Basophils, high‐affinity IgE receptors, and CCL2 in human anaphylaxis

Background The role of basophils in anaphylaxis is unclear. Objective We sought to investigate whether basophils have an important role in human anaphylaxis. Methods In an emergency department study we recruited 31 patients with acute anaphylaxis, predominantly to Hymenoptera venom. We measured expression of basophil activation markers (CD63 and CD203c); the absolute number of circulating basophils; whole‐blood FCER1A, carboxypeptidase A3 (CPA3), and L‐histidine decarboxylase (HDC) gene expression; and serum markers (CCL2, CCL5, CCL11, IL‐3, and thymic stromal lymphopoietin) at 3 time points (ie, during the anaphylactic episode and in convalescent samples 7 and 30 days later). We recruited 134 patients with Hymenoptera allergy and 76 healthy control subjects for comparison. We then investigated whether the changes observed during venom‐related anaphylaxis also occur during allergic reactions to food in 22 patients with peanut allergy undergoing double‐blind, placebo‐controlled food challenge to peanut. Results The number of circulating basophils was significantly lower during anaphylaxis (median, 3.5 cells/&mgr;L) than 7 and 30 days later (17.5 and 24.7 cells/&mgr;L, P < .0001) and compared with those in patients with venom allergy and healthy control subjects (21 and 23.4 cells/&mgr;L, P < .0001). FCER1A expression during anaphylaxis was also significantly lower than in convalescent samples (P ≤ .002) and control subjects with venom allergy (P < .0001). CCL2 levels (but not those of other serum markers) were significantly higher during anaphylaxis (median, 658 pg/mL) than in convalescent samples (314 and 311 pg/mL at 7 and 30 days, P < .001). Peanut‐induced allergic reactions resulted in a significant decrease in circulating basophil counts compared with those in prechallenge samples (P = .016), a decrease in FCER1A expression (P = .007), and an increase in CCL2 levels (P = .003). Conclusions Our findings imply an important and specific role for basophils in the pathophysiology of human anaphylaxis.


INTRODUCTION
Anaphylaxis is a potentially life-threatening, rapidly-progressing systemic allergic reaction that may lead to death due to airway obstruction or vascular collapse following exposure to allergens (including insect venom, foods and medication). 1 Mast cell activation is postulated to have a pivotal role in anaphylaxis, 2 and an increase in serum mast cell tryptase can confirm the diagnosis. 1 However, in individuals experiencing anaphylaxis, it is not unusual to find normal serum tryptase in the context of increased plasma histamine, [3][4][5] suggesting that anaphylaxis may also involve basophil activation. However, there is little published data demonstrating a direct contribution of basophils to IgE-mediated anaphylaxis in humans.
Mast cells enter tissues as immature progenitors, where they undergo the final stages of their development and remain resident in-situ for weeks/months. In contrast, basophils typically mature in hematopoietic tissues and subsequently circulate in the blood, with a half-life of less than one week. 6 Local allergen challenge studies in humans have demonstrated an influx of basophils to inflammatory sites within several hours of allergen exposure, demonstrating the existence of mechanisms for basophil recruitment from the circulation to the site of allergen exposure. [7][8][9] Both mast cells and basophils may rapidly secrete histamine and similar (but not necessarily identical) mediators and cytokines following IgE cross-linking. 2 In murine studies, basophils contribute to IgG-mediated anaphylaxis. 10 In contrast, human basophils cannot be activated through IgG receptors, and their function is inhibited by IgGmediated triggering via FcγRIIb receptors; moreover, they lack protease-activated receptors and antigen-presenting functions. 11,12 We hypothesized that basophils play an important role in human anaphylaxis, and specifically that: (1) basophils are activated during human anaphylaxis; (2) there is a basophil migration during anaphylaxis; and (3) basophil-related biomarkers may be useful to confirm anaphylaxis. We addressed our hypotheses in a series of inter-linked studies. First, in an emergency department (ED) study we investigated the up-regulation of CD63 expression (the most commonly used basophil activation marker 13 ) during and after anaphylaxis (predominantly caused by hymenoptera venom allergy). We monitored the absolute numbers of circulating basophils, the corresponding whole blood gene expression of FcεRI, carboxypeptidase A3 (CPA3) and L-histidine decarboxylase (HDC), and serum levels of the major basophil chemotactic factors, including the CCR2 ligand CCL2, and the CCR3 ligands CCL11 and CCL5. 14, 15 We also measured T cell-derived IL-3 (an important basophil priming and growth factor), and epithelial cell-derived thymic stromal lymphopoietin (TSLP) which promotes IL-3-independent basophil development and activation. 6,16,17 We then proceeded to assess whether the changes seen during venom-related anaphylaxis also occur during allergic reactions to food under the controlled setting of a double-blind placebo-controlled oral food challenge (DBPCFC) in peanut-allergic individuals.

