Allergen-specific human IgE isolated through an allergen-agnostic pipeline—understanding immune response and allergen recognition

TL;DR

This paper introduces a new method to isolate human IgE antibodies from allergic individuals to better understand immune responses and allergen recognition.

Contribution

A novel allergen-agnostic pipeline for generating and characterizing allergen-specific human IgE antibodies from immune repertoires.

Findings

The pipeline successfully identified high-affinity IgE antibodies against four grass pollen allergens.

The method enables the study of allergen-specific humoral immunity development in allergic individuals.

The approach allows for streamlined antibody development and characterization from immune repertoires.

Abstract

Allergy, characterised by antibody responses of the IgE isotype, is a major health concern. The set of monoclonal human IgE used for studying the molecular mechanisms of allergies is limited. Single-cell sequencing offers opportunities to establish novel antibodies for researching, diagnosis, and treatment of allergies. We describe and exploit a pipeline for generating recombinant IgE directly from the immune repertoires of allergic subjects. It uses single-cell sequencing of IgM– B cells of bone marrow and peripheral blood in an allergen-agnostic manner, combined with high-throughput transcriptome sequencing to identify clonotypes populating the IgE repertoire. Immunochemical and immunoprecipitation analyses are used to deconvolute the specificity of identified antibodies. High-affinity antibodies were raised against four grass pollen allergens, antibodies that illustrated aspects of the development of allergen-specific humoral immunity. The pipeline provides a streamlined approach for the development and characterisation of native allergen-specific antibodies as they occur in allergy and during allergy desensitisation.

Identification by single cell sequencing, NGS and specificity deconvolution and definition of allergen-specific human IgE from allergic subjects’ immune repertoire.

Figures

Schematic outline of study process.Pipeline for discovery of human monoclonal allergen-specific antibodies from sample collection, cell sorting, single-cell sequencing and bulk immunoglobulin variable domain-transcriptome sequencing to bioinformatic analysis, recombinant antibody production, followed by determination of allergen specificity and immunochemical analysis.

Immunochemical analysis to define antigen specificity.Binding assays to determine the specificity of allergen-specific antibodies for allergens. A Both variants of clonotype 212579 as IgG specifically bound recombinant Phl p 5, as determined by ELISA. The other antibodies did not bind any of the tested allergens. B Variants of 212579 and 257199 both bound grass pollen extracts but no other extracts, as determined using ImmunoCAP technology. C Variants of 212579 and 257199 as IgG recognised nPhl p 4 and rPhl p 11, respectively, as determined by ImmunoCAP. D Variants of 222124 as IgG-bound grass

Full text

Introduction

Antibodies of the IgE isotype bind to the high-affinity Fc receptor (FcεRI) on mast cells and basophils through their Fc domain. Upon encountering their specific antigen (allergen) they trigger a release of proinflammatory mediators from the effector cells that mediate the allergic reaction^1^. Inadequately treated allergies impact the quality of life of affected individuals^2^ and represent a societal economic burden^3–5^. In-depth studies of allergen-specific antibodies are a focal point that specifically targets the core of the disease-causing process, but also of allergen-specific antibodies of other subclasses that may prevent allergic disease.

The development and advancement of next-generation sequencing (NGS) with specific reference to the adaptive immune receptor repertoire^6^, combinatorial library technology^7,8^, human hybridoma technology^9,10^, and more recently single-cell sequencing^11,12^, and combinations of such technologies^13,14^, now allow us, despite the relative scarcity of IgE in biological fluids and the rarity of IgE-producing cells, to describe and understand human immune repertoires associated to allergic disease as well as their evolution at substantially increasing depth. For instance, the path towards the development of the IgE-producing phenotype of cells of the B cell lineage was a matter of controversy^15^, but has through the use of NGS been suggested to proceed through an IgG1-producing intermediate^16^. Only very recently, studies involving advanced single-cell sequencing and in-depth transcriptomic analysis have demonstrated the existence and molecular features of an IgG1^+^ memory B cell subset containing allergen-specific clones with the propensity to switch to high-affinity IgE^17–20^. Altogether, studies of allergen-specific repertoires have the potential to develop an understanding of the evolution of allergen-specific immune responses in health, disease, and under therapeutic intervention.

To further expand the understanding of human IgEs and to advance our ability to describe their role in allergic disease, we have now incorporated a diversity of omics technologies to further enable the isolation of allergen-specific antibodies from cells of allergic subjects. Single-cell sequencing allows us, in contrast to many other methods used in the past, to capture native human allergen-specific antibodies in full. NGS, immunoprecipitation/mass spectrometry (IP/MS), immunochemical analysis, and bioinformatics enable the isolation and characterisation of allergen-specific antibodies encoded in cells of bone marrow (BM) and peripheral blood mononuclear cells (PBMC). This study focuses on grass pollen allergens. Furthermore, the pipeline is agnostic to the source of allergens, to a prior conception of the specificity of IgEs to be isolated, and to the availability of labelled allergens for cell sorting. As such it complements existing single-cell sequencing technologies that largely focus on initial cell identification based on antibody recognition of defined allergens. The resulting data are used to exemplify paths of development of allergen-specific antibodies in subjects with allergic rhinitis.

Results

The diversity of the IgE repertoire of a given individual is, as outlined in past studies^21^, strongly suggesting that the population recognises a small number of antigens (allergens). As a subject’s allergy profile is commonly known through immunochemical analysis and/or skin prick testing, and as many allergens are thoroughly described, we conceived that it would be possible to isolate allergen-specific IgE through the integrated use of single-cell sequencing technology and conventional high-throughput sequencing of the transcriptome encoding the immunoglobulin repertoire and, subsequently, to deconvolute their allergen specificity. The approach allows for the identification of such clones independently of whether the specific antibody is expressed on the antibody-producing cells’ surface or not, or whether there is access to a suitably labelled allergen for the identification of relevant allergen-specific clones, or not. Instead, deconvolution of the specificity depends on the availability of a diversity of medium/high-throughput downstream binding assays. We built a screening platform based on six established interrelated pillars (Fig. 1). Firstly, cell sorting and subsequent single-cell VDJ sequencing were used to identify paired H and L chain sequences of class-switched cells of the B cell lineage. Secondly, bulk H chain-encoding transcriptome sequencing was used to determine the immunoglobulin germline gene repertoire of each subject and to identify heavy chain clonotypes that make up the IgE repertoire. Thirdly, a bioinformatic pipeline was employed to define clones with potential allergen specificity. Fourth, donor immunoglobulin germline gene inference was performed to enable correct gene assignment and valid analysis of mutational patterns of identified clones. Fifth, allergen-specific antibodies were produced by recombinant antibody methodology, and sixth, identification/confirmation of the target antigen was achieved by immunochemical and IP/MS analysis.Fig. 1. Schematic outline of study process.Pipeline for discovery of human monoclonal allergen-specific antibodies from sample collection, cell sorting, single-cell sequencing and bulk immunoglobulin variable domain-transcriptome sequencing to bioinformatic analysis, recombinant antibody production, followed by determination of allergen specificity and immunochemical analysis.

Cell sorting and subsequent single-cell VDJ sequencing, and immunoglobulin repertoire NGS

BM and blood samples were collected from six subjects experiencing seasonal allergic rhinitis symptoms during the pollen season in June of 2018. Sera were tested to confirm IgE reactivity to allergens at the time of sample collection. For example, donor 32 showed substantial levels of allergen-specific IgE, while other subjects demonstrated reactivity to one or several grass pollen allergens (Supplementary Fig. 1).

