Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Apr 2.
Published in final edited form as: Anal Chem. 2024 Apr 12;96(16):6209–6217. doi: 10.1021/acs.analchem.3c05253

Structural characterization of a pathogenic antibody underlying vaccine-induced immune thrombotic thrombocytopenia (VITT)

Son N Nguyen 1,a, Si-Hung Le 1,a, Daniil G Ivanov 1, Nikola Ivetic 2, Ishac Nazy 2, Igor A Kaltashov 1,*
PMCID: PMC13040464  NIHMSID: NIHMS2160902  PMID: 38607319

Abstract

Vaccine-induced immune thrombotic thrombocytopenia (VITT) is a rare but dangerous side effect of adenoviral-vectored COVID-19 vaccines. VITT had been linked to production of auto-antibodies recognizing platelet factor 4 (PF4). Here we characterize anti-PF4 antibodies obtained from a VITT patient’s blood. Intact-mass measurements indicate that a significant fraction of these antibodies represents a limited number of clones. MS analysis of large antibody fragments (the light chain, and the Fc/2 and Fd fragments of the heavy chain) confirms the monoclonal nature of this component of the anti-PF4 antibodies repertoire and reveals the presence of a mature complex biantennary N-glycan within the Fd segment. Peptide mapping using two complementary proteases and LC-MS/MS was used to determine the amino acid sequence of the entire light chain and over 98% of the heavy chain (excluding a short N-terminal segment). The sequence analysis allows the monoclonal antibody to be assigned to the IgG2 subclass and verify that the light chain belongs to the λ-type. Incorporation of enzymatic de-N-glycosylation into the peptide mapping routine allows the N-glycan in the Fab region of the antibody to be localized to the framework 3 region of the VH domain. This novel N-glycosylation site is a result of a single mutation within the germline sequence. Peptide mapping also provides information on lower-abundance (polyclonal) components of the anti-PF4 antibody ensemble, revealing the presence of all four subclasses (IgG1-IgG4) and both types of the light chain (λ and κ). This case study demonstrates the power of combining the intact, middle-down and bottom-up MS approaches for meaningful characterization of ultra-low quantities of pathogenic antibodies extracted directly from patients’ blood.

Introduction

One of the unintended consequences of the massive vaccination campaign during the COVID-19 pandemic was the discovery of a rare but extremely dangerous side effect of adenoviral (Ad) vectored vaccines, vaccine-induced immune thrombotic thrombocytopenia (VITT).1 Although VITT incidence rates are generally very low (ranging from 3 to 16 cases per million doses for ChAdOx1, and three- to four-fold lower for human Ad-vectored vaccines2), some studies reported notably higher numbers (e.g., 5 cases per 132,686 vaccine doses in Norway3). Furthermore, VITT is associated with very high mortality rates4 and has become one of the factors contributing to the vaccine hesitancy phenomenon and undermining the COVID-19 vaccination campaign globally.5–7 Lastly, the close association of VITT with a specific delivery vector questions the safety of other Ad-vectored vaccines (both existing and those at the development stage), an alarming prospect given the growing popularity of this platform.8

Despite having several distinct features, such as the unusual anatomical localization of thromboses,9, 10 clinical presentation of VITT is strikingly similar to another immune-mediated blood disorder, heparin-induced thrombocytopenia (HIT).11 At the molecular level, both VITT and HIT are associated with the emergence of antibodies recognizing a small chemokine, platelet factor 4 (PF4).3, 12 Since the anti-PF4 antibodies play a central role in HIT pathogenesis, structural characterization of VITT-associated anti-PF4 antibodies may shed light on the etiology of this vaccination side effect, as well as offer viable treatment and prophylaxis options. The initial mapping of the VITT antibodies’ epitopes on the PF4 surface using alanine scanning revealed their significant overlap with the heparin-binding sites,13 suggesting that the clonality of these antibodies is restricted. Subsequent work by other groups confirmed that the anti-PF4 antibodies in VITT patients are generated by a very limited number of B-cells.14, 15 However, the full sequences of the anti-PF4 antibodies have not been reported in those studies.

Recent advances in mass spectrometry have enabled de novo sequencing of antibodies isolated from complex biological matrices,16–19 although a range of technical issues remain to be addressed prior to this method becoming a routine tool in clinical analysis.20 The purpose of this study is complete structural characterization of a monoclonal anti-PF4 antibody extracted from a VITT patient’s blood (including not only the amino acid sequencing, but also localization of all glycans – including those outside of the canonical N-glycosylation site within the CH2 IgG domain21) with an overarching goal of generating a molecularly defined model of VITT pathogenesis.22

Materials and Methods

Extraction and purification of anti-PF4 antibodies from a VITT patient’s serum.

The plasma sample used in this study was obtained from a clinically diagnosed VITT patient referred for diagnostic testing to the McMaster Platelet Immunology Laboratory; the study was approved by the Hamilton Integrated Research Ethics Board. Anti-PF4 antibodies were affinity-purified from the pool of IgGantibodies extracted from the patient’s plasma using a protein G column.23 See Supplementary Material for a more detailed description of the entire extraction/purification procedure.

Mass spectrometry.

The anti-PF4 antibodies were buffer-exchanged to 150 mM CH3CO2NH4 followed by their Intact-mass measurements with a Synapt G2 HDMS (Waters Corp., Milford, MA) hybrid quadrupole/time-of-flight mass spectrometer equipped with a NanoLockSpray ion source. Charge state assignment was carried out using limited charge reduction,24 and these values were used as input for deconvolution with UniDEC.25

Mass analyses of the Fc/2 and Fd segments of the heavy chain and the light chain were carried out with a SolariX 7 (Bruker Daltonics, Billerica, MA) Fourier transform ion cyclotron resonance (FT ICR) MS in the LC/MS mode. The large fragments were generated by digesting the antibody sample with IdeZ (ThermoFisher, Waltham, MA) followed by disulfide reduction with DTT (Millipore Sigma, St. Louis, MO). Antibody de-N-glycosylation was carried out with PNGase F (New England Biolabs, Ipswich, MA) prior to the IdeZ digestion step. The fragments were separated on the AdvanceBio RP-mAb column (Agilent Technologies, Santa Clara, CA) followed by on-line FT ICR MS measurements. Amino acid sequencing and glycan localization were carried out by digesting the antibody samples with trypsin or chymotrypsin (New England Biolabs, Ipswich, MA) followed by LC-MS/MS analysis of proteolytic fragments with an Orbitrap Fusion (Thermo, San Jose, CA) LC/MS system using HCD to fragment peptide ions.

Results and Discussion

Intact-mass analysis of anti-PF4 antibodies in the clinical sample reveals the presence of an abundant monoclonal antibody with an anomalously high mass.

