Allergen-specific IgG⁺ memory B cells are temporally linked to IgE memory responses

Abstract

Background:

IgE is the least abundant immunoglobulin and tightly regulated, and IgE-producing B cells are rare. The cellular origin and evolution of IgE responses are poorly understood.

Objective:

The cellular and clonal origin of IgE memory responses following mucosal allergen exposure by sublingual immunotherapy (SLIT) were investigated.

Methods:

In a randomized double-blind, placebo-controlled, time course SLIT study, PBMCs and nasal biopsy samples were collected from 40 adults with seasonal allergic rhinitis at baseline and at 4, 8, 16, 28, and 52 weeks. RNA was extracted from PBMCs, sorted B cells, and nasal biopsy samples for heavy chain variable gene repertoire sequencing. Moreover, mAbs were derived from single B-cell transcriptomes.

Results:

Combining heavy chain variable gene repertoire sequencing and single-cell transcriptomics yielded direct evidence of a parallel boost of 2 clonally and functionally related B-cell subsets of short-lived IgE⁺ plasmablasts and IgG⁺ memory B cells. Mucosal grass pollen allergen exposure by SLIT resulted in highly diverse IgE and IgG_E repertoires. These were extensively mutated and appeared relatively stable as per heavy chain isotype, somatic hypermutations, and clonal composition. Single IgG_E⁺ memory B-cell and IgE⁺ preplasmablast transcriptomes encoded antibodies that were specific for major grass pollen allergens and able to elicit basophil activation at very low allergen concentrations.

Conclusion:

For the first time, we have shown that on mucosal allergen exposure, human IgE memory resides in allergen specific IgG⁺ memory B cells. These cells rapidly switch isotype, expand into short-lived IgE⁺ plasmablasts, and serve as a potential target for therapeutic intervention.

Graphical Abstract

Allergic diseases are typically lifelong, and even in the absence of allergen exposure, for this phenomenon to occur requires some form of immunologic memory. Current concepts of the cellular origin of IgE memory are primarily based on murine studies using various strains of transgenicmice.¹ It has been reported that IgG⁺ memory B cells are able to induce antigen-specific IgE memory responses when transferred into naive hosts.^2,3 Although these studies do not exclude the possibility of long-lived IgE⁺ memory B cells, they confirm the importance of indirect isotype switching that leads to allergen-specific IgE responses. In contrast, 1 study reported a transfer of IgE memory responses by a subset of IgE⁺ B cells,⁴ although the subset was later rectified to contain a mixed population of IgG⁺ and IgE⁺ B cells.⁵ In general, studies have confirmed that IgE⁺ B cells have an impaired ability to enter germinal centers (GCs), leading to short-lived plasmablasts and absence of affinity maturation.^6,7 Similarly, IgE⁺ B cells are predisposed to differentiate into short-lived plasmablasts.^6,8 A more recent finding, obtained by using a murine model of peanut allergy, showed that allergen-specific IgG response precedes IgE response,⁹ and expansion of allergen-specific IgG1⁺ memory B cells was accompanied by bone marrow reconstitution with IgE⁺ plasmablasts in mice rechallenged with allergen 9 months after sensitization.³ Taken together, mouse studies have provided convincing evidence for the role of IgG⁺ memory B cellsin maintaining IgE memory responses. However, these findings have not yet been confirmed in individuals with allergy. A recent study utilizing a validated and highly sensitive PCR-based methodology failed to identify IgE⁺ memory B cells in patients with allergy,⁹ and heavy chain variable gene (VH) repertoire sequencing data are consistent with indirect switching to IgE from primarily IgG-expressing B cells in humans.¹⁰

Moreover, observations from several clinical trials of grass pollen sublingual immunotherapy (SLIT) have shown an increase in titers of IgE antibodies in serum that peaks in the first weeks of treatment, followed by a gradual decline over time.^11–13 We therefore hypothesized that the transient increase in serum IgE titer during SLIT coincides with a clonal boost of migratory allergen-specific B cells in blood, as previously demonstrated in a study of tetanus toxoid vaccinations.¹⁴ Here, we investigate the cellular and clonal origin of IgE memory responses by using next-generation sequencing of total antibody VH repertoires in combination with cell sorting techniques and single B-cell transcriptomics.