CD19^+^, IgM^–^, CD3^–^, CD11c^–^, CD14^–^, and CD56^–^ cells were sorted by flow cytometry and processed through the 10X Genomics Chromium pipeline. In parallel, bulk antibody heavy chain encoding transcripts found in BM and PBMC were sequenced using NGS. The output of the bioinformatic processing pipeline is summarised in Supplementary Tables 1–3. Candidate genes (Table 1) that by themselves encoded IgE as defined by single-cell sequencing, or that were part of a clonotype that also carried IgE members identified through sequencing of the bulk transcriptome, were considered for further protein production and evaluation. The donor origin of each clone was primarily identified by barcodes of clonally related sequences identified by bulk sequencing, but also confirmed in 5/6 cases by sample-specific cellular barcodes (Supplementary Fig. 2), and in one case by the unique presence of the relevant antibody heavy chain-encoding allele in the inferred genotype of the donor. In the case of clone 212579, the clonotype-associated allergen-biotin/streptavidin hashtag counts suggested that the antibody was specific for Phl p 5 (Supplementary Fig. 3). Other identified clones did not associate with any particular barcoded allergen (Phl p 1, Phl p 2, Phl p 5, or Bet v 1) and they were, as subsequently shown, specific for allergens not included among the labelled allergens used during the single-cell sorting and sequencing process.Table 1. Annotation of germline gene origin and mutational status of IgE clones identified by single-cell sequencingClone^a^Donor^b^Clone by single-cell sequencingGrass pollen allergen specificityIGHV^c^IdentityIGHDIGHJ^c^IGLVIdentityIGLJIsotypeTissue20211831IgG2PBMCUnknownIGHV5-510196.9%IGHD6-1301IGHJ402IGLV6-570296.6%IGLJ30221257932IgG1BMGroup 5IGHV3-4803^d^99.3%IGHD3-1001 or IGHD3-1602IGHJ602IGKV3-200198.6%IGKJ50122212432IgEBMGroup 3IGHV3-90192.2%IGHD5-1802IGHJ302IGKV1-3301 or IGKV1D-3301^e^95.1%IGKJ30125719932IgEPBMCGroup 11IGHV3-3018 or IGHV3-30-501^e^87.8%IGHD5-501 or IGHD5-1801IGHJ602IGLV1-4001^f^97.7%IGLJ30211696534IgEPBMCGroup 4IGHV3-490593.4%IGHD6-1901IGHJ402^g^IGLV3-210397.9%IGLJ201, IGLJ30127689234IgEPBMCUnknownIGHV1-6901 or IGHV1-69D01^e^94.9%IGHD2-2102IGHJ602IGKV1-50398.6%IGKJ201^a^Sequences of heavy and light chain variable domains are available from GenBank: accession numbers: PQ539330- PQ539341.^b^The donor was primarily identified by presence of clonally related reads in sequence data describing of the IgE-encoding transcriptome in PBMC or BM. Donor origin was in the case of clone 276892 not confirmed by sample-specific barcode counts (Supplementary Fig. 2).^c^Confirmed by inference analysis to be present in lymphocyte donor’s genotype.^d^Donor 32 was the only subject of this study that carried allele IGHV3-4803 in its genotype.^e^Duplicated alleles with identical sequences that cannot be differentiated by transcriptome sequencing.^f^IGLV1-4002 shows the same identity to the rearranged sequence but this allele is not commonly present in human genomes (no cases found among 398 investigated subjects of diverse geographic origin) (https://vdjbase.org; accessed on August14^th^, 2025).^g^IGHJ403 showed the same identity as IGHJ402, but was through germline gene inference shown not to be expressed neither in the donor’s genotype nor in a large cohort (n = 99) of Norwegian subjects (http://vdjbase.org).

Gene cloning, protein expression, and deconvolution of antigen specificity of recombinant antibodies

To enable analysis of the reactivity pattern of the clones with an IgE component, codon-optimised genes encoding the variable domain of the heavy and light chain genes were produced as the human IgG1 subclass following transient transfection of eukaryotic cells. Protein products incorporating major sequence variants identified during bulk sequencing of the heavy chain transcriptome (Supplementary Fig. 4) were generated in the same manner, incorporating the light chain sequence identified for the clonotype through the use of single-cell sequencing. Three clones (116965_A, 212579_B, and 257199_A) were also produced as human IgE antibodies. The sequence encoding the heavy chain constant domains of these clones was exchanged for a codon-optimised sequence encoding the constant domains of IgE.

The determination of the specificity of human antibodies that make up the IgE population is simplified by the limited number of likely targets. The allergenic profile of the lymphocyte donors was known in part, a fact that allows us to focus our antigen search domain to a smaller set of allergens. However, the path to the identification of the specificity of each clonotype differed. Immunochemical analysis of binding of recombinant IgG to a small set of recombinant allergens initially demonstrated that both variants of clone 212579 specifically recognised Phl p 5.0101 (Fig. 2A), in agreement with the finding that the cell from which this antibody originated also specifically enriched barcodes associated with biotinylated rPhl p 5 (Supplementary Fig. 3). Further assessment of binding to allergen mixtures of grass, tree, and weed pollen as well as mite extracts, as defined in ImmunoCAP assays, showed that variants of clones 212579 and 257199 strongly bound a grass pollen extract mixture (Fig. 2B). Immunoprecipitation, trypsin digestion, and subsequent mass spectrometry analysis suggested that clonotypes 212579 and 257199 immunoprecipitated allergens of group 5 and group 11, respectively (Fig. 3). Immunoprecipitation thus supported the specificity assignment of 212579 as established by immunoassay and barcode analysis. Antibody 212579 immunoprecipitated peptides of both Phl p 5.01 and Phl p 5.02 isoallergens (Supplementary Fig. 5). Further analysis of the specificity of variants of 212579 using ImmunoCAP confirmed their ability to also bind Phl p 5.0201 (Phl p 5b) (Fig. 2C). ImmunoCAP analysis also confirmed the specificity of 257199 to Phl p 11.Fig. 2. Immunochemical analysis to define antigen specificity.Binding assays to determine the specificity of allergen-specific antibodies for allergens. A Both variants of clonotype 212579 as IgG specifically bound recombinant Phl p 5, as determined by ELISA. The other antibodies did not bind any of the tested allergens. B Variants of 212579 and 257199 both bound grass pollen extracts but no other extracts, as determined using ImmunoCAP technology. C Variants of 212579 and 257199 as IgG recognised nPhl p 4 and rPhl p 11, respectively, as determined by ImmunoCAP. D Variants of 222124 as IgG-bound grass pollen extracts of grasses belonging to the BOP but not the PACMAD clade, while clone 116965 recognised the extract of Holcus lanatus, as determined by ImmunoCAP; ND: not determined. E Clone 116965, 212579, and 257199 as IgE bound Phl p 4, Phl p 5, and Phl p 11, respectively, as determined by ELISA. F Clone 116965, at a concentration of 200 kU IgE/L, bound nPhl p 4 but not a range of other allergens, including cross-reactive carbohydrate determinants. G Clone 222124_A bound specifically to recombinant Phl p 3 by ELISA (titrated in steps of 5-fold dilution), without any signs of cross-reactivity to the unrelated allergen Phl p 5 while clone 212579 specifically bound Phl p 5. H Phl p 4-specific antibody 212579_A binds poorly to rPhl p 4 produced in E. coli. However, it binds to well-folded (Supplementary Fig. 6) rPhl p 4.0201 N69Q N388Q produced in P. pastoris, a protein that lacks sites for N-linked glycosylation. The deglycosylated allergen does not non-specifically interact with the Phl p 11-specific antibody 257199_A. Binding was assessed by ELISA and antibodies were titrated in steps of 5-fold dilution. I Recombinant Phl p 4-specific IgE antibody 116965 shows higher signals when analysed on ImmunoCAP carrying nPhl p 4 compared to natural timothy allergen extracts. Box-plots (A, E, G, H) span the range of measured signals and intermediate values are shown as circles within the boxes.Fig. 3. Immunoprecipitation/mass spectrometry to define antigen specificity.IP/MS was used to demonstrate the specificity of allergen-specific antibodies exposed to timothy extract. A Hierarchical clustering of log2-normalised MS protein level intensity data of tryptic digests derived from proteins immunoprecipitated with biologically independent variants of the herein isolated antibodies identifies the specificity of each clone set. The heatmap was based on matches to Phl p proteins in the allergen.org database. B Principal Component Analysis (PCA) of protein intensities identified by IP/MS demonstrates the separation of reactivity of the members of allergen-specific antibodies 116965 (red), 212579 (olive), 222124 (green), 257199 (blue), and negative control antibodies (magenta), as defined by IP/MS.