The anti-PF4 IgG molecules extracted from a VITT patient’s blood were subjected to intact-mass analysis under near-native conditions in solution and aggressive desolvation in the gas phase. This allowed the spectral crowding (typically observed in the low m/z regions of mass spectra of highly heterogeneous proteins acquired under denaturing conditions26, 27) to be avoided, while minimizing the presence of signal corresponding to non-covalent adduct ions.28 Nevertheless, the anti-PF4 antibody mass spectrum is highly convoluted (Figure 1A), and its deconvolution without applying any restrictions yields ambiguous results. To resolve this ambiguity, the limited charge reduction29 was applied to multiple ionic populations selected within narrow m/z windows, revealing the presence of three distinct components (Figure 1B). The identified charge states (z =23–30) were used to restrict the search space for UniDEC deconvolution, yielding a distribution shown in Figure 1A inset. The deconvoluted mass distribution has a clearly defined bimodal character, with its lower part having an appearance that is typical of polyclonal antibodies (in fact, it juxtaposes well with a deconvoluted mass distribution of anti-PF4 antibodies extracted from a HIT patient’s blood, which are known to be polyclonal30 – see the purple trace in the Figure 1A inset). In contrast to this part of the distribution, the higher-mass component of the signal has a well-defined shape typical of monoclonal antibodies.31 Another notable feature of the high-mass component of the anti-PF4 antibody signal is the anomalously high mass of the corresponding IgG molecules. Indeed, while the appearance of the lower-mass component of the antibody signal is consistent with the masses expected for the two most common subclasses (IgG1 and IgG2) carrying light chains of both λ- and κ-types32 and a single biantennary glycan of a complex type per each heavy chain (see the data fitting in Figure 1A), the notably higher mass of the component representing the antibodies with restricted clonality must invoke different interpretations. One possibility is that this component is represented by an IgG molecule of a less common subclass - IgG3 – which features an elongated hinge region with multiple O-glycans.33 Another possible explanation of the mass anomaly is the presence of additional N-glycans within these molecules outside of the canonical N-glycosylation site within the CH2 domain.34

Figure 1.

Figure 1.

Open in a new tab

Intact mass analysis of anti-PF4 IgG antibodies extracted from the blood of a VITT patient (A) and limited charge reduction measurements (B) carried out to facilitate the ionic charge state assignments. The inset shows deconvoluted mass spectra of the anti-PF4 antibodies extracted from the VITT patient’s blood (black trace) and HIT-associated anti-PF4 antibodies as a reference (purple). The color-filled curves show the results of data fitting with three components: red, monoclonal (obtained by convolution of mass distributions of its light chain and Fc/2 and Fd fragments of the heavy chain, see Figure S1 in Supplementary Material); blue, polyclonal λ-component; and orange, polyclonal κ-component.

Mass profiling of large antibody fragments reveals the presence of a fully mature biantennary N-glycan of the complex type within the Fab segment of the VITT patient-derived anti-PF4 mAb.

IdeZ digestion of the anti-PF4 antibody sample followed by disulfide reduction and LC-MS analysis of the resulting fragments yielded mass distributions of the intact light chain, as well as the Fd and Fc/2 fragments of the heavy chain (the former comprising the VH, CH1 domains and the hinge region, and the latter comprising the CH2 and CH3 domains, see Figure S1 in Supplementary Material). Interestingly, fragmenting the entire anti-PF4 antibody ensemble to the level of these segments resulted in a dramatic reduction of the relative abundance of the polyclonal component of the signal. In fact, the mass distribution of the reduced light chain is dominated by a single peak at 22,678±2 Da, a mass consistent with the λ-type.32 Unlike the light chain, mass distributions of both Fd and Fc/2 fragments of the antibody reveal some heterogeneity (see the blue traces in the corresponding panels of Figure S1). While it might be explained away by invoking the notion of a limited number of clones contributing to the overall signal, we note that the spacing patterns between the adjacent peaks are consistent with the notion of limited heterogeneity due to the presence of glycans. This pattern is fully expected for the Fc/2 segment (where the “internal” placement of the N-glycan within the CH2 domain frequently results in incomplete – and heterogeneous – glycosylation due to the steric constraints imposed upon the relevant enzymes by the antibody architecture). The two abundant peaks in the Fd mass distribution are also separated by a mass increment corresponding to a HexNAc residue, but it is not clear whether this pattern is indicative of the presence of a non-canonical N-glycan chain within this segment of the heavy chain or reveals the O-glycosylation within the hinge region, typical of IgG3 antibodies.33 Repeating these measurements using stressed anti-PF4 antibodies obtained in the course of plasmapheresis (red traces in Figure S1) had no effect on the measured mass distribution of the light chain, while revealing a few additional glycoforms within the Fd and Fc/2 segments, consistent with the limited degradation of the carbohydrate chains (but insufficient to determine the Fd glycan type).

The nature of the Fd glycan was determined by de-N-glycosylation of the antibody with PNGase F prior to its processing with IdeZ and disulfide reduction. As expected, the PNGase treatment had no effect on the mass distribution of the light chain (black trace in Figure S1). It also confirmed the presence of an immature biantennary N-glycan of the complex type within the Fc/2 segment (whose mass distribution collapsed to a single peak at 23,792±3 Da). Importantly, the mass distribution of the Fd segment also collapsed to a single peak (25,422±3 Da), not only confirming the presence of an N-glycan within this segment, but also revealing the identity of the major glycoform as a sialylated complex biantennary N-glycan based on the close match of the mass difference between the most abundant isoform of Fd before and after de-glycosylation (2,350±3 Da) and the calculated mass of the Fuc1Hex5HexNAc4NeuAc2 chain (2,350.9 Da). We also note that the second most abundant Fd glycoform is a result of an addition of a fifth HexNAc residue to the complete biantennary chain, the so-called “bisecting” GlcNAc, consistent with the previous reports of the structure of N-glycans found in Fab segments of IgG molecules.21

These measurements clearly confirm the monoclonal nature of the homogeneous high-mass component of the anti-PF4 antibodies extracted from the VITT patient’s blood sample, which will be subsequently referred to as VITT mAb. In addition, these measurements reveal the presence of a biantennary complex N-glycan within the Fab segment of VITT mAb, a feature that appears to correlate with the emergence of monoclonal antibodies in several auto-immune disorders,35–38 as well as placental immune evasion.39 In addition, Fab glycosylation has a potential to impact a range of physical and biochemical properties of IgGs (such as stability),40, 41 many of which may have important – although not completely understood – clinical implications.21 Importantly, convolution of the mass distributions of the large fragments of VITT mAb shown in Figure S1 produces a mass distribution that juxtaposes with the high-mass component of the anti-PF4 VITT antibodies (Figure 1), confirming that the molecular entity represented by this component is indeed VITT mAb.

