A Soluble Form of the High Affinity IgE Receptor, Fc-Epsilon-RI, Circulates in Human Serum

Soluble IgE receptors are potential in vivo modulators of IgE-mediated immune responses and are thus important for our basic understanding of allergic responses. We here characterize a novel soluble version of the IgE-binding alpha-chain of Fc-epsilon-RI (sFcεRI), the high affinity receptor for IgE. sFcεRI immunoprecipitates as a protein of ∼40 kDa and contains an intact IgE-binding site. In human serum, sFcεRI is found as a soluble free IgE receptor as well as a complex with IgE. Using a newly established ELISA, we show that serum sFcεRI levels correlate with serum IgE in patients with elevated IgE. We also show that serum of individuals with normal IgE levels can be found to contain high levels of sFcεRI. After IgE-antigen-mediated crosslinking of surface FcεRI, we detect sFcεRI in the exosome-depleted, soluble fraction of cell culture supernatants. We further show that sFcεRI can block binding of IgE to FcεRI expressed at the cell surface. In summary, we here describe the alpha-chain of FcεRI as a circulating soluble IgE receptor isoform in human serum.


Introduction
Allergic patients are commonly characterized by high serum IgE and high IgE-receptor expression on effector cells of the innate and adaptive immune system [1,2]. In humans, three different IgE-receptors have been described: CD23, galectin-3 and FceRI [1,2]. CD23, also known as FceRII, is a low affinity IgE receptor and the classical IgE receptor on B cells. Galectin-3, formerly known as epsilon binding protein (eBP), is another low affinity IgE receptor; its role in allergy is rather poorly defined [3,4]. FceRI, the high-affinity receptor for IgE, induces activation of mast cells and basophils via IgE-antigen complexes during the acute phase of an allergic response [5,6]. In rodents, FceRI is constitutively expressed on the surface of basophils and mast cells as a tetrameric receptor composed of the ligand-binding alpha-chain, one betachain and a pair of disulphide-linked gamma-chains. Humans can express a trimeric version of FceRI lacking the beta-chain on eosinophils and antigen presenting cells, such as dendritic cells and Langerhans cells [6,7]. Additionally, expression of FceRI on bronchial and intestinal epithelial cells was described in humans [8,9]. Serum IgE binding stabilizes surface FceRI leading to the upregulation of receptor levels in allergic patients [10,11,12].
In addition to the transmembrane forms, CD23 and galectin-3 are found as soluble proteins in human serum. Soluble CD23 (sCD23) is a modulator of IgE responses in vivo and is generated by cleavage of membrane CD23 from the surface of B-cells [13]. sCD23 has been demonstrated to enhance IgE production [14,15,16] and several reports show that high serum levels of sCD23 correlate directly with the severity of allergy and asthma [17]. Along this line, successful immune therapy is accompanied by a drop in sCD23 levels in the serum of allergic patients [18]. The role of sCD23 in modulating IgE production and its potential for monitoring allergic responses has been discussed for more than two decades [13,19,20]. However, sCD23 is currently approved as a prognostic parameter only for B-cell chronic lymphocytic leukemia [21,22,23]. Interestingly, soluble galactin-3 is also a common marker for tumor burden [4,24]. Why the production of these soluble IgE receptors is induced during malignant diseases is an interesting scientific question that has yet to be resolved. Thus, our limited understanding of the in vivo role of sCD23 and soluble galectin-3 highlights the need for continued research on soluble factors that modulate serum IgE responses in the context of an allergic response.
FceRI is an activating immune receptor of the immunoglobulin superfamily, which includes the Fc receptors CD16, CD32, CD64 and CD89 [6,25,26]. FceRI shares key structural characteristics and signaling features with these Fc receptors. For most IgE, IgG and IgA Fc receptors, soluble isoforms are found in humans. FceRI, however, has so far not been reported as a soluble IgE receptor in human serum [1,6].
Here we describe a soluble form of the FceRI alpha-chain (sFceRI). In human serum, this sFceRI is found as both a free form and bound to its ligand IgE. We show that IgE-mediated cell activation induces the release of sFceRI in vitro and that the soluble form of the receptor can inhibit binding of IgE to FceRI at the cell surface.

