Comprehensive glycan analysis of twelve recombinant human erythropoietin preparations from manufacturers in China and Japan

Journal of Pharmaceutical and Biomedical Analysis 153 (2018) 214–220

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis journal homepage: www.elsevier.com/locate/jpba

Short communication

Comprehensive glycan analysis of twelve recombinant human erythropoietin preparations from manufacturers in China and Japan Ben Cowper a,∗,1 , Xiang Li b,1 , Lei Yu b , Yong Zhou b,∗ , W.H. Fan b , C.M. Rao b a b

National Institute for Biological Standards & Control (NIBSC), Blanche Lane, South Mimms, Hertfordshire, EN6 3QG, United Kingdom National Institute for Food and Drug Control (NIFDC), Tiantan Xili No. 2, 100050, Dongcheng District, Beijing, China

a r t i c l e

i n f o

Article history: Received 23 October 2017 Received in revised form 16 February 2018 Accepted 20 February 2018 Available online 24 February 2018 Keywords: Erythropoietin Glycosylation Biosimilars Mixed mode chromatography Sialylation

a b s t r a c t Recombinant, human, erythropoietin (rhEPO) is a glycoprotein hormone which is prescribed throughout the world to treat anaemia caused by chronic kidney disease or chemotherapy. rhEPO is at the forefront of the recent emergence of biosimilar medicines, with numerous products now available worldwide. Due to its complex glycosylation profile, which has a crucial influence upon biological activity, therapeutic rhEPO preparations must be closely monitored to ensure consistency, safety and efficacy. Here, we have compared twelve rhEPO preparations from eleven manufacturers in China and one in Japan, measuring in vivo biological activity and exploring its relationship with glycosylation through sialic acid content determination, isoform distribution via capillary electrophoresis (CE), O-glycan profiling, and N-glycan mapping using a novel anion-exchange/hydrophilic interaction chromatography-mass spectrometry (AEX/HILICMS) approach. We observed differences between glycosylation profiles, including the varying occurrence of sialic acid O-acetylation, extension of N-glycan antennae with N-acetyllactosamine units, and the distribution of sialic acids across multi-antennary structures. The presence of unusually high levels of suspected penta- and hexa-anionic N-glycans in several samples is consistent with elevated rhEPO isoform acidity, which is reflected by slightly elevated in vivo bioactivities. This aside, the observed differences in glycosylation profile do not appear to have a significant influence upon biological activity in mice. Nonetheless, with the continued emergence of biosimilars, the study highlights the importance of monitoring glycosylation profiles in biological medicines, in order to detect and account for divergence between products, as well as the presence of unusual or unexpected glycans. © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction Erythropoietin (EPO) is a glycoprotein hormone which stimulates red blood cell production (erythropoiesis). Recombinant human EPO (rhEPO) preparations are commonly prescribed globally to treat anaemia caused by chronic kidney disease or chemotherapy. The first therapeutic rhEPO preparation, epoetin alfa, was licensed in the EU in 1988, and in the USA a year

Abbreviations: rhEPO, recombinant human erythropoietin; PEG, poly(ethylene) glycol; AEX, anion-exchange; HILIC, hydrophilic interaction chromatography; MS, mass spectrometry; CE, capillary electrophoresis; CRS, chemical reference substance; LC, liquid chromatography; PNGase F, peptide N-glycosidase F; MWCO, molecular weight cut off; LacNAc/LN, N-acetyllactosamine; HPAEC-PAD, high pH anion exchange chromatography with pulsed amperometric detection; CHO, Chinese hamster ovary; NANA/Neu5Ac, N-acetylneuraminic acid; NGNA/Neu5Gc, N-glycolylneuraminic acid; KDN, Ketodeoxynonulosonic acid. ∗ Corresponding authors. E-mail addresses: [email protected] (B. Cowper), [email protected] (Y. Zhou). 1 Ben Cowper and Xiang Li contributed equally to this work.

later. Numerous EPO products have since been licensed worldwide, including the sub-types epoetin beta, delta, and theta, which possess distinct glycosylation profiles [1,2], and second generation modified EPOs darbepoetin (hyperglycosylated) and Mircera (PEGylated). Since the expiry of originator epoetin alfa patents a number of biosimilars have also emerged in various markets [3–5] giving rise to a complex global network of rhEPO products (Table 1). At the present time, there are more than ten manufacturers of licensed rhEPO products in China alone. EPO contains three N-glycosylation sites (Asn-24, Asn-38, Asn83) and one O-glycosylation site (Ser-126). The production of therapeutic EPO preparations through recombinant gene expression in mammalian cell culture gives rise to a complex pattern of glycosylation, including a wide range of bi-, tri- and tetraantennary structures [6,7]. Glycan composition has a strong influence over EPO pharmacokinetics, with removal of negativelycharged branch-terminal sialic acids causing a marked decrease in plasma half-life and abolishment of in vivo biological activity [8,9], due to enhanced clearance by asialoglycoprotein receptors in the liver [10]. The glycan profile is therefore a critical quality

https://doi.org/10.1016/j.jpba.2018.02.043 0731-7085/© 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4. 0/).

