated that EPO proteins typically contain a high level of sialic acid and tetra-antennary glycans. Given the complexity of the glycoform composition and intrinsic instability of sialic acids, we utilized two complementary glycan derivatization methods for released glycan analysis to increase the reliability of the database. Initially, a sialic acid derivatization approach, adapted from Wuhrer’s study,⁴⁶ was implemented. This method enables differentiation between α2,3- and α2,6-linked sialic acids, as α2,3-linked sialic acids form stable amide bonds, whereas α2,6-linked sialic acids are converted into ethyl esters under the derivatization conditions. This linkage-specific derivatization strategy facilitates a more precise characterization of the glycan profile. We then utilized reduction and permethylation to cross-validate the glycoforms, with a particular focus on those low-abundance species.

In the sialic acid derivatized N-glycan spectra, as shown in Figure 2, significant differences in glycan profiles were observed between darbepoetin alfa and EPO-XL. For darbepoetin alfa, the tetra-antennary H₇N₆F₁S₄ glycan was the most abundant, followed by the triantennary H₆N₅F₁S₄ and tetra-antennary structures containing a N-acetyllactosamine (LacNAc) motif, such as H₈N₇F₁S₄. In contrast, EPO-XL predominantly exhibited larger glycan structures compared to darbepoetin alfa. The majority of glycans in EPO-XL were tetra-antennary with extended LacNAc units, including compositions such as H₈N₇F₁S₄, H₉N₈F₁S₄, H₁₀N₉F₁S₄, as well as even larger structures like H₁₁N₁₀F₁S₄ and H₁₂N₁₁F₁S₄. While O-acetylation was observed on most glycans in darbepoetin alfa, it appears to be less abundant in EPO-XL. Notably, all sialic acids in these N-glycans were identified as α2,3-linked, a characteristic commonly associated with proteins expressed in CHO cells. Figure S1 demonstrates the reproducibility of the glycan profiles and experimental results, highlighting the robustness of our analysis.

Figure 2.

MALDI-TOF MS spectra of sialic acid derivatized N-glycans released from (a) darbepoetin alfa and (b) EPO-XL, respectively. [Legend: H, hexose; N, N-acetylhexosamine; F, fucose; S, N-acetylneuraminic acid (sialic acid, Neu5Ac); O-Ac, O-acetylation; α−2,3, α2,3-linked sialic acid. The number following the condensed code represents the quantity of theses residues. The structures in SFGN are putative glycan structures based on MS/MS analysis and biosynthetic rules.]

To enhance the accuracy of the glycan database, we conducted additional released glycan characterization using reduction and permethylation as a complementary validation method, as illustrated in Figure S2. The two derivatization strategies detected nearly identical types of N-glycans. RPLC-MS/MS analysis of reduced permethylated glycans showed that most of the glycans were sialylated and core-fucosylated. The detection of the characteristic internal fragments B_5α′/Y_8α′ at m/z 899.4567 (mass error: 3.56 ppm), as shown in Figure S3, which contain two hexoses and two HexNAc residues, further supports the presence of LacNAc motifs. Given the low abundance of N-glycolylneuraminic acids (Neu5Gc, less than 5%) and the universal presence of fucosylation, we included only Neu5Ac in the glycan database to eliminate potential ambiguities caused by overlapping glycan motifs of identical mass (Fuc + Neu5Gc = Hex + Neu5Ac). Cross-validation of glycan components between these two derivatization strategies significantly enhanced the reliability and comprehensiveness of the glycan database. Additionally, we applied beta-elimination to release O-glycans from both samples, enabling the construction of an O-glycan database, as shown in Figure S4. The validated N- and O-glycans were then integrated into the glycan database, as summarized in Table S1, for both EPO-XL and darbepoetin alfa, respectively. This experimentally cross-validated glycan database establishes a robust reference for subsequent glycopeptide identification using bioinformatic tools, significantly reducing the search space and minimizing potential false positives.

Enzymatic Digestion Optimization for Site-Specific Glycosylation Analysis.