While the aberrant production of monoclonal antibodies is known to occur in a range of autoimmune disorders, understanding both the molecular mechanism of VITT pathogenesis and the etiology of this dangerous immune response is not possible without detailed structural characterization of the VITT mAb. While the amino acid sequence of the complementarity defining regions (CDRs) within the VL and VH domains of the antibody are obviously the major determinant of its antigen affinity and specificity, other structural aspects are important as well. For example, identification of the antibody subclass (IgG1, IgG2, IgG3 or IgG4) is important vis-à-vis clarifying the specificity of its interaction with a range of relevant Fcγ receptors,42 most notably FcγRIIa on the platelet surface. Furthermore, identification of the N-glycosylation site within the Fab segment of the monoclonal antibody may catalyze the search for molecular mechanisms enabling selection of the aberrant clone producing these pathogenic IgG molecules.43

The VITT mAb light chain type is assigned as λ based on its CL domain sequence.

Amino acid sequencing of and N-glycan localization within the VITT mAb was carried out using peptide mapping and LC-MS/MS analysis of the proteolytic fragments. Trypsin and chymotrypsin were used to produce two complementary peptide maps, which facilitated the fragment peptide alignment (Figure S2). The PEAKS algorithm44 was used for the initial peptide sequence assignment, which were then verified/corrected manually (vide infra). Since the proteolytic fragments were derived from the entire ensemble of anti-PF4 antibodies (i.e., VITT mAbs were not separated from the polyclonal component prior to the proteolytic processing), the resulting peptide fragment mixtures contained a lower-abundance, but extremely diverse background representing the polyclonal anti-PF4 antibodies. Typically, these ions’ abundance fell 1–2 orders of magnitude below that of the peptide ions representing the VITT mAb, which allowed an intensity-based criterion to be used as a filter for proteolytic fragments having similar lengths. However, since the peptide length and the number of incorporated basic residues are the major determinants of the ionic signal intensity, the VITT mAb sequence assembly was carried using a hybrid approach that uses the notion of contigs,45 while at the same time taking the advantage of the availability of gene sequences (IMGT database46) that can be used as candidate scaffolds. The utility of these candidate scaffolds varies significantly across the sequence, as the constant regions (CL, CH1, CH2 and CH3) of IgG molecules, as well as their hinge regions, exhibit minimal variability, and the only source of sequence variation is the existence of multiple subclasses.47 For example, the IgG light chains can belong to either λ- or κ-subclass, which defines the sequence variation space in the light chain outside of its variable region (the number of the gene-based candidate scaffolds for the CL domain sequence assembly). The number of heavy chain subclasses is higher (IgG1 through IgG4 in humans), and the number of corresponding genes in the IMGT database46 defines the sequence assembly space for the CH1, CH2 and CH3 domains and the hinge region. Lastly, even though multiple germline sequences are available for the variable regions (VL and VH), their use as candidate templates for the sequence assembly is limited by the fact that these are targets of somatic hypermutation. However, the latter does not alter the sequence uniformly across the entire polypeptide chain, but is mostly localized within the CDRs, while the framework regions (FRs) separating CDRs are usually affected to a significantly lesser extent. Therefore, the germline sequences can also be used to guide the sequence assembly based on contigs45 within the variable domains of the antibody as well.

As mentioned above, the VITT mAb sequence assembly and verification outside of the variable domain is relatively straightforward due to the availability of the gene sequences.46 Indeed, while the mass measurement of the VITT mAb light chain (Figure S1) suggests that it belongs to the λ-type (based on the mass distributions of the λ- and κ-light chains compiled by Barnidge et al.32), it is the abundant proteolytic fragments (both tryptic and overlapping chymotryptic fragments) derived from the CL domain of the antibody (Figure S2) that allows its type to be assigned as λ without any ambiguity based on the sequence alignment with the corresponding gene IGLC246 (Figure S3).

The hinge region structure determination allows the monoclonal anti-PF4 antibody subclass to be identified as IgG2.

The constant regions of the heavy chain exhibit more diversity due to the existence of four distinct sub-classes (IgG1 through IgG4), and while the most dramatic variability is exhibited by the hinge region, differences also exist within the CH1, CH2 and CH3 domains.46 Figure 2A shows the fragment ion spectrum of the peptide that was tentatively identified by PEAKS as part of the CH1 domain, although a large number of abundant fragment ions were not automatically identified/assigned in the corresponding mass spectrum. A manual inspection of this spectrum revealed the presence of clusters of high-intensity peaks spaced by 18.011±0.001 Da, a mass difference corresponding to multiple (up to three) losses of H2O molecules. While such large numbers of H2O molecules eliminated from fragment ions are relatively rare and are usually not included in search algorithms, we note that the tentative assignment of this peptide as SLSSVVTVPSSNF provides an explanation of this phenomenon. Indeed, this peptide contains a large number of residues with the side chains terminating in hydroxyl groups, and the extent of H2O elimination for each fragment is clearly correlated with the number of such residues (serine and threonine) it encompasses (Figure 2A). The two C-terminal residues of this peptide are unique to the IgG2 subclass, allowing VITT mAb to be assigned to this subclass.

Figure 2.

Figure 2.

Open in a new tab

Representative LC-MS/MS of a chymotryptic (A) and a tryptic (B) proteolytic fragments. The former (peptide SLSSVVTVPSSNF) is derived from the VITT mAb CH1 domain and is unique to the IgG2 subclass. Ionic peaks corresponding to b-fragments are labeled with blue triangles (overlaid white triangles represent the H2O loss), and those corresponding to y-fragments are labeled with red diamonds (an overlaid white diamond represents the NH3 loss). The inset shows the ionic signal of the precursor ion without collisional activation. The stacked arrows in the peptide ion fragmentation scheme indicate multiple H2O losses for a given b- or y-ions. The fully carbamidomethylated tryptic fragment KCCVECPPCPAPPVAGPSVFLFPPKPK (B) covers the hinge region and the N-terminal part of the VITT mAb CH2 domain (unique to the IgG2 subclass). Ionic peaks corresponding to b- and y-fragments are labeled with blue triangles and red diamonds, respectively. The zoom-in 837–854 m/z region shows the detail of the partially overlapping isotopic distributions of the two most abundant fragment ions, and the framed inset on the left shows the ionic signal of the precursor ion without collisional activation. All fragment masses match the calculated ones to 10 ppm, and the sequence coloring scheme in both panels represents Fd (orange) and Fc (green) regions correspondingly.

Further evidence allowing the VITT mAb to be classified as IgG2 comes from the sequence of the hinge region. For example, an abundant ion corresponding to a fully-carbamidomethylated peptide spanning the entire hinge region and the N-terminal part of the CH2 domain (Figure 2B) reveals the amino acid sequence and the high linear density of cysteine residues characteristic of this subclass. The fragment ion assignment in this case is relatively straightforward due to the paucity of hydroxyl-containing residues (and hence the absence of fragment ions generated by H2O elimination), and the presence of multiple proline residues (giving rise to abundant fragments). In this case the manual inspection of the fragment ion spectrum does not yield any additional features beyond those generated automatically by the search algorithm and confirms all PEAKS assignments.

