Evaluation of the impact of antibody fragments on aggregation of intact molecules via size exclusion chromatography coupled with native mass spectrometry

ABSTRACT

Aggregates are recognized as one of the most critical product-related impurities in monoclonal antibody (mAb)-based therapeutics due to their negative impact on the stability and safety of the drugs. So far, investigational efforts have primarily focused on understanding the causes and effects of mAb self-aggregation, including both internal and external factors. In this study, we focused on understanding mAb stability in the presence of its monovalent fragment, formed through hinge cleavage and loss of one Fab unit (referred to as “Fab/c”), a commonly observed impurity during manufacturing and stability. The Fab/c fragments were generated using a limited IgdE digestion that specifically cleaves above the IgG1 mAb hinge region, followed by hydrophobic interaction chromatographic (HIC) enrichment. Two IgG1 mAbs containing different levels of Fab/c fragments were incubated under thermally accelerated conditions. A method based on size exclusion chromatography coupled with native mass spectrometry (SEC-UV-native MS) was developed and used to characterize the stability samples and identified the formation of heterogeneous dimers, including intact dimer, mAb-Fab/c dimer, Fab/c-Fab/c dimer, and mAb-Fab dimer. Quantitative analyses on the aggregation kinetics suggested that the impact of Fab/c fragment on the aggregation rate of individual dimer differs between a glycosylated mAb (mAb1) and a non-glycosylated mAb (mAb2). An additional study of deglycosylated mAb1 under 25°C accelerated stability conditions suggests no significant impact of the N-glycan on mAb1 total aggregation rate. This study also highlighted the power of SEC-UV-native MS method in the characterization of mAb samples with regard to separating, identifying, and quantifying mAb aggregates and fragments.

KEYWORDS:

• Aggregation
• aggregation rate
• fragments
• heterodimers
• monoclonal antibody
• native mass spectrometry
• size-exclusion chromatography

Introduction

Monoclonal antibodies (mAbs) are now commonly used as treatments for a wide range of diseases.^1–3 Mab aggregates are critical product-related impurities that require holistic understanding and close monitoring during manufacturing and stability. For example, aggregates formed during manufacture and storage may affect the stability and safety of mAb-based therapeutics for patients, diminishing their biological efficacy and inducing unwanted immune responses.^4–6 Therefore, it is crucial to have a comprehensive understanding of mAb aggregation to control and minimize the impact of aggregates on drug safety.

Over decades of research, various causes and pathways leading to the formation of mAb aggregates have been identified.^5–8 Typically, aggregation is driven by the aggregation-prone features of the protein, such as post-translational modifications, disulfide bonds subject to scrambling or cross-linking, hydrophobic or charged regions.^6,7 Chemical alterations at the amino acid level, such as oxidation, deamidation and hydrolysis, can also facilitate aggregation.^7,9 Hydrophobic sequence segments in conserved parts of crystallizable fragment (Fc) and antigen binding fragment (Fab) domains are suggested to contribute to mAb aggregation by allowing close contact and association between mAb molecules via hydrophobic interactions.^6,10 Attraction between complementary electrostatic patches could destabilize protein molecule and foster their association.^6,11 External stimuli such as temperature, light, pH, and ionic strength may induce the formation of more of these aggregation-prone features and exacerbate the inter-molecular interactions.^7,12–14

MAb fragments, as well as aggregates, are product-related impurities that often co-purify with mAb and form during manufacturing and stability. These fragments may possess similar or different aggregation-prone features compared to their intact mAbs, depending on their sizes, domains, and modifications.^4,6,13,15 Several reports have discussed the self-association of isolated Fab and Fc fragments and their possible mechanisms.^16–18 Nelson et al. observed spontaneous self-dimerization of Fab at 4°C that could be reversed by amino acid osmolyte.¹⁶ Latypov reported Fc aggregation in an acidic environment that was possibly due to association between unfolded CH2 region.¹⁸ Zhang et al. described that co-formulating an IgG1 and its Fab alleviated the degradation of the IgG1, likely due to reduced IgG1 unfolding by the interaction with Fab.¹⁹ Despite the fact that mAb fragments, aggregates and intact mAb are typically present in the same solution, the impact of fragments on aggregate formation has not been thoroughly investigated.

A plethora of analytical technologies are currently available for characterizing mAb aggregates, such as size exclusion chromatography (SEC), analytical ultracentrifugation (AUC), light scattering techniques including dynamic light scattering (DLS), and multi-angle (static) light scattering (MALS), and spectroscopy-based techniques.^7,20–22 SEC with UV detection is considered the gold standard for aggregate separation and quantification, while AUC provides an estimation of the molecular weight of species. DLS and MALS act by measuring the light scattering of protein particles for characterizing size and structure of protein aggregates.^23,24 DLS measures the hydrodynamic radius of species and could be used to evaluate protein aggregation propensity, while MALS, usually coupled with SEC, can determine the molecular weight of aggregates. Circular dichroism and infrared spectroscopy are used for detailed structural analysis of protein aggregates.

As protein aggregation can arise from multiple pathways, resulting in complexed aggregate profiles, a technology that can characterize aggregate heterogeneity with high accuracy, requires minimum sample preparation and consumption, and of high-throughput nature is highly desired. Native mass spectrometry (MS), which uses gentle ionization and enables accurate detection of higher order structures in a native-like state, has been frequently used for studying protein aggregates.^21,25 By coupling native MS with SEC (SEC-native MS) and using a volatile buffer system compatible with both SEC and MS, this technology allows separation, characterization, and quantification of protein aggregates through a single method. Compared with SEC-MALS and other technologies discussed above for characterizing protein aggregates, SEC-native MS provides molecular weight detection with higher resolution and sensitivity, offering richer information on aggregate identity and heterogeneity.²⁶ So far, SEC-native MS has used for numerous purposes,^27–30 including studying protein-related complexes such as protein-protein, protein-ligand, protein-lipid, and protein-metal to illustrate the interactions between molecules,^31,32 determining drug-to-antibody ratios for antibody-drug conjugates,^33,34 and profiling process- and product-related size variants of protein-based therapeutics.^35,36

In this work, we focused on studying mAb aggregation in the presence of its Fab/c fragment, which is the monovalent mAb fragment resulting from the cleavage in upper hinge and loss of one Fab arm, by SEC-UV-native MS. The results indicated the presence of Fab/c led to formation of heterogeneous dimers, including mAb-Fab/c, Fab/c dimer, and mAb dimer, but their formation kinetics varied from mAb to mAb. This work also highlighted the capabilities of the SEC-UV-native MS method for studying mAb stability, as it allows for the identification and quantification of low-level aggregates, providing an in-depth characterization of aggregate structural heterogeneity.

