Introduction

Erythropoietin (EPO) is an essential glycoprotein for hormone red cell production. Recombinant human Erythropoietin (hrEPO) is used for treatment of anemia resulting from chronic kidney disease. EPO has three N-glycosylation sites and one O-glycosylation site which leads to high heterogeneity, making glycoform profiling challenging. Multiple formulations of hrEPO have been studied extensively by different approaches, including capillary electrophoresis (CE) and nanoflow liquid chromatography (nLC) coupled to advanced mass spectrometry (MS) analysis. Although CE-MS has been used to characterize intact EPO glycoforms, quantitative mapping of each glycosylation site by CE-MS has not been previously accomplished. Here, we report that sheathless CE- and nLC-based separations coupled to Orbitrap middle-down MS provide complementary information and allow for comprehensive site-specific glycan mapping of rhEPO.

Methods

Reduced and alkylated rhEPO (Erythropoietin-Alpha, ProSpec, NJ) was digested with LysC (Roche, CA).  Digested rhEPO was analyzed on an LTQ Orbitrap Elite using CESI-MS or LC/MS coupled with high resolution high mass accuracy (HRAM) FT scanning of ETD or HCD MS2 fragment ions. Low flow sheathless CESI (Beckman Coulter, CA) employed both neutral coated and bare fused silica capillaries with integrated porous tip ESI sprayers. ProsightPC 3.0, Protein Deconvolution 3.0, Proteome Discoverer 2.0 (Thermo Fisher Scientific) and Byonic (ProteinMetrics, CA) software packages were used for data analysis. A custom EPO glycan database was generated using proteinase K (Roche) digest and WAX RP-nLC-MS/MS analysis on Orbitrap Fusion. SimGlycan (PremierBiosoft, CA) was used for glycan identification.

Preliminary results

To generate the EPO glycan database, we used proteinase K to preserve site localization information. Short proteinase K glycopeptides were separated by a novel mixmode WAX-RP column. (Thermo Fisher), detected by HCDpdCID analysis on an Orbitrap Fusion and identified by SimGlycan.  Identified glycans were used to compile the custom glycan database for glycopeptides identified by Byonic or Prosight PC.

To perform complete quantitative glycan site-specific mapping of rhEPO, we employed limited Lys-C digest to yield one glycosylation site per peptide fragment of 3-10 kDa size for middle-down analysis. This enabled more complete sequencing of glycoforms compared with top-down analysis. The rhEPO glycopeptides were well separated from non-glycosylated peptides by CESI and resolved over 20 min in a 50 min long run. Glycoform resolution was mostly based on differences in the number of sialic acid residues. The two predominant O-glycosylated peptides (N-acetylhexosamine-hexose with one or two sialic acids) migrated as completely resolved peaks. Due to high efficiency separation, we identified 14 glycoforms on Ser¹²⁶ by CESI-MS but only 4 by LC-MS using a Magic C18 AQ nano-LC column.  Although our method demonstrated high efficiency separation and detection sensitivity of glycopeptides, we were initially unable to identify double N-glycosylated peptides containing Asn²⁴ and Asn³⁸. Ultimately, we determined our EPO sample protein sequence was different from the one provided by the manufacturer. Using HRAM ETD and HCD fragmentation in combination with manual de novo sequencing and wild card searches (Byonic), we were able to identify the actual sequence of the EPO standard and N-linked glycopeptides. These results suggested that CESI-HRAM middle-down is a sensitive and efficient method for characterization of proteins or biopharmaceuticals with multiple glycosylated endogenous proteoforms.