IgE clonality and aggregation of insulin affect IgE-mediated activation of sensitized basophils

Abstract

Background

Skin prick testing and specific IgE determination do not discriminate between sensitization and clinically relevant insulin allergic reactions, whereas the basophil histamine release assay has proven useful in identifying clinically relevant food and drug allergic reactions.

Objective

We investigated IgE-mediated allergic reactions to insulin by studying the effect of insulin dimerization, superstructures (aggregates), and IgE clonality on activation of sensitized basophils.

Methods

Three humanized IgE molecules directed against distinct insulin epitopes were developed, and basophils, alone or in combination, were sensitized with these antibodies. Insulin in monomeric or dimeric form and superstructures in solution or coupled to a Sepharose matrix were used to stimulate histamine release. Different insulin structures were tested on basophils sensitized with sera from 3 diabetic patients with allergic reactions and insulin-specific IgE during insulin treatment.

Results

Monoclonal humanized IgE molecules triggered basophil histamine release when insulin was dimerized, presented in a superstructure, or coupled to a Sepharose matrix—but not when insulin was monomeric. Monomeric insulin only induced histamine release when basophils were sensitized with a specific pair of IgE molecules. In basophils sensitized with sera from insulin-allergic patients containing insulin-specific IgE, only insulin in dimer formation, presented in a superstructure or coupled to a Sepharose matrix, could induce robust histamine release. Free monomeric insulin could not.

Conclusions

This study illustrates why diagnostic methods for insulin allergy are difficult to correlate to clinical symptoms. Insulin aggregation and IgE clonality may be important factors in developing clinical symptoms of insulin allergy and should be addressed to improve the diagnostic outcome of insulin allergy testing.

Adverse reactions to insulin are observed in 0.1% to 3.0% of diabetic patients, with type I hypersensitivity reactions being the most common.¹ Testing patients for specific IgE against insulin has, however, shown to have low positive predictive value because it is relatively common for diabetic patients to develop IgE against insulin without showing clinical symptoms (asymptomatic allergy).2, 3, 4

A diagnostic supplement to IgE determinations based on cellular tests like the allergen-induced basophil activation test and the basophil histamine release assay (BHRA) has been documented to diagnose allergic reactions, including to drugs.5, 6, 7 However, a challenge to the diagnostic use of the BHRA is that many drugs are small molecules that can only cross-link cell-bound IgE when bound to other proteins (haptens) or in a multimeric form.⁸ Hence, using a monomeric drug form in BHRA may not simulate the physiologic conditions eliciting a clinical allergic reaction.

This study investigated if basophils sensitized with humanized monoclonal insulin-specific IgEs (IgE_Ins) against different epitopes, alone or in combination, can activate the cells to release histamine; this may serve as a positive control in BHRA when screening patient sera for functional IgE_Ins. We then tested differences in activation of basophils sensitized with humanized IgE_Ins and challenged with either monomeric, dimeric, or superstructures (aggregates) of insulin, previously shown to be immunogenic both in vitro and in vivo,⁹ as well as cellulose-coupled insulin (ImmunoCAP). Finally, we applied the different insulin preparations on basophils sensitized with sera from patients with anti-insulin IgE.

Our aim was to study why IgE-sensitized basophils rarely respond to the insulin molecule, and whether this is due to the relatively small molecular weight of monomeric insulin.

Methods

Humanized IgE molecules to insulin

Three humanized monoclonal IgE molecules (IgE_Ins) directed against distinct epitopes on the insulin molecule were received from Novo Nordisk (Table I).¹⁰ The antibodies are recombinant antibodies with a human IgE Fc region and the Fab region from a mouse, produced in HEK293 cells in Opti-MEM media. The 3 IgE_Ins demonstrated antibody binding to human insulin in the ImmunoCAP (data not shown).

Table I.