VL and VH sequencing provides structural information on VITT mAb CDR regions.

Amino acid sequencing of the VITT mAb variable regions (both VL and VH) can be somewhat assisted by the availability of the germline sequences of the IGLVX (X= 1–11), IGLJX (x = 1–7), IGHVX (X=1–8) and IGHJX (X=1–6) genes and their alleles,46 which provide valuable reference points for mapping the CDR regions. At the same time, one needs to be mindful of the possibility of somatic hypermutations affecting not only the CDR segments, but also the framework regions (FRs) and V/C junctions, which requires identification and isolation of the specific population of the autoantibody-expressing B-cells to perform the antibody sequencing.. As the anti-PF4 repertoire was isolated from the blood and limited clonality was confirmed vis intact MS analysis,, CDR sequences were determined by de novo sequencing assisted by using the germline gene sequences46 to verify CDR placement/assignment.

An example of using de novo sequencing as a means of identifying the LCDR3 (usually the most variable part within the VL domain)48 is presented in Figure 3A. Despite the large size of this peptide (spanning - in addition to the entire LCDR3 – the C-terminal part of the FR3 segment and most of the VL/CL junction region), abundant fragment ions provide a nearly-complete coverage across the entire sequence, which allows its placement to be made readily. In this case, the y-ions ladder covers the entirety of the peptide sequence with only two gaps, at y25 and y27 positions. The small number of b-ions detected/assigned automatically and verified manually do not fill these gaps, either. While the sequence ambiguity created by one of these gaps (due to the absence of b4/y25 fragments) is small and affects only the order of two residues in the sequence (GD vs. DG), the uncertainty created by another gap (b1/y27 fragments) provides the following four options based on the measured difference between the masses of the precursor ion and the shortest detected b-fragment: VE, LD, EV and DL (here we assume that L and I are identical residues, since the ion activation methods used in this work do not allow a distinction between them to be made). However, careful examination of the fragment ion peaks unassigned by the PEAKS algorithm reveals the presence of three groups of internal ions (see Figure 3A). The largest of these groups creates a ladder spanning the EAGDEGDYY segment at a single residue-level resolution, providing the amino acid sequence information for both gaps. Therefore, a combination of the de novo sequencing algorithms with manual data processing allows the entire amino acid sequence of this peptide to be determined. Interestingly, among the four deviations form a germline sequence within this antibody segment, one (A7G) falls outside of the CDR3 and is located in the FR3, where mutation frequency is generally much lower compared to CDR’s.48 This highlights the need to exercise caution when using gene sequences to facilitate the CDR sequencing/localization work. Nevertheless, the close match of the rest of the N-terminal sequence of this peptide to that of FR3, the presence of a cysteine residue and the match of the C-terminal segment of the peptide to the six residue-long part of the VL/CL junction sequence, as well as the canonical 11 residue length of the CDR segment allow the latter to be confidently assigned as LCDR3.

Figure 3.

Figure 3.

Open in a new tab

LC-MS/MS analysis of tryptic peptides covering the carbamidomethylated LCDR3 (A) and HCDR3 (B) regions of the VITT mAb (the framed insets in both panels show the ionic signal of the intact peptide ions selected as precursors for MS/MS measurements in each case). Ionic peaks corresponding to b- and y-fragments are labeled with blue triangles and red diamonds, respectively. Insertion of an open symbol indicates either H2O or NH3 loss from the corresponding fragment ion. Internal fragments are designated with gold, olive and teal symbols (placement of all internal ions within the peptide sequence is indicated with the appropriately colored double-headed arrows, with dotted lines indicating ambiguous assignments). All fragment masses match the calculated ones to 10 ppm. The zoom-in 333.5–340.5 m/z region in panel B shows the detail of partially overlapping isotopic distributions of two representative fragment ions, as well as one of the internal fragments critical for establishing the amino acid order within the four-residue long N-terminal segment). The peptide sequence coloring scheme follows that shown in Figure S2.

Likewise, HCDR3 (usually the most variable segment of the entire antibody)48 is presented by a peptide spanning its entire length, and also including the VH/CH junction region and a short N-terminal segment of the CH1 domain (Figure 3B). The identification of the CDR3 completely relies on the identification of this tryptic peptide, as no complimentary overlapping CDR3 region were detected in the chymotrypsin-. While the automatically generated sequence shows complete coverage in this case, the abundance of the b-fragments used to identify the four N-terminal residues was determined to be extremely low upon the manual inspection of the data, and these fragment ions were discarded, leaving uncertainty vis-à-vis the N-terminal segment composition. Another sequence ambiguity is due to the gap in the y-fragments ladder between y17+ (the largest detected singly charged y-ion of sufficient abundance) and y202+. The measured mass difference between these two fragments (after correction for the different protonation states) is 333.133 Da, which agrees with the DAF sequence suggested by the de novo sequencing algorithm well within 10 ppm (333.132 Da), but also permits another option, EGF, as well as permutations of both (other isobaric di- and tri-peptide residues can be ruled out, since their masses fall far outside of the acceptable 10 ppm range). Once again, careful inspection of the fragment ion spectrum reveals a plethora of abundant unassigned peaks that can be readily interpreted as internal fragments (labeled with gold and olive dots in Figure 3B). The most abundant group of internal fragments covers the PGLEDAFDLWGQ segment with a single-residue resolution (see the golden ladder on top of the peptide sequence in Figure 3B). This internal ladder allows the DAF/EGF ambiguity to be resolved with high confidence, confirming the sequence within the corresponding gap region.

The second group of internal fragments covers the APGLEDAFD segment, also at a single-residue resolution (although one of the fragments in this series, APGLED, is indistinguishable from its isomer belonging to the first group, PGLEDA). The ladder confirms that the residue preceding proline is indeed alanine. Furthermore, two additional unique internal fragments AAPGLE and AAPGL allow the sequence to be further extended (with another alanine residue). This leaves only two possibilities, having either FA or AF as the N-terminal dipeptide residue (placing SM or MS – dipeptides isobaric to AF – at the N-terminus of the peptide would make the measured peptide ion mass fall outside of the 10 ppm range). Therefore, the only uncertainty in the peptide sequence shown in Figure 3B is the ordering of the first two amino acid residues. Lastly, one other issue with determining the HCDR3 structure is the absence of proteolytic peptides that span both the C-terminal segment of FR3 and the N-terminal segment of HCDR3. However, we note that the HCDR3 segment shown in Figure 3B comprises thirteen residues, which is a canonical length for HCDR3.49

The Fab N-glycan resides within the FR3 segment of the heavy chain and is a result of a single mutation.