B. Cowper et al. / Journal of Pharmaceutical and Biomedical Analysis 153 (2018) 214–220

215

Table 1 A summary of originator and regional biosimilar EPO products. 1 Epoetin theta was not statutorily registered as a biosimilar by the EMA, but is often clinically considered to be a biosimilar of epoetin alfa [5,25]. 2 All EPO products in China were not registered as biosimilar or originator, and are not classified by INN. INN = International Non-proprietary Name. INN

Type

Region

Brand names

Epoetin alfa

Originator

Worldwide

Eprex , Erypo , Epogen , Procrit , ESPO

Biosimilar

EU

Abseamed , Binocrit , Epoetin Alfa Hexal

India

®

Comments ®

®

®

®

®

®

®

®

®

®

®

Epoetin beta Epoetin delta Epoetin kappa Epoetin lambda Epoetin theta

Originator Originator Biosimilar Biosimilar Originator1

Worldwide EU Japan Australia EU

Ceriton , Epofer , Epofit , Eporec , ® ® ® ® Epotin , Relipoietin , Repoitin , Wepox ® NeoRecormon ® Dynepo ® Epoetin alfa BS JCR ® ® Aczicrit , Gradicrit ® ® Biopoin , Eporatio

Epoetin zeta Methoxy PEG epoetin beta Darbepoeitin alfa

Biosimilar Originator Originator

EU Worldwide Worldwide

Retacrit , Silapo ® Mircera ® ® Aranesp , Nesp

-

Biosimilar –

India China2

®

®

®

®

®

Actorise , Cresp , Darbatitor , ® ® ® ® Epoetin , Sepo , Renogen , Eposino

attribute which is closely monitored in therapeutic rhEPO products. Glycan composition is strongly influenced by cell culture process variables [11], and so the growing number of global rhEPO manufacturers places increased importance upon glycan analysis, to ensure consistency, safety and efficacy of products and batches. Previous studies have highlighted the differences in glycosylation which are observed between rhEPO products [2,12–14], including variances in O-acetylation of sialic acid and N-acetyllactosamine (LacNAc) extension of glycan antennae. Glycan profiles can be characterised using a variety of techniques [15], including capillary electrophoresis (CE), isoelectric focusing (IEF), and mass spectrometry (MS), of intact glycoproteins or fragmented glycopeptides. Detailed structural information is also commonly obtained through chromatographic analysis of enzymatically-released N-glycans, with a variety of approaches available. Due to the crucial influence of anionic sialic acid over rhEPO bioavailability, it can be useful to separate rhEPO N-glycans on the basis of charge, enabling rapid identification of the relative quantities of multi-sialylated structures. This can be achieved using high performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) [16], however the use of high pH and salt conditions necessitate the use of a specialised instrument, and prevent routine coupling to a mass spectrometer for N-glycan identification. Alternatively released N-glycans can be fluorescently-labelled for analysis using conventional HPLC(MS) instrumentation [15]. This usually involves anion exchange (AEX) or hydrophilic interaction chromatography (HILIC), the latter being particularly useful due to its use of MS-compatible mobile phases [14]. However HILIC provides a polarity-based separation, and therefore structures with different charges are not always fully segregated. Here, we present a comprehensive glycan analysis of eleven rhEPO preparations acquired from manufacturers in China, and one additional sample from Japan. Samples were subjected to N-glycan mapping, using a novel anion-exchange/hydrophilic interaction chromatography-mass spectrometry (AEX/HILIC-MS) method. This method conveniently allows for AEX-type charge-based N-glycan separation under HILIC-type MS-compatible conditions, enabling resolution and simple identification of individual structures. Intact O-glycan characterisation, capillary electrophoresis (CE) and sialic acid content determination were also performed, and the in vivo bioactivity of each preparation was also measured, in order to explore the relationship between bioactivity and rhEPO glycan composition.

Same product, licensed under multiple brand names. Same production technology transferred between manufacturers. Multiple products.

No longer marketed. Reference product is epoetin alfa. Reference product is epoetin alfa. Same product, licensed under multiple brand names. Reference product is epoetin alfa. Same product, licensed under multiple brand names. Multiple products. Multiple products.

2. Materials & methods 2.1. Reagents The recombinant EPO for physicochemical tests CRS was acquired from the European Directorate for the Quality of Medicines (EDQM, Strasbourg, France). This is a lyophilised preparation containing approximately 100 ␮g erythropoietin, with epoetin alfa and beta present in equal amounts. rhEPO preparations S01-S12 were generously donated to the National Institute for Food and Drug Control (NIFDC), China, by eleven manufacturers from China and one manufacturer from Japan. Samples were anonymised and identified via randomly assigned codes S01–S12. Samples S01, S02, S05, S06, S08, S09 and S10 were manufactured in bioreactors. Samples S03, S04, S07, S11 and S12 were manufactured using roller bottles. GlycoworksTM RapiFluor-MSTM reagents were purchased from Waters (Milford, MA, USA). Ammonium formate, urea, methylcellulose, N-acetylneuraminic acid (Neu5Ac, NANA) and trifluoroacetic acid were purchased from Sigma-Aldrich (St Louis, MI, USA). Neuraminidase was purchased from New England Biolabs (Ipswich, MA, USA). Ketodeoxynonulosonic acid (KDN) was purchased from Worthington Biochemical (Lakewood, NJ, USA). pI markers 3.59 and 5.85 were purchased from Protein Simple (San Jose, CA, USA). Pharmalytes 3–10 and 2.5-5 were purchased from GE Healthcare (Chicago, IL, USA). Acetonitrile (OptimaTM , LC/MS grade) was purchased from Fisher Scientific (Hampton, NH, USA). 2.2. N-glycan preparation RapiFluor-MS (RFMS) labelled N-glycans were prepared using Waters GlycoworksTM RapiFluor-MSTM kit reagents and protocols (Waters, Milford, MA, USA). rhEPO preparations were buffer-exchanged into water and concentrated to 2 mg/ml using 10,000 MWCO spin concentrators (Millipore, Burlington, MA, USA). Volumes of 7.5 ␮l were labelled with RFMS according to the manufacturers’ instructions. N-glycans were purified using a Waters HILIC ␮Elution plate, with elution in 3 × 30 ␮l 200 mM ammonium formate in 5% (v/v) acetonitrile. Samples were then directly analysed via LC–MS. 2.3. Intact O-glycan preparation N-deglycosylated rhEPO samples were prepared using Waters GlycoWorksTM reagents (Waters, Milford, USA). A 4 ␮l volume of a