Clone 116965 weakly bound one (Holcus lanatus) of several ImmunoCAP-carrying grass pollen mixture extracts (Fig. 2D). As few allergens of Holcus lanatus are molecularly defined, mass spectrometry data obtained by immunoprecipitation of Holcus lanatus pollen extracts were annotated using a peptide database largely composed of proteins of other grasses, as the sequences of few allergens of Holcus lanatus are known. It identified peptides of group 4 allergens as being substantially enriched. Subsequent IP/MS of extracts of pollen of Phleum pratense confirmed the specificity to be Phl p 4 (Fig. 3). The binding of this antibody to nPhl p 4 was subsequently confirmed by ELISA (Fig. 2E). Natural Phl p 4 is a glycoprotein and antibodies specific for this allergen may solely or in part be dependent on recognition of the allergen’s carbohydrates. Indeed, some other proteins co-immunoprecipitated to a minor extent with group 4 allergens. Some degree of co-precipitation is common in immunoprecipitation assays, but co-precipitation may also suggest that cross-reactive carbohydrate determinants^22^ may be targeted by antibody 116965. Further characterisation of the binding of 116965_A in the form of an IgE showed that while it at 200 kU/L strongly bound nPhl p 4, it neither bound to plant carbohydrates displayed on MUXF3 CCD (bromelain) nor to horse-radish peroxidase or to the mammalian carbohydrate nGal-alpha-1,3-Gal (alpha-Gal), as determined by ImmunoCAP assays (Fig. 2G). Although antibody 116965 only weakly recognised Phl p 4.0201 produced intracellularly in Escherichia coli as determined by ELISA, it bound with high affinity to a variant of the allergen, produced in P. pastoris, in which the sites for N-linked glycosylation had been removed by substitution of asparagine for glutamine (Fig. 2H, Supplementary Fig. 6). Collectively these assays are highly predictive of recognition by antibody 116965 of the protein component and not of carbohydrate post-translational modifications of Phl p 4.

While variants of clone 222124 failed to recognise any investigated recombinant timothy allergens as assessed by ImmunoCAP (Fig. 2C), they remained positive for pollen extracts of several grasses (Fig. 2D). Further IP/MS analysis using an extract of Phleum pratense pollen demonstrated that the 222124 antibody immunoprecipitated the group 3 allergen (Fig. 3), a specificity subsequently confirmed by the antibody’s recognition of recombinant Phl p 3 (Fig. 2F). The antibody was, however, negative for binding to the related allergen Phl p 2 (Fig. 2A, C).

In contrast to the successful identification of the specificity of the four clonotypes above, no target could be identified for antibodies 202118 and 276892 by immunoassay (Figs. 2, 3), or by grass pollen extract IP/MS. Nevertheless, a combination of methodologies allowed us to identify the specificity of two-thirds of clonotypes encoding IgE, as identified by a combination of single-cell sequencing and bulk transcriptome sequencing.

The affinity of allergen-antibody interaction

The outcome of antigen-antibody interactions is dictated by numerous factors such as the complexity of the response and the affinity of the interaction^23^. Surface plasmon resonance analysis technology was used to define the affinity of antibodies for pure natural (Phl p 4) or recombinant (Phl p 3, Phl p 5, and Phl p 11) allergens. No binding of antibodies to irrelevant allergens was observed and a control antibody did not bind immobilised allergens (Supplementary Fig. 7). In all cases, we could demonstrate high affinity in the sub-nM region and a slow dissociation rate constant with a half-life of the interaction estimated to be in the range of 18 min to 4 h (Table 2, Supplementary Fig. 8). Investigated variants for which the heavy chain had been identified through its similarity to its single-cell counterpart by sequencing of bulk transcripts similarly retained high affinity to the allergen in question (Table 2), even when combined with the light chain of the variant isolated by single-cell sequencing, suggesting that clonal evolution had not progressed in ways that depended on extensive co-evolution of the heavy and light chains.Table 2. Reaction rate constants and dissociation constants of antibody binding to recombinant allergens as determined using surface plasmon resonance technologyCloneAllergen specificityk_A_ (M^−1^ s^−1^)k_D_ (s^−1^)K_D_ (M)222124_APhl p 3.01015.5 × 10^6^0.85 × 10^−4^0.15 × 10^−10^222124_BPhl p 3.01013.4 × 10^6^1.6 × 10^−4^0.46 × 10^−10^222124_CPhl p 3.01014.5 × 10^6^0.45 × 10^−4^0.10 × 10^−10^116965_APhl p 4.01010.9 × 10^6^6.5 × 10^−4^7.0 × 10^−10^212579_APhl p 5.01010.9 × 10^6^2.3 × 10^−4^2.5 × 10^−10^212579_BPhl p 5.01010.9 × 10^6^0.99 × 10^−4^1.1 × 10^−10^257199_APhl p 11.01011.9 × 10^6^3.9 × 10^−4^2.1 × 10^−10^257199_BPhl p 11.01011.5 × 10^6^1.5 × 10^−4^1.0 × 10^−10^257199_CPhl p 11.01011.8 × 10^6^4.4 × 10^−4^2.5 × 10^−10^Sensorgrams are shown in Supplementary Fig. 8.

Mutational analysis of allergen-specific clonotypes

Annotation of allergen-specific variable domain-encoding genes was performed to enable initial studies of such antibodies and their development in vivo. Genes were annotated by IgBLAST using a validated reference data set^24^ of germline genes. The heavy chain germline gene repertoire of each subject was also inferred from IgM-encoding transcriptomes to ensure the presence of each proposed germline gene in the genotypes of the subject from which antibodies of known specificity were derived (Supplementary Fig. 9A, B). The appropriateness of the inference of alleles of IGHV genes of donor 32 was further defined by haplotyping^25^ based on the heterozygosity of germline gene IGHJ4 in its genotype. Such analysis demonstrated that sequence reads supporting the inference of two different alleles of a gene, as expected, were distributed differentially to the two haplotypes of the genome (Supplementary Fig. 9C) in support of their validity. The relevance of each allele in relation to previously published germline repertoires of human subjects in a nearby geographic region was also confirmed (Supplementary Table 4). In this way, we were able to define with high likelihood the germline gene origin of each antibody (Table 1), a methodology that strengthens the analysis of the evolutionary history of each clone. For several of the antibodies, multiple clonal variants were observed by NGS and the clonal development could be defined (Fig. 4), exemplifying the co-existence of identical clones in different tissues (BM/peripheral blood) and multiple divergent and co-existing subclone branches. As an extension, subsequent experimental studies have demonstrated the potential for the naïve repertoire to seed high affinity allergen-specific unmutated sequences into the responding pool of clonotypes and to define the role of unique allelic diversity for the generation of the high affinity phenotype of clone 212579 specific for the group 5 grass pollen allergens^26^.Fig. 4. Evolution of allergen-specific antibodies.Phylogenetic trees, generated based on bases found in the gene from the beginning of CDR1 to the end of the sequence encoded by the IGHV-gene, describing the evolution of clones belonging to the Phl p 11- (antibody 257199) (A), Phl p 5- (antibody 212579) (B), Phl p 3- (clonotype antibody 222124) (C) specific clonotypes, and clonotype 276892 of undefined specificity (D). The germline allele is shown on top. Sequences identified by single-cell sequencing and NGS are shown in green and blue, respectively. Sequences found among cells in BM and PBMC are shown as diamonds and circles, respectively. White triangular symbols represent unidentified common intermediate progenitors. Numbers represent the distance between adjacent clones as computed by the IgPhyML. Unless specified otherwise clones are of IgE isotype.