To achieve unambiguous site-specific analysis, it is crucial to produce glycopeptides containing a single N-glycosylation site, as peptides with multiple N-glycosylation sites pose substantial challenges for both separation and identification. EPO-XL and darbepoetin alfa share an identical amino acid sequence, each harboring five potential N-glycosylation sites and one characterized O-glycosylation site at Ser126. Among the five N-glycosylation sites, Asn24, Asn30, and Asn38 (EAENITTGCNETCSLNENI TVPDT), as well as Asn83 and Asn88 (GQALLVwNSSQVw-NETLQLHVD), are located in close proximity. Additionally, these sites lack cleavage sites for commonly used specific proteases. As shown in Table 1, in silico digestion of hyperEPO using conventional proteases, including trypsin, chymotrypsin, and hybrid enzyme approaches (such as trypsin/Asp-N, trypsin/Glu-C, trypsin/chymotrypsin, and Glu-C/chymotrypsin), failed to produce glycopeptides that each containing only a single glycosylation site while simultaneously covering all five N-glycosylation sites.

Table 1.

Peptides Generated from hyperEPO Proteins through In-Silico or Experimental Enzymatic Digestion

Enzyme	Condition	Asn24	Asn30	Asn38	Asn83	Asn88
Trypsin	in silico	EAENITTGCNETCSLNENITVPDTK			GQALLVNSSQVNETLQLHVDK
Chymotrypsin	in silico	EAKEAENITTGCNETCSL		NENITVPDTKVNF	VNSSQVNETL
Trypsin/Asp-N	in silico	EAENITTGCNETCSLNENITVP			GQALLVNSSQVNETLQLHV
Trypsin/Glu-C	in silico	NITTGCNE		NITVPD	GQALLVNSSQVNE
Trypsin/Chymotrypsin	in silico	EAENITTGCNETCSL		NENITVPDTK	VNSSQVNETL
Glu-C/Chymotrypsin	in silico	NITTGCNE		NITVPD	VNSSQVNE
Thermolysin		LLEAKEAENITTGC	/	LNENITVPDTK	LVNSSQVNETLQ
	1:1 16 h 37 °C			ENITVPDTKV	GQALLVNSSQV
				LNENITVPDTKVNF	ALLVNSSQVNETL
		LLEAKEAENITTG	CN	LNENITVPDTK	NS
	1:1 6 h 65 °C	LLEAKEAENITT		LNENITVPDTKVN	GQALLVNSSQVNE
				LNENTTVPDTKVNF	ALLVNSSQVNETL
	1:1 16 h 65 °C	/	/	LNENITVPDTKV	/	QVNETLQLHVDK
Pronase		EAKEAEN	ITTGCNETCS	NITVPDTK	VNSSQVNETL
	1:1 4 h 37 °C	ENITTGCNETCS		ENITVPDTK		NETL
		EAKEAENITT		NITVPDTKVN
		NITTGC	NETC	NITVPDTK	NSSQVNETL
	1:1 48 h 37 °C	ENITTG	NETCS	ENITVPDTK		NET
		AENITTGC	CNETS	ENITVPD
Proteinase K		EAENITT	CNETC	NENITVPDTK	LVNSS	QVNET
	1:1 4 h 37 °C	KEAEN	GCNETC	NITVPDTK		NETL
		EAEN	NETC	ENITVPDTK	VNSSQVNETL
		EAENITTGCN		NENI
		EAENITT	CNETC	NENITVPD	LVNSS	QVNET
		NITT	GCNETC	NENITVPDTK	VNSSQ	QVNETL
	1:1 48 h 37 °C	EAENIT	NETC	NENITVPDT		VNET
		AENITT	CNET	NITVPD		VNETL
		AENIT
	1:1 72 h 37 °C			Same with 48 h

^*No missed cleavages were allowed in the in silico digestion. A slash (/) indicates that no glycopeptides or peptides were detected at this site. The potential glycosylation sites are in red italic font.

We then explored nonspecific proteases to generate peptides of uniform length (5−10 amino acids) with high reproducibility. To optimize this process, we systematically investigated various proteases under different reaction conditions. The glycopeptide identification results for selected experimental conditions are summarized in Table 1. Among these, thermolysin, characterized by its low substrate specificity, was found to be suboptimal for the digestion of hyperEPO. As shown in Table 1, the cleavage efficiency at the densely positioned Asn83 and Asn88 sites was often incomplete. Additionally, peptides generated at Asn24 and Asn38 were excessively long, while peptides at Asn30 were either too short or frequently undetectable. However, under optimized conditions, both proteinase K and pronase demonstrated the potential to produce short peptides containing a single glycosylation site. Comparatively, the peptide sequence generated by proteinase K was more consistent across runs, irrespective of digestion time, typically ranging from 5 to 10 amino acids, as shown in Figure S5. In contrast, pronase displayed greater variability and was more likely to produce peptides consisting of 4 amino acids or fewer. For hyperEPO, the optimized proteinase K digestion condition (enzyme/sample ratio of 1:1, 37 °C for 48 h) effectively generated peptides of a single glycosylation site to cross all potential N- and O-glycosylation sites, thus enabling subsequent site-specific glycan profiling. Moreover, glycopeptides with relatively short amino acid sequences typically exhibit a higher HILIC enrichment efficiency, which enhances the recovery of O-glycopeptides and facilitates the simultaneous detection of both N- and O-linked glycopeptides.