Results

Stability characterization of two IgG1 mAbs by SEC-UV-native MS

According to ICH Q5C, stability of mAb drug substances and drug products is typically assessed by incubating them at both desired long-term storage conditions and accelerated conditions (e.g., higher temperature) in order to establish an expiration date. Stability data from accelerated conditions can also provide faster evaluation and prediction of the long-term stability, as well as early identification of potential degradation pathways. One method of characterizing stability is SEC, which separates high molecular weight species (HMWs), intact mAb and low molecular weight species (LMWs) and quantitates them based on UV chromatogram. By integrating SEC with native MS, accurate identification of these size variants could be achieved. An SEC-UV-native MS method was developed on a SCIEX TripleTOF 6600 mass spectrometer coupled with a Shimadzu UHPLC system. A salt-based buffer containing 25 mM ammonium acetate was used due to its high volatility and compatibility with both SEC and MS. The source temperature, de-clustering potential, and collision energy of the mass spectrometer were optimized to achieve gentle ionization and detection of mAb aggregates in a native-like state.

Two mAbs, referred to as mAb1 and mAb2, with respective concentration of 150 mg/mL and 200 mg/mL, were used for stability characterization. Although both are IgG1 isotype, they differ in amino acid sequence and glycosylation (mAb1 is glycosylated and mAb2 is aglycosylated). Both mAbs were incubated at 25°C for 3 months before analysis by the SEC-UV-native MS method. The mAb1 T0 sample was analyzed by SEC to contain 0.6% HMWs, and an LMW peak (LMW1) with ~ 1.4% abundance (Figure 1a). After 3-month incubation at 25°C, the sample (referred to as 3M25C) was found to contain 1.8% of HMWs and 2.1% of LMW1. Another LMW peak, referred to as LMW2, was detected in the 3M25C sample with 0.4% abundance (Figure 1a).

Figure 1. SEC-UV chromatograms of mAb1 (a) and mAb2 (b) stability samples at T0 and after 3-month incubation at 25°C (“3M25C”). Inserts in panel (a) and (b): enlarged area to display the HMW peaks.

SEC chromatogram of mAb1 with an insert showing enlarged area of the HMW peaks. SEC chromatogram of mAb2 with an insert showing enlarged area of the HMW peaks.

Native MS coupled with SEC was able to provide accurate mass information for these detected species, which led to their identification (Table 1). The main peaks, LMW1 and LMW2, were measured with mass values of 145,092 Da 98,960 Da, and 46,159 Da, respectively, and identified as intact mAb1 (theoretical mass: 145,097 Da), Fab/c fragment (theoretical mass: 98,958 Da) and Fab fragment (theoretical mass: 46,157 Da), all with mass error of less than 50 ppm. The masses of Fab/c and Fab suggested that major cleavage occurred in the upper hinge region at KSCDKTH/TCPPC, by comparison with their theoretical mass values (Table 1). A small percentage of Fab/c and Fab fragments generated from the cleavage at different sites were found coeluting with major fragments in LMW1 and LMW2, respectively (Table. S1). Four different HMW species were detected in the mAb1 3M25C sample, i.e., HMW1 at 6.3 min, HMW2 at 6.5 min, HMW3–1 at 6.6 min, and HMW3–2 at 6.7 min (the last two species coeluting under Peak 3), with respective masses of 290.5 kDa, 244.5 kDa, 198. 5 kDa, and 191.0 kDa, and were determined to be mAb1 dimer, mAb1-Fab/c dimer, Fab/c dimer, and mAb1-Fab dimer (Table 1). The detected masses of these aggregates were typically a few hundred dalton different from the predicted masses, which is likely due to incomplete desolvation during the ionization process, i.e., stripping of solvent molecules and salt adducts from protein complex, a typical observation for native MS.^25,37 The low abundance (2.1%) of these aggregates also affected the detection accuracy even when the sensitivity of the instrument had been optimized.

Table 1. Major species detected by SEC-UV-native MS in mAb1 stability samples at T0 and after 3-month incubation at 25°C (“3M25C”).

Sample name	Retention time (min) and peak naming	Detected mass (Da)	Assigned identity	Theoretical mass^# (Da)	% UV peak area
T0	6.3 (HMW)	290.5 k	mAb1 dimer	290,194	0.6
	7.1 (main peak)	145,094	mAb1	145,097	98.0
	7.7 (LMW1)	98,950	Fab/c*	98,958	1.4
3M25C	6.3 (HMW 1)	290.5 k	mAb1 dimer	290,194	1.8 (total HMWs)
	6.5 (HMW 2)	244.5 k	mAb1-Fab/c	244,054
	6.6 (HMW 3–1)	198.5 k	Fab/c dimer	197,915
	6.7 (HMW 3–2)	191.0 k	mAb1-Fab	191,254
	6.9 (main peak)	145,092	mAb1	145,097	95.6
	7.7 (LMW1)	98,960	Fab/c*	98,958	2.1
	8.6 (LMW2)	46,159	Fab*	46,157	0.4

*Masses of Fab/c and Fab fragments were calculated based on cleavage site KSCDKTH/TCPPC.

^#All the theoretical masses of mAb1 and Fab/c were calculated based on two copies of FA2 glycan were included.

