Unambiguous identification and quantification of galactose-α-1,3-galactose as a critical quality attribute with mass spectrometry and exoglycosidases

ABSTRACT

Alpha-galactosylation, galactose-α-1,3-galactose (α-Gal), is always regarded as a critical quality attribute due to its potential to provoke immunogenic responses in patients. Consequently, monitoring alpha-galactosylation in therapeutic proteins is essential, but current analytical techniques fall short in terms of identification sensitivity and quantification accuracy. Specifically, the released glycan assay by hydrophilic interaction liquid chromatography-fluorescence-mass spectrometry, the gold standard for glycan separation/identification, faces challenges due to ambiguities with isomeric glycan structures. To address these challenges, we developed a comprehensive analytical method that enhances both the identification sensitivity and quantification accuracy for α-Gal. We developed an integrated workflow that combines an advanced mass spectrometry technique – parallel reaction monitoring triple-stage mass spectrometry – with exoglycosidase sample treatment. This approach generates structurally specific signature ions, enhances identification sensitivity, enables the separation of α-Gal from its isomers, and improves quantification accuracy. By employing this more sensitive analytical approach without ambiguity in assignment for common glycan structures found in monoclonal antibodies, the safety of therapeutic proteins can be better assured, effectively minimizing the risk of α-Gal-induced immunogenicity.

Introduction

Glycosylation significantly influences the safety, immunogenicity, bioactivity and pharmacokinetics (PK) of biotherapeutics,¹ making it a critical quality attribute (CQA) that must be assessed to ensure product quality, safety, and efficacy.^2,3 Although glycosylation typically has many forms, the incorporation of galactose-α-1,3-galactose (α-Gal) is particularly notable due to its immunogenic properties. α-Gal (Figure S1) is a common glycan found in many mammals, excluding primates, and is especially prevalent in murine cell lines and Chinese hamster ovary (CHO) cells, which are widely used in the production of therapeutic proteins.⁴ α-Gal present in food and biotherapeutics can trigger immunoglobulin E (IgE)-mediated anaphylactic responses in humans, leading to severe reactions known as α-Gal syndrome.^5–7

The α-Gal on biantennary glycans can exist as mono-α-galactosylated glycans (mono-α-Gal) or bi-α-galactosylated glycan (bi-α-Gal) (Figure S1). Research indicates that bi-α-Gal in both the antigen-binding fragment (Fab) and the crystallizable fragment (Fc) regions can bind IgE,^8–10 and high levels of anti-α-Gal antibodies circulate in humans. Consequently, even low levels of α-Gal ( < 1%) can pose a risk of hypersensitivity, compromising patient safety.⁸ Moreover, although the immunogenicity of α-Gal in the Fc regions is not fully understood,² its impact on binding FCɾRIII receptor and anti-α1–3-Gal IgE antibodies underscores the need for comprehensive and effective α-Gal characterization. It is essential to minimize the presence of this epitope during biologics development and manufacturing. A detailed glycan profile,¹¹ including α-Gal analysis, is required during the process development of therapeutic proteins and when filing biologics applications with health authorities. Thorough glycan characterization^12,13 and proper control of α-Gal are crucial for ensuring proper safety and efficacy of biotherapeutics.

Bosques et al. utilized a functional ortholog of galactosyltranferase to demonstrate the presence of α-Gal in the biotherapeutic fusion protein abatacept (Orencia, CTLA4-IgG).⁴ Similarly, Qian et al. used on orthogonal matrix-assisted laser desorption/ionization hybrid quadrupole–quadrupole time of flight (oMALDI Qq-TOF) to characterize oligosaccharides in the monoclonal antibody cetuximab.⁹ Such studies emphasize the importance of advanced analytical tools and reliable detection techniques in glycan characterization. Released glycan hydrophilic interaction liquid chromatography-fluorescence-mass spectrometry (HILIC-FLR-MS) is one such tool, characterizing released glycans with MS1 identities. However, HILIC-FLR-MS cannot easily distinguish α-Gal from hybrid and bisecting glycan isomers (Figure 1(a)), leading to misidentification and overestimation of α-Gal levels. Tandem mass spectrometry (MS2) fragmentation has been used to generate oxonium signature ions, aiding in the differentiation of α-Gal. Previous studies have utilized the relative ratio of these signature ions, predominantly through high-energy collision dissociation (HCD), to identify α-Gal.¹⁴ For instance, She et al. demonstrated that the relative ratios of MS2 signature ions are indicative of α-Gal identification (Figure S2).¹⁵ However, in our evaluation of α-Gal data, MS2 pattern discrepancies were observed between collision-induced dissociation (CID) and HCD fragmentations, as well as among different charge states (Figure S3). Furthermore, the low MS2 signal intensity posed concerns regarding the reliability and quality of the analysis. Therefore, these methods fall short of definitively distinguishing α-Gal from isomeric hybrid and bisecting glycans. Nuclear magnetic resonance (NMR) can distinguish structural isomers,¹⁶ but its effectiveness is limited by low resolution and ambiguities caused by overlapping signals.

