Intact and Native Mass Analysis of Glycoproteins Marshall Bern 1 Yong J. Kil 1 Tomislav Caval 2 Vojtech Franc 2 Albert J.R. Heck 2 1 Protein Metrics Inc. 2 Biomolecular Mass Spectrometry and Proteomics, Utrecht University Contact: [email protected] Introduction www.proteinmetrics.com Results: EPO Methods – Experimental Methods – Computational Conclusions Intact mass analysis provides a simultaneous and quantitative view of a protein’s major proteoforms, including variations in glycosylation. Intact mass spectra vary with solution conditions: • Denaturing conditions give fewer adducts, higher charge states, and stronger overall signal. • Native conditions preserve noncovalent binding and original folded structure of the proteins, which results in fewer and lower charge states. Lower charge state provides more space between adjacent charge states, allowing separation of protein ions coming from a heterogeneous mass distribution. After data acquisition, m/z spectra are deconvolved to neutral mass spectra. This computational step is prone to errors and artifacts, especially for the complex signals of glycoproteins. For the past 25 years, almost all charge deconvolution has been done with the MaxEnt algorithm (Ferrige et al, 1992), which is offered in various implementations by MS instrument vendors. Here we demonstrate the utility of native MS and a new “parsimonious” charge deconvolution algorithm for the analysis of complex glycoproteins. Properdin / Factor P Cetuximab References and Acknowledgments Protein Metrics Intact mass software can deconvolve wide m/z and mass ranges with minimal artifacts. Colored dots connect neutral masses to m/z peaks and let the user validate masses visually. Assignments may be made manually or auto- matically from calculated masses or mass deltas. IdeS and DTT digested Cetuximab shows LC at ~23 kDa, Fc/2 at ~25 kDa, Fd at ~27 kDa, and Fc at ~51 kDa. We analyzed the following samples: • Recombinant human erythropoietin EPO BRP • Purified human Properdin / Factor P • Cetuximab (trade name: Erbitux) We chose these targets for their complex glycosylation. Cetuximab is, to our knowledge, the only therapeutic mAb currently on the market with Fab glycosylation. Native MS was performed by direct infusion on an Exactive Plus Orbitrap instrument with extended mass range (EMR) (Thermo Fisher Scientific) using a standard m/z range of 500-10,000. LC-MS/MS was performed on an Orbitrap Fusion with EThcD fragmentation. More details are described in a previous publication (Yang et al, 2016). Like other charge deconvolution programs (MaxEnt, Bayesian, UniDec), Protein Metrics Intact starts with an initial guess of the charges in the m/z spectrum, computes an initial m spectrum along with charge histogram for each mass, and then iteratively improves the m spectrum and charge histogram until together they produce a computed m/z spectrum close to the observed one. MaxEnt aims for a high-entropy m spectrum to improve resolution, but can also produce artifacts. The new algorithm splits the problem in two: charge inference, which aims for a minimal or parsimonious set of masses, and Richardson-Lucy peak sharpening for resolution. Thermo Exactive Plus EMR was used for native MS analysis Color indicates neutral monosaccharide composition; number gives # NeuAc’s Charge states slightly overlap in native MS spectrum of EPO. Too tall Too small Previous work (Yang et al, 2016) gave a comprehensive analysis of rhEPO glycosylation with native mass spectrometry and glycopeptide profiling using trypsin and GluC digests. Complex native mass spectra ( black borders) have mildly overlapping charge states. Charge deconvolution was performed by Bayesian Protein Reconstruct tool from BioAnalyst (Sciex) (orange). Here we reanalyze the data with Protein Metrics Intact ( blue). Overall agreement between Sciex Bayesian deconvolution and PMI Intact is good, but PMI Intact’s peak intensities agree more closely with the major charge state in the m/z spectrum, as shown at lower right. Correctness of intensities of masses below 28 kDa is hard to judge due to overlap around m/z 3100. PMI Intact Deconvolution Sciex Deconvolution PMI Intact Sciex 28,326 too small? Too tall? Unfolded Native monomer Dimer Properdin is a component of the Complement system with thrombospondin repeats, C-mannosylation, O-fucosylation, and N-glycosylation. Previous work (Yang et al, 2016) gave an analysis of properdin relying on an m/z spectrum rather than a neutral mass spectrum due to the difficulty of deconvolution. Zoom of wide- range Thermo deconvolution Zoom of wide- range PMI deconvolution Monomer z = 14+ Narrow mass range with salt adducts No adducts Deconvolution with wide m/z and mass ranges (right) surveys the proteoforms. Thermo Deconvolution (lower left) gives incorrect peak intensities and widths. Deconvolution with a narrow m/z and mass range (lower right) gives assignable monomer masses. Interestingly, tri-antennary N-glycans are observed only on proteoforms with 15 C- mannosylations. Intact mass analysis is routinely used to check primary sequence, heavy / light chain pairing, and glycosylation in both intact and reduced mAbs. Zoom of 26 – 28 kDa Fd Exactive EMR spectrum Zoom of 23 – 28 kDa mass range Fc with G0F G1F + C-terminal Lys LC Fd Zooms of the neutral mass spectrum show ordinary mAb glycosylation on the Fc, but more complex N-glycans on the heavy chain variable region (Fd), including Gal-α-Gal, antennal fucosylation, and tri-antennary glycans. The Fd mass spectrum is in close agreement with previous work (Janin- Bussat et al). Notice, however, that the Orbitrap spectrum resolves the 27,525 Da mass, and the peaks at 27,688 and 27,831 are mislabeled (100 Da off) in Janin-Bussat, missing the antennal Fuc. QTOF mass spectrum from Janin-Bussat et al. For the past 25 years, charge deconvolution has been performed almost exclusively by some version of MaxEnt. High-resolution native MS and the need to analyze more complex molecules motivated the development of a new deconvolution algorithm with the following advantages: • Wider applicability (any MS instrument, any molecule type) • Fewer and smaller algorithmic artifacts • Optional peak sharpening for greater fidelity to raw data Although the samples were chosen primarily for technology development, the studies revealed some novel characteristics of properdin and Cetuximab. • Properdin includes tri-antennary glycans, seemingly only on proteoforms with 15 C-mannosylations, evidence of correlated PTMs. • Cetuximab Fab glycosylation includes antennal fucose Ferrige, A.G.; Seddon, M.J.; Green, B.N.; Jarvis, S.A.; Skilling, J.; Staunton, J. Disentangling electrospray spectra with maximum entropy, Rapid Comm Mass Spec, 1992. Yang, Y.; Liu, F.; Franc, V.; Halim, L. A.; Schellekens, H.; Heck, A. J. R. Hybrid mass spectrometry approaches in glycoprotein analysis and their usage in scoring biosimilarity. Nat. Commun, 2016. Janin-Bussat, M.-C., et al. Cetuximab Fab and Fc N-Glycan Fast Characterization Using IdeS Digestion and Liquid Chromatography Coupled to Electrospray Ionization Mass Spectrometry. In Glycosylation Engineering of Biopharmaceuticals: Methods and Protocols; Beck, A., Ed.; Humana Press: Totowa, NJ, 2013; pp 93–113. A.J.R.H acknowledges support from the Netherlands Organization for Scientific Research (NWO) funding the large-scale proteomics facility Proteins@Work. We thank Genmab (Utrecht) for providing us with mAbs including Cetuximab and Daclizumab.