For mAb2 samples, 0.8% HMWs and 1.9% of LMW1 were detected at T0 besides the main peak. After a 3-month incubation at 25°C, HMWs increased to 2.8%, LMW1 to 2.1%, and 0.6% LMW2 was observed (Figure 1b, Table 2). The main peaks, LMW1 and LMW2, were measured with mass values of 144,442 Da, 97,124 Da, and 47,335 Da, respectively, and determined to be intact mAb2 (theoretical mass: 144,441 Da), Fab/c (theoretical mass: 97,120 Da) and Fab (theoretical mass: 47,336 Da) with mass errors less than 50 ppm. Fab/c and Fab were confirmed to originate from the cleavage at KSCDKTH/TCPPC (Table 2), the same site as for mAb1 fragments, and coeluted with a few minor species from cleavages at different sites in the hinge region (Table. S2). Only mAb2 dimer was identified in the HMWs peak in comparison with the heterogeneous aggregates detected in the mAb1 sample.

Table 2. Major species detected by SEC-UV-native MS in mAb2 stability samples at T0 and after 3-month incubation at 25°C (“3M25C”).

Sample Name	Retention time (min) and Peak Naming	Detected mass (Da)	Assigned identity	Theoretical mass (Da)	%UV peak area
T0	6.6 (HMW)	289.4 k	mAb2 dimer	288,882	0.8
	7.5 (main peak)	144,441	mAb2	144,441	97.4
	8.0 (LMW1)	97,118	Fab/c*	97,123	1.8
3M25C	6.5 (HMW)	289.4 k	mAb2 dimer	288,882	2.8
	7.4 (main peak)	144,442	mAb2	144,441	94.5
	8.0 (LMW1)	97,124	Fab/c*	97,123	2.1
	9.3 (LMW2)	47,335	Fab*	47,336	0.6

*Fab/c and Fab fragments were identified with cleavage site KSCDKTH/TCPPC.

Accurate identification of Fab/c and Fab fragments in both mAb stability samples demonstrated these two fragments, which resulted from hinge fragmentation, are likely common degradation products under thermal stress. Meanwhile, the presence of heterogenous aggregates in the stability sample of mAb1 suggested that Fab/c and Fab fragment could contribute to aggregate formation, and the impact may vary for different mAbs.

Production, purification, and characterization of Fab/c fragment

To further investigate the impact of fragments on mAb stability, an experiment was designed to evaluate the stability profiles of mAb, Fab/c, and their mixture. To obtain a large amount of the Fab/c fragment with high purity, a limited enzymatic digestion was conducted using IgdE enzyme, which specifically cleaves at KSCDKT/HTCPPCP in the hinge region of IgG1 antibodies (Figure 2A). Notably, this site is only one amino acid away from the major cleavage site (KSCDKTH/TCPPC) observed for the mAb stability samples incubated at 25°C (Table 1). The reaction protocol was optimized to limit the digestion to approximately one cleavage per molecule, so that the major digested product contained only Fab/c and Fab. The digests were characterized by a hydrophobic interaction chromatography (HIC) method where four components, Fab, Fc, Fab/c, and the intact mAb, were baseline separated for both mAbs (Figure 2b). The HIC methods were later scaled up to semi-preparative scale, which enabled fraction collection of Fab/c and mAbs (Figure. S1). The two collected fractions of each mAb were later purified and concentrated by tangential flow filtration to approximately 100 mg/ml and exchanged into the mAb formulation buffer.

Figure 2. Generation of Fab/c fragment to study Fab/c associated aggregation. (A) Limited IgdE enzymatic digestion of mAb1 and mAb2 to generate Fab/c fragments; (b) separation of digests by hydrophobic interaction chromatography; (C) stability study on the collected mAbs and Fab/c fractions from HIC and their mixture followed by SEC-UV-native MS characterization.

A schematic describing the limited IgdE digestion of mAb1 and mAb2 where cutting site is KSCDKT/HTCPPCP and the digest contain Fab/c and Fab fragment. HIC chromatograms of mAb1 IgdE digest and mAb2 IgdE digest showing four components, Fab, Fc, Fab/c, and left mAb, were separated. A schematic illustrating the design of the stability study on purified Fab/c fraction, mAb fraction and their 1:1 mixture of two mAbs. Briefly, they were placed on thermal acceleration for three months and monitored by the SEC-UV-native MS method at four time points.

For mAb1 digest, the collected Fab/c and mAb1 fractions were characterized by SEC and DLS for purity and self-interaction propensity. The SEC analysis showed that Fab/c fraction was 81.7% pure with 13.1% undigested mAb1, and mAb1 fraction was 89% pure (Figure. S2A). Less than 0.4% of aggregated species were detected in each fraction. DLS analysis, which is commonly used to assess inter-molecular association between proteins, indicated that mAb1 and Fab/c fractions exhibited comparable levels of self-association propensity (Figure. S2B). After purification and concentration to 100 mg/mL, mAb1 fraction, Fab/c fraction, along with a 1:1 mixture of the two, were placed in accelerated (25°C) and stressed (40°C, mAb1 only) conditions for the stability study. The stability samples were collected at the time points of two weeks, one month, two months, and three months for analysis by the SEC-UV-native MS method (Figure 2c).

Characterization of mAb1 stability samples containing Fab/c fragments by SEC-UV-native MS

The SEC-UV chromatograms of mAb1 fraction, 1:1 mixture, and Fab/c fraction after 3-month incubation at 25°C presented HMW peaks and the peaks corresponded to intact mAb1, Fab/c, and Fab (Figure 3a). Fc fragment was found coeluting with the Fab fragment in the Fab/c fraction and 1:1 mixture of mAb1 and Fab/c, which was likely from further degradation of Fab/c. Fragments (Fab/c, Fab, Fc) that resulted from different cleavage sites in upper hinge were also detected coeluting with their respective major species (Table. S3). The HMW cluster appeared in three partially resolved UV peaks and were detected with masses of 290.5 kDa, 244.5 kDa,197.5 kDa, and 191.1 kDa in the order of elution, with the last two species coeluting, and were determined to be mAb1 dimer, mAb1-Fab/c, Fab/c dimer, and mAb1-Fab (Figure 3b). The identified HMWs and fragments were highly similar to the species identified in mAb1 stability samples as shown in Figure 1a and Table 1, suggesting the IgdE-generated Fab/c fragment was a suitable surrogate for studying mAb-fragment interactions.