METHODS

Clinical trial samples

The study (NCT02005627) was conducted at a single academic center, Imperial College London, and included recruitment of 40 adult patients with moderate-to-severe seasonal allergic rhinitis (see Fig E1 and Table E1 in this article’s Online Repository at www.jacionline.org for subject characteristics). The trial was a randomized double-blind, placebo-controlled, time course sublingual immunotherapy study (Grazax, ALK-Abelló, Hørsholm, Denmark). The trial protocol¹⁵ and amendments were approved by the relevant ethics committees and institutional review boards. Written informed consent was obtained from all participants.

RNA extraction from PBMCs, sorted cells, and nasal biopsy samples

For the sampling time points baseline, 4 weeks, 8 weeks, 16 weeks, 7 months, and 12 months after initiation of SLIT treatment, total RNA was purified from 20 million PBMCs and nasal biopsy samples by using the RNeasy Mini kit (Qiagen, Hilden, Germany) following the recommendations of the supplier. From sorted B cells, RNA was isolated by using the RNeasy Mini kit if the sample contained more than 500,000 cells; otherwise, the RNeasy Micro kit was used.

Immunoglobulin heavy chain sequencing and annotation

Amplification of the heavy chain V(D)J region, library preparation, and high-throughput sequencing were performed by iRepertoire Inc (Huntsville, Ala). The resulting sequences were trimmed and filtered for sequence quality, and paired-end reads were joined by using PEAR v0.9.7 software.¹⁶ Identical sequences were collapsed by using fastx_collapser, a part of the FASTX Toolkit v0.0.14 (http://hannonlab.cshl.edu/fastx_toolkit/index.html). Singleton sequences were discarded from further analysis. Isotype was assigned on the basis of the first 17 nucleotides of the constant region, and annotation of V family, J family, complementarity-determining region 1 (CDR1), CDR2, and CDR3 was performed by using migmap v0.9.8 (https://github.com/mikessh/migmap).¹⁷ PCR crossover events were removed by discarding sequences that contributed with less than 5% to a given CDR3-defined clonotype. VH sequences were clustered into clonal families by using the DefineClones tool of the Change-O package v0.3.9.¹⁸ Sequences were assigned to the same clonal family if they had the same V and J family annotation, if the CDR3 region had the same length, and if the sequence identity between the CDR3 regions was greater than 90% on the nucleotide level. An IgE clonotype was defined as a clonal family that contains more than 50 IgE transcripts at a given time point. A more sensitive transcription cutoff was chosen to define IgG clonotypes clonally related to IgE (IgG_E): a clonal family was required to contain at least 10 IgE and at least 10 IgG transcripts at any time point.

Flow cytometry

For B-cell fluorescence-activated cell sorting (FACS), PBMCs were stained with CD3 fluorescein isothiocyanate, CD19 phosphatidylethanolamine, IgD PerCP/Cy5.5, CD38 phycoerythrin–cyanine 7, CD138 allophycocyanin, and CD27 Pacific Blue. The Live/Dead Fixable Aqua Dead Cell Stain Kit (ThermoFisher, Waltham, Mass) was used to ensure sorting of viable cells. Naive B cell (CD19⁺CD27⁻IgD⁺), memory B cell (double memory cells, CD19⁺CD27⁻IgD⁻; classical memory cells, CD19⁺CD27⁺IgD⁻; and IgM memory cells, CD19⁺CD27⁺IgD⁺) and plasmablast (CD19^+/low CD27⁺CD38⁺CD138⁻) populations were sorted into separate tubes. To prepare for single-cell transcriptomics, single memory B cells from patient D04 were gated (as IgE⁺CD19⁺CD4⁻CD8⁻) and sorted directly into 96-well PCR microtiter plates. Staining for surface IgE appeared unspecific, likely reflecting surface-bound IgE complexes on non-IgE memory B cells expressing the low-affinity CD23 (FcεRII) receptor.