11.8 11.3 10.8 10.6 10.7 11.2 14.5 10.1 10.3 10.2 10.9 13.4 N.D – – – – – – – – 2.6 – – – 5.9

>5.0 4.88–4.98

1.6 7.1 – 7.7 – 7.7 – – 10.0 – 3.8 – 6.3 26.5 30.2 17.6 15.6 3.4 26.1 6.3 3.1 27.0 13.8 28.8 13.7 18.8

4.65–4.75 4.45–4.54

38.7 28.8 47.1 23.0 23.0 30.6 32.4 21.7 24.7 39.5 32.6 30.1 26.7 23.1 21.4 24.5 26.9 36.6 22.9 37.5 34.5 17.9 29.1 23.3 36.4 26.1

4.24–4.32 4.09–4.18

8.7 10.6 8.3 20.8 26.9 10.9 21.0 23.9 11.2 13.2 9.7 19.3 15.0 1.5 1.9 2.0 5.0 8.5 1.6 2.9 10.9 4.9 3.3 1.8 1.4 1.2

3.90–4.02 3.76–3.85

– – 1.5 2.4 1.6 – – 4.9 2.6 0.5 – – – – – – – – – – 1.0 – – – – –

<3.7 Z

350 350 356 365 376 350 372 363 345 360 357 370 358 – – 0.9 – 1.3 – – 4.3 2.6 – – – –

−6 −5

5.7 3.2 12.8 3.9 18.8 5.4 4.3 23.5 16.2 9.0 4.8 0.3 1.2 54.0 55.9 49.4 67.6 52.6 53.1 70.3 43.8 42.2 53.6 57.3 74.9 64.4

S4 S3

27.6 31.1 24.0 19.5 19.3 29.8 19.3 17.7 23.5 26.8 28.8 19.9 26.7 10.6 7.7 9.7 8.4 5.5 9.7 5.5 6.2 9.3 9.0 8.0 4.5 6.9

rhEPO samples were diluted to approximately 0.33 mg/ml with sodium phosphate buffer (5 mM, pH 7.2). A 20 ␮l sample volume was transferred to screw capped vials, along with 20 ␮l neuraminidase (1U/100 ␮l) and incubated for 1 h (±15 min) at 37 ◦ C. A 50 ␮l volume of ketodeoxynonulosonic acid (KDN) internal standard working solution (0.04 mM) and 70 ␮l of purified water was added to the digested sample. 5-acetylneuraminic acid (NANA) calibration standards were prepared by mixing 20, 40, 60, 80 and 100 ␮l NANA standard solution (12 ␮g/ml, 95%) with 50 ␮l KDN working solution (0.04 mM) and 90, 70, 50, 30 and 10 ␮l water respectively. Samples were analysed via high pH anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD), using a Thermo ICS-5000 ion chromatography system equipped with an anion-exchange column (CarboPac PA-1) and guard column (CarboPac PA-100) (Thermo Fisher Scientific, Waltham, MA, USA). Both the column and detector were maintained at 20 ◦ C. 20 ␮l sample/standard volumes were injected. Isocratic elution was

S2

2.6. Sialic acid content determination

2.1 2.0 3.2 0.8 2.4 2.0 0.5 4.4 6.1 1.5 1.1 0.4 0.8

A master mix (MM) was prepared fresh daily containing 4 M urea, 0.35% methylcellulose and 4% Pharmalytes mixture (Pharmalytes 3–10:Pharmalytes 2.5-5, 1:3). rhEPO samples were prepared using 100 ␮l of MM with 0.5 ␮l each of pI markers 3.59 and 5.85, at a target concentration of 0.2 mg/ml. Samples were vortexed briefly to ensure complete mixing and centrifuged at 10,000 rpm for 3 min to remove air bubbles. CE was performed using an iCE3 whole-capillary system, an FC-coated column cartridge, and pI markers 3.59 and 5.85 (ProteinSimple, CA, USA). The anolyte and catholyte solutions were 80 mM phosphoric acid and 100 mM sodium hydroxide respectively. Samples were focused for 1 min at 1.5 kV, followed by 6.5 min at 3 kV, and A280 images of the capillary were taken using iCE analysis software.