Figure 3. Characterization of three mAb1 stability samples containing Fab/c fragments after a three-month incubation at 25°C by the SEC-UV-native MS method. a) SEC-UV chromatograms of mAb1 fraction, 1:1 mixture and Fab/c fraction at time point of three months (“3M25C”); insert panel: enlarged area to display the HMW peaks; b) mass spectra of HMW peaks, with the deconvoluted masses and the corresponding dimer species shown; the charge states highlighted in blue were used for the construction of XICs and peak area calculation; the charge states highlighted in orange were interference peaks from mAb1 dimer; c) plots of total HMWs in mAb1 fraction, 1:1 mixture and Fab/c fraction based on UV peak area; d) plots of major species in mAb1 fraction, e) 1:1 mixture and f) Fab/c fraction based on UV peak area.

The total amount of aggregates at the end of the 3-month incubation was quantified to be 1.1%, 1.0%, and 0.9% in mAb1 fraction, 1:1 mixture and Fab/c fraction, respectively, compared with ~ 0.4% of HMW at T0 (Figure 3c). Abundance changes for the major components, i.e., mAb1, Fab/c and Fab and Fc, were also plotted as shown in Figure 3d–f. Generally, there was a continued trend of fragment formation during the stability incubation in all three fraction samples. In mAb1 fraction, mAb1 peak decreased from 89.0% to 80.1% during the three-month incubation while the Fab/c and Fab peak showed 5.0% and 3.1% increase, respectively (Figure 3d), indicating mAb1 further degraded into Fab/c and Fab during stability. This was also supported by the MS data where Fab and Fc fragments with different cleavage sites than the IgdE digestion site were identified, indicating additional fragments were formed through heat-degradation rather than enzyme cleavage (Table. S3). In the Fab/c fraction, the abundance of Fab/c decreased from 81.7% to 61.5%, while that of Fab and Fc fragment increased 11.6% within the first month and plateaued afterward (Figure 3f). In the 1:1 mixture, only Fab/c component showed 6.6% abundance decrease within the first two weeks, while the mAb1 component had minimal change (Figure 3e). These findings indicated that Fab/c was more susceptible to degradation compared with mAb1 under the same stress condition.

The peak areas of three partially separated HMW peaks corresponding to different dimer species were plotted to further illustrate the formation of each species (Figure 4). Overall, as the percentage of Fab/c fragment increased in the mixture (from Figure 4a–c), coeluting Fab/c dimer and mAb1-Fab (HMW3 represented in green), and mAb1-Fab/c heterodimer (HMW2 represented in blue), became more dominant, while mAb1 dimer (HMW1 represented in black) decreased. Specifically, in the mAb1 fraction, mAb1 dimer, coeluting Fab/c dimer, and mAb1-Fab, and mAb1-Fab/c dimer showed a similar pattern with comparable levels of increase around 0.3% by the end of 3 months (Figure 4a). Considering the starting concentration ratio of mAb1 and Fab/c in this fraction was around 90:10, the formation tendency of Fab/c-associated dimers is comparatively higher than that of mAb1 dimer. In the Fab/c fraction, coeluting Fab/c dimer and mAb1-Fab had the largest increase of 0.4% followed by mAb1-Fab/c (0.15%), while mAb1 dimer was barely detectable during the stability study (Figure 4c). In the 1:1 mixture, where the starting concentration of mAb1 and Fab/c were comparable, mAb1-Fab/c dimer had a 0.4% increase, which was most abundant among the three fraction samples. The mixture of Fab/c dimer and mAb1-Fab dimer increased 0.3%, whereas mAb1 dimer showed minimal change (Figure 4b). By comparing the three fraction samples, it was clear that Fab/c-associated dimer species, including Fab/c dimer and mAb1-Fab/c dimer, exhibited a faster aggregate formation compared with mAb1 dimer.

Figure 4. Plots of dimer species formation based on UV peak area in (a) mAb1 fraction, (b) 1:1 mixture and (c) Fab/c fraction. Plots of dimer species formation based on XIC peak area in (d) mAb1 fraction, (e) 1:1 mixture and (f) Fab/c fraction.

Read the detailed description of this figure

Plots of HMW1, HMW2, and HMW3 based on UV peak area in mAb1 fraction. All three HMWs increased with a comparable level and in a similar pattern. Plots of HMW1, HMW2, and HMW3 based on UV peak area in 1:1 mixture. HMW1 almost remained the same level throughout the incubation. HMW2 and HMW3 increased at a comparable rate. Plots of HMW1, HMW2, and HMW3 based on UV peak area in Fab/c fraction. HMW1 was barely detected and remained unchanged throughout the incubation. HMW3 increased significantly followed by HMW2. Plots of mAb1 dimer, mAb1-Fab/c dimer, Fab/c dimer, and mAb1-Fab based on XIC peak area in mAb1 fraction. mAb1 dimer and mAb1-Fab showed a similar trending. Similar observation was for mAb1-Fab/c and Fab/c dimer. The increase rate of the first two was higher than the second two. Plots of mAb1 dimer, mAb1-Fab/c dimer, Fab/c dimer, and mAb1-Fab based on XIC peak area in 1:1 mixture. The ranking of the increase rate from high to low is Fab/c dimer, mAb1-Fab/c, mAb1-Fab, mAb1 dimer. Plots of mAb1 dimer, mAb1-Fab/c dimer, Fab/c dimer, and mAb1-Fab based on XIC peak area in Fab/c fraction. The ranking of the increase rate from high to low is Fab/c dimer, mAb1-Fab/c, mAb1-Fab, mAb1 dimer. Fab/c shows a much higher increase rate than the other three.