Study participants
Emergency Department (ED) study: We prospectively recruited 31 patients (13 female, age 18-79 years) presenting with an acute episode of anaphylaxis to the ED of the University Hospital Golnik, Slovenia (June-August 2011; July-November 2013). Severity of reactions was graded according to Mueller criteria. 18 We collected blood samples during the reaction (at presentation to the ED) and in convalescent samples seven and/or 30 days after the anaphylactic episode (Table E1).

Hymenoptera venom allergic controls and healthy subjects:
We recruited two groups of control participants for comparisons: (1) 134 patients (49 females, age 23-67 years) with confirmed venom anaphylaxis from whom blood samples were obtained at least two months after the last sting reaction, and prior to initiation of venom immunotherapy; and (2) 76 healthy controls (47 females, age 17-79 years).
To assess for possible confounding by treatment with corticosteroids and its effect on basophil activation, absolute cell count, FcRI expression and soluble markers, 17 healthy subjects received a single dose of 64 mg of oral methylprednisolone and were monitored for up to 24 hours after the treatment (Table E2).
Peanut allergy study: We recruited 22 peanut-allergic individuals (Table E3) in whom peanut allergy was confirmed by the DBPCFC (details in the Online supplement). Blood samples were collected prior to challenge, at cessation of challenge due to the onset of objective symptoms 19 (but prior to administration of any treatment), and 2-4 hours post-challenge.
Ethical approval was obtained from the Slovenian National Medical Ethics Committee (ED study and control participants), and the London Central Research Ethics Committee (peanut allergy study). All subjects provided written informed consent.

Statistical analysis
The distribution of data was assessed using the D'Agostino and Pearson test. We used appropriate non-parametic and parametric tests for comparisons between the groups, including Wilcoxon's signed-rank test, Mann-Whitney U-test, t-test with a Welch correction and Pearson correlation. Data are expressed as the median unless otherwise stated. We compared the performance of basophil-related biomarkers in discriminating between patients with anaphylactic reactions and those without using receiver operating characteristic (ROC) curve analysis. Analyses were performed using GraphPad Prism (GraphPad Software, La Jolla, CA, USA).

Study participants
ED study and controls: Table E1 and Figure E1show detailed information on demographic characteristics, clinical and emergency treatment and sampling data of 31 ED patients. The reaction was caused by an insect sting in 28 patients. The median time from the onset of symptoms to sample collection was 105 minutes (range 20 minutes to 5 hours, Figure E1). Convalescent samples were collected from 28 patients seven days after the anaphylactic episode, and from 23 patients after 30 days (Table E1); two patients provided samples 24 hours after the acute episode.
We measured basophil activation and counts in all ED patients and controls, and serum tryptase in all ED patients and venom-allergic controls (Table E4). We ascertained gene expression in 15, chemokines and IL-3 in 17, and TSLP in 14 ED patients, and analyzed FcεRI expression in 37 venom-allergic controls, and CCL2 in 71 healthy controls (Table E4).
Peanut allergy study: Basophil counts were determined in 22 peanut allergic patients prior to, and during both active and placebo arms of the DBPCFC. CCL2 levels (n=22) and FcεRI expression (n=12) were ascertained during the active arm of the DBPCFC.

Basophil markers in ED study and controls
Basophil activation: The percentage of CD63-activated basophils in ED patients during anaphylactic episodes was low (median 3.8%). These values were marginally higher compared to seven (median 2.9%; P=0.01) and 30 days later (median 2.9%, Fig. 1A; P=0.05).
Only four patients had >5% activated basophils, and only one exhibited an activation of >10%. This was mirrored by a small, but significantly higher percentage of CD63-activated basophils during anaphylaxis compared to venom-allergic controls (median 3.1%, P=0.01), or healthy controls (median 2.4%, P=0.001, Fig. 2A). Expression of the activation marker CD203c correlated highly with that of CD63 (Fig. E2).
Using a binary cut-off of 11.4 µg/L, tryptase was increased during the acute episode in 22/31   Serum CCL2 levels: CCL2 increased significantly at the time of objective symptoms during the active arm of the DBPCFC compared to baseline levels (P=0.003, Figure 6A). CCL2 levels returned to baseline within two hours of the onset of symptoms (Fig. 6B); the rate of increase in CCL2 was significantly greater in the active compared to the placebo arm of the DBPCFC (P=0.008; Fig. 6B).