S1

2.5. Capillary electrophoresis (CE)

S01 S02 S03 S04 S05 S06 S07 S08 S09 S10 S11 S12 CRS

N-glycan and intact O-glycan samples were analysed via LC–MS using a Waters Acquity H–Class Bio UPLC system equipped with a Single Quadruple Detector (SQD2) (Waters, Milford, USA). RFMSlabelled N-glycan separations were achieved by injection of 1 ␮l sample volumes onto a GlycanPac AXH-1 column (2.1 × 150 mm, 1.9 ␮m; Thermo Fisher Scientific, Waltham, MA, USA) maintained at 30 ◦ C. Gradient elution was employed using 100 mM ammonium formate-acetonitrile-water (2:70:28 to 15:57:28, v/v/v, over 40 mins), at a flow rate of 0.4 ml/min. Fluorescence was monitored at 265/425 (Ex/Em). Intact O-glycan separations were achieved by injection of 1 ␮l sample volumes onto a Waters Acquity UPLC Glycoprotein Amide 300 Å column (2.1 × 150 mm, 1.7 ␮m; Waters, Milford, MA, USA) maintained at 45 ◦ C. A gradient was employed using 0.1% trifluoroacetic acid-0.1% trifluoroacetic acid in acetonitrile (23:77 to 33:67, v/v, over 20 min), at a flow rate of 0.2 ml/min. Fluorescence was monitored at 280/320 (Ex/Em). Mass spectrometry data was acquired in ES+ mode, with a capillary voltage of 3 kV, cone voltage of 40 V, source temperature of 150 ◦ C, desolvation temperature of 350 ◦ C and desolvation gas flow of 800 l/hr. Data was acquired and analysed using Empower 3.1 software (Waters, Milford, USA).

Capillary electrophoresis (RPA%)

2.4. Liquid chromatography-mass spectrometry (LC–MS)

N-glycan mapping (RPA%)

5% RapiGestTM SF Surfactant solution (in Glycoworks Rapid Buffer) was added to 10.2 ␮l of water and 5 ␮l of rhEPO (at 2 mg/ml in water). Samples were incubated at 90 ◦ C for 3 mins, then allowed to cool, before addition of 0.8 ␮l Glycoworks Rapid PNGase F. Reactions were incubated at 50 ◦ C for 5 mins, allowed to cool, and then directly analysed via LC–MS.

Sialic acid (mol/mol EPO)

B. Cowper et al. / Journal of Pharmaceutical and Biomedical Analysis 153 (2018) 214–220 Table 2 N-glycan mapping, capillary electrophoresis (CE) and sialic acid content determination data for each rhEPO preparation. N-glycan mapping: relative peak areas (RPA%) of mono- (S1), bi- (S2), tri- (S3) and tetra- (S4) sialylated, and penta- (−5) and hexa- (−6) anionic N-glycans, and corresponding Z-numbers. The conventional Z-number equation [21](Hermentin, et al., 1996) has been expanded to account for the presence of penta-anionic N-glycans; Z = (S0%RPA x 0) + (S1%RPA × 1) + (S2%RPA x 2) + (S3%RPA × 3) + (S4%RPA x 4) + (−5 %RPA × 5). Asialylated N-glycans were not detected. Copies of individual chromatograms are included in Supplementary Information A. All values reported are means derived from triplicate analysis, summarised in Supplementary Table 1. “-“ = not detected. Capillary electrophoresis: relative peak areas (RPA) for peaks corresponding to different pI ranges. Copies of individual electropherograms are included in Supplementary Information B. All values reported are means derived from triplicate analysis, summarised in Supplementary Tables 2a-m. “–“ = not detected. Sialic acid content: all values reported are means derived from triplicate analysis, summarised in Supplementary Table 3. N.D = not determined.

216

B. Cowper et al. / Journal of Pharmaceutical and Biomedical Analysis 153 (2018) 214–220

217

Fig. 1. N-glycan mapping of released RFMS-labelled N-glycans from the Ph. Eur. erythropoietin for physicochemical tests CRS via AEX/HILIC. The approximate elution times of mono-, bi-, tri- and tetra-sialylated, and penta-anionic oligosacchairdes are annotated, as are structures which were identified via mass spectrometry (see Supplementary Information A).

performed using a mobile phase blend of sodium hydroxide (0.5 M)-sodium acetate (1 M)-water (30:12.5:57.5, v/v/v) and a flow rate of 1.0 ml/min. Sialic acid contents were determined by comparison with standardized NANA solutions. 2.7. In vivo bioassays In vivo bioassays were carried out according to the in vivo normocythaemic mouse bioassay protocol of the Chinese Pharmacopeia [17]. Assays were calibrated using the Chinese national standard for erythropoietin, which is itself calibrated against the WHO International Standard for erythropoietin, recombinant for bioassay, coded 11/170 [18]. Eight-week-old BALB/c mice were allocated to standard and sample groups in a fully randomized order, with five mice per treatment group. The standard and test rhEPO sample were diluted to appropriate concentrations with saline containing 0.1% (w/v) bovine serum albumin. A single dose of 7.5, 15 or 30 IU rhEPO/0.2 ml per mouse was injected subcutaneously on day 1. On day 4, blood was taken from the orbital venous sinus of each mouse and reticulocytes were counted using a R-500 Hematology Analyzer (Sysmex). Biological activity was calculated by comparing test sample response to the standard using a parallel line method. This study was approved by the Ethics Committee of National Institute for Food and Drug Control. 3. Results & discussion 3.1. N-glycan resolution and identification via AEX/HILIC-MS N-glycan analysis was performed using “mixed-mode” [19] anion-exchange/hydrophilic interaction (AEX/HILIC) chromatography, enabling separation of released N-glycans on the basis of both

