Browse the corpus

Adenoviral Inciting Antigen and Somatic Hypermutation in VITT. BACKGROUND: Vaccine-induced immune thrombocytopenia and thrombosis (VITT) is a rare prothrombotic complication that occurs after adenoviral vector-based vaccination against coronavirus disease 2019; in rare cases, it can also occur after natural adenovirus infection. VITT is mediated by platelet-activating antibodies against the highly cationic protein platelet factor 4 (PF4). The underlying inciting antigen trigger and immunopathogenesis remain unknown. METHODS: We used antibody proteomics to determine the amino acid sequences of anti-PF4 antibodies from 21 patients with VITT and sequenced the genes encoding the immunoglobulin light-chain hypervariable region from 100 patients with VITT. To identify an adenoviral trigger, we used the antigen-binding fingerprints of anti-PF4 and anti-adenovirus protein antibodies to identify a shared serum clonotype and subsequently used adenovirus protein peptides and recombinant anti-PF4 VITT antibodies to map the mimicking linear epitope. RESULTS: Genomic and proteomic profiling of VITT antibodies revealed a shared immunoglobulin light-chain allele, IGLV3-21*02 or *03, harboring a critical somatic hypermutation, K31E. Only antibodies purified against adenoviral core protein VII (pVII) contained anti-PF4 species matching the VITT fingerprint; antibodies against intact virions or other adenoviral proteins did not. Cross-reactive IgGs were mapped to a basic linear epitope on pVII. A pathogenic anti-PF4 VITT antibody, back-mutated to germline (K31), lost its prothrombotic activity in vitro and in vivo and preferentially bound pVII, a finding that directly supported the role of the hypermutation in the antigenic shift from adenovirus pVII to PF4. CONCLUSIONS: The results of our study indicate that VITT occurs when, in persons with immunoglobulin light-chain allele IGLV3-21*02 or *03, a specific somatic hypermutation develops that affects antibodies that recognize a specific epitope on the adenoviral core protein pVII, which results in misdirection of antibody targeting toward PF4. (Funded by Deutsche Forschungsgemeinschaft and others; German Clinical Trials Register number, DRKS00025738; EU Post-Authorization Study Register number, EUPAS45098.).

Anti-platelet factor 4 (PF4) antibodies cause the severe adverse reaction, vaccine-induced immune thrombocytopenia and thrombosis (VITT).1–3 VITT is specific for adenovirus vector-based Covid-19 vaccines, ChAdOx1 nCoV-19-S [AstraZeneca] and Ad26.COV2.S [Johnson & Johnson/Janssen].4 The common feature of these two vaccines is the adenovirus particle,5 albeit derived from distinct adenovirus serotypes: ChAdOx1 is a chimpanzee adenovirus, and Ad26 a human adenovirus. Remarkably, VITT also occurs rarely after natural adenovirus infection,6–8 suggesting that a highly conserved structure within different adenovirus subtypes potentially induces the anti-PF4 response. This notion is corroborated by the uniform structure of the anti-PF4 antibodies induced in the VITT immune response. The anti-PF4 antibodies isolated from sera obtained from patients with VITT after vaccination or natural adenovirus infection9 are typically monoclonal or oligoclonal.10,11 They exhibit light and heavy chains with nearly identical (“stereotyped”) molecular fingerprints among patients. This allowed us to generate recombinant versions of these antibodies, enabling detailed in-vitro and in-vivo studies.12,13

ation or natural adenovirus infection9 are typically monoclonal or oligoclonal.10,11 They exhibit light and heavy chains with nearly identical (“stereotyped”) molecular fingerprints among patients. This allowed us to generate recombinant versions of these antibodies, enabling detailed in-vitro and in-vivo studies.12,13 Given that high-titer anti-PF4 antibodies appear as early as 5 days post-vaccination, they likely form as part of a secondary (boosted) immune response. Considering the virtually identical fingerprints between the anti-PF4 antibodies after vaccination and adenovirus infection, we hypothesized that anti-PF4 antibodies likely emerge from a memory anti-adenoviral response. Here we pinpoint the antigen inducing VITT as a cross-reactive determinant shared between PF4 and the highly conserved, highly abundant adenoviral core protein VII (pVII), which binds to the viral genome and mediates its nuclear transport.14 Importantly, pVII shares similarities to PF4 in both charge and binding properties to natural ligands like DNA.14 Strongly stimulated B cells undergo somatic hypermutation, introducing single base changes into the rearranged variable regions of IgG at a rate of approximately one mutation per kilobase per generation,15 driving affinity-based selection.16 We demonstrate that a specific genetic background (IGLV3-21*02/*03 allele) together with a specific somatic hypermutation causes a striking shift in antibody specificity from pVII to PF4, resulting in pathogenic “VITT antibodies” capable of forming platelet-activating PF4-IgG immune complexes.