Improved Site-Specific Glycosylation through Derivatization.

With the experimentally cross-validated glycan database and the optimized proteinase K digestion protocol, we successfully obtained the site-specific glycosylation profile of hyperEPO proteins. As shown in Figure S6, intact glycopeptide analysis enabled the unambiguous identification of five N-glycosylation sites, facilitating site-specific glycan profiling. However, compared to the released glycan results in Figure 2, trisialylated glycans appeared more abundant in the glycopeptide analysis, whereas tetra-sialylated glycans were less prominent, as shown in the left pie charts of Figure 3. One possible explanation for this discrepancy is that a higher number of sialic acids in native glycopeptides can lead to reduced ionization efficiency. Additionally, since hyperEPO proteins are highly sialylated, the loss of labile sialic acids during MS ionization or ion transfer was observed, as demonstrated in Figure S7 of the Supporting Information, despite the application of mild ionization conditions.

Figure 3.

Comparative analysis of N- and O-glycopeptides without derivatization (left) and with derivatization (right). [Legend: H, hexose; N, N-acetylhexosamine; F, fucose; S, N-acetylneuraminic acid (sialic acid, Neu5Ac). The number following the condensed code represents the quantity of theses residues.]

To address these challenges and enhance site-specific glycan profiling, we implemented a streamlined glycopeptide derivatization strategy to improve the stability of sialic acids and increase the ionization efficiency of sialylated peptides. This strategy consists of two sequential steps: dimethylation and amidation, with a reaction scheme illustrated in Figure S8. The dimethylation step, adapted from Lu’s protocol,⁴⁷ specifically protects primary amine groups, such as the peptide N-terminus and lysine (K) side-chain amine residues, from undesired side reactions during amidation. Each primary amine undergoes dimethylation, resulting in a mass shift of +28.03 Da. Subsequently, amidation modifies carboxyl groups on sialic acids, acidic amino acid side chains, and peptide C-termini. The amidation reaction, derived from Wuhrer’s research,⁴⁶ effectively stabilizes sialic acids of different linkages by converting their carboxyl groups into amides, thereby preventing sialic acid loss during MS analysis. After amidation, a mass shift of −0.984 Da per carboxyl group is introduced. To streamline the workflow and enhance efficiency, we optimized the protocol into a one-tube dual-derivatization strategy, allowing both reactions to proceed sequentially in the same centrifuge tube without the need for desalting. Details of the reaction are provided in the Supporting Information. The effectiveness and efficiency of this one-tube dual-derivatization strategy were investigated by using IgG trypsin digests. As shown in Figure S9, all peptides and glycans were effectively modified, with sialylated peptides exhibiting significantly improved ionization after derivatization. This dual-derivatization approach ensures that both primary amines and carboxyl groups are properly modified, resulting in improved glycopeptide stability and enhanced detection sensitivity, particularly for sialylated peptides.

We then applied the optimized derivatization strategy to EPO-XL to compare the glycoforms observed in underivatized and derivatized glycopeptides. The distribution of four dominant glycan groups detected in the glycan-level analysis, including three N-glycans (H₇N₆F₁S_x, H₈N₇F₁S_x, H₉N₈F₁S_x) and one O-glycan (H₁N₁S_x), is illustrated in Figure 3. For N-glycopeptides, the pie chart revealed that in underivatized glycopeptides, glycans carried between two and four sialic acids, with glycans containing three sialic acids accounting for 45%−49% of each N-glycan type. After derivatization, the proportion of glycans with four sialic acids increased significantly, with tetra-sialylated glycoforms accounting for 68%−97%, aligning more closely with the released glycan data. A similar trend was observed for O-glycopeptides, where derivatization enhanced the proportion of two sialic acids on H₁N₁S_x, increasing from 80% in the underivatized samples to 94% after derivatization. The shift in glycoform patterns clearly highlights the importance of protecting sialic acids when sialylated glycopeptides are studied, ensuring an accurate representation of glycan distribution at the peptide level. This further underscored the robustness of our derivatization strategy in preserving sialic acids and enhancing the accuracy of the site-specific glycosylation analysis.