Detection of a soluble form of FceRI alpha (sFceRI) in human serum
To give a definitive answer whether a soluble form of the alpha chain of FceRI exists in humans, we performed immunoprecipitation experiments to isolate this protein from serum. Sera from patients with normal IgE levels and elevated IgE were run over IgE-columns. Eluates from these columns were analyzed with the FceRI alpha-chain specific mAb 19-1 by Western blot [12]. The IgE used for these precipitation is commonly used for detection of FceRI [12,27] and has a chimeric immunoglobulin containing the human IgE heavy chain and a murine Fab-anti NP fragment (referred to as cIgE from here on). Columns were prepared by coupling cIgE to NP sepharose. 10 ml serum was run over a gravity column packed with 0.5 ml beads. Figure 1A shows a representative positive (right lane) and a negative serum (left lane). A soluble form of FceRI-alpha (sFceRI) was precipitated as a protein of ,40 kDa ( Figure 1A). The higher molecular weight bands of the Western blot shown in Figure 1A ($130 kDa) are a result of cross-reactivity of the secondary anti-mouse antibody used for immunoblotting and the precipitating cIgE. Only the lowmolecular weight protein from the serum precipitate is recognized specifically by the anti-FceRI alpha-chain specific mAb 19-1. Since this antibody recognizes the IgE binding epitope of FceRI alpha [12,28], these data show a soluble non-IgE bound form of the receptor in human serum. When the mAb 19-1 was replaced with an isotype control antibody, the sFceRI band was no longer detected (data not shown). Individuals with normal to moderately elevated IgE levels tested strongest positive in the immunoprecipitation assay. In such sera the sFceRI is likely still available for precipitation, whereas in patients with elevated IgE the soluble receptor is mainly bound to serum IgE and therefore cannot be immunoprecipated by the mAb19-1.
To perform a more detailed molecular characterization of sFceRI, we next compared sFceRI-alpha precipitated from serum to FceRI-alpha precipitated from the cell surface of MelJuso-ac cells. As expected for a soluble form, sFceRI has a lower molecular weight than the surface expressed protein. Unlike transmembrane FceRI-alpha, which forms a multimeric complex with the common FcR-gamma chain (also called FceRI-gamma) [1,6], sFceRI was not associated with FceR-gamma ( Figure 1B). This finding confirms that sFceRI is likely a soluble serum protein that is distinct from the membrane multimeric form of the receptor [6,28].
In summary, this set of results show that the Fc-portion of human IgE can interact with a soluble alpha-chain protein in serum and that this serum sFceRI does not have the molecular characteristics of the multimeric membrane-associated FceRI.

Detection of sFceRI in human serum by ELISA
To allow for semi-quantitative analysis of sFceRI levels in human serum, we next established a sandwich-ELISA system (schematic model in Figure 2A). For this ELISA, the anti-FceRIalpha mAb CRA1 was used as the capture antibody. This mAb binds the stalk region of the alpha-chain and is expected to capture a serum alpha-chain without interfering with the IgE-binding epitope [29]. Levels of serum sFceRI dropped when comparing pre-and post-immunoprecipitation samples ( Figure 2B), confirming the specificity of our ELISA. As a further control, capture and detection antibodies were omitted, which consistently resulted in a loss of sFceRI signal (data not shown). We were also able to detect sFceRI in plasma with this ELISA, (data not shown). Using a small collective of atopic pediatric patients (5 boys, 3 girls, mean age 10.3+/22.7 years; detailed patient characteristics are found in Table S1), we established that this ELISA is a feasible method for detection of sFceRI in a larger patient set. In summary, this set of data describes a novel ELISA for the detection of serum sFceRI. Any conclusions about the clinical relevance of this finding are however not possible based on the small patient collective.