Detailed descriptions of methods are given in the Supplementary Appendix. Patients, who provided written informed consent, as described,17 and were reported previously,10,17–19 presented with typical VITT or pre-VITT after ChAdOx1 nCoV19-S or Ad26.COV2.S vaccination according to the revised Brighton collaboration case definition4 (Table 1), or after adenovirus infection.9 All patients had platelet-activating anti-PF4 antibodies detected by anti-PF4/heparin enzyme-linked immunosorbent assay (ELISA), with strong reactivity against PF4 alone in a chemiluminescence assay,7 and by the PF4-induced platelet activation assay (PIPA).2 Furthermore, baseline repository plasma was available from two individuals who had donated blood several weeks before Covid-19 vaccination and subsequent development of VITT.

assay (ELISA), with strong reactivity against PF4 alone in a chemiluminescence assay,7 and by the PF4-induced platelet activation assay (PIPA).2 Furthermore, baseline repository plasma was available from two individuals who had donated blood several weeks before Covid-19 vaccination and subsequent development of VITT. We affinity-purified antibodies from sera obtained from patients with VITT. We used ChAdOx1 virion-coated magnetic beads to isolate antibodies binding to ChAdOx1 surface antigens. We next purified anti-PF4 and anti-core protein VII (pVII) antibodies from these sera using PF4-coupled and pVII-coupled magnetic beads, and, by the same approach, anti-penton, anti-protein IIIa (pIIIa), anti-protein V (pV), and anti-protein VI (pVI) antibodies (all adenovirus proteins).10 We separated the heavy and light chains of these affinity-purified antibodies by SDS-PAGE under reduced conditions, analyzed peptides by tandem mass spectrometry, and determined peptide sequences by de novo sequencing and International ImMunoGeneTics database matching using Peaks studio XPro software (Bioinformatics Solution Inc., Waterloo, ON, Canada). Using the amino acid sequences of anti-PF4 antibodies obtained by mass spectrometry, we generated two recombinant anti-PF4 IgG1 antibodies by reverse engineering12 (CR22046, CR23004). In both recombinant antibodies, we also engineered back-mutations to the germline by exchanging amino acid glutamic acid at position-31 of the light chain hypervariable region to germline lysine (E31K) (E31K-CR22046 and E31K-CR23004). For CR22046, we also generated a second antibody variant, changing the first aspartic acid at position-50 of the DDSD motif in the light chain complementarity-determining region 2 (LCDR2) encoded by IGLV3-21*02/*03 to tyrosine encoded by the alleles IGLV3-21*01 or *04 (D50Y-CR22046) (Figure S1). Recombinant antibodies were characterized by the same methods as the original sera. Human serum contains proteins that may interfere with PF4-platelet interaction,20 so we dissolved them in normal (anti-PF4 antibody-negative) human serum.

to tyrosine encoded by the alleles IGLV3-21*01 or *04 (D50Y-CR22046) (Figure S1). Recombinant antibodies were characterized by the same methods as the original sera. Human serum contains proteins that may interfere with PF4-platelet interaction,20 so we dissolved them in normal (anti-PF4 antibody-negative) human serum. DNA purified from peripheral EDTA blood was sequenced on an Illumina NovaSeq 6000 S4 Flowcell using 150bp paired-end reads. The genes encoding the hypervariable region of the immunoglobulin light chain were retrieved from the genome. Microtiter plates were coated with adenovirus vectors ChAdOx1-S, Ad26.COV2.S, and recombinantly produced adenovirus proteins (GenScript): penton, pIIIa, pV, pVI, and pVII, and a peptide library of the ChAdOx1 pVII protein (Mimotopes, Melbourne, Australia; peptide sequences in Supplementary Figure S2) to test binding of antibodies from VITT sera and the recombinant anti-PF4 antibodies and their back-mutated counterparts. The recombinant antibody CR22046 or its back-mutated variant E31K-CR22046 were given intravenously (1.5μg/g bodyweight/day) for 5 consecutive days to a transgenic mouse line expressing human PF4 and FcγIIa receptors with an additional knockout for mouse PF4 on a C57BL/6J background (B6(Cg);SJL-Pf4tm1 Tg(PF4) Tg(FCGR2A)).21–23