Group 3 pollen allergen-specific antibody recognises its target in extracts of grasses of the BOP but not the PACMAD clades of grasses

Grasses (Poaceae) are dominated by members of two separate clades, PACMAD and BOP^27^. Antibody 222124, specific for group 3 allergens recognised, as outlined above, antigens of extracts of a range of, but not all, grasses (Fig. 2D). Extracts recognised by this antibody were all of the BOP clade of grasses while pollen extracts of grasses of the PACMAD clade were not recognised. Thus, either group 3 allergens are absent from the grasses of the PACMAD clade, or immunochemically divergent from such allergens identified in several grasses of the BOP clade. A BLAST search of the genome assembly of Phragmites australis demonstrated the absence of any protein-coding sequence of this species of the PACMAD clade with a sequence identity to Phl p 3.0101 above 34% (Supplementary Fig. 10). This confirms the likely absence of proteins that display cross-reactive properties to Phl p 3-specific antibodies in the proteome of Phragmites australis.

Allergen-specific antibodies may reach a high-affinity state with minimal somatic hypermutation

Antibody 212579 recognises different isoallergens of the group 5 allergen of Phleum pratense. The clonotype was found as an IgE by NGS of bulk transcriptome sequencing but was, as outlined above (Fig. 4B), also identified as an IgG1 by single-cell sequencing. While the IgE was substantially mutated, the IgG1 carried very few gene mutations and consequently very few substitutions in its heavy and light chain variable domains. It thus demonstrates that an allergen-specific, high-affinity (sub-nM) character can be readily reached from a non-mutated naïve B cell with limited involvement of somatic hypermutation. This antibody also illustrates the co-existence of IgG1 and IgE variants of the same clonotype in allergic subjects even in the absence of allergen-specific immunotherapy.

Group 11 pollen allergen—a candidate for stereotypic antibody responses

Clone 257199 specific for the group 11 grass pollen allergen was, as outlined above, encoded by a precursor that had had been created from germline gene IGHV3-30. Past studies have identified two clonally unrelated antibodies specific for Phl p 11 using combinatorial library technology^21^. All three clones derive from the same allele of this germline gene, suggesting the existence of a recurrent stereotypic antibody response to group 11 grass pollen allergens.

Group 4 pollen allergen-specific antibody—epitope representation in standard grass pollen extract immunochemical assays

Clone 116965, specific for Phl p 4, binds natural allergen with high affinity. It did bind to natural Phl p 4 represented on allergen component ImmunoCAP g208 as an IgE. It also, as an IgG1, bound group 4 allergens in extracts of pollen of Phleum pratense and Holcus lanatus. Nevertheless, as IgG1 it only bound weakly to one (Holcus lanatus) of nine tested grass pollen extract-carrying ImmunoCAPs of the initial screen. Further testing of ImmunoCAP reactivity at higher IgE concentrations demonstrated comparably low levels of binding to extract-based tests in particular at high concentrations of IgE (Fig. 2I). This binding profile suggests that group 4 allergens might not be highly represented on pollen extract-based clinical assays. Nevertheless, the binding activity of IgE at high concentration was found to extracts of pollen of four grasses of the grass BOP clade but only very limited binding was seen to extracts of pollen of two grasses of the PACMAD clade of grasses (Supplementary Fig. 11). Although group 4 pollen allergens have not been recorded in the IUIS allergen nomenclature database (https://www.allergen.org) for the two tested grasses of the PACMAD clade, a BLAST search identified sequences with identity to Phl p 4.0101 up to 70% (Phragmites australis), 69% (Cynodon dactylon), and 64% (Sorghum bicolour) in putative proteins of grasses of the PACMAD clade. Group 4 allergens are thus likely present in such grasses, although with a sequence that might not allow cross-reactivity to antibody 116965.

Human allergen-specific antibodies isolated by single-cell sequencing of human B-cells—potential for developability

Allergen-specific antibodies may find utility in therapeutic and diagnostic applications. However, as with all proteins, antibodies harbour a range of liabilities that can affect their stability and homogeneity, factors that may affect their usefulness in certain applications. The Therapeutic Antibody Profiler^28^ offers an opportunity to in silico assess human allergen-specific antibodies derived from the present pipeline for developability potential based on five biophysical parameters. Only one (clone 257199) of the four antibodies had a score in the test’s amber flag region (Supplementary Table 5), suggesting that antibodies developed from human allergen-specific antibody repertoires through single-cell sequencing have the potential for further development as therapeutic drugs. The herein-developed proteins were also assessed for other liability factors. For instance, some germline genes encode sequence motifs prone to N-linked glycosylation and hypermutation may introduce yet other such sequence motifs. Low-risk deamidation motifs were present in all heavy chains and lambda light chains of antibodies 257199 and 116965. The heavy chain of antibody 222124 carried a high-risk deamidation site within complementarity determining region 3 (CDR3). The lambda light chain of 257199 had mutated in a way that introduced an N-linked glycosylation site in residue 57 in CDR2 through an N- > S substitution in residue 66. Although these native human antibodies in general showed overall good developability characters in silico, the analytical approach identifies sites of liability that might be addressed if such antibodies are to be further developed for particularly demanding applications.

Discussion

Allergen-specific human antibodies of the IgE isotype are central to the establishment of allergic conditions and diseases. However, when these antibodies are present as isotypes other than IgE, they can hold the capacity to resolve allergic conditions by blocking allergens from binding to IgE on effector cells^1^. Their key roles in allergy make allergen-specific antibodies of all isotypes important objects of study but the rarity of IgE-producing cells represents a major obstacle to their isolation. Specific human antibodies may be developed by chimerization or humanisation of murine antibodies^29^, or directly generated from immunised mice carrying human immunoglobulin loci^30^. Even more relevant antibodies representative of human allergen-specific immune responses may be generated by selection from combinatorial antibody libraries derived from the IgE-encoding repertoire by phage display^7,8^ or by human hybridoma technology^9^. Single-cell VDJ sequencing, either alone, in combination with additional sequencing technologies as described herein, or in combination with transcriptome sequencing to capture the differentiation stage of the cell producing each antibody^31–34^, offers opportunities to capture such antibodies directly from the human repertoire. As the technologies and their throughput have developed, they hold the capacity to be important contributors to the growing antibody resource that defines human allergen-specific humoral immunity. Furthermore, such antibody specificities hold the potential to be used for passive immunotherapy against disease, opportunities that are already explored in clinical studies^35,36^. The present study identifies four fully human, high affinity antibodies specific for grass pollen allergens in allergic subjects. Access to such antibodies was used to explore allergen-specific recognition of allergens, to exemplify the link of specificities harboured in BM and peripheral blood, to connect antigen-specific antibodies of diverse isotypes, to identify potentially novel stereotypic antibody responses to allergens, and through use of specific antibodies study allergen presence in diverse plant species and materials.

In the present study, we developed a pipeline that enables the capture of human IgE as found in allergic subjects in an allergen-agnostic manner, i.e., without prior assumptions of the specificity of the antibodies to be discovered. The pipeline exploits the ability of single-cell sequencing to capture the sequence of paired heavy and light chains of individual cells of the B cell lineage in combination with the high information content of high-throughput sequencing. The system is compatible with, but does not require, the incorporation of barcoded allergens for the direct determination of allergen specificity based on the accumulation of such barcodes in droplets associated with a particular heavy/light chain sequence combination, as evidenced by the discovery of a group 5 allergen-specific clone. Importantly, the Phl p 5-specific antibody identified in the present study was identified as an IgG clonotype, but its association with an IgE response was defined by the clonotype’s presence in the IgE-encoding transcriptome as defined by bulk sequencing. Single-cell sequencing allowed us to identify the paired heavy and light chain sequences that enabled us to recreate the intact allergen-specific binder, albeit one that differed by some somatic hypermutation events from that of the identified IgE-encoding transcripts.