Figure 1.

MS2 and MS3 fragment ions of an α-Gal example and its isomers. (a). The InstantPC-labeled released glycans from fusion1 using HILIC-FLR-MS showed co-eluted α-Gal isomers, along with an example spectrum of α-Gal MS2. (b). The presence of MS3 fragment ion at m/z 325.11 can confirm an α-Gal identity unambiguously in therapeutic proteins.

To address these limitations, we integrate parallel reaction monitoring with triple-stage mass spectrometry (PRM-MS3) to enhance α-Gal isomer differentiation and identification sensitivity. Initially developed for targeted proteomics,^17,18 PRM selectively isolates precursor ions with a specific mass-to-charge ratio (m/z) in the quadrupole, fragments them, and detects the fragment ions in the orbitrap.¹⁹ The specific trapping capabilities of the quadrupole-ion trap/orbitrap instrument distinguish analyte fragment ions from background signals, thereby improving the selectivity for low abundant analytes.²⁰ Meanwhile, MS3, previously utilized in isobaric labeling multiplexed proteomics and single-cell proteomics,^21,22 offers well-evaluated parameters that can be leveraged for glycan studies. Although the combination of PRM and MS3 has not been extensively applied in the characterization of therapeutic biologicals, these techniques hold promise in enriching the low abundant signature MS2 fragment ions (m/z 528.19) from α-Gal and generating structurally unique MS3 diagnostic ions (m/z 325.11) in therapeutic proteins. We hypothesize that the PRM-MS3 technique will effectively overcome analytical challenges, providing a robust assay for unambiguous α-Gal identification. α-Gal (trans) and β-galactose (β-Gal, cis) differ in the glycosidic bond orientation at the anomeric carbon (C1) relative to the CH2OH group at C5 (Figure S1).²³ To the best of our knowledge, mass spectrometry is unable to distinguish this orientation difference.

Sialidase A, α1–3,6 galactosidase and α1–2,3,6 mannosidase are exoglycosidases that serve as valuable tools for the study of N-linked glycans. Sialidase A catalyzes the hydrolysis of all non-reducing terminal N-Acetylneuraminic acid (NANA) and N-Glycolylneuraminic acid (NGNA),²⁴ thereby simplifying the analysis of α-Gal by eliminating the heterogeneity associated with sialic acid groups. The enzyme α1–3,6-galactosidase catalyzes the hydrolysis of terminally linked α1–3 and α1–6 galactose residues, facilitating the confirmation of α-Gal identity by generating a retention-time shift following the removal of terminal α-galactoses. According to Ren et al., treatment with exoglycosidase facilitates the quantitative characterization of α-Gal, where the quantitative difference of α-Gal peak area before and after α1–3,4,6 galactosidase treatment indicates the abundance of α-Gal.²⁵