We affinity-purified antibodies from sera obtained from patients with VITT. We used ChAdOx1 virion-coated magnetic beads to isolate antibodies binding to ChAdOx1 surface antigens. We next purified anti-PF4 and anti-core protein VII (pVII) antibodies from these sera using PF4-coupled and pVII-coupled magnetic beads, and, by the same approach, anti-penton, anti-protein IIIa (pIIIa), anti-protein V (pV), and anti-protein VI (pVI) antibodies (all adenovirus proteins).10 We separated the heavy and light chains of these affinity-purified antibodies by SDS-PAGE under reduced conditions, analyzed peptides by tandem mass spectrometry, and determined peptide sequences by de novo sequencing and International ImMunoGeneTics database matching using Peaks studio XPro software (Bioinformatics Solution Inc., Waterloo, ON, Canada). Using the amino acid sequences of anti-PF4 antibodies obtained by mass spectrometry, we generated two recombinant anti-PF4 IgG1 antibodies by reverse engineering12 (CR22046, CR23004). In both recombinant antibodies, we also engineered back-mutations to the germline by exchanging amino acid glutamic acid at position-31 of the light chain hypervariable region to germline lysine (E31K) (E31K-CR22046 and E31K-CR23004). For CR22046, we also generated a second antibody variant, changing the first aspartic acid at position-50 of the DDSD motif in the light chain complementarity-determining region 2 (LCDR2) encoded by IGLV3-21*02/*03 to tyrosine encoded by the alleles IGLV3-21*01 or *04 (D50Y-CR22046) (Figure S1). Recombinant antibodies were characterized by the same methods as the original sera. Human serum contains proteins that may interfere with PF4-platelet interaction,20 so we dissolved them in normal (anti-PF4 antibody-negative) human serum.

DNA purified from peripheral EDTA blood was sequenced on an Illumina NovaSeq 6000 S4 Flowcell using 150bp paired-end reads. The genes encoding the hypervariable region of the immunoglobulin light chain were retrieved from the genome.

Microtiter plates were coated with adenovirus vectors ChAdOx1-S, Ad26.COV2.S, and recombinantly produced adenovirus proteins (GenScript): penton, pIIIa, pV, pVI, and pVII, and a peptide library of the ChAdOx1 pVII protein (Mimotopes, Melbourne, Australia; peptide sequences in Supplementary Figure S2) to test binding of antibodies from VITT sera and the recombinant anti-PF4 antibodies and their back-mutated counterparts.

The recombinant antibody CR22046 or its back-mutated variant E31K-CR22046 were given intravenously (1.5μg/g bodyweight/day) for 5 consecutive days to a transgenic mouse line expressing human PF4 and FcγIIa receptors with an additional knockout for mouse PF4 on a C57BL/6J background (B6(Cg);SJL-Pf4tm1 Tg(PF4) Tg(FCGR2A)).21–23

We determined the amino acid sequences of the heavy and light chains of purified anti-PF4 antibodies from 21 patients with VITT after either ChAdOx1 nCoV-19-S (n=18) or Ad26.COV2.S Covid-19 (n=3) vaccination. All antibodies possessed similar molecular fingerprints with the common “ED” motif in the heavy chain and restricted IGLV3-21*02/*03 light chain, consistent with Ig light chain genotyping (Table 1; Figure 1A). Notably, a striking shared somatic mutation was observed in the light chains across different patients, which resulted in an exchange of a positively charged lysine (K) to a negatively charged glutamic acid (E) mutation at position-31 (K31E) (Table 1; Figure 1A). We confirmed through germline immunoglobulin G sequencing in 18 subjects with available DNA that this E31 was a somatic hypermutation, i.e., not present as a germline polymorphism. This finding was confirmed by genome sequencing of an additional 82 patients with VITT,17 all of whom had the IGLV3-21*02/*03 allele encoding lysine at position-31 (K31) in their germline sequence of the hypervariable light chain region.