As the UV chromatograms were not able to fully separate the HMW species, native MS data were used to illustrate the aggregation kinetics for each species, especially the coeluted ones. In brief, three most abundant charge states were selected for each dimer to construct extracted ion chromatograms (XICs) (Figure 3b) and the summed peak areas were plotted. Co-eluting Fab/c dimer (197.5 kDa) and mAb1-Fab (191.1 kDa) were successfully differentiated and plotted separately along with mAb1 dimer and mAb-Fab/c dimer for mAb1 fraction (Figure 4d), 1:1 mixture (Figure 4e), and Fab/c fraction (Figure 4f). In general, the aggregate formation trends obtained from MS were consistent with the UV data. MS peak area, however, should not be directly correlated with the absolute abundance of proteins because the ionization efficiency of different HMW species may be different. For example, Fab/c dimer had much higher MS signal intensity than mAb1 dimer, but they showed similar UV intensities. Interestingly, aggregation formation between mAb1 and Fab (magenta dots in Figure 4d–f) was found at comparable rate in all three sample sets, although the starting mAb1% varied from 90% to 20% in the mixture. Since the Fab fragment was continuously produced through degradation of mAb1 and Fab/c in the stability study, an increased amount of Fab fragment might be driving the mAb1-Fab dimer formation in different mixtures.

Stability samples with 40°C incubation were also analyzed in a similar fashion. As expected, the aggregation rates of all dimer species were higher at 40°C than at 25°C in all three sample sets (Figure. S3). Fab/c-associated dimer species, i.e., mAb1-Fab/c and Fab/c dimer, exhibited higher aggregation rate in 1:1 mixture and Fab/c fraction at 40°C, which correlated with the observation at 25°C.

Study the impact of Fab/c fragment using mAb2

We also studied mAb2 using the same workflow to further investigate the fragment-associated aggregation (Figure 2). After IgdE digestion and HIC fractionation, the isolated mAb2 fraction was 76% pure with 13% of Fab/c, while Fab/c fraction was 80% pure and contained 5% mAb2. Fab and Fc fragments (5–7%) were also detected in the two fractions following fractionation. A low-level (<2%) HMW species of 149 kDa was detected and suspected to be related to enzymatic digestion and purification process as its abundance remained comparable before and after thermal stress in both fractions (Figure. S4, Table. S4).

Three stability sample sets, i.e., mAb2 fraction, 1:1 mixture and Fab/c fraction, each at 100 mg/mL, were assessed for stability at 25°C. After a three-month incubation at 25°C, SEC-UV chromatograms of all stability samples presented a broad HMW peak, a main peak, and three LMW peaks. (Figure 5a) Based on the detected masses, the main peak was identified as intact mAb2, and the three LMW peaks were Fab/c, Fab, and Fc fragments, respectively, all from cleavage at KSCDKT/HTCPPC in the hinge region (Table. S4). A time-course plot of UV peak area for each major species (Figure. S5) showed that the degradation pattern of mAb2 fractions was very similar to those of mAb1, where formation of Fab and Fc fragments continued. By taking advantage of native MS, three HMW species with masses of 289.2 kDa, 242.0 kDa, and 194.9 kDa were detected and assigned as mAb2 dimer, mAb2-Fab/c, and Fab/c dimer (Figure 5b). The total HMWs showed a steady increase during the three months at 25°C and did not plateau, and the HMW increase was 0.6%, 0.5%, and 0.2% in mAb2 fraction, 1:1 mixture, and Fab/c fraction, respectively, suggesting the Fab/c fragment had a smaller aggregation formation tendency than mAb2 (Figure 6a–c).

Figure 5. Characterization of three mAb2 stability samples containing Fab/c fragments after three-month incubation at 25°C by the SEC-UV-native MS method. a) SEC chromatograms of mAb2 fraction, 1:1 mixture and Fab/c fraction; insert panel: enlarged area to display the HMW peaks; b) Mass spectra of detected dimer species under the HMW peak; deconvoluted masses and cartoon illustrated identifications are shown on the right side; the charge states labeled and highlighted in blue of each dimer species were used for the construction of XICs and peak area calculation; the charge states highlighted in orange were interference peaks from mAb2 dimer.

Figure 6. Plots of total dimer species change during 25°C stability incubation for (a) mAb2 fraction, (b) 1:1 mixture and (c) Fab/c fraction based on UV peak area. Plots of formation of each dimer species in (d) mAb2 fraction, (e) 1:1 mixture, and (f) Fab/c fraction based on mass spec XIC peak area.

Read the detailed description of this figure

A plot of total dimer species vs. incubation time based on UV peak area in mAb2 fraction. The increase throughout the incubation was steady with a level change of 0.6%. A plot of total dimer species vs. incubation time based on UV peak area in 1:1 mixture. The increase throughout the incubation was steady with a level change of 0.5%. A plot of total dimer species vs. incubation time based on UV peak area in Fab/c fraction. The increase throughout the incubation was steady with a level change of 0.2%. Plots of mAb2 dimer, mAb2-Fab/c, and Fab/c dimer based on XIC peak area in mAb2 fraction. The ranking of the increase rate from high to low is mAb2 dimer, mAb2-Fab/c, Fab/c dimer. Plots of mAb2 dimer, mAb2-Fab/c and Fab/c dimer based on XIC peak area in 1:1 mixture. The ranking of the increase rate from high to low is Fab/c dimer, mAb2-Fab/c, mAb2 dimer. Plots of mAb2 dimer, mAb2-Fab/c and Fab/c dimer based on XIC peak area in Fab/c fraction. Fab/c dimer increased significantly while mAb2-Fab/c only increased slightly. mAb2 dimer was barely detected and remained at minimal level.

To analyze the aggregation kinetics of three coeluting HMWs separately, XICs were constructed for each of them using the same process as for mAb1 sample sets (Figure 6d–f). Similar to the mAb1 study, as the percentage of Fab/c increased in the fraction samples (from Figure 6d–f), less self-aggregation of mAb2 was observed, while Fab/c aggregation intensified significantly, with the formation of mAb2-Fab/c heterodimer being most abundant in the 1:1 mixture. In contrast to the mAb1 study results, the mAb2-Fab dimer was not detected in mAb2 stability sample sets, although there were 5–10% of Fab fragment present in all three sample sets. This might suggest that the association interaction between mAb and Fab fragment is stronger in mAb1 than in mAb2.