Human insulin epitopes for insulin-specific antibodies as determined by X-ray crystallography

Antibody	Epitope insulin A chain	Epitope insulin B chain	PDB ID
HUI-001∗	(2), 3, 5, (6), 8-12, (14), 15, 17-18	3, 5	8ONK (Online Repository)
HUI-018	(6), 7-8, 10-14, (16), 17	(4), 5-7, 10, 14, (17-18)	6Z7W10
S1	3, 13	1-2, 4-10, 12-13, 16-17, (21), 24, 26, (27-28)	8ONI (Online Repository)

Residue numbers denote residues within 4 Å from Fab. Residues within parentheses denote detected variations in antibody–insulin complex.

PDB, Protein Data Bank (www.wwpdb.org).

^∗HUI-001 epitopes were observed when bound to insulin in ternary complex with S1. If not complexed, HUI-001 epitopes could vary slightly.

Serum samples from 3 diabetic patients

Serum samples were obtained from 3 patients with boosted anti-insulin IgE after administration of an insulin analog (Table II). Samples were obtained before (baseline) and after (postdosing) administration of the insulin analog. All 3 patients provided informed consent for their sera be used for this purpose.

Table II.

ImmunoCAP IgE measurements of patient serum samples

Patient no.	Visit	Total IgE (kUA/L)	hIns sIgE (kUA/L)
5101	Baseline	>100	0.12
	After dosing	>100	3.13
5104	Baseline	>100	1.33
	After dosing	>100	43.82
5110	Baseline	>100	0.03
	After dosing	>100	87.3

Patient serum samples were tested by ImmunoCAP assay for sIgE toward human insulin; 1 kU_A/L = 2.4 ng/mL.

hIns, Human insulin; sIgE, serum-specific IgE.

Insulin

Lyophilized Novo Nordisk human insulin (5808 kDa molecular weight) was reconstituted in phosphate-buffered saline (PBS) at a stock concentration of 100 μmol.

Insulin dimer

A covalent dimer insulin analog was included as the smallest structure potentially able to cross-link two IgE molecules directed against the same unique insulin epitope. This dimer is formed by two insulin monomers with deleted B30 Thr and with B25 Phe replaced by cysteine, creating a disulfide bond between the cysteines at B25. The lyophilized dimer was reconstituted in PBS to a concentration of 100 μmol. The dimer was demonstrated in ImmunoCAP to be immunoreactive to the 3 humanized IgE_Ins (data not shown).

Insulin superstructures

Four insulin superstructures (aggregates) were studied: small and large particulates, and spherulite β-sheet and α-helical (Table III).⁹ The particulate samples consist of spherical, compact particles with a primarily α-helical secondary structure. Small particulates are nano-sized (200-300 nm), while the large particulates are a mixture of nano-sized and micron-sized (3-5 μm) particulates. The spherulite samples consist of flexible, 15-20 μm–sized spherical particulates. The spherulite secondary structure of the samples differ where one type is mainly β-sheet, and the other type is primarily α-helical. Production and characterization followed reported procedures for particulates¹¹ and spherulites.¹²

Table III.

Type, size, secondary structure, and appearance of insulin superstructures

Type		Size	Secondary structure	Appearance
Particulate	Small	200-300 nm	Primarily α-helical	Spherical, dense
	Large	3-5 μm + 200-300 nm
Spherulite	β-Sheet	15-20 μm	Primarily β-sheet	Spherical, fluffy
	α-Helical		Primarily α-helical

Simplification of table by Thorlaksen et al;⁹ see this study for images of particles.

Samples containing the superstructures were stored at 4°C and used within 2 weeks of being produced. To prepare samples for use in cell assays, they were dialyzed against PBS to exchange the harsh solvents used during preparation according to the procedure reported in Thorlaksen et al.⁹