Detection of N-glycosylated peptides was carried out by using reporter ions corresponding to mono- and di-saccharides in the fragment ion spectra.50 Specifically, the following reporter ions were used: m/z 204.09 (HexNAc), m/z 274.10 (NeuAc-H2O), m/z 292.11 (NeuAc) and m/z 366.14 (Hex/HexNAc or HexNAc/Hex). In addition to identifying four glycopeptides representing the canonical N-glycosylation site within the CH2 domain of the antibody (see Figure S4 in Supplementary Material), application of this filter allowed another glycopeptide to be identified (Figure 4A). A careful examination of its fragment ion spectrum revealed – in addition to the abundant low-m/z reporter ions – a characteristic glycan chain fragmentation pattern terminating at an ion peak with a measured mass of 978.52 Da, likely representing the intact “base” peptide with a completely stripped carbohydrate chain. De-glycosylation of the antibody results in a complete disappearance of the glycopeptide signal (Figure 4B), while a novel peptide is detected whose mass corresponds to the deamidated version of the putative carbohydrate-free peptide at 978.52 Da (vide supra). This novel doubly charged ion (m/z 490.25) displays a prominent signal in the chromatogram of the de- glycosylated sample, while being absent in the digest of the intact antibody (Figure 4). Collisional activation of this peptide ion generates a distinct fragmentation pattern, from which its amino acid sequence can be readily deduced (LTLSDDTSK). The underlined Asp residue satisfies the N-glycosylation site criterion when converted back to asparagine, Asn-Xxx-Ser/Thr. Aligning this peptide with the available gene sequences allows the matching sequence within the FR3 segment of IGHV2–26 to be identified (LTLSKDTSK), in which a single Lys/Asn mutation gives rise to a novel N-glycosylation site.

Figure 4.

Figure 4.

Open in a new tab

Localization of the N-glycosylation site within the Fab region of the VITT mAb by LC-MS/MS analysis of the tryptic fragment LTL(N*/D)TSK derived from the antibody before (maroon) and after (olive) PNGase F treatment. Panels A and B show TICs of the glycopeptide LTLN*TSK (m/z 1110.45) and its de-N-glycosylated form (m/z 490.25) in both samples, respectively. Panels C and D represent fragment ion spectra for these two peptides, and the framed insets show the molecular ions corresponding to these peptides in the MS1 spectra (which were used as the precursor ions for MS/MS measurements). The gray arrow in panel C identifies the ionic signal of the peptide from which the entire carbohydrate chain was stripped in the gas phase following collisional activation. The labeled low-mass ions (below m/z 400) in the same mass spectrum were used as “carbohydrate markers” when searching for glycopeptides in the entire LC-MS/MS data set.

The polyclonal component of the VITT anti-PF4 antibodies is diverse and represented by all sub-classes of the heavy (IgG1 through IgG4) and light (both λ and κ) chains.

The lower-abundance peptide fragments representing the polyclonal component of the antibody sample provide an opportunity to evaluate the diversity of antibodies outside of the monoclonal component. Multiple low-abundance fragments can be readily aligned with the IGKC gene,46 revealing the presence of the κ-type light chains. Furthermore, lower-abundance peptide fragments mapping to the hinge region in the heavy chain reveal the presence of all four subclasses of the latter (IgG1 through IgG4). While the quantitation of the fractions of these subclasses based on the ionic signal intensity is a dubious proposition, especially given the dramatic variation of the length and composition of the corresponding peptides, the minor subclass-specific variations within the peptides encompassing the canonical N-glycosylation site within the CH2 domain provide a viable opportunity vis-à-vis relative quantitation of the sub-classes. Indeed, the corresponding tryptic peptide in IgG2 (EEQFN*STFR) is distinct from both IgG1 (EEQYN*STYR), IgG3 (EEQYN*STFR) and IgG4 (EEQFN*STYR) (N-glycosylation sites are marked with a star). Since the Phe/Tyr substitutions are not expected to change the ionization efficiency, the signal intensity of the de-glycosylated peptides can be used as a measure of the relative abundance of these sub-classes. This analysis (Figure S5) confirms the prevalence of the IgG2 subclass in the antibody sample (over 80%), while the IgG1 content is only 10% (followed by a single-percentage point contribution from IgG3 and IgG4 combined). The sub-class distribution diagrams shown in the Figure S5 are also noteworthy for allowing the limitations of this approach to the subclass quantitation to be defined. Indeed, while the 10:87:3 IgG1/IgG2/(IgG3+IgG4) ratio reported in the Figure S5 is based on the abundance of the fully protonated molecular ions (MH44+), a notably lower fraction of IgG2 (73%) is obtained when the calculations are based on the intensity of Fe3+ adducts (MHFe4+). The Fe3+ adducts of carbohydrate ions have previously been reported by Klein and Zaia,51 and the presence of two additional hydroxyl groups within the glycopeptide derived from IgG1 (compared to IgG2) should increase the binding affinity by adding two ligands, therefore decreasing the relative abundance of ions representing the IgG2-derived peptides.

Regardless of the procedure used to calculate the fractions of different IgG subclasses within the anti-PF4 antibodies, IgG2 clearly is the most abundant class in this sample. While the low combined fraction of IgG3 and IgG4 in the antibody is not surprising (they typically account for ca. 10% of the total IgG content), the most abundant subclass in the human sera is IgG1,52 accounting for ca. 65% of the total. The dramatic distortion of the IgG2/(total IgG) balance observed in our work for the anti-PF4 IgGs extracted from a VITT patient’s blood is noteworthy, since the IgG2s are known to be involved primarily with recognizing polysaccharides,47 a class of biopolymers notably lacking within its presumed antigen, PF4. While the VITT mAb clearly is one of the primary causes of this misbalance, it remains to be seen what the subclass distribution is among the polyclonal anti-PF4 antibodies, and whether it has any relevance vis-à-vis VITT etiology.

Conclusions

The MS-based characterization of the antibodies extracted from the clinical samples (such as the patients’ blood) remains a non-trivial task that at present cannot be performed in an automated fashion. The presence of the fragments of the background clones and the low quantity of material requires extensive validation of reported findings. Characterization of the pathogenic anti-PF4 antibodies derived from a VITT patient’s not only provides a case study of how such analysis can be performed, but also reveals several intriguing structural features of these antibodies, which are likely to be highly relevant vis-à-vis defining both etiology and pathogenesis of VITT.. First, the subclass distribution is heavily biased towards IgG2, and in fact a significant proportion of the entire antibody ensemble is a monoclonal IgG2 molecule, VITT mAb. This is surprising, since the primary target of IgG2 is bacterial polysaccharides, and carbohydrate chains are notably absent within the VITT mAb’s presumed antigen (a small chemokine PF4). Carbohydrates (O-glycans) are present within the fiber proteins of Ad vectors,53 but their role in Ad immunogenicity remains obscure. Second, the amino acid sequences of both LCDR3 and HCDR3 (the two most variable segments in the light and heavy chains, respectively) show a significant similarity of the sequences reported earlier for an unrelated set of VITT patients.15 This similarity is particularly noteworthy for HCDR3, since it is produced by recombination of three genes, VH, D and J, and is the most variable antibody segment even in the absence of somatic hypermutations. Lastly, VITT mAb has a fully mature biantennary N-glycan within its VH domain (FR3). While the presence of N-glycans in IgG molecules outside of the CH2 domain is currently viewed as uncommon,34 recent reports suggest that this structural feature is shared by several autoimmune disorders.54 In fact, it may serve as a molecular trigger that activates non-canonical clone selection pathways and allows it to evade elimination by the “self/not-self” control machinery,36 leading to the onset of an autoimmune disorder

Supplementary Material

supplementary information

Supporting Information: additional experimental details, VITT mAb sequence and MS data.