Serum levels of sFceRI correlate with serum IgE levels in patients with elevated IgE
Due to the absence of a recombinant sFceRI protein for the generation of a standard curve, the specific, blanked OD was used for semi-quantitative analysis of serum sFceRI levels. To investigate the occurrence of sFceRI and potential associations Figure 1. A soluble form of the high affinity IgE receptor, FceRI, is found in human serum. A. Immunoprecipitations from a negative (first lane, Serum 1) and a positive (right lane, Serum 2) serum specimens. Soluble FceRI (sFceRI) was precipitated from serum with IgE-loaded NIP-beads and eluted with non-reducing Laemmli sample buffer. Eluates were separated on 12% non-reducing SDS-PAGE gels, transferred to PVDF membranes and probed with anti-FceRI-alpha mAb 19-1 followed by peroxidase (HRP)-conjugated goat-anti-mouse IgG for detection of precipitated a-chain. B. Comparison of sFceRI from serum (upper left blot) with FceRI precipitated from the cell surface of MelJuso-ac cells (upper right blot). In the low molecular weight range, blots were re-probed with an anti-FceRI-gamma polyclonal serum. sFceRI does not associate with the gamma chain (lower panel). Molecular weight is given in kDa. doi:10.1371/journal.pone.0019098.g001 with serum IgE in pediatric patients, we screened sera from a cohort of 119 children (56 boys, 63 girls, mean age 10.6+/25.3 years) with a wide range of normal and elevated serum IgE levels. Patients were categorized based on the specifications given in the Materials and Methods section into individuals with normal or elevated IgE. We found a weak correlation between serum IgE and sFceRI in patients with elevated IgE (n = 32, rho = 0.291, p = 0.106, Spearman's rank correlation, Figure 2C). In children with normal IgE levels, sFceRI was also found, but no correlation to serum IgE was detected (n = 87, rho = 0.02, p = 0.874, Spearman's rank correlation, Figure 2D). Interestingly, patients with highest levels of sFceRI did not show elevated serum IgE levels ( Figure 2D). In an independent experiment, we confirmed that sFceRI itself did not interfere with detection of IgE and vice versa (data not shown). One limitation of the analyzed patient cohort is that it does not contain healthy individuals, because truly healthy children do not undergo upper GI tract endoscopy for diagnostic purposes. Our patients are controls with regards to the noninflamed esophageal tissue, but show clinical symptoms of yet unclassified nature. For conclusions on the clinical relevance of our findings it is therefore important to study sFceRI in sera from truly healthy controls.

sFceRI circulates as a free or an IgE-complexed protein in human serum
The alpha-chain of FceRI has a high-affinity-binding site for IgE [6]. It is thus likely that sFceRI exists as a preformed complex with IgE in human serum. By omitting the IgE incubation step in our ELISA and detection with the anti-human IgE-HRP conjugate, our method allowed for the detection of sFceRI-IgE complexes in serum. Subtracting the signal without the in vitro IgE incubation step from the signal with the IgE incubation step allowed us to determine how much sFceRI was complexed to IgE in vivo. We randomly selected 14 sera that were positive for sFceRI and found that in human serum, sFceRI is present as both a free and an IgE-bound protein ( Figure 2E).