ImmunoCAP

Insulin preparations were biotinylated using the LYNX Rapid Plus Biotin Antibody Conjugation Kit (LNK273B). The biotinylated insulins were then filtered through a prewashed PD-10 column (Cytiva Lifesciences, 17085101) by centrifuging at 1000 × g for 2 minutes. The concentrations of the biotinylated insulins were determined by a NanoDrop One spectrophotometer (Thermo Fisher Scientific, ND-ONE-W), then stored in 50% glycerol at −20°C. The streptavidin o212 CAPs (Thermo Fisher Scientific, 14-5320-01) and control Human Insulin c73 CAPs (Thermo Fisher Scientific, 14-4347-01) were prewashed using the “Pre-Wash ImmunoCAP” mode on the Phadia 100 Immunoassay Analyzer (Thermo Fisher Scientific, 12-3500-01). A total of 50 μL of the 50 μg/mL insulin preparation was added to the streptavidin CAPs and incubated on the ImmunoCAP carousel, with the lid closed, at 37°C for 30 minutes. Anti-IgE ImmunoCAP (Thermo Fisher Scientific, 14-4417-01) was used as a positive assay control. The matrix negative reference serum (Novo Nordisk) was used to dilute the humanized IgE_Ins.

Passive sensitization of buffy coat donor basophils

Buffy coats obtained from the blood bank (National University Hospital, Copenhagen) were stripped of membrane-bound IgE, as described previously.^13,14 The stripped buffy coats were sensitized with 3 monoclonal IgE_Ins targeting different insulin epitopes, either individually or in combination (Table I). Before sensitization, the monoclonal antibodies were mixed with a pool of negative reference serum and incubated with stripped buffy coats in a final concentration of 168 ng/mL IgE_Ins. Stripped buffy coats were also sensitized with sera (final dilution 25%) from 3 patients obtained before and after boosting of anti-insulin IgE. All sensitization experiments were performed for 1 hour at 37°C.

Allergen-induced BHRA

The BHRA has previously been described elsewhere.^13,14 In brief, the insulin preparations were diluted in 12 concentrations (3.5-fold dilution factor, starting from 100 μg/mL down to 0.0004 μg/mL and a negative buffer control) with PIPES [piperazine-N,N′-bis(2-ethanesulfonic acid)] buffer. Then the preparations were pipetted into the histamine plates (RefLab, RLA210) containing an inbuilt histamine standard curve. The passively sensitized cells were then pipetted into each well and incubated for 1 hour at 37°C. After incubation, the plates were washed to remove cells and interfering substances. To calculate the percentage of histamine release, a sample was lysed using 7% HClO₄. Histamine release (% BHRA) was then calculated as follows:

$$\frac{\text{Released}\text{histamine}\left(\text{ng}/\text{mL}\right)}{\text{Total}\text{histamine}\text{content}\left(\text{ng}/\text{mL}\right)}\times 100=\%\text{BHRA}$$

Preparation of insulin-coupled ImmunoCAPs

Insulin ImmunoCAPs were prepared according to the manufacturer’s instructions (Thermo Fisher Scientific), where streptavidin ImmunoCAPs were coupled to the different biotinylated insulins included in the study. Commercial total IgE ImmunoCAP (Thermo Fisher Scientific, 14-4509-01) containing anti-IgE antibody was included as positive control in CAP-BHRA.

Challenge of insulin-ImmunoCAP with basophils

The CAP-BHRA has been described previously.¹⁴ The donor basophils were passively sensitized as above. The sensitized cells were applied to the insulin ImmunoCAP and incubated in a humidified atmosphere for 30 minutes at 37°C. The ImmunoCAPs were then centrifuged at 300 × g for 5 minutes, supernatants were applied to glass fiber plates (RefLab, RLA210), and histamine was detected as previously described.^13,14

To calculate the histamine released, the mean histamine release of the background values (basophils could release up to 10% histamine on uncoupled, streptavidin-coated CAPs) were subtracted from histamine released using insulin-coated ImmunoCAPs. The background-normalized values of the duplicate CAPs were averaged for the mean histamine release of the sensitized basophils to the coupled insulins.

Total cellular histamine content

Cellular histamine was measured by lysis of 400 μL cell suspension using 120 μL 7% HClO₄ (RefLab, RLA704), followed by vortexing and incubation for 30 minutes at 37°C. The samples were centrifuged at 12,500 × g and supernatants pipetted onto a Quantitative plate (RefLab, RLA218), followed by quantification of histamine, as described previously.^13,14

Area under the curve

The area under the curve (AUC) calculations were based on the logarithmic allergen concentration–response curves for each passive sensitization in response to challenge with the insulin allergens.