Predictors of anaphylaxis
As indicated by the estimated area under the ROC curve (AUROC), CCL2 and FcεRI expressions were the most accurate readouts in discriminating between patients with anaphylactic reactions from those without, followed by basophil counts and tryptase levels: (for further details, see Online supplement). With a cut-off of >334 pg/µL, the estimated sensitivity and specificity of CCL measurements were 94% and 96% respectively, compared with 93% and 92% for FcεRI expression (cut-off<0.2) and 87% and 81% for basophil counts (cut-off>12 cells/µL).

DISCUSSION
Our study demonstrated a substantial (~80%) reduction in circulating basophils during anaphylactic reactions to hymenoptera venom. Decreased gene expression of FcεRI, CPA3 and HDC confirmed the flow cytometry data. We also observed an increase in CCL2, which correlated with a decrease in circulating basophils. We replicated these findings in peanutallergic individuals experiencing allergic reactions during DBPCFC to peanut. Compared to the reactions in the emergency department, which were generally more severe, we observed more modest (but nonetheless significant) changes at the time of objective symptoms during the peanut challenges. Taken together, these data suggest that anaphylaxis induces a rapid and considerable basophil migration. The mechanism of anaphylaxis-related basophil migration appears to be selective, because no significant changes were seen for lymphocytes, PMNs, or chemotactic factors which may affect other effector cells such as eosinophils (e.g. CCL5 and CCL11).

Limitations
The nature of the management of anaphylaxis (including administration of high-dose corticosteroids) makes it difficult to exclude the potential confounding by treatment and draw an unequivocal interpretation of the decrease in basophils in the ED setting. In our ED study, 94% of patients received methylprednisolone, and 42% epinephrine. Corticosteroids have a well-described effect on blood leucocytes, including an increase in circulating neutrophils and decrease in lymphocytes and basophils. 24,25 The kinetics of the response of various leukocytes to corticosteroid administration varies, with neutrophilia and lymphopenia preceding the onset of basopenia, 25 which was confirmed in our study. Compared to healthy controls who received oral corticosteroids, the reduction in blood basophils (but not lymphocytes or PMNs) was much greater and occurred at an earlier time in patients with acute anaphylaxis, suggesting that the changes in basophils were not related to treatment. Moreover, we replicated the observed changes in basophil markers in the controlled setting of peanut-allergic individuals undergoing DBPCFC, where the study design allowed for blood sampling both prior to challenge and before any treatment. This avoids the issue of confounding by treatment (both with corticosteroids and epinephrine), and allows comparison with pre-reaction samples (something not possible in the ED setting). We acknowledge that two previous reports failed to detect a change in absolute basophil counts following food challenge. 26,27 However, these studies involved fewer patients experiencing only mild allergic symptoms, and used methods for basophil detection less sensitive and specific than that employed in our study.
Several cytokines and chemokines are involved in basophil migration, with the CCR2 ligand CCL2 and the CCR3 ligand CCL11 eliciting the most potent migratory responses. 15 However, there is a difference in the cellular specificity of these chemokines. CCR2 is virtually undetectable on human eosinophils 28 , and thus CCL2 fails to induce eosinophil migration, which is not the case for the CCR3 ligands CCL5 and CCL11. 29 Therefore, CCL2-mediated migration may represent a unique mechanism for the selective migration of human basophils in allergic reactions. However, in the present study we could not determine the cellular sources of CCL2 during acute reactions.
We could not answer the question of whether anaphylaxis is associated with extensive activation and degranulation of circulating basophils. Patients with anaphylaxis present to ED up to hours after onset of symptoms, and it takes additional time to obtain informed consent and perform venipuncture. In our study, the median time between the onset of symptoms and sample collection was 105 minutes, which is comparable to previous ED studies. 4,30,31 Plasma histamine levels, which correlate with anaphylactic symptoms, 32,33 typically peak within 5-10 minutes after the onset of anaphylaxis and subsequently decrease to baseline within one hour as a result of rapid catabolism. Consequently, the relatively modest increase in CD63 expression on basophils (a marker of basophil degranulation) may represent an underestimate of the peak basophil activation during acute reactions. In a recent open food challenge study of delayed responses to meat in patients sensitized to galactose-alpha-1,3-galactose, expression of CD63 on was reported for >15% of basophils in 9/12 patients at the onset of symptoms. 34 This is consistent with our data, which also supports more extensive basophil activation (typically up to 20% of basophils expressing CD63 and CD203c) during peanutallergic reactions. 35 In our ED study, only one of 31 predominantly venom-allergic patients had >15% CD63-activated basophils, despite the fact that the majority (24/31) experienced anaphylactic reactions of Mueller grade III-IV severity (with bronchospasm, airway obstruction, hypoxemia or hypotension, and collapse). Whether this difference is due to the unavoidable delay in sampling following onset of symptoms in the ED compared to the challenge setting, or a difference in the extent of basophil activation for venom versus foodinduced allergic reaction, is unknown. It is most likely that we detected only those basophils that remained in the circulation following the acute reaction (approximately 20% of the normal level of basophils), and not the basophils that had migrated out of the circulation.