Failing to account for digestion efficiency, however, can lead to an underestimation of α-Gal levels, as undigested α-Gal is excluded from analysis. This omission can expose patients to low levels of α-Gal and raise the risk of anaphylaxis. The enzyme α1–2,3,6 mannosidase catalyzes the hydrolysis of terminally linked α1–2, α1–3, and α1–6 mannose residues,²⁶ assisting in the separation of α-Gal from its hybrid and bisecting glycan isomers and improving the accuracy of quantification. Our preliminary results showed poor digestion efficiency for glycans attached to proteins or peptides. Moreover, fluorescence labeling can impair enzyme activity on fluorescence-labeled released glycans, highlighting the need for improved digestion efficiency. This study focuses on optimizing several aspects of the enzyme treatment process, including the enzyme application step, the enzyme-to-protein ratio, digest time, and the combination of various enzyme reactions to reduce sample preparation time. Efficient exoglycosidase treatments can aid in confirming the presence of α-Gal and allow more accurate quantification. Therefore, we propose the inclusion of all three described exoglycosidases, sialidase A, α1–3,6 galactosidase and α1–2,3,6 mannosidase, to improve both the efficiency and effectiveness of quantification.

In this study, we introduce a comprehensive method that offers a more reliable identification and accurate quantification of α-Gal. This approach allows us to differentiate α-Gal isomers and achieves unprecedented low limits of detection (LOD) and quantification (LOQ) at 0.1%. Enabling more effective identification and quantification of α-Gal provides significant value to the development of therapeutic proteins and the biopharmaceutical industry by mitigating safety and immunogenicity risks through reduced exposure to α-Gal.

Materials and methods

Materials

The monoclonal antibody 1 (mAb1, IgG1), mAb2 (IgG1), mAb3 (IgG4), fusion protein-1 (fusion1) and fusion protein-2 (fusion2) were produced in-house (Genentech, South San Francisco, CA) (Figure S1). Advancebio Gly-X N-Glycan Prep with InstantPC Kit, sialidase A (GK80040), InstantPC-labeled G2F with mono-α-Gal (GKPC-404) and InstantPC-labeled Man-6 controls (GKPC-104) were purchased from Agilent (Santa Clara, CA). All α1–3,6 galactosidase (P0731S), α1–2,3,6 mannosidase (P0768S) and α1–6 mannosidase (P0727S) were obtained from New England Biolabs (Ipswich, MA). α-1,3-galactosyltransferase (7887-GT) was purchased from R&D systems (Minneapolis, MN). All reagent solutions were prepared in LCMS-grade water. The Acquity HILIC glycan BEH amide column (2.1 mm × 150 mm, 1.7 μm, 130 Å) was obtained from Waters (Milford, MA). LCMS-grade Acetonitrile (ACN) and formic acid (FA) were from Fisher Scientific (Waltham, MA).

Released glycan InstantPC labeling

For released glycan analysis, Advancebio Gly-X N-Glycan prep with InstantPC kit (Agilent) was used according to manufacturer recommendations. In short, samples were diluted to 2 mg/mL protein with water, denatured with 2 μL denaturant, and incubated at 90°C for 3 min. After cooling to room temperature, the samples were incubated with 2 μL PNGaseF working solution at 50°C for 5 min. The materials were then labeled with 5 μL of InstantPC dye solution at 50°C for 1 min. Buffer exchange to the eluent was performed with the cleanup plate. The samples were either analyzed directly or stored at −70°C for 3 weeks.

HILIC-PRM-M3 method

Vanquish HPLC systems (Thermo Fisher Scientific) and HILIC glycan BEH amide columns (2.1 mm × 150 mm, 1.7 μm, 130 Å) (Waters) were used for the glycan separation. The fluorescence-labeled released glycan materials were injected (injection amount equals to glycans from 4 μg antibody) onto the system. The gradient elution was conducted over a 91-min period with the column temperature controlled at 60°C. The solvent program utilized 50 mM ammonium formate, pH 4.4 (Solvent A) and 0.1% FA in ACN (Solvent B), with a flow of 400 μL/min (0–74.8 min: 75–59% B, 74.8–83 min: 59–5% B, 83–84 min: 5% B, 84–84.5 min: 5–75% B, 84.5–91 min: 75% B). An Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific) was coupled to the HPLC system. MS source conditions were: sheath gas flow 30 L/min, aux gas flow 10 L/min, spray voltage 3500 V, capillary temperature 320°C. For PRM, doubly charged ions at m/z 1105.94 and triply charged ions at m/z 737.62 were selected and subjected to MS2 fragmentation using CID in the ion trap. Subsequently, the singly charged MS2 ions at m/z 528.19 were isolated and further fragmented by MS3 using CID in the ion trap. Both MS and MS2 ions were isolated using a 2 m/z window and fragmented with 30% CID collision energy and 20 ms CID activation time. The automatic gain control (AGC) target for MS3 ions was set to 3e4, and the maximum ion injection time was at 1000 ms. The detector for MS3 ions was the ion trap with 1 microscan.