Study the impact of N-glycan on mAb1 stability

A further stability study to assess the impact of N-glycosylation was conducted. PNGase F was used to remove N-glycans on mAb1. The deglycosylated mAb1 and the control (untreated mAb1) were concentrated to 150 mg/ml and incubated at 25°C for up to three months prior to the online SEC-native MS analysis. The amount of HMW species in deglycosylated mAb1 was comparable to the control at each time point, suggesting N-glycans had negligible impact on the total aggregation rate of mAb1 (Figure 7). The native MS identified intact dimer as the dominate dimer species. mAb1-Fab dimer was commonly detected in both deglycosylated mAb1 and control samples, while the Fab/c-Fab/c dimer was only detected in control sample. (Table S6)

Figure 7. SEC-UV chromatograms of a) deglycosylated mAb1 and b) control sample at T0, one-month (“1 M”), two-months (“2 M”) and three-month (“3 M”) time points of incubation at 25°C. The two post monomer peaks with retention times of approximately 19.5 min and 23.0 min were the fab/c fragment and fab fragment, respectively. C) a plot of %HMWs vs. incubation time (month) based on UV peak areas of deglycosylated mAb1 and control sample.

It appears that for mAb1, the N-glycosylation has negligible impact on HMW aggregate formation at intact mAb level, while it has some impact on the formation of Fab/c-containing HMW aggregates. The result suggested that the N-glycan in the Fab/c fragment could play a role in protein/protein interaction and promote the formation of Fab/c-containing dimers. This observation agrees with earlier experiments that the N-glycan-containing Fab/c of mAb1 has a greater tendency to form aggregated species in both the mixing study and the real-time stability study.

Discussion

The formation of mAb fragments is commonly observed during manufacturing and stability storage, yet their interaction with mAb and the impact on aggregation have not been fully understood through systematic studies. As fragments possess structural similarities and differences (e.g., exposed regions) compared to the intact mAb, the presence of mAb fragments in the same environment may influence mAb aggregation profiles. In this work, we investigated the potential impact of Fab/c, a predominant fragment resulting from cleavage in the upper hinge, on the thermal stability of mAbs by identifying fragment associated HMWs and monitoring their aggregation kinetics. A workflow was developed using limited IgdE digestion and HIC fractionation to generate large quantities of Fab/c fragments followed by SEC-native MS analysis of the stability samples. Our study suggests that the presence of Fab/c fragment leads to the formation of heterogeneous dimers for the two mAbs studied, including mAb-Fab/c, Fab/c dimer, in addition to mAb dimer. mAb-Fab dimer was observed for mAb1 samples. In general, when Fab/c was present at a higher percentage in a mAb solution, more Fab/c associated dimers, i.e., mAb-Fab/c and Fab/c dimer, were observed, while formation of mAb dimer slowed down. Using SEC coupled with native MS, the accurate identification of individual dimer species can be achieved, and the kinetics of dimer formation can be measured, with sensitivity and precision that surpasses most other technologies.^6,38

Aggregation of mAbs can be affected by multiple structural features of the mAb, including charge patches, hydrophobicity, post-translational modifications (PTMs), and glycosylation. Several studies demonstrated that deglycosylation can reduce mAb stability and increase mAb aggregation.^39–41 Wada et al. reported that different N-glycans such as core fucosylation, terminal galactose, sialylation, and mannosylation, had distinct effects on the thermal stability of IgG1 mAbs,⁴² highlighting the complex role played by glycans on mAb stability. Our deglycosylation study suggests that the impact of N-glycans could be complicated and might need to be examined case by case because, for mAb1, N-glycans appear to have no significant impact on the total aggregation rate but contribute to the formation of Fab/c containing dimer under the 25°C accelerated stability conditions. The sequence and PTMs determine many aggregation-prone features of mAbs, such as hydrophobic patches, electrostatic motifs, and disulfide bonds subject to scrambling or crosslinking.⁶ These previous studies on full-length mAbs could potentially be applied to predict the aggregation of mAb fragments. To fully understand the impact of structural features, a full panel assessment of mAb fragments with different isotypes, and glycans, among others, would be required.

Although the Fab/c fragment used in this study was generated through enzymatic digestion, it showed a high degree of structural similarity to the Fab/c fragment detected in real stability samples. Therefore, the results presented here could provide insight into the Fab/c-involved aggregation as it relates to real mAb stability. The detection of mAb-Fab in mAb1 samples added another layer of information on fragment-associated mAb aggregation. These findings highlight potential concerns that the presence of mAb fragments could pose to the stability of mAb therapeutics. It was reported the cleavage in the upper hinge region, which is the source of Fab/c and Fab fragments, is mostly affected by pH, with acidic (<5) and basic pH (>8) increasing hinge cleavage. Additionally, metal ions, such as copper and iron, were observed to catalyze the cleavage reaction when present in the formulation.¹⁵ Therefore, formulation development, among other process development, plays a critical role in controlling hinge cleavage and subsequent fragment formation.

Our study also touched upon the degradation of mAbs and Fab/c fragments, and the results indicated Fab/c were more susceptible to degradation than mAb under the evaluated stability conditions, likely due to its impaired structural integrity. Besides thermal stress, we suspect that the residual IgdE enzyme from HIC enabled fractionation, though at trace levels if present, may also contribute to the degradation process, suggested by a larger decrease of mAb1 and mAb2, 9% and 14%, in their respective HIC fractions compared to 2% and 3% degradation in undigested mAb1 and mAb2 samples at the end of the 3-month incubation (Figure. S6) Although this inconsistency did not affect the conclusions on the aggregation kinetics of the two mAbs, it highlights the necessity of optimizing the purification method for further investigation into the stability of Fab/c fragment.