charge and hydrophilicity. Multi-sialylated structures are sequentially separated into “clusters” in order of increasing negative charge, akin to a conventional anion-exchange chromatography method, which allows the relative quantities of multi-sialylated structures to be determined (Table 2). Within each charge cluster, N-glycans are further resolved according to their degree of branching and the occurrence of O-acetylated sialic acid and Nacetyllactosamine extensions (Fig. 1). In recent years sophisticated mass spectrometric approaches have enabled direct detection of these modifications in rhEPO samples [12,13]. However this method provides a novel approach to chromatographic separation of individual N-glycan structures which are fluorescently-labelled, which is comparable to that achieved through alternative high pH anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD) methods [16]. A further advantage of AEX/HILIC is that elution is achieved using a HILIC-type gradient, utilising volatile mobile phase conditions, which allows for direct coupling to a mass spectrometer and identification of individual N-glycan structures. Bi-, tri- and tetra-sialylated oligosaccharides were detected and identified via mass spectrometry in all rhEPO samples (Supplementary Information A). Although it was not possible to formally identify mono-sialylated N-glycans via mass spectrometry, the appearance of a peak cluster prior to the bi-sialylated N-glycans in all samples is strongly indicative of their presence. Asialylated/neutral N-glycans were not detected. The appearance of additional peak clusters following the tetra-sialylated N-glycans is suggestive of the presence of penta-anionic (visible in all samples) and hexa-anionic structures (visible in samples S03, S05, S08 and S09). Although it was not possible to formally identify these peaks via mass spectrometry, due to the poor m/z signal intensity of overlapping structures, they are likely to represent additionally sialylated (e.g. penta-sialylated) and/or sulphated glycans [16].

218

B. Cowper et al. / Journal of Pharmaceutical and Biomedical Analysis 153 (2018) 214–220

Fig. 2. The in vivo bioactivity (in International Units per milligram, IU/mg) of each rhEPO preparation (green) plotted alongside the relative peak areas (RPA%) of the peak corresponding to isoforms with pI 3.90-4.02 during capillary electrophoresis (CE) (blue) and the peak cluster corresponding to penta-anionic oligosaccharides during N-glycan mapping (red). %RPA values have been “normalised” to a 0–1 scale to facilitate overlaid plotting. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.2. Sialic acid content and distribution The sialic acid content varies slightly between samples, with an observed range of 10.1–14.5 mol sialic acid per mol EPO (Table 2); all above the minimum of 10 mol/mol EPO specified in the Chinese Pharmacopoeia (ChP) monograph for Recombinant Human Erythropoietin Injection (CHO cell) [17] and European Pharmacopoeia monograph for Erythropoietin concentrated solution [20]. However capillary electrophoresis (CE) and N-glycan mapping data reveal significant differences between samples in terms of the molecular distribution of sialic acid. For example, CE analysis shows that samples S05 and S08 are enriched with acidic isoforms, including elevated levels of a peak with pI ∼3.98 (8.5% and 10.9% relative peak area, RPA, respectively, compared with ≤5% in the other samples) with S08 also possessing elevated acidic isoforms at pI ∼3.83 and ∼3.69 (Table 2). S05 and S08 are also deficient in more basic isoforms, with a peak at pI 4.74 present at only 3.4 and 3.1% RPA respectively, compared to >6% in the other samples. These findings are supported by the N-glycan mapping data for S05 and S08, which possess unusually high levels of penta-anionic oligosaccharides (18.8 and 23.5% RPA respectively), and evidence of hexa-anionic structures, which together will enhance the overall isoform acidity of each rhEPO preparation. This observation is consistent with a strong overall correlation between the relative peak areas of acidic isoforms (pI < 4.0) via CE, and the presence of highly acidic penta/hexa-anionic N-glycans (Table 2 & Fig. 2). Additionally, samples which are enriched with more basic isoforms (pI ≥ 4.65), such as S02, S06, S09 and S11, generally possess reduced levels of pentaanionic N-glycans along with elevated levels of less acidic mono-, bi- and/or tri-sialylated N-glycans. Sample S09 is unique amongst the twelve preparations, as it presents CE peaks at pI ∼3.76 and ∼5.08 (both 2.6% RPA) and therefore has a particularly wide isoform distribution, which is supported by the presence of both elevated mono-sialylated (6.1% RPA) and penta-anionic (16.2% RPA) N-glycans. Sample S03 is also unusual in possessing elevated pentaand hexa-anionic N-glycans without significant enrichment of low pI (<4.0) peaks via CE. It is likely that their presence is offset by the relatively high levels of mono- and bi-sialylated structures, which together contribute to a narrow, more neutral isoform distribution via CE.