We determined the amino acid sequences of the heavy and light chains of purified anti-PF4 antibodies from 21 patients with VITT after either ChAdOx1 nCoV-19-S (n=18) or Ad26.COV2.S Covid-19 (n=3) vaccination. All antibodies possessed similar molecular fingerprints with the common “ED” motif in the heavy chain and restricted IGLV3-21*02/*03 light chain, consistent with Ig light chain genotyping (Table 1; Figure 1A). Notably, a striking shared somatic mutation was observed in the light chains across different patients, which resulted in an exchange of a positively charged lysine (K) to a negatively charged glutamic acid (E) mutation at position-31 (K31E) (Table 1; Figure 1A). We confirmed through germline immunoglobulin G sequencing in 18 subjects with available DNA that this E31 was a somatic hypermutation, i.e., not present as a germline polymorphism. This finding was confirmed by genome sequencing of an additional 82 patients with VITT,17 all of whom had the IGLV3-21*02/*03 allele encoding lysine at position-31 (K31) in their germline sequence of the hypervariable light chain region. This negatively charged glutamic acid at position-31 of the light chain, formed by somatic mutation, together with the negatively charged “ED” binding motif of the VITT antibody heavy chain hypervariable region (Figure 1A), forms a strongly negatively charged paratope, facilitating strong binding to positively charged PF4 (Figure 1B). In contrast, in three healthy individuals developing non-pathogenic anti-PF4 antibodies after ChAdOx1 nCoV19-S vaccination,24 the resulting antibodies were low-titer and non-platelet-activating. Furthermore, antibody sequencing revealed evidence of multiple antibody clonotypes lacking both the typical IGLV3-21*02/*03 allele and the K31E mutation.

y individuals developing non-pathogenic anti-PF4 antibodies after ChAdOx1 nCoV19-S vaccination,24 the resulting antibodies were low-titer and non-platelet-activating. Furthermore, antibody sequencing revealed evidence of multiple antibody clonotypes lacking both the typical IGLV3-21*02/*03 allele and the K31E mutation. From the pre-vaccination plasma of two individuals who subsequently developed VITT after ChAdOx1 nCoV19-S vaccination, we isolated non-pathogenic, polyclonal anti-PF4 antibodies. In both individuals we identified the prototypical anti-PF4 light chain encoded by IGLV3-21*02 expressing the germline-encoded lysine at position-31 (K31). In one individual, a rare somatic variant with a K31E mutation was observed, present in 2% of the LCDR1 peptides detected in the proteomic analysis. In contrast, their VITT anti-PF4 antibodies generated in the acute-phase were dominated by the typical K31E mutation (Table 1), absent in the germline DNA of these two individuals.

dividual, a rare somatic variant with a K31E mutation was observed, present in 2% of the LCDR1 peptides detected in the proteomic analysis. In contrast, their VITT anti-PF4 antibodies generated in the acute-phase were dominated by the typical K31E mutation (Table 1), absent in the germline DNA of these two individuals. We next generated two recombinant VITT IgG antibodies (CR22046 and CR2300413) from patients developing VITT after ChAdOx1 nCoV19-S or Ad26.COV2.S, respectively (Figure 1A). Both recombinant antibodies possessed activity profiles identical to typical VITT antibodies,13 i.e., they bound PF4 (Figure 1C) but not PF4/heparin complexes (Figure S3), and activated platelets in a PF4-dependent fashion (Figure 1D). Additionally, dosing of CR22046 (n=15) led to a rapid and pronounced platelet count decrease (−79.3%) (Figure 1E) and thromboses in 13/15 hPF4-hFcγRIIa-transgenic mice (86.7%), including cerebral venous sinus and splanchnic vein thromboses (Figure 1F).23