Statistical analysis

All statistical analyses were performed by GraphPad Prism v9.0.1 software (GraphPad Software). When using basophil donors for the BHRA and CAP-BHRA, the histamine release was expressed as percentage histamine release of the total cellular histamine content. Mixed effect analysis, with Geisser-Greenhouse correction, was also used to determine significance between response curves. After the concentrations were transformed to log₁₀ format, the AUC of the dose–response curve was determined. First, normal distribution was assessed by the Shapiro-Wilk normality test. The baseline for the AUC was 10 units, as 10% of histamine release is the cutoff for a positive response in the BHRA. Thus, only the AUC of the positive values was calculated. Multiple responses were compared by 2-way ANOVA with Geisser-Greenhouse correction. Multiple comparisons were made by Tukey multiple comparisons test. If comparing two responses and the datasets were normally distributed, a paired t test was used. Otherwise, Wilcoxon matched-pairs signed-rank test was used.

Results

Effect of insulin superstructures on histamine release from basophils sensitized with monoclonal IgE_Ins

To evaluate the role of insulin as superstructures (aggregates) presented to basophils, 4 types of insulin superstructures were tested on basophils sensitized with monomeric anti-insulin IgE (HUI-001) in BHRA (Table III). The challenge was carried out at 12 concentrations; the results are expressed as the percentage of total histamine release (Fig 1). A dose-related histamine release response was observed for small particulates, whereas only a small response, almost reaching cutoff, was seen with large particulates (Fig 1). The largest structures, spherulite β-sheet and α-helical, induced no response. Monomeric insulin did not induce histamine release. Because small particulates gave the best response in BHRA, this superstructure was included in subsequent experiments.

Fig 1.

Response of HUI-001–sensitized basophils when challenged with insulin superstructures in BHRA. HUI-001–sensitized basophils were challenged with insulin superstructures (Table III) and with unstressed hIns as negative control, and measured in BHRA. Results depict means and SDs of dose–response curves of 3 basophil donors normalized to their total histamine. Dotted line indicates cutoff for positive response in BHRA. hIns, Human insulin; HR, histamine release.

Effect of IgE clonality, insulin dimerization, and insulin aggregation on basophil histamine release

Donor basophils were passively sensitized with the 3 monoclonal antibodies, where HUI-001 is directed against an epitope with amino acid residues primarily on the insulin A chain, HUI-018 against an epitope containing residues on both the A and B chains, and finally S1, which has an epitope with residues primarily on the insulin B chain (Table I).

The 3 insulin IgEs were selected on the basis of their binding to different parts of the insulin molecule, as illustrated by X-ray crystallography (see the Online Repository available at www.jaci-global.org).¹⁰ The crystal structures show that the HUI-001 and HUI-018 epitopes overlap, and that neither of their epitopes overlap with the insulin dimerization surface (Fig 2).^10,15 The pair HUI-001 and S1 was cocrystallized with insulin; this structure illustrates how HUI-001 and S1 can bind simultaneously to insulin. Furthermore, the structures including S1 Fab shows that the S1 epitope overlaps with the insulin dimerization surface.

Fig 2.

Ribbon diagrams with disulfide bonds represented as sticks of (A) human insulin monomer from PDB entry 6S34 with insulin A chain in light blue and B chain in dark blue,(B) insulin dimer from PDB entry 3U4N¹⁵ where the other insulin monomer of the dimer is shown in dark gray and the engineered disulfide between monomers is displayed in pink, and (C) crystal structures of human insulin in complex with HUI-018 in green (PDB entry 6Z7W, chains E and F),¹⁰ S1 in gray (PDB entry 8ONI, chains H and L; see Online Repository available at www.jaci-global.org), and HUI-001 in pink (PDB entry 8ONK, chains J and K; see Online Repository), where insulins of complexes have been superimposed on their B chain α-helices (residues 9-19). Insulin parts of Fab complexes have been excluded from illustration; only human insulin in same view as in (A) is displayed for clarity. Ribbon diagrams of Fabs have been supplemented by transparent surface representation. PDB, Protein Data Bank (www.wwpdb.org).