Acknowledgments

This work was supported by a grant R01 GM112666 from the National Institute of Health. The experimental work was carried out using shared instrumentation in the Mass Spectrometry Core facility at UMass-Amherst, RRID:SCR_019063.

References

  • (1).Cines DB; Bussel JB SARS-CoV-2 Vaccine-Induced Immune Thrombotic Thrombocytopenia. N. Engl. J. Med 2021, 384 (23), 2254–2256, DOI: 10.1056/NEJMe2106315 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (2).Herrera-Comoglio R; Lane S Vaccine-Induced Immune Thrombocytopenia and Thrombosis after the Sputnik V Vaccine. N. Engl. J. Med 2022, 387 (15), 1431–1432, DOI: 10.1056/NEJMc2210813 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (3).Schultz NH; Sørvoll IH; Michelsen AE; Munthe LA; Lund-Johansen F; Ahlen MT; Wiedmann M; Aamodt AH; Skattør TH; Tjønnfjord GE; Holme PA Thrombosis and Thrombocytopenia after ChAdOx1 nCoV-19 Vaccination. N. Engl. J. Med 2021, 384 (22), 2124–2130, DOI: 10.1056/NEJMoa2104882 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (4).Kehr S; Berg P; Müller S; Fiedler SA; Meyer B; Ruppert-Seipp G; Witzenhausen C; Wolf ME; Henkes HH; Oberle D; Keller-Stanislawski B; Funk MB Long-term outcome of patients with vaccine-induced immune thrombotic thrombocytopenia and cerebral venous sinus thrombosis. NPJ vaccines 2022, 7 (1), 76, DOI: 10.1038/s41541-022-00491-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (5).Zidan A; Noureldin A; Kumar SA; Elsebaie A; Othman M COVID-19 Vaccine-Associated Immune Thrombosis and Thrombocytopenia (VITT): Diagnostic Discrepancies and Global Implications. Semin. Thromb. Hemost 2023, 49 (1), 9–14, DOI: 10.1055/s-0042-1759684 [DOI] [PubMed] [Google Scholar]
  • (6).Adu P; Poopola T; Medvedev ON; Collings S; Mbinta J; Aspin C; Simpson CR Implications for COVID-19 vaccine uptake: A systematic review. J. Infect. Public Health 2023, 16 (3), 441–466, DOI: 10.1016/j.jiph.2023.01.020 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (7).Abad N; Messinger SD; Huang Q; Hendrich MA; Johanson N; Fisun H; Lewis Z; Wilhelm E; Baack B; Bonner KE; Kobau R; Brewer NT A qualitative study of behavioral and social drivers of COVID-19 vaccine confidence and uptake among unvaccinated Americans in the US April-May 2021. PloS one 2023, 18 (2), e0281497, DOI: 10.1371/journal.pone.0281497 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (8).Coughlan L; Kremer EJ; Shayakhmetov DM Adenovirus-based vaccines-a platform for pandemic preparedness against emerging viral pathogens. Mol. Ther 2022, 30 (5), 1822–1849, DOI: 10.1016/j.ymthe.2022.01.034 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (9).Salih F; Schonborn L; Endres M; Greinacher A Immunvermittelte Sinus- und Hirnvenenthrombosen: VITT und prä-VITT als Modellerkrankung [Cerebral Venous and Sinus Thrombosis with Immunological Pathogenesis: VITT and Pre-VITT]. Aktuelle Rheumatol 2022, 47, 490–501, DOI: 10.1055/a-1936-3123 [DOI] [Google Scholar]
  • (10).Lippi G; Favaloro EJ Cerebral Venous Thrombosis Developing after COVID-19 Vaccination: VITT, VATT, TTS, and More. Semin. Thromb. Hemost 2022, 48 (1), 8–14, DOI: 10.1055/s-0041-1736168 [DOI] [PubMed] [Google Scholar]
  • (11).Chen L; Pavord S Clinical picture of VITT. Semin. Hematol 2022, 59 (2), 76–79, DOI: 10.1053/j.seminhematol.2022.02.001 [DOI] [PubMed] [Google Scholar]
  • (12).Scully M; Singh D; Lown R; Poles A; Solomon T; Levi M; Goldblatt D; Kotoucek P; Thomas W; Lester W Pathologic Antibodies to Platelet Factor 4 after ChAdOx1 nCoV-19 Vaccination. N. Engl. J. Med 2021, 384 (23), 2202–2211, DOI: 10.1056/NEJMoa2105385 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (13).Huynh A; Kelton JG; Arnold DM; Daka M; Nazy I Antibody epitopes in vaccine-induced immune thrombotic thrombocytopenia. Nature 2021, 596, 565–569, DOI: 10.1038/s41586-021-03744-4 [DOI] [PubMed] [Google Scholar]
  • (14).Kanack AJ; Bayas A; George G; Abou-Ismail MY; Singh B; Kohlhagen MC; Splinter NP; Christ M; Naumann M; Moser KA; Smock KJ; Grazioli A; Wen R; Wang D; Murray DL; Padmanabhan A Monoclonal and oligoclonal anti-platelet factor 4 antibodies mediate VITT. Blood 2022, 140 (1), 73–77, DOI: 10.1182/blood.2021014588 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (15).Wang JJ; Armour B; Chataway T; Troelnikov A; Colella A; Yacoub O; Hockley S; Tan CW; Gordon TP Vaccine-induced immune thrombotic thrombocytopenia is mediated by a stereotyped clonotypic antibody. Blood 2022, 140 (15), 1738–1742, DOI: 10.1182/blood.2022016474 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (16).Guthals A; Gan Y; Murray L; Chen Y; Stinson J; Nakamura G; Lill JR; Sandoval W; Bandeira N De Novo MS/MS Sequencing of Native Human Antibodies. J. Proteome Res 2017, 16 (1), 45–54, DOI: 10.1021/acs.jproteome.6b00608 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (17).Peng W; Pronker MF; Snijder J Mass Spectrometry-Based De Novo Sequencing of Monoclonal Antibodies Using Multiple Proteases and a Dual Fragmentation Scheme. J. Proteome Res 2021, 20 (7), 3559–3566, DOI: 10.1021/acs.jproteome.1c00169 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (18).Melani RD; Des Soye BJ; Kafader JO; Forte E; Hollas M; Blagojevic V; Negrão F; McGee JP; Drown B; Lloyd-Jones C; Seckler HS; Camarillo JM; Compton PD; LeDuc RD; Early B; Fellers RT; Cho BK; Mattamana BB; Goo YA; Thomas PM; Ash MK; Bhimalli PP; Al-Harthi L; Sha BE; Schneider JR; Kelleher NL Next-Generation Serology by Mass Spectrometry: Readout of the SARS-CoV-2 