Identification a Novel O-glycosylation Site at Ser120 in EPO-XL.

Our protocol allows the simultaneous detection of N- and O-glycopeptides. To accurately determine glycosylation sites and glycoforms, we employed two complementary tandem mass spectrometry techniques: stepped collision energy HCD (sceHCD) and electron transfer dissociation with HCD supplemental activation (EThcD). Intriguingly, in addition to the previously reported O-glycosylation site at Ser126 in EPO proteins, we unambiguously identified a new site at Ser120 in EPO-XL.

The O-glycosylated AISPPDAA (residues 118−125) glycopeptide was consistently detected in EPO-XL. In both sceHCD and EThcD MS/MS spectra of the derivatized AIS-(HexNAcHexNeuAc)PPDAA peptide, abundant oxonium ions were observed, including peaks at m/z 291.119 and 273.108 from amidated Neu5Ac. The sceHCD spectrum, shown in Figure 4a, exhibits abundant Y₀, Y₁, Y₂, b, and y ions, enabling glycan composition analysis and peptide sequencing. The oxonium ion-triggered EThcD spectrum (Figure 4b) displayed numerous c-type fragments containing the glycan motif, which were essential to nail down the glycosylation site. Notably, since O-glycosylation predominantly occurs on serine (S) and threonine (T) residues, the presence of a single serine residue within AISPPDAA significantly facilitated site identification without ambiguity. Moreover, the native AIS-(HexNAcHexNeuAc) PPDAA glycopeptide was also identified using both HCD and EThcD analyses, as shown in Figure S10, further supporting the identification of a novel O-glycosylation site at Ser120.

Figure 4.

Identification of a novel O-glycosylation site at Ser120 in the EPO-XL protein via (a) sceHCD and (b) EThcD MS/MS analysis of the derivatized glycopeptide AIS(HexNAcHexNeuAc)PPDAA. Glycan fragments are shown in red, peptide fragments are in blue, and those containing the glycan motif are in bold. Italicized labels in the figure indicate amino acid residues or oxonium ions modified by derivatization, including dimethylation of the peptide N-terminus, amidation of the peptide C-terminus, and amidation of sialic acid. The peak assignments are provided in Tables S2 and S3 in the Supporting Information.

Since O-glycosylation at Ser120 has not been previously reported for EPO proteins, we systematically analyzed recombinant human erythropoietin and darbepoetin alfa using our method. However, in both of these two EPO proteins, we did not detect glycosylation at Ser120, nor any other novel glycosylation sites. This finding suggests that the Ser120 glycosylation site is unique to EPO-XL. The functional significance of this discovery is under investigation and will not be discussed further here.

Comparative Site-Specific Glycan Profiling of Darbepoetin Alfa and EPO-XL.

Based on our methodology advancements, we conducted a detailed site-specific N- and O-glycosylation comparison between the commercial drug darbepoetin alfa injection and a higher-potency novel analogue, EPO-XL, as shown in Figure 5.

Figure 5.

Site-specific N- and O-glycosylation analysis of derivatized darbepoetin alfa and EPO-XL. Glycan abundance was calculated by the sum of glycopeptides with same glycans in the same site, and relative abundance was normalized at each site. [Legend: H, hexose; N, N-acetylhexosamine; F, deoxyhexose; S, N-acetylneuraminic acid; G, N-glycolylneuraminic acid. The number following the condensed code indicates the quantity of the corresponding residue.]

For N-glycosylation, tetra-antennary glycans were consistently observed across all glycosylation sites in both hyperEPO proteins, with similar glycoform distributions within each protein. However, distinct glycosylation patterns were observed between the two. EPO-XL was enriched with H₈N₇F₁S_X, H₉N₈F₁S_X, and H₁₀N₉F₁S_X, whereas darbepoetin alfa exhibited a higher abundance of H₇N₆F₁S_X. Moreover, differences in glycan heterogeneity were evident. EPO-XL displayed a more uniform glycan distribution, predominately featuring high-abundance LacNAc tetra-antennary species across all five sites. In contrast, darbepoetin alfa exhibited greater structural diversity with bi-, tri-, and tetra-antennary glycans. For example, triantennary H₆N₅F₁S₃ glycan was present at all N-glycosylation sites in darbepoetin alfa but was nearly absent in EPO-XL.