The presence of penta-anionic structures in rhEPO preparations is not commonly addressed in the literature. For example, they were not accounted for in the original proposed calculation of “Z-number”, a hypothetical summarisation of the relative abundances of multi-sialylated N-glycans [21]. However their presence in small quantities is reported in some therapeutic rhEPO products [16,22,23], and they are detected here, at 1.2% RPA, in the Ph. Eur. EPO for physicochemical tests CRS, a 50:50 mixture of originator products epoetin alfa and beta (Fig. 1 & Table 2). This correlates with the isoform distribution of the Ph. Eur. CRS, in which low and high pI forms are relatively deficient and prevalent respectively. With the exception of one sample, S12, all of the rhEPO preparations analysed contain significantly elevated penta-anionic N-glycans compared to the Ph. Eur. CRS. Hexa-anionic N-glycans have yet to be reported in literature concerning therapeutic rhEPO products, and are not detected in the Ph. Eur. CRS. However they are suspected to be present in several samples here, likely due to further sialylation or sulfation of penta-anionic structures. Nine of the twelve rhEPO preparations also contain elevated monosialylated N-glycans compared to the Ph. Eur. CRS, and there are also notable differences observed in bi-, tri-, and tetra-sialylated N-glycan levels in many samples. Overall there is a wide range of N-glycan distributions across the sample set, which are not necessarily reflected by their calculated Z-numbers (Table 2). For example, the S09 Z-number is very similar to S01, S02 and S06, but their multi-sialylated N-glycan distributions differ significantly. Therefore, whilst the Z-number provides a convenient summary of overall sialylation, it does not account for different sialic acid distributions and may not provide sufficiently in-depth information to compare biosimilar rhEPO preparations. Analysis of N-deglycosylated samples via HILIC-MS has also enabled characterisation and comparison of intact O-glycosylation (Supplementary Information C). Mono-, bi- and asialylated structures have been identified in all cases, and whilst their relative quantities vary slightly between rhEPO preparations, none differ significantly from the Ph. Eur. CRS. Considering the relatively small influence of the single O-glycan over total rhEPO sialylation, it is unlikely that the small differences observed have a significant impact upon drug pharmacokinetics.

B. Cowper et al. / Journal of Pharmaceutical and Biomedical Analysis 153 (2018) 214–220

219

3.3. O-acetylated and N-acetyllactosamine extended N-glycans The AEX/HILIC method developed for rhEPO N-glycan analysis achieves resolution of individual structures within multi-sialylated peak clusters, which are identifiable via mass spectrometry (Supplementary Information A). Previous studies have demonstrated that rhEPO N-glycans are susceptible to modification, most notably through O-acetylation of sialic acid, extension of one or more antennae with an additional N-acetyllactosamine (LacNAc, LN) disaccharide, and the presence of N-glycolyl- (Neu5Gc, NGNA) rather than N-acetyl- (Neu5Ac, NANA) neuraminic acid (i.e. sialic acid) [12,13]. N-glycans bearing Neu5Gc are believed to be present at low levels (∼1% total sialic acid) in rhEPO preparations [11] and have not been detected here, however multiple O-acetylated sialic acids and/or anntennal LacNAc extensions are present in all samples. Analysis of the Ph. Eur. CRS provides the N-glycan distribution of a prototypical therapeutic rhEPO preparation (Fig. 1). In each peak cluster the most prominent peak is the basic multi-antennary core structure containing one sialic acid per antenna (i.e. S2NA2F, S3NA3F, S4NA4F). Further structures with one or more asialylated antennae (e.g. bi-sialylated tri-/tetra-antennary, S2NA3F/S2NA4F, and tri-sialylated tetra-antennary, S3NA4F) are also detected in reduced quantities. O-acetylated sialic acid is present on bi-, triand tetra-sialylated N-glycans, with the latter bearing as many as four O-acetyl modifications. LacNAc extensions are also detected in each peak cluster; present on up to three antennae of tetrasialylated, tetra-anntennary N-glycans. When compared with the Ph. Eur. CRS, the various rhEPO preparations possess a wide range of N-glycan profiles (Fig. 3 & Supplementary Information A). Samples S02 and S04 most closely resemble the Ph. Eur. CRS, whereas some samples contain comparatively increased O-acetylated sialic acid (S01, S09, S10, S11), or LacNAc-extended structures (S03, S05, S06, S09). Samples S07 and S12 are comparatively deficient in both modifications, whilst S09 is completely deficient in O-acetylated N-glycans. Numerous samples also contain elevated levels of bi/tri-sialylated structures with one or more additional asialylated antennae (S04–06, S08–11). The differences observed between the rhEPO preparations in terms of antennarity, sialylation, O-acetylation and LacNAcextension likely derive from differences in their respective manufacturing processes. Although all rhEPO preparations were produced in Chinese hamster ovary (CHO) cells, small differences or changes in pH, temperature, culture method or media can still have a significant influence over the glycosylation profile of a recombinant protein [11]. The twelve preparations were produced using a mixture of roller bottles and bioreactors, however there are no particularly strong correlations between the system used and resulting glycosylation profile. O-acetyl sialic acid and LacNAcextended structures are elevated and depleted in preparations produced using either method. Penta-/hexa-anionic N-glycans are most prevalent in samples S05, S08 and S09, which were produced using a bioreactor, however these structures are also visible in sample S03, which was produced in roller bottles. Therefore it appears that these factors are more strongly influenced by the precise composition of the culture medium. 3.4. Trends in glycosylation and biological activity The close relationship between rhEPO biological activity and glycosylation has long been established [8–10]. The in vivo bioactivity of each preparation has been determined here (Supplementary Table 4), in order to evaluate the impact of the observed variation in glycosylation upon the rhEPO structure-activity relationship. Bioactivities ranged from 100,905–122,957 IU/mg, suggesting that the significant divergences in N-glycan profiles of the twelve preparations are not reflected in major differences in in vivo bioactivity