Antibody Repertoire. J. Proteome Res 2022, 21 (1), 274–288, DOI: 10.1021/acs.jproteome.1c00882 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (19).de Graaf SC; Hoek M; Tamara S; Heck AJR A perspective toward mass spectrometry-based de novo sequencing of endogenous antibodies. mAbs 2022, 14 (1), 2079449, DOI: 10.1080/19420862.2022.2079449 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (20).Beslic D; Tscheuschner G; Renard BY; Weller MG; Muth T Comprehensive evaluation of peptide de novo sequencing tools for monoclonal antibody assembly. Brief. Bioinform 2023, 24 (1), 1–12, DOI: 10.1093/bib/bbac542 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (21).van de Bovenkamp FS; Hafkenscheid L; Rispens T; Rombouts Y The Emerging Importance of IgG Fab Glycosylation in Immunity. J. Immunol 2016, 196 (4), 1435–1441, DOI: 10.4049/jimmunol.1502136 [DOI] [PubMed] [Google Scholar]
  • (22).Ivanov DG; Ivetic N; Du Y; Nguyen SN; Le SH; Favre D; Nazy I; Kaltashov IA Reverse Engineering of a Pathogenic Antibody Reveals the Molecular Mechanism of Vaccine-Induced Immune Thrombotic Thrombocytopenia. J. Am. Chem. Soc 2023, 145 (46), 25203–25213, DOI: 10.1021/jacs.3c07846 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (23).Huynh A; Arnold DM; Smith JW; Elliott TD; Ivetic N; Kelton JG; Nazy I The role of fluid-phase immune complexes in the pathogenesis of heparin-induced thrombocytopenia. Thromb. Res 2020, 194, 135–141, DOI: 10.1016/j.thromres.2020.06.012 [DOI] [PubMed] [Google Scholar]
  • (24).Abzalimov RR; Kaltashov IA Electrospray ionization mass spectrometry of highly heterogeneous protein systems: Protein ion charge state assignment via incomplete charge reduction. Anal. Chem 2010, 82 (18), 7523–7526, DOI: 10.1021/ac101848z [DOI] [PubMed] [Google Scholar]
  • (25).Marty MT; Baldwin AJ; Marklund EG; Hochberg GK; Benesch JL; Robinson CV Bayesian deconvolution of mass and ion mobility spectra: from binary interactions to polydisperse ensembles. Anal. Chem 2015, 87 (8), 4370–4376, DOI: 10.1021/acs.analchem.5b00140 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (26).Bobst CE; Sperry J; Friese OV; Kaltashov IA Simultaneous Evaluation of a Vaccine Component Microheterogeneity and Conformational Integrity Using Native Mass Spectrometry and Limited Charge Reduction. J. Am. Soc. Mass Spectrom 2021, 32 (7), 1631–1637, DOI: 10.1021/jasms.1c00091 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (27).Kafader JO; Melani RD; Schachner LF; Ives AN; Patrie SM; Kelleher NL; Compton PD Native vs Denatured: An in Depth Investigation of Charge State and Isotope Distributions. J. Am. Soc. Mass Spectrom 2020, 31 (3), 574–581, DOI: 10.1021/jasms.9b00040 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (28).Kaltashov IA; Ivanov DG; Yang Y Mass spectrometry-based methods to characterize highly heterogeneous biopharmaceuticals, vaccines, and nonbiological complex drugs at the intact-mass level. Mass Spectrom. Rev 2024, 43, 139–165, DOI: 10.1002/mas.21829 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (29).Yang W; Ivanov DG; Kaltashov IA Extending the capabilities of intact-mass analyses to monoclonal immunoglobulins of the E-isotype (IgE). mAbs 2022, 14 (1), 2103906, DOI: 10.1080/19420862.2022.2103906 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (30).Huynh A; Arnold DM; Kelton JG; Smith JW; Horsewood P; Clare R; Guarne A; Nazy I Characterization of platelet factor 4 amino acids that bind pathogenic antibodies in heparin-induced thrombocytopenia. J. Thromb. Haemost 2019, 17 (2), 389–399, DOI: 10.1111/jth.14369 [DOI] [PubMed] [Google Scholar]
  • (31).Pawlowski JW; Bajardi-Taccioli A; Houde D; Feschenko M; Carlage T; Kaltashov IA Influence of glycan modification on IgG1 biochemical and biophysical properties. J. Pharm. Biomed. Anal 2018, 151, 133–144, DOI: 10.1016/j.jpba.2017.12.061 [DOI] [PubMed] [Google Scholar]
  • (32).Barnidge DR; Dasari S; Ramirez-Alvarado M; Fontan A; Willrich MA; Tschumper RC; Jelinek DF; Snyder MR; Dispenzieri A; Katzmann JA; Murray DL Phenotyping polyclonal kappa and lambda light chain molecular mass distributions in patient serum using mass spectrometry. J. Proteome Res 2014, 13 (11), 5198–5205, DOI: 10.1021/pr5005967 [DOI] [PubMed] [Google Scholar]
  • (33).Damelang T; Rogerson SJ; Kent SJ; Chung AW Role of IgG3 in Infectious Diseases. Trends Immunol 2019, 40 (3), 197–211, DOI: 10.1016/j.it.2019.01.005 [DOI] [PubMed] [Google Scholar]
  • (34).Maverakis E; Kim K; Shimoda M; Gershwin ME; Patel F; Wilken R; Raychaudhuri S; Ruhaak LR; Lebrilla CB Glycans in the immune system and The Altered Glycan Theory of Autoimmunity: a critical review. J. Autoimmun 2015, 57, 1–13, DOI: 10.1016/j.jaut.2014.12.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (35).Vergroesen RD; Slot LM; Hafkenscheid L; Koning MT; van der Voort EIH; Grooff CA; Zervakis G; Veelken H; Huizinga TWJ; Rispens T; Scherer HU; Toes REM B-cell receptor sequencing of anti-citrullinated protein antibody (ACPA) IgG-expressing B cells indicates a selective advantage for the introduction of N-glycosylation sites during somatic hypermutation. Ann. Rheum. Dis 2018, 77 (6), 956–958, DOI: 10.1136/annrheumdis-2017-212052 [DOI] [PubMed] [Google Scholar]
  • (36).Visser A; Hamza N; Kroese FGM; Bos NA Acquiring new N-glycosylation sites in variable regions of immunoglobulin genes by somatic hypermutation is a common feature of autoimmune diseases. Ann. Rheum. Dis 2018, 77 (10), e69, DOI: 10.1136/annrheumdis-2017-212568 [DOI] [PubMed] [Google Scholar]