, and activated platelets in a PF4-dependent fashion (Figure 1D). Additionally, dosing of CR22046 (n=15) led to a rapid and pronounced platelet count decrease (−79.3%) (Figure 1E) and thromboses in 13/15 hPF4-hFcγRIIa-transgenic mice (86.7%), including cerebral venous sinus and splanchnic vein thromboses (Figure 1F).23 To understand the functional role of K31E in VITT pathogenesis, we examined the recombinant antibodies CR22046 and CR23004, back-mutated to express the germline lysine (K) residue at position-31 of the hypervariable region of the light chain (E31K-CR22046 and E31K-CR23004). Both back-mutated recombinant antibodies showed a major reduction of binding to PF4 in the chemiluminescence test and required very high concentrations (≥120 μg/mL) to induce platelet activation in the functional PF4-dependent assay (Figure 1C&1D). In-vivo, the back-mutated E31K-CR22046 caused thromboses only in 3/12 mice (Figure 1F) and exhibited the same platelet count profile as the negative control (Figure 1E). We conclude that K31E represents a crucial somatic hypermutation, which is required to produce the highly reactive anti-PF4 antibodies characteristic of VITT. To determine the impact of the IGLV3-21 light chain allele, we substituted the IGLV3-21*02/*03 allele aspartic acid (D) at position-50 with tyrosine (Y) as encoded by the IGLV3-21*01 or *04 allele (D50Y-CR22046). This resulted in loss of PF4 binding and platelet activation (Figure 1C&1D), compared with CR22046.

To understand the functional role of K31E in VITT pathogenesis, we examined the recombinant antibodies CR22046 and CR23004, back-mutated to express the germline lysine (K) residue at position-31 of the hypervariable region of the light chain (E31K-CR22046 and E31K-CR23004). Both back-mutated recombinant antibodies showed a major reduction of binding to PF4 in the chemiluminescence test and required very high concentrations (≥120 μg/mL) to induce platelet activation in the functional PF4-dependent assay (Figure 1C&1D). In-vivo, the back-mutated E31K-CR22046 caused thromboses only in 3/12 mice (Figure 1F) and exhibited the same platelet count profile as the negative control (Figure 1E). We conclude that K31E represents a crucial somatic hypermutation, which is required to produce the highly reactive anti-PF4 antibodies characteristic of VITT. To determine the impact of the IGLV3-21 light chain allele, we substituted the IGLV3-21*02/*03 allele aspartic acid (D) at position-50 with tyrosine (Y) as encoded by the IGLV3-21*01 or *04 allele (D50Y-CR22046). This resulted in loss of PF4 binding and platelet activation (Figure 1C&1D), compared with CR22046. Following the hypothesis that pathogenic anti-PF4 antibodies arise from a misdirected memory response to a shared adenoviral antigen, we next sought to determine whether VITT-related antibody species are present in the antibody response to the adenoviral vector. First, we did not detect any matched light chain or heavy chain motifs unique to pathogenic anti-PF4 clonotypes in antibodies isolated from intact ChAdOx1 virion particles from VITT serum. Further, neither human anti-PF4 antibodies (purified from VITT sera) nor recombinant anti-PF4 antibodies bound to ChAdOx1 or Ad26 virion coated on ELISA plates (Figure S4). These findings argue against an inciting antigen on the adenoviral surface.

tibodies isolated from intact ChAdOx1 virion particles from VITT serum. Further, neither human anti-PF4 antibodies (purified from VITT sera) nor recombinant anti-PF4 antibodies bound to ChAdOx1 or Ad26 virion coated on ELISA plates (Figure S4). These findings argue against an inciting antigen on the adenoviral surface. Next we examined responses to adenovirus core proteins, focusing on pVII as a candidate, given its high abundance and similarities in charge and polyanion binding properties to PF4.14 Immunoglobulin G purified from four VITT sera against recombinant ChAdOx1 pVII protein was found to be oligoclonal but contained a clonotypic species that matched anti-PF4 antibody fingerprints from the same patient with VITT (Figure 2A). Antibodies purified against each protein (pVII and PF4) revealed reciprocal cross-reactivity (Figure 2B1&2). Anti-pVII antibodies detected in a healthy vaccinee were non-cross-reactive with PF4 (Figure 2B2 “healthy donor”). Furthermore, human recombinant anti-PF4 VITT antibodies also bound to pVII (Figure 2B3). Importantly, the recombinant antibody E31K-CR22046, in which we back-mutated the sequence at position-31 to the germline sequence, lost reactivity to PF4 while showing enhanced binding to pVII compared with its K31E counterpart (Figure 2B3). In contrast, none of our controls—the purified anti-penton, anti-pIIIa, anti-pV or anti-pVI antibodies—contained clonotypic species that matched anti-PF4 fingerprints from the same patient with VITT.