Interpretation
Recent reports have implicated a specific effector role for basophils in acute allergic responses. 21,[36][37][38] Studies which used oral food or nasal allergen challenge responses in omalizumab-treated adults with peanut 37 or cat allergies 36 have suggested that acute reactions may be basophil, rather than mast cell, dependent. Decreases in the basophil allergen responses following venom immunotherapy reflect the induction of tolerance to sting challenges. 21 A recent study in peanut allergic children suggested that an in vitro basophil activation test at baseline may correlate with reaction severity at subsequent food challenge. 38 However, these in vitro studies could not confirm whether basophil activation actually contributes to the acute allergic reactions, or is a surrogate marker of mast cell or overall IgE responsiveness. Thus, studies investigating human basophils during allergic reactions in vivo are required. However, such studies in a controlled challenge setting are difficult due to the general consensus that patients who may experience severe anaphylactic reactions should be excluded. Moreover, reaction severity at challenge is generally limited by the controlled nature of the challenge (where allergen exposure is stopped at onset of objective symptoms).
We therefore combined an ED-based study in venom allergy, which focused on basophil migration and/or activation during more severe anaphylaxis, with a study of peanut-allergic reactions during DBPCFC in which patients tended to experience less severe reactions. Data from this latter study in peanut-allergic subjects corroborated the findings from the ED study.
One interesting question which remains unanswered is when and where basophil activation occurs. Anti-IgE, anti-FcRI or allergen stimulation of basophils also promote their migration and adherence to endothelial cells. 39,40 However, these stimuli may enhance basophil adherence to the vascular endothelium and migration at concentrations which are lower than the threshold required for basophil degranulation and histamine release. 39,40 Therefore, IgEmediated basophil migration may be induced without basophil degranulation. This suggests that basophils may be activated after migration, or partly in circulation and partly after migration, or may even migrate without activation. The different clinical severities and endorgan patterns of anaphylaxis 1,2 and the finding that serum mast cell tryptase is often within normal limits 3,4 suggest that local rather than generalized mast cell and/or basophil degranulation may predominate in some individuals. Additional studies are required to confirm these speculations.
The short time frame within which the reduction in circulating basophils occurred, coupled with previous findings that basophils are the granulocytes most resistant to apoptosis, 41 suggest that anaphylaxis induces a prompt basophil migration rather than elimination by apoptosis. We did not observe a change in serum IL-3 or TSLP. This suggests that it is unlikely that basophil migration during anaphylaxis is related to changes in basophil development or homeostasis, a process which is IL-3-elicited for basophils that operate in an IgE-dependent manner, or TSLP-elicited for basophils that operate in a non-IgE-dependent manner. 6 Our results are consistent with a recent study which demonstrated no changes in CCL11 or IL-3 during anaphylaxis. 30 Risk assessment of individuals with anaphylaxis is hampered by limitations in laboratory tests to confirm the diagnosis, and predict its severity. 42,43 Currently, the only readily available laboratory test to confirm the diagnosis of anaphylaxis is the measurement of total tryptase in serum/plasma. 1,2 However, even when blood sampling is optimally timed, tryptase levels are often within the normal limits, particularly for food-induced reactions. 3,4 In our study of predominantly venom-induced reactions, a diagnostic increase in the total tryptase was seen in 71% of the individuals with anaphylaxis, which is comparable to other reports. 30 While other mediators have been proposed as potential biomarkers, 30,31,44-46 these have not exhibited sufficient diagnostic utility or technical reproducibility to be routinely used. 1,2 Our results indicate that CCL2, FcεRI expression and basophil counts may potentially be useful biomarkers of anaphylaxis. However, a substantially broader assessment is required to validate these methods and replicate the findings.

Conclusions
Our data suggest a substantial migration of circulating basophils during anaphylaxis, which correlates with a significant increase in serum concentration of the major basophil chemotactic factor CCL2. These findings suggest an important and specific role for basophils in the pathophysiology of human anaphylaxis.