Single-cell transcriptomics

The assay used to capture whole mRNA transcripts was adapted from the Smart-seq2 protocol.^19,20 Briefly, mRNA was captured by using poly-dT oligos and directly reverse-transcribed into full-length cDNA by using the described template-switching LNA oligo.^19,21 Whole transcriptome cDNA was amplified by PCR. Quality and quantity of cDNA amplification were assessed by capillary electrophoresis using Fragment Analyzer (Advanced Analytical Technologies, Inc, Ames, Iowa) and fluorescent double-stranded DNA intercalating dye-based assay (Picogreen, Invitrogen, Carlsbad, Calif). Before sequencing, all libraries were purified by using AMPure XP beads (Beckman Coulter Life Sciences, Indianapolis, Ind)(0.9:1). Samples were sequenced on the Illumina sequencing platform HiSeq2500 (Illumina, San Diego, Calif). Libraries generating a total of 172 million uniquely mapped reads (a median of ~1.8 million total uniquely filtered mapped reads per cell).

Single-cell RNA sequencing data were mapped against the human hg19 reference genome and United States Cancer Statistics gene models by using TopHat software (v1.4.1, library-type fr-unstranded). The single-cell RNA sequencing data were integrated with the single-cell data from Croote et al²²by using the R library Seurat.²³

Antibody expression and characterization

Recombinant IgE antibodies were transiently expressed in HEK293 suspension cultures (Freestyle 293, Thermo Fisher Scientific, Waltham, Mass). Expression plasmids were custom made at Genscript (Piscataway, NJ). Recombinant IgE antibodies were screened for specificity by surface plasma resonance (Biacore 3000, GE Healthcare, Marlborough, Mass). Basophil activation assays were done as previously described.²⁴

Statistical analysis

P values were calculated by a 2-sample Wilcoxon test with R open source software.

RESULTS

Sublingual allergen immunotherapy activates 2 subsets of IgE⁺ and IgG⁺ B cells of common clonal origin

To evaluate the IgE repertoire development, antibody responses were investigated by use of next-generation sequencing of VH repertoires amplified by PCR in 21 subjects at baseline and after 4 weeks of grass SLIT-tablet (for information about sequencing depth see Table E2 in this article’s Online Repository at www.jacionline.org). A cluster and isotype assignment analysis identified 998 IgE clonotypes derived from the week 4 samples during SLIT. Of these IgE clonotypes, 22% clustered together with a minor population of IgG, indicating common clonal origin, as shown in Fig 1, A. This defines a specific subset of the IgG repertoire that we here call IgG_E and that is likely to share antigen-specificity with the IgE repertoire.

FIG 1.

Early clonal development of the IgE memory response at baseline and after 4 weeks during grass SLIT. A, Transcript level and isotype distribution of the 100 most frequent IgG_E clonotypes. Each vertical bar in the large panels indicates the transcriptional level of individual clonotypes and is colored according to isotype distribution. Data are sorted according to clonotype transcript levels at week 4. Compressed panels below show the isotype distribution within each clonotype. Horizontal placement indicates clonal relationship between time points. B, Total levels of IgE and IgG_E transcripts of individual donors before and during treatment. C, Transcript levels of individual IgE clonotypes identified at both baseline and week 4 and for the whole set of identified IgG_E clonotypes. D, Number of IgE and IgG_E clonotypes identified per donor.

The level of IgE transcripts per sample at baseline was low and increased after 4 weeks of SLIT in accordance with a boost of migrating IgE⁺ B cells (Fig 1, B). IgG_E transcripts also increased in response to SLIT, although to a lower level than IgE transcripts (Fig 1, B). Similarly, transcripts of individual IgE clonotypes (Fig 1, C), identified in both the baseline and week 4 samples, increased in response to SLIT, and the same transcriptional increase was observed for individual IgG_E clonotypes (Fig 1, C). In addition, the number of identified IgE and IgG_E clonotypes increased from baseline to the 4 week time point (Fig 1, D). Most of the IgE clonotypes shared between the baseline and week 4 samples were already switched to IgE at baseline (Fig 1, A and see Fig E2 in this article’s Online Repository at www.jacionline.org) indicating a precommitment to the IgE lineage before allergen exposure. The level of somatic hypermutations (SHMs) in IgE repertoires was similar to those of IgG, IgG_E, and IgA (see Fig E3, A in this article’s Online Repository at www.jacionline.org), which is in agreement with sequential isotype switching from IgM to IgG and then IgE. Furthermore, the average level of SHMs in the IgE and IgG_E repertoires did not increase, even within individual IgE clonotypes (Fig E3, B), despite the daily high-dose administration of grass pollen tablet for 4 weeks. This indicates that switching from IgG to IgE happens without further affinity maturation.