Exoglycosidase treatment

The InstantPC-labeled released glycan samples were dried in a SpeedVac, each resuspended in water, and split three ways for subsequent reactions. Following the operation suggestion of the enzyme vendor, the pH for α1–3,6 galactosidase (New England Biolabs) was adjusted to 5.5 with the presence of 10 mg/mL bovine serum albumin in the reaction. The pH for α1–2,3,6 mannosidase (New England Biolabs) was adjusted to 4.5 with the presence of 2 mM Zn2+ in the reaction. For both enzymes, the samples were incubated at 37°C overnight. The exoglycosidase digestion efficiency in this study was calculated using the following formula of relative area (RA):

[(RA_α-Gal in sialidase A only – RA_α-Gal in sialidase A/galactosidase co-treatment) + (RA_α-Gal in sialidase A only – RA_α-Gal in sialidase A/mannosidase co-treatment)]/RA_α-Gal in sialidase A only.

Released glycan HILIC-FLR-MS method

Acquity H-class UPLC systems (Waters) and HILIC glycan BEH amide columns (2.1 mm × 150 mm, 1.7 μm, 130 Å) (Waters) were used for released glycan analysis. The released glycan materials were injected (injection amount equals to glycans from 0.4 μg antibody) onto the system. Gradient elution (91 min) was conducted at column temperature of 60°C with 50 mM ammonium formate pH 4.4 (Solvent A) and 0.1% FA in ACN (Solvent B) at 400 μL/min flow rate (0–74.8 min: 75–59% B, 74.8–83 min: 59–5% B, 83–84 min: 5% B, 84–84.5 min: 5–75% B, 84.5–91 min: 75% B). The excitation wavelength of the fluorescence detector was 285 nm, and the emission wavelength was set to 345 nm. A QE Orbitrap Mass spectrometer (Thermo Fisher Scientific) was hyphenated to the UPLC-FLR system. Data was acquired in the 500–2200 m/z range under positive ion mode. MS source conditions were sheath gas flow 30 L/min, aux gas flow 6 L/min, spray voltage 3500 V, capillary temperature 320°C. Full MS scans were acquired at a resolution of 70,000. AGC target was set to 3e6, and the maximum ion injection time was at 200 ms.

Data analysis

Full MS data for released glycan HILIC-FLR-MS were processed by the Biomap module in PMI-Byos (Protein Metrics, Inc.), with InstantPC (261.15 Da) as the glycan label and the MS1 tolerance at 10 ppm. The isotopic pattern of released glycans were manually verified. PRM-MS3 extracted ion chromatograms (XIC) were analyzed manually with Xcalibur (Version 4.1), and the MS3 mass tolerance was set to 300 ppm for data from the ion trap, corresponding to a tolerance of 0.1 m/z. The signal-to-noise (S/N) ratio of all MS3 ions, the MS3 XIC area under the curve (AUC) of m/z 325.11, MS3 scan numbers bearing m/z 325.11, and the MS3 signal intensity of m/z 325.11 were used to estimate which parameters perform better.

Results and discission

Method development

Unambiguous α-Gal identification

During the glycan characterization, hybrid and bisecting glycan isomers of α-Gal were observed, posing challenges for both identification and quantitation of α-Gal with traditional released glycan assays by HILIC-FLR-MS (Figure 1(a)). The MS2 fragment ion with m/z 528.19 is critical for unambiguous identification of α-Gal and distinction from hybrid and bisecting glycans. Specifically, when the MS2 ion with m/z 528.19 originates from α-Gal, both fragment ions with m/z 325.11 and m/z 366.14 are observed in the MS3 spectrum (Figure 1(b)). In contrast, if the MS2 ion with m/z 528.19 arises from other isomers, only the MS3 ion with m/z 366.14 is expected. Therefore, the m/z 325.11 ion in the MS3 spectrum serves as a signature ion for α-Gal. The S/N ratio of all MS3 ions at the corresponding retention time was monitored to estimate the abundance of the precursor MS2 ions at m/z 528.19. Various parameters for PRM-MS3 were systematically evaluated, including ion charge states, fragmentation modes, AGC targets, and maximum injection times, to optimize the condition for detecting the α-Gal MS3 signature ions with m/z 325.11. The number of MS3 spectra, the signal intensity and XIC AUC for the MS3 ion with m/z 325.11 were all leveraged as criteria to determine the optimal conditions of this method.