Three specific binders, the allergens of which were not incorporated in the barcoding strategy, were defined by the pipeline as deconvolution of the specificity, which could be achieved through the use of a panel of immunochemical and immunoprecipitation-based methods, methods that can be selected by a priori knowledge of the study subjects’ clinically defined allergy. All allergen-specific IgEs identified by this target-agnostic pipeline demonstrated a sub-nanomolar affinity for their respective antigen, suggesting that native human IgEs commonly show high affinity for their target antigen, in agreement with past studies^37^. Importantly, we were able to isolate antibodies against an allergen for which a recombinant protein was not readily commercially available to support immunochemical analysis, cell sorting, and clonal identification based on barcoded antigens during bioinformatic analysis. Such a binder, specific for group 3 grass pollen allergens, would have escaped detection with a strategy dependent on sorting or identification of allergen-specific antibodies using labelled allergens. The use of multiple specificity-defining methodologies was important as there is limited availability of recombinant reagents and immunochemical assays in clinical use to define specificity to some allergens. In that context, immunoprecipitation and mass spectrometry served as a critical tool to determine the specificity of the antibody. Assays defining carbohydrate-specific IgE were used to confirm the possible involvement of recognition of the carbohydrate moiety of Phl p 4. The antibody (clone 116965) did, however, not recognise carbohydrates in assays designed to measure antibodies specific for cross-reactive carbohydrate determinants of proteins^38^ or a non-glycosylated mutant of Phl p 4. This strongly suggests that its specificity is not directed towards post-translational modifications of allergens, but to the protein itself. Altogether, a combination of high-throughput/content technologies allowed us to isolate and characterise allergen-specific antibodies as they occur in allergic human subjects.

The pipeline was able to identify binders to four different grass pollen allergens. To our knowledge, the Phl p 3-specific IgE defined in this study is the first reported human monoclonal antibody specific for this allergen. Despite the sequence similarity of group 2 and 3 allergens, it does not cross-react to Phl p 2, an antigen that in part carries epitopes shared with Phl p 3^39^. It thus represents a unique reagent to investigate the nature and outcome of antibody-allergen interaction in the absence of recognition of the cross-reacting allergen Phl p 2 even in complex allergen mixtures. It identifies cross-reacting allergens in pollen extracts of several grasses of the BOP clade of grasses (Anthoxanthum odoratum, Dactylis glomerata, Festuca pratensis, Lolium perenne, Bromus inermis, and Holcus lanatus). In contrast, extracts of pollen of grasses of the PACMAD clade, i.e., Cynodon dactylon, Phragmites communis, and Sorghum halepense, did not contain evidence of antigens that cross-reacted with this antibody. The present study of the antibody to group 3 allergen demonstrates the diversity of presence and/or immunoreactivity of allergens of phylogenetically distinct species^27^, and highlights the need for further studies of the presence and cross-reactivity of such allergens in diverse grasses that in many cases may be incompletely characterised concerning their proteome and allergome.

In contrast to Phl p 3, multiple human monoclonal antibodies have in the past been generated to Phl p 4^21^, Phl p 5^8,13,14,21,40,41^, and Phl p 11^21^. While Phl p 4 and Phl p 5 are targeted by antibodies of diverse immunoglobulin germline gene origin, all three defined, clonally unrelated Phl p 11-specific antibodies have an origin in rearrangements of germline gene IGHV3-30, in this case allele IGHV3-30*18. We postulate that this IGHV gene encodes a product with important sequence features that allow for binding to Phl p 11. Future studies of specific antibodies targeting Phl p 11 will define whether this specificity represents yet another example of a response featuring stereotypic antibodies as a dominant aspect in allergen recognition, in addition to the previously defined stereotypic responses to Phl p 2^13,42,43^ and peanut allergen Ara h 2^11,33,44^. Moreover, to what extent a dominance of such stereotypic responses affects the clinical outcome of exposure to such allergens remains to be determined. We postulate that a dominance of stereotypic immune responses opens up for the development of vaccine strategies that promote the development of protective immunity targeting epitopes recognised by such common rearrangements.

The use of the present pipeline offers insight into paths through which antibodies evolve during the allergic immune response as we, in addition to sequences derived by single-cell sequencing, have access to partial sequences of additional variants both in peripheral blood and BM. It is known from past studies that IgE commonly develops in paths shared by other isotypes. It has been suggested that such shared linages in particular develop from IgG1-producing cells^16^. Indeed, the Phl p 5-specific antibody 212579 was isolated as an IgG1 antibody from a cell in BM that appears in an evolutionary path that allows for downstream differentiation into IgE production. This clone, derived from rearrangements involving germline genes IGHV3-48*03 and IGKV3-20, demonstrated high (sub-nM) affinity despite its low level of mutation. Its clonally related IgE-encoding sequence was more mutated, but its recreated recombinant product retained a similar level of affinity. The precise phenotypes of the respective cells producing these antibodies were not determined through the herein-used analysis pipeline, but we hypothesise that these cells may represent examples of recently described^17–19^ phenotypic variants of allergen-specific IgG1 memory B cells and further differentiated IgE-producing cells. Multiple sub-lineages of IgE clonotypes 257199 and 222124 were also seen in other tissue locations as that of the original clone isolated by single-cell sequencing, illustrating the ability of single clonotypes to populate different tissue locations in agreement with past studies^13,14^. Combinations of these high-end technologies will develop our understanding of the evolution of humoral immunity to allergens over time, as it occurs during the development and maintenance of the disease state and during long-term disease-modifying treatment like allergen-specific immunotherapy. Studies employing NGS enriched by knowledge of the identity of allergen-specific clonotypes as defined by combinatorial library technology have already been able to demonstrate longevity and evolution of allergen-specific clonotypes during several years of allergen-specific immunotherapy^13,14^, in agreement with studies identifying long-lived IgE-producing plasma cells in bone marrow^45^. As the throughput of single-cell technologies expands, we will be able to capture the true nature of events accompanying successful or failed allergen-specific immunotherapy.

In summary, in the present study, we isolated and studied four human antibodies against grass pollen allergens that belong to clonotypes involving variants of the IgE isotype. These antibodies, isolated through a pipeline independent of the antibodies’ affinity for the antigen during the cell sorting process, showed high (sub-nM) affinity for their target allergen, but highly variable levels of somatic mutation relative to their unmutated precursors. Overall, the approach/the present pipeline provides insight into diverse pathways of development and evolution of allergen-specific antibodies and their allergen-recognition profiles.

Methods

Pipeline strategy

A global experimental strategy, summarised in Fig. 1, was devised to isolate of human IgE from allergic subjects using a combination of single-cell VDJ sequencing and bulk heavy-chain sequencing together with downstream molecular analysis that allowed for the assessment of identified antibodies. Single-cell sequencing allowed for the identification of paired immunoglobulin heavy and light chain variable domains that were encoded by the IgE repertoire. Bulk transcriptome sequencing enabled the determination of the presence of specific clones in the transcriptome and their encoded isotypes. In addition, due to its greater read depth, transcriptome sequencing identified additional heavy chain sequence variants of the allergen-specific clones isolated by single-cell sequencing. Bioinformatic tools enabled sequence analysis and clonal evolution analysis. Recombinant antibody technology provided antibody products for downstream analysis using various immunochemical techniques, including those to identify the target antigen of isolated antibodies. Finally, immunoprecipitation in combination with mass spectrometry allowed for the identification of the molecular specificity of antibodies, even some for which standard immunoassays using recombinant allergens were not available.