Stable composition of IgE repertoires during sublingual allergen immunotherapy

The effect of grass SLIT-tablet on IgE repertoire development during 1 year of treatment was investigated by analyzing longitudinal samples from 3 patients selected for high levels of IgE transcripts at week 4 (D04, D07, and D16). VH transcripts for individual clonotypes tended to decline after 4 weeks of treatment, indicating a reduced number of peripheral IgE⁺ B cells (Fig 2, A). Nonetheless, IgE repertoires remained diverse for more than 6 months, and each time point had a substantial fraction of private as well as shared IgE clonotypes (Fig 2, B). The IgE isotype was conserved for most clonotypes throughout the 6 months of treatment. The IgE repertoire isolated at week 4 (ie, at the peak of the serologic IgE response) yielded the highest number of clonotypes, which consistently constituted 51% to 52% of the repertoires at later time points (Fig 2, B, top row). This consistent resampling rate indicates a relatively fixed grass tablet–induced IgE repertoire. Sampling at later time points led to a progressively less efficient sampling of the IgE repertoire that is likely explained by a lower number of IgE-producing B cells in the blood samples (ie, contraction of the IgE repertoire), as evident by the gradual drop in IgE transcripts over time.

FIG 2.

Longitudinal development of the IgE memory response during 1 year of grass SLIT. A, Transcript level and isotype distribution of all IgE clonotypes from 3 donors at multiple time points, as indicated. Clonotypes are sorted according to transcript levels at week 4. Horizontal placement indicates clonal relationship between time points. B, Proportion of IgE clonotypes that are shared between a reference time point (full dark purple pie) and all other time points during 6 months of SLIT. In each row a different time point serves as reference and arrows indicate the direction of comparison. Dark purple coloring denotes the proportion of shared IgE clonotypes, and pie size is log-proportional to the IgE repertoire size at the given time point. Clonotypes from 3 donors were pooled. C, Longitudinal analysis of the total transcript levels of IgE and IgG_E. Each line represents the response of a single subject. D, The average frequency of SHMs for IgE clonotypes that were identified at a minimum of 3 time points (upper panel) and, in the case of IgG_E, of all clonotypes belonging to the IgG_E repertoires at each time point (lower panel). Each dot represents the average nonsynonymous mutation rate among transcripts contained in any particular clonotype. BL, Baseline.

The parallel trajectories of the total levels of IgE and IgG_E VH transcripts further support simultaneous activation and coevolution of 2 clonally related populations of IgE⁺ and IgG_E⁺ B cells (Fig 2, C). For both repertoires, SHM levels remained constant during treatment, indicating no further affinity maturation (Fig 2, D). Thus, despite the daily exposure to allergen over the course of 1 year of SLIT, the cellular IgE memory response, composed of proliferating IgE⁺ and IgG_E⁺ B cells, appeared relatively stable, with no signs of isotype switching, clonal skewing, or further mutagenesis.

Nasal and blood IgE repertoires are clonally related

To understand the relationship between antibody repertoires in blood and the nasal mucosa, VH repertoire sequencing was performed on nasal biopsy samples from 7 donors that were collected at baseline and after 1 year of treatment. The nasal biopsy samples contained a larger fraction of IgA transcripts than did the matching samples collected from blood (Fig 3, A).

FIG 3.

The IgE repertoire of the nasal mucosa from 7 patients with grass allergy at baseline and after 12 months of grass SLIT-tablet treatment. A, Isotype distribution profiles of pooled nasal antibody repertoires from the 7 donors. The corresponding data from pooled blood PBMC repertoires of these 7 donors at baseline is shown for comparison (left column). B, Transcript levels of total IgE of individual donors in nasal biopsy samples and blood. C, Number of IgE clonotypes per nasal sample in comparison with matched blood baseline from the same donors. D, Transcript levels of individual IgE clonotypes in nasal biopsy samples and blood (at baseline) as indicated. E, Pooled analysis of the overlap of IgE repertoire from the 7 selected donors. Overlap in IgE clonotype use (dark colors) of the nasal repertoires and the blood at baseline (upper row) or at the blood IgE peak point of 4 weeks into SLIT (lower row). The total number of identified IgE clonotypes at each time point is indicated in brackets.