sequence, lost reactivity to PF4 while showing enhanced binding to pVII compared with its K31E counterpart (Figure 2B3). In contrast, none of our controls—the purified anti-penton, anti-pIIIa, anti-pV or anti-pVI antibodies—contained clonotypic species that matched anti-PF4 fingerprints from the same patient with VITT. We next mapped the binding site of the anti-PF4 IgGs purified from VITT sera using overlapping pVII synthetic peptides. Strongest binding occurred to a 15-mer peptide reflecting a single linear epitope in the middle of the ChAdOx1 pVII (Figure 3A). This finding was fully recapitulated by our human recombinant anti-PF4 antibodies. Consistent with all other experiments, the back-mutated anti-PF4 antibody E31K-CR22046 showed increased binding to the 15-mer peptide (Figure 3B). The identified 15-mer peptide (RYARAKSRRRRIARR) is highly conserved in ChAdOx1 and Ad26 (Figure 3C; 3D) and predicted to form an alpha-helix analogous to the PF4 epitope in VITT25 (Figure S5).

We next generated two recombinant VITT IgG antibodies (CR22046 and CR2300413) from patients developing VITT after ChAdOx1 nCoV19-S or Ad26.COV2.S, respectively (Figure 1A). Both recombinant antibodies possessed activity profiles identical to typical VITT antibodies,13 i.e., they bound PF4 (Figure 1C) but not PF4/heparin complexes (Figure S3), and activated platelets in a PF4-dependent fashion (Figure 1D). Additionally, dosing of CR22046 (n=15) led to a rapid and pronounced platelet count decrease (−79.3%) (Figure 1E) and thromboses in 13/15 hPF4-hFcγRIIa-transgenic mice (86.7%), including cerebral venous sinus and splanchnic vein thromboses (Figure 1F).23 To understand the functional role of K31E in VITT pathogenesis, we examined the recombinant antibodies CR22046 and CR23004, back-mutated to express the germline lysine (K) residue at position-31 of the hypervariable region of the light chain (E31K-CR22046 and E31K-CR23004). Both back-mutated recombinant antibodies showed a major reduction of binding to PF4 in the chemiluminescence test and required very high concentrations (≥120 μg/mL) to induce platelet activation in the functional PF4-dependent assay (Figure 1C&1D). In-vivo, the back-mutated E31K-CR22046 caused thromboses only in 3/12 mice (Figure 1F) and exhibited the same platelet count profile as the negative control (Figure 1E). We conclude that K31E represents a crucial somatic hypermutation, which is required to produce the highly reactive anti-PF4 antibodies characteristic of VITT.

the back-mutated E31K-CR22046 caused thromboses only in 3/12 mice (Figure 1F) and exhibited the same platelet count profile as the negative control (Figure 1E). We conclude that K31E represents a crucial somatic hypermutation, which is required to produce the highly reactive anti-PF4 antibodies characteristic of VITT. To determine the impact of the IGLV3-21 light chain allele, we substituted the IGLV3-21*02/*03 allele aspartic acid (D) at position-50 with tyrosine (Y) as encoded by the IGLV3-21*01 or *04 allele (D50Y-CR22046). This resulted in loss of PF4 binding and platelet activation (Figure 1C&1D), compared with CR22046.

Following the hypothesis that pathogenic anti-PF4 antibodies arise from a misdirected memory response to a shared adenoviral antigen, we next sought to determine whether VITT-related antibody species are present in the antibody response to the adenoviral vector. First, we did not detect any matched light chain or heavy chain motifs unique to pathogenic anti-PF4 clonotypes in antibodies isolated from intact ChAdOx1 virion particles from VITT serum. Further, neither human anti-PF4 antibodies (purified from VITT sera) nor recombinant anti-PF4 antibodies bound to ChAdOx1 or Ad26 virion coated on ELISA plates (Figure S4). These findings argue against an inciting antigen on the adenoviral surface.