Clinical samples

Six subjects experiencing symptoms of allergic rhinitis during the season and with a positive skin prick test (Soluprick, ALK, Hørsholm, Denmark) to these pollen allergens (Table 3) were recruited. Blood and BM samples were collected (by clinical routine) from all subjects during mid-June 2018 when levels of grass pollen were high (Supplementary Fig. 12). The study was performed according to the Declaration of Helsinki, approved by the regional ethical board at Lund University (Dnr. 2018/327), and written informed consent was obtained from all study participants. All ethical regulations relevant to human research participants were followed.Table 3. Description of subjects investigated in this studyDonorSexAgeSkin prick test positive31Male38Birch; timothy; mugwort32Male25Birch; timothy; mugwort; cat; dog33Male22Birch; timothy34Male24Timothy; dust mite35Male28Timothy36Male28Birch; timothy

Allergens

Recombinant Phl p 1.0102, Phl p 2.0101, Phl p 5.0101, Art v 1.0101, and Bet v 1.0101 were obtained from Biomay (Vienna, Austria). Recombinant Phl p 11.0101 and native Phl p 4 (nPhl p 4) were kindly provided by Dr. Jonas Lidholm (Thermo Fisher Scientific, Uppsala, Sweden). Phl p 3.0101 was produced as described^39^. Soluble, recombinant Phl p 4.0201 (residues 8-508) produced in E. coli was obtained from Antibodysystems (Paris, France). Phl p 4.0201 N69Q N338Q, lacking sites for N-linked glycosylation, was produced in Pichia pastoris and purified as described^46^. The folded state of the latter non-glycosylated allergen was determined by circular dichroism on a ChirascanPlus spectropolarimeter (Applied Photophysics Ltd, Leatherhead, UK). Sample was prepared at a concentration of 0.15 mg/ml in a 20 mM phosphate buffer (pH 8.0) containing 150 mM NaF. Measurements were performed at 20 °C using a 1 mm path length quartz cuvette. CD spectra were recorded over the wavelength range of 260–195 nm with a 1 nm data pitch. Spectrum represents the average of three scans and was baseline-corrected by subtracting the buffer spectrum recorded under identical conditions. The data were smoothed using Savitzky-Golay smoothing algorithm and the results were expressed as a mean residue ellipticity at the given wavelength. Pollen extracts of Phleum pratense and Holcus lanatus were obtained from Stallergenes Greer (Baar, Switzerland). Biotinylated recombinant Phl p 1, Phl p 2, Phl p 5b, and Bet v 1 were obtained from Astra Biotech (Berlin, Germany).

Cell isolation

Mononuclear cells were isolated from peripheral blood and BM by density centrifugation on Ficoll-Paque (Cytiva, Uppsala, Sweden). Cells for flow cytometry and single-cell sequencing were cryopreserved in RPMI 1640 containing 10% foetal bovine serum and 10% DMSO and stored in liquid nitrogen before use.

RNA purification and transcriptome sequencing

RNA was purified from freshly isolated peripheral blood or BM mononuclear cells using RNAeasy Mini/Midi prep kit (QIAGEN, Hilden, Germany). RNA extraction was performed according to the manufacturer’s instruction, but with centrifugating for 30 s instead of the suggested 15 s or 7 min instead of the suggested 5 min. Extracted RNA was stored at −80 °C. Immunoglobulin heavy chain genes were PCR amplified using Biomed-2 upstream primers^47^ and the constant region (IgA, IgE, IgG, and IgM-specific) downstream primers using a semi-nested amplification strategy^43^. Adaptors and barcodes suitable for sequencing using the MiSeq technology were introduced during this process. Samples were pooled and MiSeq (Illumina, San Diego, CA, USA) 2× 300-cycle paired sequencing was performed at the Center for Translational Genomics (CTG) at Lund University (Lund, Sweden).

Cell sorting and single-cell sequencing

Cells were thawed, blocked with ChromPure mouse IgG whole molecule (Jackson ImmunoResearch, West Grove, PA, USA), and labelled with barcoded cell hashing antibodies (Supplementary Table 6). Subsequently, the cells were stained with fluorochrome-labelled antibodies (Supplementary Table 7) and viability stain 620 (BD Biosciences, Franklin Lakes, NJ, USA), and biotinylated allergens preincubated with barcoded Streptavidin PE (Supplementary Table 6). Cells were washed and CD19^+^, IgM^–^, CD3^–^, CD11c^–^, CD14^–^, and CD56^–^ cells were sorted into a separate tube, using FACSAria IIu (BD Biosciences). An example of the gating strategy is shown in Supplementary Fig. 13. Sorted cells were pooled into four samples (peripheral blood samples from subjects 31, 34, and 36 in one pool; peripheral blood samples from subjects 32, 33, and 35 in one pool; BM samples from subjects 31, 33, and 36 in one pool; BM sample from subject 32, 34, and 35 in one pool). Pooled samples were used for single-cell gene expression and immune profiling library preparation, employing the 10X Genomics platform (Pleasanton, CA, USA), and subsequently sequenced at the Center for Translational Genomics (CTG) at Lund University (Lund, Sweden).

Pre-processing of single-cell sequencing data

Data defining antibody-encoding sequences of cells captured by single-cell sequencing were identified using the Cell Ranger (v. 3.1.0) software (10X Genomics). Paired-end fastq files were filtered and processed with the Cell Ranger V(D)J tool’s standard pipeline (described at https://www.10xgenomics.com/support/cn/software/cell-ranger/8.0/algorithms-overview/cr-5p-vdj-algorithm) to identify antibody sequences, using the human reference dataset GrCh38 v 3.1.0 as the reference database. In short, likely erroneous reads were filtered out, adaptors and primer were trimmed and full-length antibody encoding transcripts were generated. Additionally, The hashing and allergen-streptavidin barcoding data (fastq files) were processed using the Cell Ranger count tool in order to generate matrices of the number of barcodes associated with each cell. No other processing, except for converting the matrices to CSV files using Cell Ranger mat2csv, was performed. However, we did not rely primarily on this data for determining neither allergen specificity nor sample origin (the latter primarily being determined using clonal relation to sequences in the NGS transcriptome data). Identified heavy and light chain antibody sequence data were filtered using an in-house script, keeping only productive, full-length sequences with the settings high_confidence = True and is_cell = True. If more than one heavy chain or more than one light chain sequence were assigned to the same cell, these sequences were discarded if they expressed different V(D)J sequences or identical V(D)J sequences of different isotypes. Cells that lacked either a heavy or a light chain were not discarded at this stage, but were however not considered candidates for synthesis and downstream analysis.

Pre-processing of antibody transcriptome sequencing data

Antibody transcriptome sequencing fastq files were processed using the Presto (0.5.11) toolkit^48^. FilterSeq.py was used to remove reads with a mean Phred quality score (q) below 20 and barcodes and primers were trimmed using MaskPrimers.py. Forward and reverse reads were combined by employing PairSeq.py followed by AssemblePairs.py. Again, FilterSeq.py was used to remove paired sequences with q < 25 and sequences were separated according to their isotype (defined through the primer previously removed) using SplitSeq.py. Any sequences that lacked a specific sequence associated with their assigned isotype (CCGACCAGCCCCAAGG for IgA, CCACCAAGGGCCCATC for IgG, ACACAGAGCCCATCCG for IgE, and a GGGA end for IgM) were discarded.