IgE transcripts were present in all nasal biopsy samples except 1, but at levels that were low and comparable to those in the blood baseline samples (Fig 3, B). In accordance, IgE repertoire diversity, ie, the number of clonotypes per sample (Fig 3, C), and the transcript level of individual IgE clonotypes (Fig 3, D) were low and comparable to the blood baseline values. There was a clear clonal relationship between blood and nasal IgE repertoires (Fig 3, E) that increased at week 4 during SLIT. Similar to the blood repertoire, the nasal repertoires contained a significant and consistent fraction of IgG_E (Fig 3, E and see Fig E4 in this article’s Online Repository at www.jacionline.org). Thus, the IgE and IgG_E memory responses in blood, induced by oral allergen provocation, appeared closely associated with the quiescent nasal IgE repertoire.

The IgE memory response contains a transcriptionally heterogeneous population of memory B cells

To understand the cellular origin of the IgE repertoire in the periphery, VH repertoires were analyzed from sorted subsets of naive B cells (CD20⁺IgD⁺CD38⁻), memory B cells (CD20⁺IgD⁻CD38⁻), and plasmablasts (CD20^lowIgD⁻CD38⁺) collected at week 4 (Fig 4, A and B). Naive B cells were evenly distributed in a multitude of small clusters of IgM or IgD isotypes, whereas plasmablasts were of relatively large-sized clusters with all isotypes (except IgD) represented in accordance with a repertoire shaped by clonal expansion. Memory B cells contained a multitude of small clusters of all isotypes but were dominated by a few large IgE clonotypes (Fig 4, B).

FIG 4.

Cellular phenotypes of IgE memory B-cell responses. A, FACS sorting of PBMCs of patient D04 collected at week 4. Cells were bulk-sorted by using phenotypic markers for naive B cells, plasmablasts, and memory B cells as indicated. B, VH repertoire sequencing of the FACS-sorted populations. Waterfall plot of the transcriptional levels and isotype distribution for individual clonotypes. Only the 100 most frequent clonotypes are shown for each sample of sorted cells.

Transcriptomic profiling of single grass-specific IgG⁺ memory B cells and IgE⁺ preplasmablasts

To address the observed transcriptional heterogeneity in the memory B-cell compartment, we performed indexed single-cell FACS sorting followed by single-cell transcriptomic profiling of single memory B cells of subject D04 week 4 sample (experimental flow outlined in Fig 5, A²²).

FIG 5.

Single-cell transcriptomic profiling of memory B cells sorted from donor D04 at week 4 during SLIT. A, Single-cell sequencing and analysis workflow. B, A pooled t-distributed stochastic neighbor embedding (tSNE) analysis combining the 93 single-cell transcriptomes of subject D04 (black dots) with a reference data set of 973 B-cell transcriptomes (gray dots).²² The 2 cells with productive and IgE rearrangements are labeled in purple and marked by red arrows, and the remaining cells expressing grass-specific antibodies are labeled in orange. C, Transcription levels of the plasmablast marker CD38 in all transcriptomes belonging to the plasma cell cluster in the tSNE analysis in comparison with the CD38 levels in plasma cells of the reference data set. D, Transcript levels and isotype distribution in VH antibody repertoires, at different time points, of individual VH clonotypes identified by both bulk repertoire sequencing and single-cell transcriptomics. E, Allergen-induced activation of basophils passively coated with 3 different mixtures of purified mAbs according to specificity. (Blue line indicates a mix of 6 Phleum pratense 5 (Phl p 5)-specific mAbs, red line indicates a mix of 6 Phl p 6–specific mAbs, and green line indicates a mix of all 11 mAbs). Overlapping gray lines show the activity of the individual 11 mAbs. NGS, Next-generation sequencing; PB, plasmablast; PC, plasma cell; RNA-seq, RNA sequencing; VL, light chain variable gene.