IgE-antigen-mediated receptor crosslinking induces the production of sFceRI from an FceRI-expressing cell line
The mechanism of sFceRI production cannot be studied using primary human cells due to limited acess to patient material. We thus took advantage of a recently established cell line that allows studying the function of trimeric FceRI in vitro. This new cell line model is based on MelJuso cells, which were stably transfected with FceRI-alpha and FceRI-gamma cDNA. The resulting MelJuso-ac cells express FceRI-alpha at the cell surface ( Figure 3A) and can bind monomeric IgE ( Figure 3B, left FACS histogram). In line with previous reports, IgE binding to MelJusoac also induces upregulation of surface FceRI-alpha, a key feature of this FC receptor ( Figure 3B, right FACS histogram). Multimeric FceRI complexes containing FceRI-alpha and FceRI-gamma subunits can be precipitated from this cell line ( Figure 3C) and receptor activation by crosslinking of FceRI induces efficient receptor internalization from the cell surface ( Figure 3D). All of these features match the characteristics of trimeric human FceRI found in the literature [10,11,12]. Thus, we used this cell model to address whether IgE-mediated activation of cell-surface FceRI induces the release of the soluble form of the receptor. MelJuso-ac cells were loaded overnight with hapten-specific cIgE. After removal of excess cIgE, surface FceRI was activated by crosslinking the receptor-bound ligand with haptenized antigen. 36 h after receptor crosslinking, sFceRI was precipitated from culture B. ELISA measurements pre-and post-immunoprecipitation with IgE-loaded NIP-beads confirmed that IgE immunoprecipitation depleted serum of sFceRI. OD, optical density at 450 nm. C. In children with elevated IgE-levels, levels of sFceRI and total IgE levels correlate. D. In children with normal serum IgElevels, sFceRI could be detected, but no correlation with total IgE levels was found. E. sFceRI circulates as a free or an IgE-complexed protein in human serum. By omitting the IgE-loading step in the ELISA protocol, circulating complexes of IgE and sFceRI were measured. The fraction of free sFceRI was then calculated as OD(total sFceRI)2OD(IgE-sFceRI complexes) = OD (free sFceRI). Graph displays the 14 patients with the highest OD (total sFceRI) with an arbitrary cut off of .0.15. doi:10.1371/journal.pone.0019098.g002 supernatants with a cIgE column and visualized by immunoblotting with mAb 19-1 and compared to sFceRI precipitated from patient serum ( Figure 3E). Next, the kinetics of sFceRI release was studied by harvesting supernatants 4, 8, 24 and 32 h after receptor crosslinking for analysis by ELISA. Accumulation of sFceRI was observed only after receptor crosslinking ( Figure 3F, left graph). sFceRI was not detected in supernatants of empty vectortransfected MelJusoØØ cells that do not express FceRI ( Figure 3F, right graph). To demonstrate that the detected protein was a soluble form of FceRI and not protein shedded with exosomes or derived from cell debris, sequential high-speed ultracentrifugation was performed to deplete the supernatants from cell debris and exosomes as established by Thery et al. [30]. sFceRI was detected in the exosome-depleted, soluble fraction after high-speed centrifugation confirming that the detected protein is a bona fide soluble form of the receptor ( Figure 3G).

sFceRI inhibits IgE loading of FceRI at the cell surface in vitro
Since we detected IgE-sFceRI complexes in serum, we speculated that sFceRI could interfere with IgE-binding to FceRI when expressed at the cell surface. If that indeed occurs, sFceRI could function as a potential modulator of IgE-mediated immune activation. We tested this hypothesis by loading FceRI-expressing MelJuso-ac cells with either a mix of cIgE and cell-culture derived sFceRI or with cIgE diluted with medium control. Cell-bound cIgE was visualized by flow cytometry with PE-conjugated hapten NP. sFceRI efficiently blocked binding of cIgE to FceRI expressed at the cell surface ( Figure 4A). Binding of cIgE was blocked in a dose dependent manner as dilution curves with supernatants from sFceRI-containing MelJuso-ac cells and control supernatants from inactivated cells demonstrated ( Figure 4B). In summary, these results show that sFceRI can interfere with binding of IgE to FceRI at the cell surface of immune cells.