Identification of single-cell-identified clones expressing IgE members

To identify clones in the single-cell data that expressed IgE members, heavy chain sequences of the single-cell sequencing data and transcriptome sequencing data were combined and searched for common clones. First, fasta files, pre-processed as described above, were separately analysed using IMGT/HighV-QUEST^49^ with IMGT/V-QUEST version 3.5.19, reference directory release 202031-2, and receptor type IGH. For single-cell sequencing data, unfiltered fasta files were used. The output files were modified using the Change-O 0.4.5 toolkit^50^, first by parsing files for each subject separately using the MakeDb tool. Each single-cell sequencing fasta file contained sequences derived from multiple donors, as samples had been pooled before library generation. Thus, each single-cell sequences fasta file was used multiple times – one for each subject from which it contained data. An in-house script was thereafter applied to discard non-functional sequences, trim framework (FRW) 1 from the sequences, and collapse all sequences identical in the remaining variable region (CDR 1 to FRW4). Clones were identified (as sequences sharing the same IGHV origin, the same CDR3 length, and similar CDR3 sequences) for each donor, by applying the Change-O tool DefineClones^50^. Initial grouping was done taking all listed germline genes into account (on a gene level, not on an allele level) and single linkage was applied for hierarchical clustering. Nucleotide hamming distance normalised by length was used, assigning any sequence with no more than a 10% distance to the same clone. Among the defined clones, those that expressed at least one IgE member, either in the transcriptome or the single-cell sequencing data, and at least one member in the single-cell sequencing data were identified. Finally, these clones were searched for candidates for synthesis and downstream analysis, likely to be associated with allergy. Such candidates were required to carry information on both the heavy and the light chain in the single-cell sequencing data and either be expressed as IgE isotype in the single-cell sequencing data or have a high proportion of IgE reads in the transcriptome sequencing data. Variant sequences of the heavy chain of all such clones were identified in the transcriptome sequencing data of the donors’ PBMC and BM samples. PCR recombination may generate products derived from two different genes^51,52^. Sequence variants (typically present as very few reads, less than 3% of common sequences of the same clonotype as defined by the CDR3 sequence) that at their 5’-end carried a sequence suggesting that it had been generated by such a process were removed from further consideration following manual analysis.

Sequences identified by single-cell sequencing were initially annotated within the Cell Ranger platform (reference database: GRCh38 v 3.1.0) but subsequently re-annotated using IgBLAST employing the curated AIRR-C Reference Set^24^ that incorporates well-documented alleles that are missing from the reference database GRCh38 (e.g., IGHV3-9), eliminates alleles that may have been defined in error^53^, and corrects truncations that may hamper proper allele annotation^24^. To support the validity of the annotation, the expressed IGHV, IGHD, and IGHJ germline gene repertoires of donors 32 and 34 that contributed allergen-specific antibodies were determined by inference using the bulk IgM-transcriptome of BM and PBMC and the Bionamic Antibody R&D Data Management platform^54^. As such transcriptome sequencing was performed using PCR products amplified with the BIOMED-2 primer set^47^, a version of the AIRR-C Reference set^24^ incorporating bases from the beginning of CDR1 of alleles of IGHV was used to guide the inference process. As a result of the sequencing and analysis process, alleles only carrying differences in the framework 1 region could not be differentiated. The validity of the inference of the IGHV genotype of donor 32 was further assessed by haplotyping^25^ performed based on the genotype’s heterozygosity of the IGHJ4 gene. The outcome was also compared with germline genes known to be expressed in a large Norwegian patient cohort, derived in Scandinavia, just as the samples of the present study. (Supplementary Table 4).

Trees of clonally related sequences derived by single-cell sequencing and NGS were generated by the BuildTrees tool within IgPhyML (IgPhyML version 1.1.5 09022, BuildTrees.py version 1.3.0 2022.12.11)^55^ of the Immcantation suite (https://immcantation.readthedocs.io/) and were based on nucleotide sequences ranging from the beginning of CDR1 to the end of the sequence encoded by the IGHV gene, excluding reads with evidence of PCR artefacts^51^.

Gene synthesis and cloning

Sequences encoding the variable domain of L and H chains of selected antibodies, identified as described above, were modified for subsequent synthesis (Supplementary Fig. 4; GenBank accession numbers: PQ539343-PQ539361), and cloned into an IgG1 expression vector to allow for transient expression in mammalian cells. First, restriction sites were inserted in the 5′ and 3′ ends, to match the cloning sites of the IgG1 vector. H chain sequences were additionally trimmed in the 3′ end, to end directly after the restriction site. Next, GenScript’s online tool GenSmart Codon Optimisation (https://www.genscript.com/gensmart-free-gene-codon-optimization.html; expression host organism: human) was used to generate codon-optimised sequences. The 5′ and 3′ ends were further modified to allow for the design of gene amplification primers of appropriate properties. Finally, the sequences were scanned for restriction sites for several restriction enzymes (BspTI, Eco91I, BshTI, Pfl23II, SfiI, NotI, EagI, AvrII, and BamH1) in other locations than those required for restriction insertion into the IgG1 vector. Where such sites were identified, they were removed by replacing a codon with another one encoding the same amino acid. These sequences were synthesised by Twist Biosciences (South San Francisco, CA, USA) and provided in the pTwist Kan High Copy vector.

The synthesised genes were cloned into human IgG1 vector using the In-Fusion HD Cloning kit (Takara Bio, Mountain View, CA, US), according to the manufacturer’s instructions. The IgG1-encoding vector was linearised using restriction enzymes (Thermo Fisher Scientific, Waltham, MA, USA). IgG1 vector was incubated for 2 h at 37°C, together with BshTI (0.5 µl/µg vector), BspTI (0.5 µl/µg vector), Eco91I (0.5 µl/µg vector), and Pfl23II (2 µl/µg vector) in 1X buffer O (Thermo Fisher Scientific). PCR amplification of synthesised genes was performed using CloneAmp HiFi PCR Premix (Takara Bio) and primers obtained from Eurofin Genomics (Ebersberg, Germany); the primer sequences were selected according to the In-Fusion HD Cloning Kit user manual. 30 cycles of 1) 98 °C for 10 s and 2) 69 °C for 30 s were used. Linearised vector fragments and PCR amplified genes were purified using NucleoSpin Gel and PCR Clean-up (MACHEREY-NAGEL, Düren, Germany), as described in the user manual, and visualised on 1.0–1.2% agarose gels. InFusion Cloning was performed by incubating 100 ng linearised IgG1 vector, 30–45 ng of VH encoding gene, and 30–45 ng of VL encoding gene in 10 µl 1X In-Fusion HD Enzyme Premix (Takara Bio) for 15 min at 50 °C. The In-Fusion reaction mixture was used to transform Stellar Competent Cells (Clontech Laboratories, Mountain View, CA, USA) as described in the In-Fusion HD Cloning Kit user manual. In short, Stellar Competent Cells were incubated with an In-Fusion reaction mixture for 30 min on ice, heat shocked for 45 s at 42 °C, and incubated for another 1–2 min on ice. Cells were thereafter incubated in SOC medium (Clontech Laboratories) at 37 °C for 1 h and plated on LB plates (100 µg/ml carbenicillin). After overnight incubation at 37 °C, individual colonies were picked and grown overnight in 2 ml 2xYT (1% glucose, 100 µg/ml carbenicillin) at 37 °C. Plasmid DNA was isolated from the overnight culture using QIAprep Spin Miniprep Kit (QIAGEN), according to the manufacturer’s instructions, but eluting for 3 min in 50 µl preheated (50 °C) nuclease-free water. Cloning results were evaluated through restriction digestion (using the method described above), visualised on 2% agarose gel, and subsequently sent to Eurofins Genomics for sanger sequencing to confirm that the IgG1 vector contained the correct inserts. Overnight cultures were also stored as glycerol stocks.

Subsequently, IgE-encoding vectors with the variable domains of 116965_A, 212579_B, and 257199_A were also constructed. A codon-optimised gene encoding the constant domains of human IgE was synthesised by Genscript (Piscataway, NJ, USA), cleaved with ClaI and Eco91I (Thermo Scientific), and cloned into initially established eukaryotic expression vectors encoding recombinant antibody specificities 116965_A, 212579_B, 257199_A to replace their gene encoding the constant domains.