Site-specific variations in N-glycosylation were also identified. For EPO-XL, Asn30, Asn38, and Asn83 were predominantly occupied by H₉N₈F₁S₄, whereas Asn24 and Asn88 were enriched in H₁₀N₉F₁S₃. Meanwhile, tetra-sialylated glycans, such as H₈N₇F₁S₄, H₉N₈F₁S₄ and H₁₀N₉F₁S₄, were consistently detected at Asn24 and Asn88, although at slightly lower abundances, compared to H₁₀N₉F₁S₃. Additionally, EPO-XL exhibited ultrahigh-LacNAc glycans, H₁₁N₁₀F₁S₄ and H₁₂N₁₁F₁S₄, which were detected at all glycosylation sites except Asn30. However, their relative abundances remained low, with H₁₁N₁₀F₁S₄ not exceeding 8% and H₁₂N₁₁F₁S₄ remaining below 1% at their respective sites. In contrast, darbepoetin alfa exhibited a different glycosylation profile. Asn24, Asn38, Asn83, and Asn88 were consistently dominated by H₇N₆F₁S₄, with respective proportions of 30.66%, 34.77%, 51.40%, and 40.63%. The only exception was Asn30, which displayed a distinct glycoform distribution with H₈N₇F₁S₄ as the major species (38.80%). Furthermore, the glycan profiles of Asn83 and Asn88 were highly similar, with H₇N₆F₁S₄ and H₈N₇F₁S₄ together accounting for over 70% of the total glycoforms. Previous studies on EPO with three canonical N-glycosylation sites (Asn24, Asn38, Asn83) have demonstrated that sialylation at the Asn38 and Asn83 sites has a greater impact on enhancing EPO activity and prolonging its half-life compared to Asn24.²⁹ Additionally, hypersialylated EPO bearing extended LacNAc structures has been associated with enhanced in vitro and in vivo bioactivity.⁴⁸ In this study, both Asn38 and Asn83 are predominantly occupied by tetra-sialylated glycans in EPO-XL and darbepoetin alfa, with EPO-XL displaying comparatively larger LacNAc structures, consistent with its improved efficacy. The functional implications of glycosylation heterogeneity at Asn30 and Asn88 remain to be fully understood, necessitating further glycoengineering and pharmacological research, for which this study offers key foundational data.

A common feature observed in both proteins was the prevalence of N-glycans with four sialic acids, which was attributed to the derivatization that effectively preserved sialic acid moieties. It is worth mentioning that acetylation (O-Ac) modifications were also retained during derivatization, revealing significant differences between the two proteins, as shown in Figure S11. Darbepoetin alfa exhibited higher levels of O-Ac than EPO-XL, with H₇N₆F₁S₄ carrying up to six O-Ac modifications, detected at all glycosylation sites except Asn30. Additionally, a small proportion of glycans containing one-two O-Ac modifications was uniquely identified in EPO-XL.

For the common O-glycosylation site, Ser126, both proteins exhibited similar glycan compositions, including H₁N₁S₁ and H₁N₁S₂, except that EPO-XL contained a small amount of Neu5Gc glycan (H₁N₁S₁G₁). In both cases, H₁N₁S₂ was the predominant glycoform, accounting for 93.74% in darbepoetin alfa and 91.78% in EPO-XL. Additionally, a novel O-glycosylation site, Ser120, was identified in EPO-XL, where only a single glycoform, H₁N₁S₁, was observed. Since the number of sialic acid residues on the EPO protein is positively correlated with its hematopoietic activity,⁴⁹ the additional sialic acid at Ser120 may contribute to the enhanced in vivo activity of EPO-XL compared with darbepoetin alfa (data not shown).

Evaluation of Glycan Content and Site Occupancy.

To assess the glycosylation ratio in these two hyperEPO proteins, we further examined the glycan content through intact glycoprotein mass analysis and site occupancy at the peptide level. Experimental details are provided in the Supporting Information. The intact mass analysis by LC-MS revealed that the average molecular weights of darbepoetin alfa and EPO-XL were 37321.37 and 40421.71 Da (shown in Figure S12), respectively, corresponding to approximately 102.9% and 119.7% glycan content. In addition, SDS-PAGE analysis of both proteins, with and without PNGase F treatment, further confirmed that EPO-XL has a higher molecular weight, with more than half of its mass attributed to glycans (Figure S13).