Molecular mimicry between the adenoviral core protein pVII and PF4, combined with somatic hypermutation transforming an anti-pVII immune response to a misdirected anti-PF4 immune response, is a fundamental pathobiological mechanism of VITT. Immune cross-reactivity between pVII and PF4 results from shared, highly positively-charged epitopes with structural similarity in alpha helixes of both proteins. Nonetheless, production of pathogenic, high-avidity anti-PF4 reactive antibodies requires additional specific genetic features. The first prerequisite is a genetic predisposition harboring the allele of IGLV3-21*02/*03 of the hypervariable region of the immunoglobulin light chain. This is supported by our finding that it was present in 99 of 100 patients with VITT (versus the expected background frequency of 20 to 60%, depending upon ethnicity26). The second prerequisite is a somatic hypermutation exchanging a positively charged lysine to a negatively charged glutamic acid at position-31 of the light chain hypervariable region of the anti-pVII antibodies. We confirm this previously described typical fingerprint of the hypervariable immunoglobulin G regions9,10 in all 21 patients with VITT investigated.

rmutation exchanging a positively charged lysine to a negatively charged glutamic acid at position-31 of the light chain hypervariable region of the anti-pVII antibodies. We confirm this previously described typical fingerprint of the hypervariable immunoglobulin G regions9,10 in all 21 patients with VITT investigated. Both IGLV3-21*02 and *03 alleles encode the identical DDSD motif in the LCDR2, critical for forming the paratope (the part of the antibody that binds to PF4). The single amino acid difference between these alleles (position-17) is not critical for paratope formation. However, when we substituted this sequence with an IGLV3-21*01 or *04 sequence (D50Y-CR22046), anti-PF4 binding capacity was lost (Figures 1C; S5). The prerequisite for alleles IGLV3-21*02 or IGLV3-21*03 may explain the lower frequency of VITT in Asia, as these two haplotypes are found in approximately 60% of Caucasian populations, but only 20% of Asian populations.26

IGLV3-21*01 or *04 sequence (D50Y-CR22046), anti-PF4 binding capacity was lost (Figures 1C; S5). The prerequisite for alleles IGLV3-21*02 or IGLV3-21*03 may explain the lower frequency of VITT in Asia, as these two haplotypes are found in approximately 60% of Caucasian populations, but only 20% of Asian populations.26 Many healthy individuals with the genetic background of IGLV3-21*02/*03 acquire anti-adenovirus antibodies.27 However, induction of VITT requires B-cells specific for the pVII RYARAKSRRRRIARR epitope and an additional, specific somatic hypermutation leading to a negatively-charged amino acid at position-31 of the immunoglobulin G light chain hypervariable region. All pathogenic anti-PF4 VITT antibodies reported before,9,10 or sequenced in this study, show either glutamic acid or aspartic acid at this position. The germline DNA of all 100 patients with VITT we investigated encoded lysine at this position, confirming that this mutation is somatic. This specific hypermutation likely occurs in few B-cells, explaining why VITT anti-PF4 antibodies are mono- or oligoclonal. Recently, pathogenic anti-PF4 antibodies in heparin-induced thrombocytopenia were also found to be monoclonal,28 but the role of antibody genetics in shaping these responses is unknown.

s somatic. This specific hypermutation likely occurs in few B-cells, explaining why VITT anti-PF4 antibodies are mono- or oligoclonal. Recently, pathogenic anti-PF4 antibodies in heparin-induced thrombocytopenia were also found to be monoclonal,28 but the role of antibody genetics in shaping these responses is unknown. In VITT antibodies, the key mutation at position-31—from a positively charged lysine residue to a negatively charged glutamic or aspartic acid—results in a striking change in antibody avidity towards PF4 and pathogenicity, as shown by the in-vitro and in-vivo experiments using our recombinant antibodies, with and without the K31E mutation. This high avidity is the prerequisite to cluster PF4, thereby forming pathogenic, platelet-activating immune complexes containing PF4.29,30 When we back-mutated IGLV3-21 residue 31 from glutamic acid to lysine, as encoded by the germline DNA (Figure 1A), the antibodies still bound to PF4 but with low binding activity and activated platelets only in the presence of PF4 at very high antibody concentrations (Figures 1C&D). Consistent with our model, the back-mutated E31K-CR22046 showed stronger binding to pVII (Figure 2B3).