Discussion
We here describe a soluble version of the FceRI-alpha (sFceRI) chain that circulates in human serum as a free protein or bound to its natural ligand, IgE. We show that sFceRI is released upon IgEantigen-mediated activation of cell surface FceRI in vitro and, maybe most interestingly, that sFceRI interferes with IgE-binding to cellular FceRI in vitro. The affinity of IgE with its high affinity receptor FceRI was defined after crystallization of the ligand with recombinant version of the alpha chain [31,32]. It is therefore highly likely that the soluble form in human serum has equally high affinity as described in the literature.
Commonly, the reagents used to detect transmembrane forms of FceRI-alpha are directed against the IgE-binding epitope of the protein. Thus, the identification of sFceRI could easily have been missed if the detection reagents were not selected carefully. We here established an ELISA system that uses a monoclonal antibody directed against the stalk region of the protein [29] to capture sFceRI and use human IgE combined with anti-IgE for detection [12]. By omitting the IgE incubation step, this ELISA also allows for an assessment of the amount of sFceRI that circulates as a preformed complex with serum IgE.
Several studies with recombinant versions of sFceRI are found in the literature [33,34]. Since the recombinant sFceRI used as a tool to interfere with allergic responses and a potential therapeutic agent, there have been some speculations about a soluble serum equivalent in the literature [8]. A single report is found in the literature that described a soluble complex of FceRI in cultures of human eosinophils [35]. Since the integrity of FceRI complexes requires the presence of cell membranes [28,36], Seminario et al. most likely described a version of the receptor that was released in an exosomal fraction rather than a bona fide soluble protein.
Based on our current understanding of the mechanism of sFceRI generation, it is fair to assume that serum sFceRI is a reflection of FceRI activation. In an independent study, we were able to confirm the observation of Liang et al. [37] showing that patients can carry substantial amounts of IgE on peripheral blood cells even in the absence of elevated serum IgE [27]. In summary, these two studies show that cells in the peripheral blood bind IgE from the serum and thereby can clear the serum of IgE. These IgE-loaded cells could be the source of sFceRI when activated. Our finding that the presence of sFceRI correlates with serum IgE supports this hypothesis. On the other hand, IgE-mediated cell activation could also account for the detection of serum sFceRI in the absence of high serum IgE levels. Whether patients that have high sFceRI are protected from allergic diseases will have to be addressed in detail. Along this line of argument, it is tempting to speculate that serum sFceRI is a predictive marker for the onset of allergies that may be detectable even before serum IgE levels are elevated. We are currently investigating this hypothesis in a prospective cohort study. sFceRI is also an excellent candidate for an efficient in vivo modulator of IgE-mediated responses. While sCD23 has to trimerize to develop considerable affinity for its ligand [1], sFceRI can bind IgE with a one-to-one ligand-receptor ratio. Additionally, the affinity of the FceRI-IgE interaction is exceptionally high and disruption of a once formed contact requires low pH, which is physiologically found only in the stomach [1,6,7]. The finding that receptor crosslinking is required for the production of sFceRI also hints at a potential negative feedback mechanism. Antigen-IgEmediated receptor crosslinking could induce shedding of sFceRI to remove IgE-binding sites from the cell surface and to terminate receptor-mediated signaling. In addition, we show here that sFceRI has the ability to prevent IgE-binding to surface expressed receptors. Thus the presence of serum sFceRI could inhibit IgEloading of effector cells of allergy in vivo.
Our in vitro studies suggest that sFceRI may also prevent IgEmediated activation of the immune system by clearing the serum of IgE in a manner comparable to omalizumab. Omalizumab is a recombinant humanized monoclonal antibody directed against serum IgE and currently approved for the treatment of severe allergic asthma [38,39,40]. Omalizumab also downregulates cell surface levels of FceRI [41].
In summary we here describe a new soluble form of human FceRI, the high affinity receptor for IgE (s FceRI), in human serum. Establishing an improved quantitative ELISA with a standard protein is now highly important, because such a quantitative method will allow for an extensive comparative analysis of serum levels in various patient groups. Such studies will be of outmost importance to draw conclusions about the clinical relevance of this new serum IgE receptor. It will be also essential to gain a better understanding of how endogenous levels of sFceRI are regulated and how sFceRI is linked to IgE-mediated immune activation in vivo.

Ethics statement
Patient sera used for this study were obtained from an ongoing prospective cohort study on the role of FceRI in the gastrointestinal tract at Children's Hospital Boston or had been previously obtained as part of routine clinical care at the Medical University of Vienna. The prospective cohort study was approved by the Investigational Review Board of Children's Hospital Boston (Harvard Medical School, Boston, MA) and patients or their legal guardians provided written informed consent. The retrospective study of patient sera was approved by the Ethics Committee of Medical University of Vienna, Vienna, Austria.