Transcriptomic data were integrated with a reference data set of 973 single cells prepared from CD19⁺ B cells²² and clustered by t-distributed stochastic neighbor embedding analysis of normalized gene expression counts (Fig 5, B). Most of the sorted memory B cells (85 of 93) had the expected naive/memory phenotype. However, 8 cells clustered as plasmablasts, indicating phenotypic heterogeneity within the population of sorted memory B cells. The expression of signature genes of the memory/naive (membrane spanning 4-domains A1 [MS4A1] and interferon regulatory factor 8 [IRF8]) and plasmablast (PR domain 1 [PRDM1] and interferon regulatory factor 4 [IRF4]) populations was consistent with that in previous reports²² (see Fig E5 in this article’s Online Repository at www.jacionline.org). In accordance with the negative selection for CD38 expression in the FACS sorting protocol, the 8 cells with a plasmablast phenotype differed from the reference population in CD38 expression (Fig 5, C). Two transcriptomes contained productive IgE transcripts, and they both clustered with the plasmablast-like subset of cells. Thus, it appears that the CD38⁻ CD27⁺ preplasmablast population is enriched in IgE⁺ cells, explaining the large number of IgE transcripts in sorted memory B cells.

For 64 single-cell transcriptomes, it was possible to retrieve the full sequence of cognate pairs of antibody heavy chain and light chain variable region genes. These cognate pairs were aligned to the total antibody VH repertoires (Fig 5, D) which allowed for selection of 11 antibody sequences based on clonal relationship to IgE and/or IgG clonotypes. An additional 5 antibodies were selected on the basis of the presence of sterile germline transcripts of the constant region of IgE (εGLT),²² which is indicative of active involvement in T_H2 inflammatory response.²⁵ Of the 16 antibodies, 14 bound to grass extract and were mostly specific for major allergens (Table I).

TABLE I.

Summary of data on mAb clones

	mAb03	mAb07	mAb13	MAB17	mAb36	mAb38	mAb40	mAb41	mAb44	mAb46	mAb58	mAb77	mAb78	mAb83	mAb93	mAb94
Single-cell transcriptome
Isotype	IgG4	IgG2	IgG1	IgE	IgG1	IgG2	IgG1	IgG1	IgG1	IgG2	IgG1	IgG1	IgG1	IgE	IgG1	IgG4
CD23 transcript count	426	347	2438	475	842	1	826	262	1275	3179	1477	494	956	0	928	780
Sterile ԑGLT	−	+	+	−	+	−	+	+	+	+	+	+	+	−	+	+
Phenotype	MEM	MEM	MEM	PB	PB	MEM	MEM	MEM	MEM	MEM	MEM	MEM	MEM	PB	MEM	MEM
Bulk sequencing
Isotypes	IgE	IgE	IgE	IgE	*	IgE	*	*	IgE/IgG	*	IgE/IgG IgE/IgG		*	IgE/IgG	IgG	IgE/IgA
Specificity
Phl p extract	+	+	+	+	+	+	+	+	−	+	+	+	−	+	+	+
Phl p 1	−	−	−	−	−	−	−	−	−	−	−	−	−	−	−	+
Phl p 5	−	−	+	+	+	+	+	+	−	+	+	+	−	+	+	−
Phl p 6	−	−	+	−	+	−	−	−	−	+	−	+	−	+	−	−

epsilonGLT, Epsilon immunoglobulin germline transcript; Phl p, Phleum pratense.

Allergen-specific clones are set in boldface.

^*mAb not clonally related to bulk antibody repertoire.

The antibodies were of high affinity and able to trigger basophil activation at a very low concentration when combined (Fig 5, E). Eleven transcriptomes of the memory/naive phenotype encoded allergen-specific antibodies of the IgG1, IgG2, and IgG4 isotypes, which in 7 cases were coexpressed as IgE according to VH repertoire sequencing, thereby demonstrating a direct link between ongoing IgE memory response and allergen-specific IgG⁺ memory B cells. Interestingly, most of the selected B-cell transcriptomes contained CD23 transcripts (15 of 16) and 10 of 12 of the εGLT⁺ transcriptomes encoded allergen-specific antibodies. Further, the coexistence of clonally related IgG_E⁺ memory B cells and IgE⁺ plasmablasts agrees with the difference in transcriptional levels (IgE >> IgG_E; Fig 1, B) and the slower kinetics in synthesis of allergen-specific IgG relative to IgE in the early phases of SLIT (see Fig E6 in this article’s Online Repository at www.jacionline.org).