Passive sensitization was performed with each of the 3 monoclonal IgE_Ins. The sensitized cells were challenged with monomeric human insulin, a covalent insulin dimer and small particulates. The challenge was carried out at 12 concentrations and the results expressed as the percentage of total histamine release. Dose–response curves are presented in Fig 3, A-C. No histamine release response to the monomeric human insulins was observed with any of the 3 monoclonal IgEs. In contrast, challenge with dimeric insulin or small insulin particulates mediated varying degree of histamine release when basophils were sensitized with each of the 3 monoclonal IgEs (Fig 3, A-C).

Fig 3.

BHRA dose–response curve of basophils sensitized with (A) HUI-001, (B) HUI-018, and (C) S1, (D) combination of HUI-001/HUI-018, (E) combination of HUI-018/S1, and (F) combination of HUI-001/S1 challenged with various insulins. Results depict means and SDs of dose–response curves of 3 basophil donors normalized to their total histamine. Any HR values over 100% were adjusted to 100%. Dotted line indicates cutoff for positive response in BHRA. HR, Histamine release.

As expected, cells sensitized with pairwise combinations of the monoclonal IgEs responded to all 3 insulin types (Fig 3, D-F). The histamine release after pairwise sensitization was approximately average of the individual monoclonal IgEs at equivalent total IgE concentrations.

Cells sensitized with a combination of HUI-001/HUI-018 or HUI-018/S1 IgE did not respond to the monomeric insulin (Fig 3, D and E). Interestingly, cells sensitized with the HUI-001/S1 IgE combination (Fig 3, F) showed a response to monomeric insulin—something that neither HUI-001 or S1 IgE was capable of on its own in equivalent total concentrations.

Histamine release from basophils sensitized with patient sera responds to insulin covalent dimers and aggregated insulin

The effect of dimerization and aggregation of insulin was examined on basophils sensitized with serum from 3 diabetic patients before (predose) and after (postdose) the development of anti-insulin IgE and allergic reactions. Despite development of insulin-specific IgE in all 3 patients (Table II), there was only minimal histamine release just above cutoff to monomeric human insulin when cells were sensitized with postdosing sera (Fig 4, B, D, and F). There were low responses to dimeric insulin in 2 of 3 patients, whereas challenge with the insulin superstructure (small particulates) induced significant responses in all 3 patients.

Fig 4.

BHRA dose–response curve of donor basophils sensitized with p1 at (A) baseline and (B) after dosing, p2 at (C) baseline and (D) after dosing, and p3 at (E) baseline and (F) after dosing. Results depict dose–responses of single basophil donor normalized to total histamine. Any HR values over 100% were adjusted to 100%. Dotted line indicates cutoff for positive response in BHRA. HR, Histamine release.

Sensitization with monoclonal IgE_Ins could induce histamine release from basophils using insulin coupled to Sepharose

The finding that certain insulin superstructures can induce histamine release from monomeric anti-insulin IgE-sensitized basophils prompted us to further explore the effect of presenting insulin in another matrix format by coupling of monomeric insulin (human insulin) molecules to a matrix (ImmunoCAPs) and incubate the CAP insulin with monoclonal anti-insulin IgE (HUI-001, HUI-018, or S1)-sensitized basophils. We found that basophils sensitized with all 3 antibodies alone induced a response when incubated with human insulin CAP (Fig 5). Control experiments using commercially coupled insulin ImmunoCAPs and anti-IgE coupled ImmunoCAPs (Thermo Fisher Scientific) induced histamine release in all sensitization experiments (Fig 5). As expected from the regular BHRA experiments (Fig 3), small particulates also gave a robust response with all 3 monoclonal antibodies (Fig 5).

Fig 5.

HR of basophils sensitized with different monoclonal insulin sIgE challenged with various insulins bound to CAPs in CAP-BHRA system. Donor basophils passively sensitized with 3 different anti-hIns IgEs (HUI-001, HUI-018, and S1) were challenged with different biotinylated insulins coupled to streptavidin CAPs. Supernatants were analyzed for histamine content. Depicted are means and SDs of 3 basophil donor HR responses with baseline removed (streptavidin CAP), after being normalized to their total histamine. Anti-IgE was used as positive control for IgE-mediated basophil degranulation. hIns, Human insulin; HR, histamine release; sIgE, serum-specific IgE.