To achieve the optimal MS3 scan numbers and signal intensity, the finalized method includes all charge states for each α-Gal ion, executes CID at 30% normalized collision energy (NCE) and an activation time of 20 ms in the ion trap (Table S1 , S2), and sets the AGC target and the maximum injection time at 300% and 1000 ms (Table S1 , S3), respectively. We also evaluated the microscans setting, which averages signal intensity and can improve data quality,²⁷ but no significant differences in data quality were observed (Table S5). Notably, the PRM-MS3 only method doubled the MS3 scan numbers compared to data-dependent acquisition (DDA)-MS3 and parallel acquisition of PRM-MS2/PRM-MS3 approaches (Table S6). To our best knowledge, MS3 has not been applied in characterizing α-Gal glycans previously, and our study is the first one utilizing the PRM-MS3 technique for this purpose.

Exoglycosidase treatment for α-Gal quantitation

To effectively quantify α-Gal, we used exoglycosidase treatments and optimized key parameters, including the execution step of exoglycosidases, protein-to-enzyme ratio, incubation time, and the combination of various exoglycosidases. Of note, exoglycosidases work optimally between pH 4.5 and 5.5, whereas InstantPC labeling of released glycans requires a pH above 5.5. When the exoglycosidase treatments were performed on released glycans prior to InstantPC labeling, the glycan solutions were adjusted to pH 4.5 for optimal enzyme activity. However, this acidic pH disrupted the subsequent InstantPC labeling step, as the free primary amine groups on the released glycans became protonated at acidic pH, blocking the lone pair electrons on nitrogen and preventing the necessary reaction for labeling. To overcome this challenge, we adjusted the workflow and performed exoglycosidase treatments after InstantPC labeling of the released glycans. (Figure 2).

Figure 2.

The optimized workflow of α-Gal characterization: HILIC-FLR-PRM-MS3 and exoglycosidase treatments with HILIC-FLR-MS on released glycans with InstantPC-labeling.

Treatment with α1–3,6 galactosidase at a protein-to-enzyme ratio of 10:1 successfully removed the terminal galactose group from the InstantPC-labeled G2F with mono-α-Gal. Similarly, treatment with α1–2,3,6 mannosidase at the same protein-to-enzyme ratio (10:1) effectively removed most of the terminal mannose groups from the InstantPC-labeled hybrid glycans. For both enzymes, the samples were incubated overnight at 37°C (Table S4). Compared with the samples being treated with sialidase A separately prior to InstantPC labeling, the InstantPC-labeled glycan samples co-treated with sialidase A and exoglycosidase after InstantPC labeling showed no significant difference in removing terminal galactosidase or mannose groups (data not shown). This finding enabled us to streamline the overall protocol, reducing the processing time from 3 d to 2 d by combining sialidase A and exoglycosidase treatments (Figure 2).

The final method was structured as a three-way experimental workflow. One portion of the InstantPC-labeled released glycan underwent only sialidase A treatment, followed by HILIC-FLR-MS analysis, serving as the control run. Additionally, the residual material of this portion was analyzed by HILIC-PRM-MS3 with CID in the ion trap for α-Gal identification. The second portion of the labeled glycans was co-treated with sialidase A and α1–3,6 galactosidase, followed by HILIC-FLR-MS analysis to confirm the presence of α-Gal. Finally, the third portion of the labeled glycans was subjected to sialidase A and α1–2,3,6 mannosidase co-treatment, followed by HILIC-FLR-MS analysis for α-Gal quantification using fluorescence detection.