In terms of methodology, the SEC-UV-native MS method enabled the identification of HMWs with structural heterogeneity, intact mAbs with various glycoforms and fragments generated from cleavage at different sites in the hinge region, which could not be achieved by other methods such as DLS or AUC. Non-covalent interactions were preserved under native MS conditions, allowing for the scrutinization of low-level heterodimers, particularly when studying co-eluting species. In the case of mAb1, Fab/c dimer and mAb1-Fab coeluted under one UV peak and were differentiated by native MS based on their deconvoluted masses. Similarly in the case of mAb2, native MS enabled the dissection of the broad peak of HMWs into three dimer species. An exploration of using MS XIC peak area was performed in this study to provide aggregation trending for each dimer. The case study of mAb2 suggested this type of analysis may be particularly valuable when dealing with co-eluting species, as XIC peak area could differentiate all three dimer species and profile their changes during stability when they could not be chromatographically separated.

The current method has some limitations. In particular, HMW species were detected with large mass errors due to limited mass resolution and incomplete desolvation of the TOF instrument. Improved signal intensity and mass resolution were achieved using an Orbitrap UHMR mass spectrometer, which confirmed the identities of the HMWs with < 80 ppm mass error detected in mAb1 fraction samples (Table. S5). HMWs with very small mass differences may not be differentiated by the current MS instrumentation. For instance, mAb2-Fc, if present in the mAb2 fractions, could not be distinguished from Fab/c dimer due to their close masses (theoretical masses of mAb-Fc and Fab/c dimer are 194,519 Da and 194,515 Da). This issue could be potentially solved using higher-resolution mass spectrometers, i.e., with advanced detection capabilities in mass spectrometers, such as extended mass range and multi-stage detection using MS/MS and ion mobility-MS, precise identification of complexed HMWs and differentiation of those with similar molar masses could be achieved.²⁵

The methodology we developed provides a promising platform for studying mAb aggregation influenced by the presence of fragments. Briefly, optimized limited IgdE digestion allowed efficient production of Fab, Fc in addition to Fab/c fragments for IgG1 mAbs, and HIC methods enabled baseline separation and fractionation of all three fragments. The SEC-UV-native MS method provided capabilities in the separation, identification, and quantification of mAb aggregates. As chromatographic and MS technologies continue to advance, this platform holds the potential to accurately characterize more mAb complexes and even study co-formulated mAb-based therapeutics, which tend to form more heterogeneous aggregates.

In conclusion, we investigated the impact of Fab/c fragment on mAb aggregation. We generated Fab/c fragments using limited IgdE digestion and prepared mAb samples containing different levels of Fab/c fragment, which were then incubated under a thermally accelerated condition for stability analysis. The results demonstrated that the presence of Fab/c fragment in mAb led to the formation of heterodimers, including mAb-Fab/c, Fab/c dimer, in addition to mAb dimers. The aggregation rate for Fab/c containing dimers correlate with the abundance of Fab/c fragment, while the aggregation patterns in terms of kinetics and type of dimers formed vary from mAb to mAb. In terms of methodology, SEC-UV-native MS is a valuable addition to the toolbox for understanding mAb stability as it allows the accurate identification and quantification of HMWs, intact mAb and LMWs in a single method. The analytical platform presented in our study also provides valuable insight into studying mAb aggregation and could be potentially expanded to investigating non-covalent interactions between mAb and other components to enhance our knowledge of mAb stability.

Materials and Methods

Materials

IgG1 mAb1 and mAb2 were produced at Biogen. mAb1 is glycosylated and the major glycoforms are core-fucosylated biantennary glycans with 0 and 1 galactose (FA2 and FA2G1); mAb2 is aglycosylated. IgdE enzyme (FabALACTICA®) and PNGaseF enzyme was purchased from Genovis (Massachusetts, US) and New England Biolabs (Massachusetts, US), respectively. LC-MS grade water and LC-MS grade ammonium acetate were purchased from Fisher Scientific. Ammonium sulfate and sodium phosphate was from J.T. Baker. Pellicon XL 50 Ultrafiltration Cassettes for tangential flow filtration, 30 kDa and 10 kDa Amicon® Ultra-15 Centrifugal Filter Units were from MilliporeSigma. The 1 ml HiTrap MabSelect SuRe LX Protein A column (Catalog# 29-2684-02) was from GE Healthcare (Massachusetts, US).

Methods

Limited digestion of mAbs with IgdE

IgdE was reconstituted into LC-MS degrade water to make the concentration of 40 IU/ul. Concentrated mAb solution (100–150 mg/ml) was mixed with IgdE solution at ratio of 100 IU enzyme per 1 mg mAb. Digestion buffer containing 100 mM sodium phosphate, pH 7.0, was added to make the final concentration of mAb 20 mg/ml. The solution was incubated at 37°C for 3 h and frozen at −70°C immediately after incubation. To control the digestion reaction so that the major products are Fab/c and Fab fragments, method development was conducted on enzyme to substrate ratio, reaction time, and temperature (data not shown).

Analytical hydrophobic interaction chromatography

Analytical HIC was conducted by injecting 40–200 µg enzymatic digests onto a Thermo Fisher ProPac, HIC-10 4.6 × 250 mm column at room temperature and separated by a Waters ACQUITY UPLC system using fluorescence detection (ex280/em340). Mobile phase A (MPA) was 1.8 M ammonium sulfate, 100 mM sodium phosphate, pH 6.0 and mobile phase B (MPB) was 100 mM sodium phosphate, pH 6.0. Flow rate was set at 0.2 ml/min. The separation gradient optimized for mAb1 digests started at 25% MPB for 5 min then followed by a steep increase from 25% to 56% within 1 min. During 6 min to 51 min, % MPB was increased gradually from 56% to 60%, then increased to 90% within 1 min. MPB was kept at 90% for 8 min, then decreased to 25%. The equilibration time was set as 40 min. The total run time was 100 min. The HIC LC gradient for mAb2 digests includes 5 min flowing of 25% MPB, a steep increase from 25% to 55% within 1 min, from 6 min to 51 min % MPB increased to 65%, then increased to 90% within 1 min. MPB was kept at 90% for 8 min to flush the column, then decreased to 25% to equilibrate. The equilibration time was set as 40 min. The total run time was 100 min.