Monomeric insulin coupled to Sepharose can induce histamine release with postdosing patient sera

When cells were sensitized with postdosing sera, there was a response to ImmunoCAPs coated with both human insulin and small particulates (Fig 6). Only baseline serum from p2 give rise to a small response in the CAP-BHRA to CAP-insulin as well as CAP small particles, which is in line with the observation that p2 had a positive result for insulin-specific IgE already at baseline, which was then boosted after dosing (Table II).

Fig 6.

HR of basophils sensitized with patient sera in presence of various insulins bound to CAPs in CAP-BHRA system. Donor basophils were passively sensitized with (A) serum samples from 3 patients as well as with their (B) initial baseline samples. Basophils were then challenged with different biotinylated insulins coupled to streptavidin CAPs. Supernatants were analyzed for histamine content by BHRA. Depicted is mean of single basophil donor’s HR response with background removed (streptavidin CAP) after being normalized to total histamine. Background HR, as measured by streptavidin CAP, ranged 4.6-13 kU_A/L, while aIgE ranged 31-44 kU_A/L. Anti-IgE was used as positive control for IgE-mediated basophil degranulation. HR, Histamine release.

Discussion

After the advent of insulin analogs, a number of new analogs were approved that differed in various degrees in amino acid sequences from human insulin.¹⁶ Alongside this, there has been growing concern regarding adverse reactions to insulin products, as the advances in drug development that can drastically improve the function but also make the new analogs further distinct from the human insulin, potentially changing their immunogenicity/allergenicity.

A diagnostic supplement to IgE determinations based on cellular tests like the basophil activation test and BHRA has been documented to diagnose clinically relevant allergies.^17,18 The basophil-based tests are also used to diagnose drug-allergic reactions,5, 6, 7 but this is often challenging because many drugs are small molecules that can only cross-link cell-bound IgE when they are in a multimeric form or bound to haptens.⁸ Cellular tests are also affected by IgE affinity and clonality, which might not be reflected in IgE assays.¹⁹ It has also been shown that IgE with a higher affinity for the allergen can induce histamine release at lower concentrations of allergen than an IgE with a lower affinity.^20,21 Further, Christensen et al²⁰ showed that IgE directed against a single epitope did not activate the basophils, whereas a mix of IgEs binding two or more different epitopes could activate the basophils.

It has been reported that up to 40% of diabetic patients without symptoms of insulin allergy show positive skin test results or specific IgE to insulin.⁴ The frequency likely depends on the immunogenicity of the particular insulin drug product, but insulin-specific IgE testing in general seems to have only marginal diagnostic value.²² Possible explanations of the inconsistencies between the relatively frequent presence of specific IgE against insulin and the less frequent allergic reactions to insulin include lack of insulin superstructures presented to the sensitized cells and/or the clonality of IgE directed against different nonoverlapping insulin epitopes not being available as a result of insulin’s small size.

To test the effect of superstructures (aggregates), we measured histamine release from basophils sensitized to humanized monoclonal IgE against different insulin structures. Dimeric insulin was included as the smallest structure potentially able to cross-link two IgE molecules directed against the same unique insulin epitope. Superstructures of insulin were included because they have previously been demonstrated to be immunogenic both in vitro and in vivo.⁹ We included commercially available cellulose-coupled insulin (ImmunoCAP) as a convenient way of simulating insulin aggregation to identify an anti-insulin IgE repertoire capable of reacting to insulin superstructures. The ImmunoCAP system was chosen because we have previously demonstrated that birch allergen Bet v 1 ImmunoCAPs were able to induce histamine release from basophils sensitized with birch-allergic individuals’ sera.¹⁴

Our findings—that both dimeric insulin and the small particulates were able to induce histamine release from basophils sensitized with a well-characterized and specific monoclonal anti-insulin IgE—is in favor of the hypothesis that insulin is more allergenic when presented to cells in a repetitive manner and with a specific set of characteristics. Because there was no differentiation between the two spherulite types, our findings suggest that the ability of superstructures to elicit an allergic response is not dependent on secondary structure but rather is dependent on the superstructure’s size. Moreover, it suggests that there may be an upper size limit, as only the small particulates, not the larger structures, provided a response (Fig 1). This may have clinical implications; first, patients sensitized to insulin are not at risk when presented with insulin monomers, and second, patients sensitized to insulin presented with covalently bound insulin dimers and/or stable insulin superstructures with specific characteristics (which may develop if insulin is improperly stored) may be at risk of developing a hypersensitivity reaction.