Exoglycosidase vendors suggest that a combination of α1–2,3,6 mannosidase and α1–6 mannosidase could enhance digestion efficiency. However, our results showed that treatment with a 1:1 mixture of α1–2,3,6 mannosidase and α1–6 mannosidase did not yield a significant improvement in removing terminal mannose groups from InstantPC-labeled glycans (data not shown). Although the optimal working pH conditions for sialidase A, α1–3,6 galactosidase, and α1–2,3,6 mannosidase are 6.0, 5.5, and 4.5, respectively, the co-treatment conditions were found to be compatible. This compatibility is attributed to the robustness of sialidase A, which operates effectively across a broader pH range of 4.5–8.0. Based on these facts, we selected a working pH of 5.5 for the co-treatment of α1–3,6 galactosidase and sialidase A, and a working pH of 4.5 for the co-treatment of α1–2,3,6 mannosidase and sialidase A.

In summary, our optimized method provides a comprehensive approach for unambiguous identification and effective quantification. This approach eliminates the need for additional analysis and enables data acquisition from a single sample preparation of InstantPC-labeled released glycans with high efficiency.

α-Gal identification with PRM-MS3

In the MS2 analysis of fusion1, which contains 4% α-Gal in its released glycan profile, the PRM-MS3 targeted analysis successfully distinguished α-Gal from isomeric hybrid and bisecting glycans by detecting the signature MS3 fragment ion at m/z 325.11 (Figure 3).

Figure 3.

The PRM-MS3 analysis of InstantPC-labeled α-Gal from fusion1. (a). The MS2 spectrum of the triply charged α-Gal ion at m/z 737.62, which has a ratio of m/z 528.19 to m/z 366.14 close to 1 and an ambiguous identity. (b). The MS3 spectrum of the MS2 precursor ion at m/z 528.19. The peak at m/z 325.11 (marked with the asterisk) was observed as the signature MS3 fragment ion of α-Gal. The retention time was 32.8 min.

As a negative control, InstantPC-labeled released glycans from mAb1 were analyzed, where no α-Gal was detected. To account for potential matrix effects from labeling reagents, salts, and other sample components, various amounts of commercially available InstantPC-labeled G2F with mono-α-Gal were spiked into the negative control material. For each technical triplicate, at least one MS3 scan at m/z 325.11 was required with a minimum S/N ratio of 3:1.²⁸ Additionally, the XIC AUC of MS3 at m/z 325.11 must be greater than zero. The limit of detection (LOD) for the PRM-MS3 analysis was determined to be 0.1% α-Gal, based on the spiked-in α-Gal levels (Table S7).

When we tried to identify InstantPC-labeled G2F with mono-α-Gal, discrepancies were observed in the relative intensity ratios of the CID MS2 fragments at m/z 528.19 to m/z 366.14 between doubly charged ions (m/z 1105.94) and triply charged ions (m/z 737.62) (Figure S3). Specifically, the relative intensity ratio (m/z 528.19 to m/z 366.14) was greater than 2 in the spectrum of the doubly charged ion, whereas it was approximately 1 in the spectrum of the triply charged ion. A similar discrepancy was observed when comparing CID and HCD MS2 spectra. Due to this variability and ambiguity, the previously proposed identification method, which relies on a relative intensity ratio >1, would fail under these circumstances,¹⁵ often necessitating additional techniques for confirmation. The HILIC-PRM-MS3 method developed in this study provides a diagnostic MS3 ion at m/z 325.11, which can reliably distinguish α-Gal from its isobaric hybrid and bisecting glycans. Compared to traditional HILIC-FLR-MS methods, our approach is more structurally specific. However, the method relies on prior determination of the specific α-Gal mass. For unknown α-Gal glycans, an initial survey run with HILIC-FLR-MS is necessary to identify the precursor MS1 mass before applying the PRM-MS3 method. This method can be applied to O-linked glycans if they are successfully released using O-Glycosidase and derivatized with 2-aminobenzamide (2-AB), with the glycan mass properly calculated. Once the precursor MS1 mass is known, users can leverage the same MS2, MS3 fragment masses, along with established parameter settings. For the negative control, the major released glycans observed are the standard biantennary mAb glycans derived from CHO cells. While the literature does not include a specific LOD for α-Gal, the lowest reported level is approximately 1%. Using our workflow, injecting released glycans from 4 μg of mAb1 achieved an LOD of 0.1%, representing a 10-fold increase in sensitivity.