DISCUSSION

We have demonstrated for the first time that the serologic increase in allergen-specific IgE titer following mucosal allergen exposure was accompanied by a cellular boost of IgE-producing plasmablasts in blood. We observed high levels of IgE transcripts in sorted populations of plasmablasts (CD19⁺CD20^lowCD27⁺CD38⁺) and identified single allergen-specific B cells with a CD38⁻ preplasmablast phenotype in sorted memory B cells (CD19⁺ CD20⁺CD27⁺CD38⁻). Stimulation of human B cells ex vivo shows the emergence of plasmablast-like IgE⁺ B cells (Blimp-1⁺IgE^highCD38) from PBMCs of patients with allergy after 5 to 7 days of coculture with allergen²⁶ and IgE⁺ B cells (CFSE^lowCD19-^midCD27high) from tonsils after costimulation with IL-4 and anti-CD40 antibody.⁸ Both of these phenotypes were compatible with IgE⁺ preplasmablast and plasmablast phenotypes. The simultaneous drop in serum titers and IgE transcripts suggests that the human IgE⁺ plasmablasts are short-lived. Together with the absence of accumulation of SHMs, this parallels observations in mice, in which IgE⁺ plasma cells were short-lived and showed reduced affinity maturation, presumably because of a transient and incomplete GC phase.⁶

Such extrafollicular formation of IgE memory responses²⁷ could explain why allergic diseases progress slowly, in particular in adulthood, on account of a slowly evolving IgE repertoire. Longitudinal studies with samples taken years apart have demonstrated that IgE repertoires in subjects with allergy are oligoclonal and persist over time.²⁸ Similarly, it has been demonstrated that IgE repertoires sampled in 2 birch pollen seasons are overlapping.²⁹ We also observed such persistence of the IgE repertoire demonstrated by the limited clonal evolution, by the overlap between SLIT induced blood IgE repertoires and nasal repertoires taken 11 months apart, and by the absence of further isotype switching of the IgE repertoire. Considering the daily exposure to high doses of allergen for 1 year, this implies that allergen exposure, as such, is not the main cause for diversification of the IgE repertoire.

At the peak of the IgE memory response, we isolated single IgG⁺ memory B cells that encoded antibodies specific for the major grass allergens and belonged to clonotypes simultaneously expressing IgE. Considering the high fraction of allergen-specific antibodies (at least 14 of 64 cognate heavy and light chain variable gene pairs) in this population of memory B cells selected solely on the basis of phenotypic markers, the observed coexpression of CD23 and εGLT appears as a potential marker for memory B cells involved in IgE responses. Both markers are known to be under STAT6 control and induced by IL-4.^30,31 Whether these “T_H2-polarized” memory B cells are present in a “quiescent” state or the result of the daily exposure to allergen during SLIT remains to be determined. Further, the high frequency of allergen specific B cells reported here contrasts with all previous studies in patients with allergy, which typically report very low prevalence of allergen-specific B cells.^22,32–36 The high level of IgE transcripts (15%) in the PBMC fraction of subject D04 was in accordance with the observation that 10 of the top 100 clonotypes in the plasmablast sorted fraction were IgE producing.

Are those IgG_E⁺ memory B cells the source of the IgE⁺ plasmablast response, and hence, the provenance of IgE memory? Several observations support this notion: (1) single-cell transcriptomic analysis showed an equal representation of IgG_E⁺ memory B cells and IgE⁺ plasmablasts and the absence of IgE⁺ memory B cells; (2) most (9 of 11) of the allergen-specific IgG_E⁺ memory B cells contained εGLT transcripts, pointing to recent exposure to T_H2 cytokines such as IL-4, and thereby, to active involvement in the ongoing allergic inflammation; (3) the upregulation of IgG_E⁺ memory B cells coincided with the increase in specific IgE titers and was misaligned with the much later increase in allergen-specific serum IgG titers, and hence, was not associated with a concurrent IgG_E⁺ plasmablast response; (5) IgE transcript levels were consistently higher than the levels for IgG_E, supporting the idea that it is preferentially isotype-switched IgE⁺ B cells that leave the memory state and differentiate into plasmablasts; and (5) the similar and constant levels of SHMs in IgE and IgG_E repertoires, even within clonotypes, indicate isotype switching outside GCs, and hence, the absence of affinity maturation. Thus, allergen-specific IgG⁺ memory B cells, which are capable of rapid extrafollicular isotype switching to IgE, are likely the progenitors of the IgE-secreting plasmablasts forming the serologic IgE memory response at the site of inflammation, as was recently proposed by Gould and Ramadani.¹ Moreover, the observation of high ratios of allergen-specific to total IgE and low levels of IgG in the nasal mucosa of patients with rhinitis corroborates these findings.³⁷