Semi-preparative hydrophobic interaction chromatography and fraction collection

A Thermo Fisher ProPac HIC-10 22 mm × 250 mm semi-preparative column and Waters Alliance HPLC was used for separation and fraction collection of Fab/c fragment and mAbs. Injection amount was ~ 20–30 mg per injection. The mobile phases were the same as analytical HIC. UV detector with wavelength set at 280 nm was used for signal detection. Flow rate was 4.57 ml/min. The column was kept at room temperature. Separation gradient for mAb1 was the same as the one for analytical scale. The separation gradient for mAb2 had minor modifications compared to the one for analytical scale to have MPB reach 57.5% at 6 min and 67% at 51 min. The collected fractions of each digest component were pooled together, respectively, before going through buffer exchange to formulation buffer using Pellicon XL 50 Ultrafiltration Cassettes (MilliporeSigma) integrated with 10 kDa membrane onto a Pellicon® Labscale TFF Systems, followed by 30 kDa and 10 kDa Amicon® Ultra-15 Centrifugal Filter Units (MilliporeSigma). The final concentration of the fraction was determined by UV280 using SoloVPE system.

SEC characterization of HIC fractions

Purity of the two purified fractions was characterized by injecting 50 µg of each sample onto a Waters ACQUITY UPLC Protein BEH SEC Column (200 A, 1.7 µm 4.6 mm × 150 mm) and separating the components by a Shimadzu UHPLC. UV at 280 nm was used for signal detection. 25 mM ammonium acetate was used as mobile phase. Flow rate was set at 0.2 ml/min.

DLS analysis of HIC fractions

Each purified fraction was serially diluted to 20 mg/ml, 10 mg/ml, 5 mg/ml, 2.5 mg/ml, and 1.25 mg/ml using formulation buffer. 20 µL of sample solution of each concentration was analyzed by DLS in triplicate using Wyatt Dyna Pro Nanostar and processed by Dynamics software. The obtained results were further presented as a plot of diffusion coefficient (cm²/s) vs. concentration (mg/ml) to calculate the kD values.

SEC-native MS analysis of mAb stability samples

One hundred micrograms of each stability sample was injected onto a Waters ACQUITY UPLC Protein BEH SEC Column (200 A, 1.7 µm 4.6 mm × 150 mm) and analyzed by a Shimadzu UHPLC coupled with a SCIEX TTOF 6600 mass spectrometer (AB Sciex, Massachusetts, US). 25 mM ammonium acetate was used as the mobile phase with an isocratic flow rate at 0.2 ml/min. UV at 280 nm was used for detection. Column temperature was set at 30°C. For MS parameters of SCIEX TTOF, collision energy (CE) was set at 6 eV, declustering potential was at 110 eV, ion spray voltage floating was 5,000 V, temperature was 350°C, mass range was 2,000–9,000 m/z, cycle time was 1.0025 second. Two ion gas, Gas 1 and Gas 2, were both set at 60 psi and curtain gas at 35 psi. Data acquisition software was Analyst TF (1.7.1). Integration and quantification of UV peaks detected by SEC were performed by MultiQuant software. Deconvolution of mass spectra data was conducted by PeakView (Version 2.2.011391) and GeneData Expressionist (Version 16).

Deglycosylated mAb1 sample preparation and the stability study

Deglycosylation of mAb1 was conducted by mixing mAb1 with PNGase F enzyme at enzyme-to-substrate ratio of 1.25 UI : 1 ug, and then incubating at a final protein concentration of 10 mg/ml in 100 mM Tris buffer, pH 7.5 at 37°C overnight. Complete N-glycan removal was confirmed by MS analysis. The residual PNGase F enzyme was then removed through affinity purification using a Protein A column. The purified deglycosylated mAb was then buffer exchanged into mAb1 formulation buffer using a 10 kDa Amicon® Ultra-4 Centrifugal Filter Unit and concentrated to 150 mg/ml. A control sample was prepared in the same way except that the PNGase F enzyme was replaced by the equal volume of 100 mM Tris buffer, pH 7.5 buffer. Both samples were incubated at 25°C for up to 3 months. Time-point samples at T0, 1 month, 2 month and 3 month were submitted for SEC-native MS analysis.

About 11 µg of protein in each sample was analyzed by SEC-native MS using an LC-MS system composed of Waters Acquity UPLC and Thermo Scientific Q Exactive UHMR mass spectrometer equipped with a nano-spray source (Thermo). A Waters BEH SEC column (4.6 mm x 300 mm, 1.7 µm, 125Å) was used for separation of components with isocratic flow of 0.1 mL/min for 40 min using 100 mM ammonium acetate, pH 6.8, as a mobile phase. The eluent from the SEC column was split at a ratio of 1 to 285 so that the effluent entered the mass spectrometer at a flowrate of 0.7 µL/min. A stainless-steel emitter (O.D. 150-μm, I.D. 30-μm, Thermo Scientific ES542) was used for nano-spray ionization. The data were acquired in the positive mode with the capillary temperature set at 250°C, the capillary voltage at 2.1 kV, in-source CID at 80 eV and HCD trapping at 3 V. MS data were acquired over 2000 to 12,000 m/z range. Protein Metrics v4.5 software was used for data deconvolution and analysis.

An alphabetical list of abbreviations

AUC	=	Analytical Ultracentrifugation
DLS	=	Dynamic Light Scattering
HIC	=	Hydrophobic Interaction Chromatography
HMWs	=	High Molecular Weight species
LMWs	=	High Molecular Weight species
MALS	=	Multiangle Light Scattering
SEC	=	Size Exclusion Chromatography
SEC-UV-native MS	=	Size Exclusion Chromatography coupled with native Mass Spectrometry
PTMs	=	Post-translational modifications.

Acknowledgments

The authors thank Bernice Yeung for her critical review and revision on the manuscript, Christopher Barton for his scientific input on data analysis and presentation, and Felicia Wang for the Protein A purification of the deglycosylation stability samples.

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/10.1080/19420862.2024.2334783

Additional information

Funding

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