Overall, the InstantPC-labeled released glycan HILIC-PRM-MS3 method developed in this study provides a sensitive and accurate tool for identifying α-Gal, even in the presence of isobaric glycans. It represents a significant advancement in α-Gal isomer analysis, achieving a low LOD at 0.1% for commonly observed biantennary α-Gal in mAbs.

α-Gal quantification with exoglycosidase treatments and HILIC-FLR-MS

In order to quantify the levels of α-Gal, HILIC-FLR-MS was utilized. Based on our observations, α-Gal co-eluted with its hybrid and bisecting glycan isomers at the same retention time. By removing the terminal galactose group, the co-treatment with sialidase A and α1–3,6 galactosidase shifted the retention time of G2F with mono-α-Gal to that of G2F. In the HILIC-FLR-MS profile, the relative area of the α-Gal peak decreased compared to the control, confirming the presence of α-Gal (Figure 4). Similarly, the co-treatment of sialidase A and α1–2,3,6 mannosidase removed terminal mannose groups from isobaric glycans, shifting these glycans to a different retention time from α-Gal.

Figure 4.

Relative area (%) results of InstantPC-labeled G2F mono-α-Gal from fusion1 by HILIC-FLR-MS. InstantPC-labeled released glycans were treated with sialidase A only, sialidase A and α1–3,6 galactosidase, sialidase A and α1–2,3,6 mannosidase, respectively.

Consequently, upon confirming the presence of α-Gal, we propose integrating the entire peak, including both the main peak and the shoulder, after overnight co-treatment with sialidase A and α1–2,3,6 mannosidase for improved quantification of α-Gal. For fusion1 expressed in a CHO cell line, the presence of α-Gal was confirmed, as the relative area of the full peak (4.1% total) decreased by 3.3% (to 0.8% total) following co-treatment with sialidase A and α1–3,6 galactosidase. Using co-treatment of sialidase A and α1–2,3,6 mannosidase, α-Gal was quantified at 3.7% (total peak). To further improve the accuracy of quantification and validate the results, various amounts of commercially available InstantPC-labeled G2F with mono-α-Gal were spiked into InstantPC-labeled released glycans from mAb1 (used as a negative control). The limit of quantification (LOQ) was determined to be 0.1% α-Gal, based on spiked-in levels, with a minimum S/N ratio of 10:1 (Table S8).²⁹ Quantification of α-Gal using the HILIC-FLR-MS assay exhibited excellent repeatability, with a relative standard deviation (RSD) of less than 5% across technical triplicates.

For the analysis shown in Figure 4, the digestion efficiency was calculated to be 87.8%, based on the exoglycosidase digestion efficiency formula mentioned above. In prior studies, α-Gal quantification was typically performed by comparing the peak area before and after galactosidase treatment.²⁵ However, this method excludes undigested α-Gal, leading to an underestimation of its quantitation, which could increase the risk of immunogenic reactions, such as α-Gal syndrome. To address this limitation, we propose quantifying α-Gal based on the peak area after mannosidase treatment, which accounts for all α-Gal, as well as a small proportion of undigested hybrid and bisecting isobaric glycans. While our approach may slightly overestimate α-Gal levels, it provides a more cautionary and patient-safety-focused assessment compared to the galactosidase-based method, which risks underestimation. Although both metrics rely on S/N ratios, they are assessed using different techniques: the LOD was determined using the PRM-MS3 method, while the LOQ was derived from the HILIC-FLR-MS signal. Typically, the LOD is defined using an S/N ratio of 3, whereas the LOQ requires an S/N ratio of 10.

In addition, we investigated the identity of the shoulder peak with α-1,3-galactosyltransferase (α1,3GalT).³⁰ To generate isobaric α-Gal glycans in vitro, we incubated released glycans from mAb2 with α1,3GalT. HILIC-FLR-MS analysis revealed G2F and the isomers of mono-α-galactosylated G1F with the same mass. Notably, the isomers of mono-α-galactosylated G1F (Figure S4a) were baseline-separated with a 1.5-min retention time difference.