Others have shown that the IgE repertoire determines effector cell degranulation in response to allergen challenge.²⁰ We therefore sensitized basophil cells with a combination of two IgE molecules with distinct epitope recognition (Fig 2 and Table I) and challenged with monomeric insulin. Indeed, it was possible to induce histamine release with insulin by combining an IgE with epitope specificity primarily to A chain (HUI-001) and an IgE with specificity to the B chain (S1) (Fig 3, F). Simultaneous binding of this pair of anti-insulin IgE was confirmed by cocrystallization. None of the other two pairs of anti-insulin IgE was able induce basophil activation to monomeric insulin as their epitopes overlap (Fig 3, D and E). This demonstrates that clonality of the IgE repertoire is important for basophil cell activation and indicates that a polyclonal response to both A and B insulin chains can result in a cellular response to two different IgE epitopes on the same insulin molecule. A restricted repertoire of IgE to insulin epitopes could explain why most patients sensitized toward insulin do not experience clinical symptoms, and only in rare cases is a combination of anti-insulin IgE specificities that can elicit histamine release in response to monomeric insulin formed.

Finally, we included sera from 3 patients both before and after the development of insulin allergy. Postdose sera showed no response to monomeric insulin and a small but insignificant response to dimeric insulin in 2 of 3 sera, whereas all sera reacted to the small particulates. This supports our finding that superstructures can activate cells sensitized with IgE with only one specificity. The patient IgE repertoire might be quite narrow: a response can only take place when sensitized cells are presented with the appropriate nano-sized structures. We saw robust responses with individual control antibodies and postdosing patient serum, but virtually no response from baseline (Figs 5 and 6). The CAP-BHRA results demonstrate this hypothesis because monomeric insulin is able to induce histamine release from all 3 postdose sera when insulin is presented in a large structure like cellulose.

To summarize, single anti-insulin IgE monoclonal antibodies and monomeric insulin can, as expected, not induce FcεR1 cross-linking and thereby not induce basophil activation (Fig 7). Increasing the clonality of anti-insulin IgEs increases the likelihood of basophil activation toward monomeric insulin. Utilizing a covalently bound insulin dimer is sufficient to induce basophil activation with both monoclonal and polyclonal anti-insulin IgE and can potentially yield a strong response. However, certain insulin aggregates seem to be the most potent basophil activator, especially in patient sera, compared to monomeric and dimeric insulin.

Fig 7.

Summary of proposed mechanisms. Stronger basophil activation is seen with increase in clonality of anti-insulin IgE and structural complexity of insulin.

In conclusion, this study illustrates why diagnostic methods for insulin allergy are difficult to correlate to clinical symptoms and that certain aggregates of insulin and IgE clonality may be important factors in developing insulin allergy and symptom elicitation. Similar mechanisms could play a role in other allergic reactions to peptide drugs or in storage proteins of legumes.

Key messages.

• •Single anti-insulin IgE monoclonal antibodies and monomeric insulin can, as expected, not induce FcεR1 cross-linking and thereby not induce basophil activation.
• •The structural presentation of insulin seems to be as important as the anti-insulin IgE repertoire of insulin-allergic patients, as can be seen by covalently bound insulin dimers and aggregated insulin particulates being potent basophil activators.
• •The CAP-HR provides a promising platform for investigating the potential of drug aggregates to induce basophil activation.

Disclosure statement

Funded by RefLab ApS and Novo Nordisk A/S.

Disclosure of potential conflict of interest: The authors are employees of either RefLab ApS or Novo Nordisk A/S.