DBPCFC to peanut
DBPCFC were conducted according international consensus criteria (PRACTALL) 1 . In brief, subjects underwent double blind placebo controlled peanut challenge (DBPCPC) over two separate days, at least 7 days apart. On each day, subjects received increasing doses, every 30 minutes, of peanut protein (or placebo) at the following doses: 3mg, 10mg, 30mg, 100mg, 300mg, 1000mg and 3000mg until stopping criteria were met (as per PRACTALL consensus 1 ).
Blood samples were collected from a venous cannula sited prior to challenge, and immediately snap-frozen or transferred without delay for flow cytometry.

IL-3 spiking experiments
For IL-3 measurements we performed spiking experiments with E. coli-derived recombinant human IL-3 protein (from R&D Systems) in which a known amount of recombinant protein was spiked into a sera sample with undetectable intrinsic IL-3 concentration (thus bellow 17 pg/mL according to our detection limit) and run in the ELISA. We successfully recovered samples spiked with 250, 125, 62.5 or 32.5 pg/mL of recombinant human IL-3 protein, but not samples spiked with known concentrations of 15.6, 7.8 or 3.9 pg/mL recombinant human IL-3 protein.
This sensitivity is within the range of the minimum detectable concentration of IL-3 (from 3.46-57.4 pg/mL) evaluated by the commercial kit manufacturer (R&D Systems; Human IL-13 Quantikine ELISA Kit).

Inter-assay coefficient of variation
We estimated an inter-assay coefficient of variation of 6.7% for the absolute basophil count and 4.8% for basophil CD63 activation by repeated measurements in five healthy control subjects.

The effect of oral corticosteroid on basophil markers and other blood cells
We followed 17 healthy subjects up to 24 hours after a single dose of 64 mg of oral methylprednisolone (Table E2).
Basophil activation: There was no significant effect of the treatment with oral corticosteroids on basophil (CD63) activation ( Figure E4A).

Circulating basophils:
We identified a small, but statistically significant decrease in the absolute number of blood basophils (from a median of 23.4 to 19.7 cells/µL; median decrease 19%, P=0.006). However, a major decrease (to 8 cells/µL, median decrease 67%, P=0.004) was observed 5 hours after methylprednisolone administration (Fig. E4B). Basophils numbers returned to normal values after 24 hours (to 22 cells/µL).

Gene expression:
We observed a small and non-significant decrease FcεRI expression 2.5-3 hours after methylprednisolone intake, followed by a substantial decrease after 5 hours which corresponded to a major decrease in basophils (median decrease 63%, P=0.006, Fig. E4C). FcεRI expression did not differ between the baseline level and 24 hours after methylprednisolone.
After 2.5-3 hours, there was a small, but statistically significant decrease in the number of blood lymphocytes (median 960 to 768 cells/µL, P=0.004, Fig. 4E). There was no difference in lymphocyte counts 5 and 24 hours after methylprednisolone compared to baseline (Fig. E4E).

Serum markers:
There was no significant effect of the treatment with oral corticosteroids on CCL2, CCL5, CCL11 or IL-3 ( Fig. E5A-D).
ED patients: In two ED patients (No. 14 and 15, Table E1) in whom we collected samples during the acute anaphylactic episode and 24 hours later, and who received emergency treatment with systemic corticosteroids, during the acute allergic reaction we observed changes in basophils, CCL2 and tryptase, but not in PMNs and lymphocytes (Fig E6). The increase in the PMNs and the decrease in the lymphocytes became evident only at the 24-hour sampling point (Fig. E6).
The decrease in basophil count and FcεRI expression, as well as the increase in tryptase and CCL2 level were also observed in two ED patients (No. 20 and 29, Table E1) who did not receive treatment with corticosteroids (Fig. E7).

Predictors of anaphylactic reactions
We compared the performance of basophil counts, basophil activation, tryptase levels, as well as CCL2 and FcεRI expressions in discriminating between patients with anaphylactic reactions and those without using a ROC curve analysis. For the control groups, we used the patients with confirmed venom allergy from whom samples were obtained at least two months after the last sting reaction, and before venom immunotherapy was initiated (134 controls for basophil counts, 13 basophil activation and tryptase level, and 37 controls for FcεRI expression) or healthy controls (54 controls for CCL2).
When         Figure E7. Basophil absolute count, whole blood FcεRI gene expression, serum tryptase and CCL2 serum concentration during the acute anaphylactic reactions to hymenoptera venom, and 7 and 30 days after the anaphylactic episode in ED patients divided according methylprednisolone treatment (patients No. 20 and 29 were not treated with methylprednisolone; Table E1). Data are presented as a person-to-person scatter plot.