The reaction also produced mono-α-galactosylated G2Fs (Figure S4b) and bi-α-galactosylated G2F (Figure S4c). Because mono-α-galactosylated G2Fs appeared as adjacent peaks without baseline separation, we hypothesized that the shoulder peak of mono-α-galactosylated G2F originated from an epimer formed during InstantPC labeling,³¹ rather than from its isomer.

Application in various therapeutic modalities and cell lines

To demonstrate the capability of the integrated workflow, we applied it to identify and quantify α-Gal in four therapeutic proteins with distinct formats or expressed in various cell lines (as detailed in Figure 5). First, we tested two mAbs (both expressed in CHO cells), expecting low α-Gal levels. One of the mAbs is IgG1 (mAb2) with the α1,3GalT gene knocked out in the cells, and the other is IgG4 (mAb3). In addition, we analyzed two fusion proteins expressed in different cell lines. Fusion1 was expressed in a CHO cell line, in which we expected a high level of α-Gal compared to standard mAbs, due to its complex glycan profile. Fusion2 was expressed in the human 293 cell line, in which we expected no α-Gal with absence of the α1,3GalT gene. From our results, the α-Gal level in mAb2 was below the LOQ. For mAb3, α-Gal was detected at 0.09%, which is at the LOQ. The level of α-Gal in fusion1 was 1.2%. For fusion2, the treatment with α1–3,6 galactosidase caused no peak area change from the control, both at 1.4%, indicating this peak is not α-Gal. The peak could be attributed to a specific bisecting glycan, which does not react significantly with α1–2,3,6 mannosidase, explaining the persistence of the peak after the treatment of α1–2,3,6 mannosidase (Figure 5(d)). All results aligned with our expectations, confirming the capability of our method in analyzing α-Gal.

Figure 5.

α-Gal quantification after the co-treatment of sialidase A and α1–2,3,6 mannosidase in various molecules and cell lines. (a). mAb2 (IgG1) expressed in CHO cell line with α1,3GalT gene knocked out (α-Gal below LOQ, green trace). (b). mAb3 (IgG4) expressed in CHO cell line (α-Gal below LOQ, blue trace). (c). Fusion1 expressed in CHO cell line (black trace). (d). Fusion2 (bispecific fusion modality) expressed in human 293 cell line (not α-Gal but a bisecting glycan, red trace).

Our study highlights the potential for widespread implementation of this method in the pharmaceutical industry. Importantly, bisecting glycans should not be present in CHO cell lines, suggesting that the co-eluted isobaric glycans observed in fusion1 are α-Gal and hybrid glycans. In contrast, for human cell lines, the co-eluted isomers are likely hybrid and bisecting glycans rather than α-Gal glycans, as the α1,3GalT gene is not expressed. Regarding bisecting N-glycans, Dang et al. identified characteristic ions of glycopeptides with bisecting structures in tandem mass spectra.³² However, these peaks were absent in our fusion2 MS2 data from the human 293 cell line. Furthermore, as this workflow analyzes released glycans, the effectiveness of the method is expected to be independent of the structural complexity of the biologics.

Overall, for molecules in early pharmaceutical development (prior to Phase III), this α-Gal characterization workflow can be integrated into the molecular assessment process, where the current control strategy primarily focuses on upstream cell line modifications and analytical clone screening. For molecules in later stages of development (post-Phase III), this workflow supports CQA evaluation, informs quality control activities, and enhances patient safety during clinical trials.

Conclusions

In conclusion, we developed a comprehensive novel method to identify and quantify α-Gal explicitly and effectively by HILIC-PRM-MS3 and exoglycosidase treatments. This method distinguishes α-Gal from its isomers without ambiguity. One set of fluorescent labeling experiments also simplifies the sample preparation. By incorporating α1–2,3,6 mannosidase besides α1–3,6 galactosidase, our method not only confirms the presence of α-Gal with 0.1% LOD but quantifies it more accurately at 0.1% LOQ by moving the isobaric hybrid and bisecting glycans to other retention times. This more sensitive and effective α-Gal analytical method enables early α-Gal characterization, triggers timely action during the development process, and ensures the safety of therapeutic proteins by preventing α-Gal related immunogenicity.

Funding Statement

The author(s) reported there is no funding associated with the work featured in this article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2025.2588410