We then looked into site occupancy by averaging the LC peak areas of multiple deglycosylated peptides at each glycosylation site. The summarized experimental results in Table 2 indicate that both darbepoetin alfa and EPO-XL exhibit high site occupancy at all five N-glycosylation sites. Darbepoetin alfa showed the highest site occupancy at Asn83, reaching 97.1%, with all other site occupancies above 91%. EPO-XL, on the other hand, had the lowest site occupancy at this site, at 86.8%, while the remaining sites ranged between 87% and 100%. Notably, we eliminated the influence of deamination on site occupancy calculations by employing ¹⁸O labeling and deamidated peptide extraction.^50,51 Nonetheless, the results reflect the high glycosylation levels of these two hyperglycosylated glycoproteins. Additionally, while a systematic assessment of O-glycosylation site occupancy remains challenging due to the absence of a universal O-glycosidase enzyme, it can be estimated by analyzing the relative abundance of O-glycopeptides and their corresponding nonglycosylated peptides. This analysis suggests a high occupancy of approximately 90% or higher. The site occupancy results align with the intact molecular data, confirming the presence of abundant glycans in both hyperEPO proteins.

Table 2.

N-Glycosylation Site Occupancy in Darbepoetin Alfa and EPO-XL

	Average Site Occupancy (%)
N-glycosylated sites	Darbepoetin alfa	EPO-XL
Asn24	91.8	93.0
Asn30	96.0	87.4
Asn38	92.8	89.8
Asn83	97.1	86.8
Asn88	93.7	100
Average	94.3	91.4

CONCLUSIONS

In this study, we presented an integrated workflow to address the challenges in site-specific glycosylation analysis of hyperglycosylated proteins, emphasizing the importance of sialic acid derivatization for the accurate assignment of sialylated glycopeptides. By utilizing an experimentally cross-validated glycan library, reliable nonspecific digestion, and a streamlined glycopeptide derivatization strategy, we successfully obtained the initial site-specific glycosylation profiles of hyperEPO proteins. Our study revealed the quantitative glycosylation differences between the commercial darbepoetin alfa drug and a higher-potency new analogue, EPO-XL. Notably, we unambiguously identified a novel O-glycosylation site at Ser120 in EPO-XL, along with simultaneous analysis of both N-and O-glycosylation. Numerous studies have confirmed that the glycoforms and glycosylation sites of EPO proteins play crucial roles in their pharmacological effects. Our findings represent significant progress in understanding the glycosylation of hyperEPO proteins, providing valuable insights for glycosylation quality control and offering guidance for EPO drug development.

While this workflow was developed using hyperEPO, it is readily adaptable for the in-depth analysis of other proteins with complex glycosylation. Given the extensive sequence diversity and glycan heterogeneity of such proteins, enzymatic digestion conditions and derivatization efficiency should be carefully optimized. Customized protocols may be needed to ensure the comprehensive characterization of all glycosylation sites, especially when these sites are located within or adjacent to proline-rich regions. For proteins bearing large and complex glycan structures, access to a well-curated glycan database is essential to reducing ambiguities in glycopeptide assignment.

Supplementary Material

Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.analchem.5c02433.

Details about materials and experimental methods. Figure S1: Relative intensities of amidated N-glycans from darbepoetin alfa and EPO-XL; Figure S2: MALDI-TOF MS spectra of reduced and permethylated N-glycans; Figure S3: sceHCD MS/MS spectra of the reduced and permethylated N-glycans in EPO-XL; Figure S4: MALDI-TOF MS spectra of reduced and permethylated O-glycans; Figure S5: Reproducibility of peptide sequences achieved under the optimized proteinase K digestion condition; Figure S6: Site-specific glycosylation analysis of EPO-XL; Figure S7: In-source dissociation of sialic acids observed in LC-MS/MS analysis; Figure S8: One-tube dual-derivatization reaction scheme for glycopeptides; Figure S9: Comparative analysis of MALDI-TOF MS spectra for native and derivatized glycopeptides; Figure S10: sceHCD and EThcD MS/MS spectra of the native glycopeptide AIS(HexNAcHexNeuAc)PPDAA in EPO-XL; Figure S11: Site-specific comparison of derivatized darbepoetin alfa and EPO-XL, featuring O-acetylated glycan structures; Figure S12: Intact mass analysis of darbepoetin alfa and EPO-XL; Figure S13: SDS-PAGE analysis for both hyper-EPO proteins before and after N-glycan removal; Table S1: N- and O-glycan databases for both darbepoetin alfa and EPO-XL; Tables S2 and S3: Peak lists of Figures 4a and 4b, respectively; Tables S4 and S5: peak areas of targeted peptides at each site of two hyperEPO proteins for site occupancy calculation (PDF)