Fig. 3. Expanded AEX/HILIC chromatograms of selected rhEPO samples (S01, S02, S06, S07 and S09), highlighting the differences observed between the glycosylation profiles of different samples. The lower chromatogram is annotated with the positions of the tri- (S3), and tetra- (S4) sialylated and penta-anionic (-5) peak clusters as well as the tetra-sialylated, tetra-antennary N-glycan (S4NA4F) and the relative positions of its O-acetylated and LacNAc (N-acetyllactosamine)-extended derivatives. Full-scale individual chromatograms for each rhEPO preparation are included in Supplementary Information A.

in mice. rhEPO in vivo bioassays are inherently variable; the ChP monographs for rhEPO products state that potencies must be within 80–140% of the stated potency [17], whilst the Ph. Eur. monograph permits confidence intervals of 64% and 156% [20]. Within these bounds, the bioactivities of the twelve rhEPO preparations can be interpreted as approximately equal. Nonetheless there does appear to be a trend between in vivo bioactivity and the presence of elevated levels of acidic N-glycans and isoforms (Fig. 2). Samples S05, S08 and S09 all possess unusually high levels of suspected penta-anionic N-glycans. The increased acidity of these preparations will theoretically lead to reduced clearance in vivo, which is indeed reflected by slightly elevated bioactivities. Otherwise, the more discrete glycan modifications observed here, such as variations in sialic acid O-acetylation and LacNAc extension, do not give rise to any particular trends in biological activity. However it should be noted that investigations into biological activity in this study were restricted to use of a mouse model in vivo bioassay from the Chinese Pharmacopoeia rhEPO monograph, which does not allow for a particularly in-depth assessment of the pharmacology of the

220

B. Cowper et al. / Journal of Pharmaceutical and Biomedical Analysis 153 (2018) 214–220

various rhEPO preparations. Incidentally, during a recent study of a proposed manufacturing change for a therapeutic rhEPO, it was observed that increases in LacNAc-extended antennae led to elevated potency in chronic kidney disease patients, suggesting that this modification may in fact have wider pharmacological implications in humans [24]. Interestingly, samples S07 and S12 possess significantly larger sialic acid contents than the other ten samples (Table 2) but this is not reflected by elevated in vivo bioactivities. Both samples are particularly enriched with tetra-sialylated N-glycans (>70% RPA), but are also relatively deficient in penta-anionic structures and low pI isoforms compared with the more bioactive samples S05, S08 and S09. Taken together, this suggests that these samples, in spite of their elevated sialic acid contents, are not sufficiently equipped to prolong in vivo circulation. This neatly demonstrates that sialic acid content alone is not a major determinant of in vivo bioactivity of rhEPO, and that the distribution of sialic acid, in particular the maximal sialylation/negative charge of tetra-antennary structures, is a critical attribute. 4. Conclusions This study has utilised several complimentary analytical techniques to demonstrate the breadth of glycosylation profiles observed in licensed rhEPO products available in China and Japan. This includes a novel AEX/HILIC-MS method, which conveniently combines simultaneous charge- and polarity-based N-glycan separation (allowing rapid quantitation of multi-charged N-glycan populations) with simple mass spectrometric identification of individual structures. The twelve preparations exhibit significant differences in terms of the occurrence of sialic acid O-acetylation, extension of N-glycan antennae with N-acetyllactosamine units, and the distribution of sialic acids across multi-antennary structures, including the suspected presence of penta- and hexa-anionic (sialylated/sulfated) oligosaccharides in some cases. Aside from slightly increased in vivo bioactivities in samples enriched with acidic isoforms/N-glycans, the observed differences do not confer significant effects upon biological activity in the mouse bioassay. Nonetheless, the varied presence and scale of glycan modifications observed in this selection of rhEPO samples demonstrates the importance of monitoring glycosylation in biological medicines, particularly in the context of the growing biosimilars market, in order to ensure continued safety and efficacy. CHO cells used to produce rhEPO are known to contain potentially harmful glycan epitopes, such as N-glycolyl sialic acid, which is immunogenic in humans [11]. Close monitoring of glycosylation in new and existing rhEPO products, using methods such as those described in this study, will hopefully ensure that significant occurrences of such harmful moieties are avoided in the future. Acknowledgements This work was financially supported by grants from the National Science and Technology Major Project (No. 2015ZX09501008-001). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.jpba.2018.02.043.