One important question remains: Do long-lived IgE⁺ memory B cells exist? The low levels of IgE transcripts at baseline in blood and nasal samples could represent a rare population of long-lived IgE⁺ memory B cells giving rise to the subsequent IgE plasmablast response, whereas long-lived plasma cells are not likely to be found in blood.³⁸ Considering the simultaneous presence of IgG_E transcripts in these baseline samples, such IgE transcripts could also be the result of homeostatic self-renewal of IgG_E⁺ memory B cells turning into IgE⁺ plasmablasts by microbial products or bystander T-cell help.³⁹ Further, such allergen-specific plasma cells in subjects with allergy have previously been identified,^8,22 and B-cell cultures from donors with allergy, but not from healthy donors, expressed IgE by T-cell bystander activation, suggesting differences in the state of activation of memory B cells in these donors.⁴⁰ In further support, mouse studies suggest that lifelong food allergy is the consequence of recurrent activation of memory B cells leading to relatively short-lived plasma cells.³

IgE class switching can occur directly from IgM to IgE or from sequential rearrangements via IgG1, IgG2, or IgG4.^7,10,41 The intermittent IgG phase allows for affinity maturation and was proposed as the mechanism involved in the production of affinity-matured IgE antibodies in memory responses.^7,42 Prior studies using deep sequencing of human IgE repertoires show that IgE VH genes are most closely related to clonal lineages of IgG, particularly IgG1, and share extensive patterns of hypermutation with this isotype.^10,33 In agreement, peanut allergen-specific antibodies isolated from antigen-specific B cells were in most cases derived from class-switched cells expressing IgG.³⁵ We found direct evidence of indirect switching by identifying single IgG1⁺, IgG2⁺, and IgG4⁺ memory B cells expressing allergen-specific antibodies that were simultaneously expressed as clonal variants in the IgE repertoire. Because IgE clonal families often were of the same lineage as IgG clonal families, showed no clonal relationship to IgM, and contained SHMs at levels comparable to those in IgG clonal families, we conclude that direct switching from IgM to IgE has an insignificant role in allergen-specific IgE memory responses. Our previous work showed indirect evidence, switching from all IgG subclasses and less from IgM,⁴³ and we now prove the inferred antibody production. As with other studies,^10,33,35 a limitation of the current study is that it cannot formally exclude the existence of IgE⁺ memory B cells given the limited sampling depth of single-cell transcriptomics. Second, the daily exposure of high allergen doses might lead to cellular dynamics different from those of the daily low-level exposure during a pollen season. Thus, IgE memory responses induced by sublingual application of SLIT-tablets might not fully represent a memory response to natural allergen exposure, considering that the end-result of SLIT is clinical tolerance.

Our findings have several clinical implications. The relatively fixed composition of the IgE repertoire during the first year of SLIT demonstrates why long-term immunotherapy is not associated with any signs of disease progression, such as de novo sensitizations. Moreover, we demonstrate that antigen exposure per se is not a driving factor for IgE repertoire diversification. Finally, the existence of a distinct population of allergen-specific IgG⁺ memory B cells, prone to isotype switching and IgE secretion, can explain the lifelong persistence of allergy and is an obvious new target for therapeutic intervention.

Key messages.

• Lifelong persistence of allergy is underscored by the existence of allergen-specific IgG⁺ memory B cells that are prone to isotype switching and secretion of IgE.

• The fixed composition of the IgE repertoire during the first year of SLIT treatment provides evidence as to why long-term immunotherapy is not associated with any signs of disease progression.

Abbreviations used

CDR

Complementarity-determining region

FACS

Fluorescence-activated cell sorting

GC

Germinal center

IgG_E

IgG clonally related to IgE

SHM

Somatic hypermutation

SLIT

Sublingual allergen immunotherapy

VH

Heavy chain variable gene