  • (37).Kissel T; Ge C; Hafkenscheid L; Kwekkeboom JC; Slot LM; Cavallari M; He Y; van Schie KA; Vergroesen RD; Kampstra ASB; Reijm S; Stoeken-Rijsbergen G; Koeleman C; Voortman LM; Heitman LH; Xu B; Pruijn GJM; Wuhrer M; Rispens T; Huizinga TWJ; Scherer HU; Reth M; Holmdahl R; Toes REM Surface Ig variable domain glycosylation affects autoantigen binding and acts as threshold for human autoreactive B cell activation. Sci. Adv 2022, 8 (6), eabm1759, DOI: 10.1126/sciadv.abm1759 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (38).Koers J; Sciarrillo R; Derksen NIL; Vletter EM; Fillié-Grijpma YE; Raveling-Eelsing E; Graça NAG; Leijser T; Pas HH; Laura van Nijen-Vos L; Braham MVJ; Buisman AM; de Jong J; Schriek AI; Tio-Gillen AP; Teng YKO; Steenhuis M; Swaneveld FH; de Taeye SW; van Gils MJ; Verschuuren J; Rutgers B; Heeringa P; Horváth B; Jacobs BC; de Leeuw K; Franssen CFM; Veyradier A; Coppo P; Gelderman KA; Marieke van Ham S; van Els C; van der Woude D; Huizinga R; Huijbers MG; Kuijpers TW; Toes REM; Bos NA; Rispens T Differences in IgG autoantibody Fab glycosylation across autoimmune diseases. J. Allergy Clin. Immunol 2023, 151 (6), 1646–1654, DOI: 10.1016/j.jaci.2022.10.035 [DOI] [PubMed] [Google Scholar]
  • (39).Gu J; Lei Y; Huang Y; Zhao Y; Li J; Huang T; Zhang J; Wang J; Deng X; Chen Z; Korteweg C; Deng R; Yan M; Xu Q; Dong S; Cai M; Luo L; Huang G; Wang Y; Li Q; Lin C; Su M; Yang C; Zhuang Z Fab fragment glycosylated IgG may play a central role in placental immune evasion. Hum. Reprod 2015, 30 (2), 380–391, DOI: 10.1093/humrep/deu323 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (40).Courtois F; Agrawal NJ; Lauer TM; Trout BL Rational design of therapeutic mAbs against aggregation through protein engineering and incorporation of glycosylation motifs applied to bevacizumab. mAbs 2016, 8 (1), 99–112, DOI: 10.1080/19420862.2015.1112477 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (41).van de Bovenkamp FS; Derksen NIL; van Breemen MJ; de Taeye SW; Ooijevaar-de Heer P; Sanders RW; Rispens T Variable Domain N-Linked Glycans Acquired During Antigen-Specific Immune Responses Can Contribute to Immunoglobulin G Antibody Stability. Front. Immunol 2018, 9, 740, DOI: 10.3389/fimmu.2018.00740 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (42).Bruhns P; Iannascoli B; England P; Mancardi DA; Fernandez N; Jorieux S; Daëron M Specificity and affinity of human Fcgamma receptors and their polymorphic variants for human IgG subclasses. Blood 2009, 113 (16), 3716–3725, DOI: 10.1182/blood-2008-09-179754 [DOI] [PubMed] [Google Scholar]
  • (43).Scherer HU; van der Woude D; Toes REM From risk to chronicity: evolution of autoreactive B cell and antibody responses in rheumatoid arthritis. Nat. Rev. Rheumatol 2022, 18 (7), 371–383, DOI: 10.1038/s41584-022-00786-4 [DOI] [PubMed] [Google Scholar]
  • (44).Tran NH; Qiao R; Xin L; Chen X; Liu C; Zhang X; Shan B; Ghodsi A; Li M Deep learning enables de novo peptide sequencing from data-independent-acquisition mass spectrometry. Nat. Methods 2019, 16 (1), 63–66, DOI: 10.1038/s41592-018-0260-3 [DOI] [PubMed] [Google Scholar]
  • (45).Mai Z-B; Zhou Z-H; He Q-Y; Zhang G Highly Robust de Novo Full-Length Protein Sequencing. Anal. Chem 2022, 94 (8), 3467–3475, DOI: 10.1021/acs.analchem.1c03718 [DOI] [PubMed] [Google Scholar]
  • (46).Lefranc MP; Ehrenmann F; Kossida S; Giudicelli V; Duroux P Use of IMGT(®) Databases and Tools for Antibody Engineering and Humanization. Methods Mol. Biol 2018, 1827, 35–69, DOI: 10.1007/978-1-4939-8648-4_3 [DOI] [PubMed] [Google Scholar]
  • (47).Vidarsson G; Dekkers G; Rispens T IgG subclasses and allotypes: from structure to effector functions. Front. Immunol 2014, 5, 520, DOI: 10.3389/fimmu.2014.00520 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (48).Murphy K; Travers P; Walport M; Janeway C Janeway’s immunobiology; Garland Science, 2012. [Google Scholar]
  • (49).Chiu ML; Goulet DR; Teplyakov A; Gilliland GL Antibody Structure and Function: The Basis for Engineering Therapeutics. Antibodies 2019, 8 (4), 55, DOI: 10.3390/antib8040055 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (50).Kaltashov IA; Wang S; Wang G Mass Spectrometry in Biopharmaceutical Analysis; De Gruyter, 2021. DOI: doi: 10.1515/9783110546187. [DOI] [Google Scholar]
  • (51).Klein J; Zaia J Relative Retention Time Estimation Improves N-Glycopeptide Identifications by LC-MS/MS. J. Proteome Res 2020, 19 (5), 2113–2121, DOI: 10.1021/acs.jproteome.0c00051 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • (52).Papadea C; Check IJ Human immunoglobulin G and immunoglobulin G subclasses: biochemical, genetic, and clinical aspects. Crit. Rev. Clin. Lab. Sci 1989, 27 (1), 27–58, DOI: 10.3109/10408368909106589 [DOI] [PubMed] [Google Scholar]
  • (53).Cauet G; Strub J-M; Leize E; Wagner E; Dorsselaer AV; Lusky M Identification of the Glycosylation Site of the Adenovirus Type 5 Fiber Protein. Biochemistry 2005, 44 (14), 5453–5460, DOI: 10.1021/bi047702b [DOI] [PubMed] [Google Scholar]
  • (54).Vletter EM; Koning MT; Scherer HU; Veelken H; Toes REM A Comparison of Immunoglobulin Variable Region N-Linked Glycosylation in Healthy Donors, Autoimmune Disease and Lymphoma. Front. Immunol 2020, 11, 241, DOI: 10.3389/fimmu.2020.00241 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supplementary information
NIHMS2160902-supplement-supplementary_information.docx (898.3KB, docx)

RESOURCES