References [1] W. Jelkmann, Biosimilar epoetins and other follow-on biologics, Am. J. Haematol. 85 (2010) 771–780. [2] Z. Shahrokh, L. Royle, R. Saldova, J. Bones, J.L. Abrahams, N.V. Artemenko, S. Flatman, M. Davies, A. Baycroft, S. Shegal, M.W. Heartlein, D.J. Harvey, P.M. Rudd, Erythropoietin produced in a human cell line (Dynepo) has significant differences in glycosylation compared with erythropoietins produced in CHO cell lines, Mol Pharmaceutics 8 (2011) 286–296. [3] A. Mikhail, M. Farouk, Epoetin biosimilars in europe: five years on, Adv. Ther. 30 (2010) 28–40. [4] W. Jelkmann, The ESA scenario gets complex: from biosimilar epoetins to activin traps, Nephrol. Dial. Transplant. 30 (2015) 553–559. [5] F. Bocquet, P. Paubel, I. Fusier, A.-L. Cordonnier, M. Sinegre, C. Le Pen, Biosimilar versus patented erythropoietins: learning from 5 years of european and japanese experience, Appl. Health Econ. Health Policy 13 (2015) 47–59. [6] J.M. Davis, T. Arakawa, T.W. Strickland, D.A. Yphantis, Characterization of recombinant human erythropoietin produced in chinese hamster ovary cells, Biochemistry (1987) 2633–2638. [7] P.H. Jensen, N.G. Karlsson, D. Kolarich, N.H. Packer, Structural analysis of Nand O-glycans release from glycoproteins, Nat. Protoc. 129 (2012) 1299–1310. [8] E. Tsuda, G. Kawanishi, M. Ueda, S. Masuda, R. Sasaki, The role of carbohydrate in recombinant human erythropoietin, Eur. J. Biochem. 40 (1990) 405–411. [9] S. Erbayraktar, G. Grtasso, A. Sfacteria, Q.-W. Xie, T. Coleman, M. Kreilgaard, L. Torup, T. Sager, Z. Erbayraktar, N. Gokmen, O. Yilmaz, P. Ghezzi, P. Villa, M. Fratelli, S. Casagrande, M. Leist, L. Helboe, J. Gerwien, S. Christensen, M.A. Geist, L.O. Pederson, C. Cerami-Hand, J.-P. Wuerth, A. Cerami, M. Brines, Asialoerythropoietin is a nonerythropoietic cytkine with broad neautoprotective activity in vivo, PNAS (2003) 6741–6746. [10] M.N. Fukuda, H. Sasaki, L. Lopez, M. Fukuda, Survivial of recombinant erythropoietin in crculations: the role of carbohydrates, Blood (1989) 84–89. [11] P. Hossler, S.F. Khattak, Z. Jian Li, Optimal and consistent protein glycosylation in mammalian cell culture, Glycobiology (2009) 936–949. [12] A. Harazono, N. Nashii, R. Kuribayashi, S. Nakazawa, N. Kawasaki, Mass spectrometric glycoform profiling of the innovator and biosimilar erythropoietin and darbepoietin by LC/ESI–MS, J. Pharm. Biomed. Anal. (2013) 65–74. [13] Y. Yang, F. Liu, L.A. Halim, H. Schellekens, A. Heck, Hybrid mass spectrometry approaches in glycoprotein analysis and their usage in scoring biosimilarity, Nat. Commun. 1339 (2016) 13397–13407. [14] R.I. Thomson, R.A. Gardner, K. Strohfeldt, D.L. Fernandes, G.P. Stafford, D.I.R. Spencer, H.M.I. Osborn, Analysis of three epoetin alpha products by LC and LC-MS indicates differences in glycosylation critical quality attributes, including sialic acid content, Anal. Chem. 645 (2017) 6455–6462. [15] L. Zhang, S. Luo, B. Zhang, Glycan analysis of therapeutic glycoproteins, MABS (2016) 205–215. [16] S. Kandzia, N. Grammel, E. Grabenhorst, H.S. Conradt, Quantitative N-glycan mapping of glycoprotein therapeutics by HPAEC-PAD: glycosylation characteristics of different recombinant human EPO products, in: T. Noll (Ed.), Cells and Culture: Proceeding of the 20th ESACT Meeting, Springer New York, 2010, pp. 867–871. [17] Chinese Pharmacopoeia Commision, Pharmacopoeia of the People’s Republic of China, vol. III, tenth ed., China Medical Science and Technology Press, Beijing, 2015. [18] C. Burns, R. Tiplady, J. Hockley, WHO International Collaborative Study of the Proposed 3rd International Standard for Erythropoietin, Recombinant, for Bioassay, WHO Expert Committee on Biological Standardisation, World Health Organisation, Geneva, 2012. [19] K. Zhang, X. Liu, Mixied-mode chromatography in pharmaceutical and biopharmaceutical applications, J.Pharm. Biomed. Anal. (2016) 73–88. [20] The Directorate for the Quality of medicines of the Council of Europe (EDQM), European Pharmacopoeia, ninth ed., Council of Europe, Strasbourg, 2017. [21] P. Hermentin, R. Witzel, E.-J. Kanzy, G. Diderrich, D. Hoffman, H. Metzner, J. Vorlop, H. Haupt, The hypothetical N-glycan charge: a number that characterises protein glycosylation, Glycobiology (1996) 217–230. [22] N. Grammel, F. Dittrich, S. Kandzia, E. Grabenhorst, H.S. Conradt, Biosimilarity of recombinant human EPO products from CHO cell lines: a carbohydrate structural view, in: T. Noll (Ed.), Cells and Culture: Proceeding of the 20th ESACT Meeting, Springer New York, 2010, pp. 861–865. [23] M. Mattio, N. Ceaglio, M. Oggero, N. Perotti, I. Amadeo, G. Orozco, G. Forno, R. Kratje, M. Etcheverrigaray, Isolation and chatacterization of a subset of erythropoietin glycoforms with cytoprotective but minimal erythropoietic activity, Biotechnol. Prog. 101 (2011) 1018–1028. [24] G. Grampp, P.L. McElroy, G. Camblin, A. Pollock, Structure-function relationships for reconbinant erythropoietins: a case study from a proposed manufacturing changhe with implications for erythropoietin biosimilar study designs, J. Pharm. Sci. (2010), http://dx.doi.org/10.1016/j.xphs.2018.01.018. [25] I. Abraham, K. MacDonald, Clinical efficacy and safety of HX575, a biosimilar recombinant human erythropoietin in the management of anemia, Biosimilars (2012) 13–25.