ine, as encoded by the germline DNA (Figure 1A), the antibodies still bound to PF4 but with low binding activity and activated platelets only in the presence of PF4 at very high antibody concentrations (Figures 1C&D). Consistent with our model, the back-mutated E31K-CR22046 showed stronger binding to pVII (Figure 2B3). The rapid onset of the IgG-mediated VITT immune response (as early as 5 days post-vaccination) strongly indicates that the anti-PF4 immune response is secondary. In two individuals we identified a plasma sample available from a few weeks before they developed VITT. Analysis of the clonotypes of the anti-PF4 antibodies in both patients indicated that the clones with the somatic hypermutations rapidly expanded by boosting after vaccination. The hypermutation likely occurred during boosting, but in one patient also a rare preformed anti-PF4 antibody clone harboring the K31E mutation may have rapidly expanded. In these patients, onset of VITT after vaccination was at day 5 and 7, respectively. A novel aspect of this work is that we have leveraged mass spectrometry-based proteomics to define molecular mimicry via clonal matching of anti-PF4 and anti-pVII repertoires. This advances the conventional approach of immune cross-reactivity by providing unambiguous evidence of mimicry at a clonal level.

The rapid onset of the IgG-mediated VITT immune response (as early as 5 days post-vaccination) strongly indicates that the anti-PF4 immune response is secondary. In two individuals we identified a plasma sample available from a few weeks before they developed VITT. Analysis of the clonotypes of the anti-PF4 antibodies in both patients indicated that the clones with the somatic hypermutations rapidly expanded by boosting after vaccination. The hypermutation likely occurred during boosting, but in one patient also a rare preformed anti-PF4 antibody clone harboring the K31E mutation may have rapidly expanded. In these patients, onset of VITT after vaccination was at day 5 and 7, respectively. A novel aspect of this work is that we have leveraged mass spectrometry-based proteomics to define molecular mimicry via clonal matching of anti-PF4 and anti-pVII repertoires. This advances the conventional approach of immune cross-reactivity by providing unambiguous evidence of mimicry at a clonal level. Our study has limitations. The critical role of the human specific IGLV3-21*02 or *03 alleles currently precludes mechanistic studies of VITT induction in animal models. This could be overcome by generating human IGLV3-21*02 or *03 transgenic mice. We cannot exclude that additional, not yet identified cofactors (e.g., adenoviral nucleic acids, proteins associated with pVII, spike protein expressed by the adenovirus vector), or other genetic or environmental factors, are required for the pathologic VITT anti-PF4 response (Figure 4). In addition, our study is specific for a misdirected immune response against adenovirus (ChAdOx1 and Ad26) pVII; anti-PF4 disorders induced by other viruses, e.g., cytomegalovirus,31 likely have different viral proteins and antibody binding epitopes involved.

quired for the pathologic VITT anti-PF4 response (Figure 4). In addition, our study is specific for a misdirected immune response against adenovirus (ChAdOx1 and Ad26) pVII; anti-PF4 disorders induced by other viruses, e.g., cytomegalovirus,31 likely have different viral proteins and antibody binding epitopes involved. The mechanisms underlying pathologic VITT antibodies may help to unravel other poorly understood pathologies, in which a boosted immune response results in pathologic autoantibodies, e.g. post-transfusion purpura in which a boosted anti-human platelet antigen 1a alloantibody response results in formation of autoantibodies destroying the autologous platelets of the patient,32,33 and red cell autoantibodies accompanying production of alloantibodies after blood transfusion.34 The results of this study are relevant for strategies to make adenovirus vector-based vaccines safer. This vaccination platform has major advantages, e.g., affordability for healthcare systems in low- and middle-income countries and rapid vaccine provision in case of a new pandemic. Identification of molecular mimicry between pVII and PF4 as a fundamental upstream immune mechanism of VITT may now pave the way for developing safer adenoviral vector-based vaccines in which pVII is replaced by non IGLV3-21*02/*03 allele activating analogues.