Antibody production

2 ml 2xYT (1% glucose, 100 µg/ml carbenicillin) were inoculated with glycerol stocks of transformed Stellar Competent cells carrying the IgG1 or the IgE vector with correct VH and VL gene insertions and incubated overnight at 37 °C. EndoFree Plasmid Maxi Kit (Qiagen) was used, according to the manufacturer’s handbook, to isolate the vector, which was subsequently transfected into HEK293 cells. The HEK293 cells (cultured in DMEM medium with 10% FCS at 37 °C and 5% CO_2_) were seeded at a density of 0.3 × 10^6^ cells/well in a 6-well plate (Corning Inc, Corning, NY, USA). At a density of 10^6^ cells/well (approximately 2 days after seeding), the medium was changed to 2 ml fresh medium. PEI in 0.15 M NaCl was added to DNA (900 ng DNA/10^6^ cells, preboiled at 62 °C for 10 min, and diluted in 0.15 M NaCl) in a 6:1 ratio and added to the cells. 1 ml of fresh medium was added after 4 h of incubation. Culture supernatant containing produced antibodies was retrieved after 3 days of culture and used, without prior purification, for deconvolution of antibody target and various immunochemical analyses.

Antibody target deconvolution—Immunochemical analysis

Immunochemical analysis was used to determine the specificity of identified recombinant antibodies. An ELISA was performed by coating 1 µg/ml recombinant allergens Art v 1.0101, Bet v 1.0101, Phl p 1.0102, Phl p 2.0101, and Phl p 5.0101 to 96-wells microtiter plates (Corning) overnight. After washing, monoclonal IgG1 antibodies were incubated with the immobilised antigen. Bound antibody was detected using horse radish peroxidase (HRP)-labelled anti-human IgG (Jackson ImmunoResearch, West Grove, PA, USA). Allergen-specific binding of serum-derived and recombinant IgE was similarly performed, using rabbit-anti-human IgE (Dako Denmark A/S) and HRP-labelled goat-anti-rabbit IgG (Zymed Laboratories, South San Francisco, CA, USA) as the detection reagent. Assays were performed using 0.05% Tween 20 and 0.5% cold water fish skin gelatine (Merck, Darmstadt, Germany), or 0.05% Tween 20 and 0.5% bovine serum albumin Cohn fraction V (Saveen & Werner, Limhamn, Sweden) in PBS as dilution buffer. SuperSignal ELISA Pico Chemiluminescent Substrate or 1-Step™ TMB ELISA Substrate (Thermo Fisher Scientific) were used as substrates. Further immunochemical assessment of recombinant antibodies of the IgG isotype was performed using ImmunoCAP technology (Thermo Fisher Scientific), initially using ImmunoCAPs carrying allergen mixes ex1, gx1, gx6, hx2, tx9, wx3, wx6, and wx7, and subsequently using defined grass pollen sets g1-g7, g10, g11, and g13, and defined allergen components (g205, g206, g208-g213, and g215). This initial allergen-specific IgG ImmunoCAP analysis was performed according to the manufacturer’s instruction at the Thermo Fisher Diagnostics ApS Service Laboratory (Alleröd, Denmark), an ISO 13485 and ISO 9001 certified laboratory. Samples were diluted 1:100 in Phadia 200 System sample IgG diluent, and results were calculated into mg_A_/L against the WHO IgG calibrator. The limit of quantification was 2 mg_A_/L. Additional ImmunoCAP analyses of patients’ sera and monoclonal antibodies of the IgE isotype were performed according to the manufacturer’s instructions at the unit of Clinical Immunology and Transfusion Medicine at Region Skåne (Lund, Sweden). This laboratory is accredited to the international standard ISO 15189 for several ImmunoCAP analyses. All ImmunoCAP methods were certified under either IVDD or IVDR.

Antibody target deconvolution—Immunoprecipitation/mass spectrometry

For some selected antibodies, an IP/MS strategy was applied to further decipher the target antigen, using Pierce MS-Compatible Magnetic IP Protein A/G Kit (Thermo Fisher Scientific) according to the manufacturer’s description. Briefly, immune complexes were formed by incubating IgG antibodies mixed with pollen extract from Holcus lanatus or Phleum pratense overnight at 4 °C, and subsequently caught on Protein A/G magnetic beads. After washing, antibodies and any bound proteins were eluted, using the IP-MS Elution Buffer, dried, reduced, alkylated, and in-solution digested with trypsin. Peptides were cleaned using C18 spin columns (BioPureSPN, The Nest Group, Ipswich, MA, USA). Peptides (approximately 800 ng per injection) were separated on an in-house packed 360 µm OD × 75 µm ID × 15 cm L, 15 µm tip SelfPack NanoLC column with emitter tip and frit (MS Wil, Aarle-Rixtel, the Netherlands) using an EASY nLC1200 (Thermo Scientific). The LC method was 3 min 5–10% B, 30 min 10–25% B, 5 min 25–40% B, 5 min 40–90% B, followed by 7 min at 90% B; with solution A being 1% formic acid, and solution B 1% formic acid in 80% acetonitrile. The LC was coupled online to a QExactive HF-X mass spectrometer operating using data-dependent acquisition in positive ion mode. The 50 min top-20 method used automatic gain control (AGC) target 3E5 for MS1 spectra and 1E5 for MS2, with maximum injection time 50 ms and 20 ms respectively. Resolution was set to 120,000 for MS1 and 15,000 for MS2, and the normalised collision energy was set to 27 for MS2 with a 1.2 m/z isolation window.

The LC-MS/MS data were processed using MaxQuant^56^ version 1.6.17.0 employing a database consisting of all Phleum pratense UniProt proteins as of 20220602, the antibody sequences, and all available timothy-derived sequences from the WHO/IUIS allergen nomenclature database (https://www.allergen.org) as of June 2022. Default settings with label-free quantifications were used for MaxQuant including fixed carbamidomethylation of cysteines, variable methionine and N-terminal protein acetylation and filtering at FDR < 0.01 at the protein and peptide levels. The mass spectrometry data have been deposited to ProteomeXChange via the PRIDE partner repository^57^ with the dataset identifier PXD056681.

The protein intensity data from MaxQuant were log2-transformed in NormalyzerDE and PCA plots were generated in OmicLoupe^58^. Hierarchical clustering and heatmaps were generated in R version 4.3.2 using the ComplexHeatmap package.

Surface plasmon resonance-based analysis of reaction rate constants and affinity

Affinity and reaction rate constants of allergen-antibody interactions were determined using a MASS-16 surface plasmon resonance-based biosensor (Bruker, Hamburg, Germany). Briefly, the human IgG Fc-specific monoclonal antibody of the Human Antibody Capture Kit (Cytiva, #BR100839) was immobilised onto a high-capacity amine sensor chip (Bruker). Recombinant allergen-specific antibody (about 100 RU) was captured by the immobilised antibodies. Antigen was flown across the surface at a flowrate of 30 µl/min. The surface was regenerated using the 3 M MgCl_2_ regeneration solution (Cytiva). Analysis was performed using the Sierra Analyser software version 3.4 (Bruker). Further analysis (extent of cross-reactivity to other allergens and binding of a Phl p 4-specific human antibody to non-glycosylated Phl p 4 N69Q N388Q produced in P. pastoris) was performed on a Biacore 1 K+ instrument (Cytiva) using the same surface chemistry (CM5 sensor chip (Cytiva)) and immobilisation and running strategy. Data were analysed using the Biacore Evaluation Software (Cytiva).

Antibody developability assessment

Protein sequence and structure features may contribute to the liabilities of the product. The Therapeutic Antibody Profiler^28^ (https://opig.stats.ox.ac.uk/webapps/sabdab-sabpred/sabpred/tap) was used to assess the isolated human allergen-specific antibodies in relation to five developability metrics. High-risk sequence motifs associated with variable domain isomerization, deamidation and glycosylation were identified using the Bionamic Antibody R&D Data Management platform (Bionamic AB, Lund, Sweden).

Statistics and reproducibility

Assays, with the exception of ImmunoCAP assays, were performed in duplicates or triplicates. Results are reported as average ± standard deviation whenever appropriate. ImmunoCAP analyses of patients’ sera and monoclonal antibodies of the IgE isotype were performed at an accredited laboratory, as outlined above.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary material Supplementary Data 1 Description of Additional Supplementary Files Reporting Summary