Antibodies and reagents
Anti-human FceRI alpha mAb 19-1 was kindly provided by Dr. J.P. Kinet (Laboratory of Allergy and Immunology, Beth Israel Deaconess Medical Center, Boston, MA) and used as previously described [12,28,36]. Anti-human FceRI alpha mAb CRA1 (clone AER-37) was purchased from eBioscience, San Diego, CA. Anti-FceRI-gamma polyclonal serum was purchased from Millipore, Billerica, MA. Chimeric IgE that contains the human Fc domain and recognizes the haptens 4-hydroxy-3-nitrophenylacetic acid (NP) and 4-hydroxy-3-iodo-5-nitrophenylacetic acid (NIP) with its Fab region (cIgE) was derived from Jw 8/5/13 cells (Serotec, Oxford, UK, kindly provided by Dr. D. Maurer, Department of Dermatology, Medical University of Vienna, Austria, [12,42]) and was used for immunoprecipitation of properly folded FceRI-alpha and for in vitro cell culture experiments. Phycoerythrin (PE)conjugated was purchased from Biosearch Technologies, Novato, CA, and used for flow cytometry analysis. Anti-mouse IgG (Fc specific, produced in goat; Sigma Aldrich, St. Louis, MO, #M3534-1 mL) was used for coating of the ELISA plates. High-IgE human serum (total IgE.2000 kU/L) was purchased from Bioreclamation, Hicksville, NY to assure the quality control of IgE used for detection of sFceRI by ELISA. Goat anti-human IgE HRP conjugated antibody (Caltag, Invitrogen, Carlsbad, CA) was used as a secondary antibody.

Patient sera
Sera from adult individuals were tested for the presence of sFceRI by immunoblot and ELISA. Sera from eight polysensitized, highly atopic children were analyzed for sFceRI by ELISA. Total serum IgE and allergen-specific IgE were measured by solid phase immunoassay (Phadia ImmunoCAPH, Pharmacia Diagnostics, Uppsala, Sweden). Total serum IgE levels are given in kU/l, specific IgE is given in kUA/l and CAP RAST classes.
Sera from 119 children were obtained from an ongoing prospective cohort study on the role of FceRI in the gastrointestinal tract. Patients between 1 and 19 years of age scheduled for an elective esophago-gastro-duodenoscopy at the Division of Gastroenterology at Children's Hospital Boston were randomly invited to participate. Subjects who used steroids in any form, immunomodulatory drugs, mast cell stabilizer, or leukotriene inhibitor within the last 3 months, as well as patients with an established diagnosis of autoimmune, inflammatory, or immunodeficiency disease were not enrolled. Total serum IgE levels were assessed according to standard procedures at Children's Hospital Boston using the Elecsys IgE II kit (Roche Diagnostics, Mannheim, Germany). IgE levels are given in kU/l. Expected normal ranges for this assay are 30 kU/l for age 0-3 years, 200 kU/l for 3-10 years, 500 kU/l for 10-14 years, and 200 kU/l for .14 years.

Immunoprecipitation and immunoblotting of sFceRI
To target the fully mature form of FceRI alpha as expressed on the cell surface, we loaded MelJuso-ac cells with cIgE (10 mg/ml in PBS) before solubilization in lysis buffer (3610 6 cells per ml; 0.5% Brij 96, 20 mM Tris, pH 8.2, 20 mM NaCl, 2 mM EDTA, 0.1% NaN 3 ) containing protease inhibitors (Complete, Roche, Genentech, South San Francisco, CA) for 30 min on ice. Immunoprecipitation was next performed with NIP-beads (Sigma) as previously described [36,42]. Proteins were eluted from beads in non-reducing Laemmli sample buffer and samples were separated on 12% non-reducing SDS-PAGE gels, transferred to PVDF membrane (Pierce, Thermo Fisher Scientific, Rockford, IL) and probed with anti-FceRI-alpha (mAb 19-1 or CRA1 for reducing conditions, both 0.5 mg/ml) followed by peroxidase (HRP)conjugated goat-anti-mouse IgG for detection of precipitated achain. For immunoprecipitation of sFceRI from serum, 2-5 ml serum was used. cIgE-loaded NIP columns were also used for purification of sFceRI from supernatants of MelJuso-ac cells prior to immunoblotting. Peroxidase activity was detected by Super-Signal chemiluminescent substrate (Pierce). Accordingly, sFceRI was precipitated from serum with IgE-coupled beads and immunoblotting was performed.
For co-immunoprecipitation of FceRI alpha and gamma chains from MelJuso-ac, cells were loaded overnight with cIgE. Cell lysates were prepared and incubated with NIP sepharose beads as described above. FceRI alpha-chain was detected with mAb 19-1, FceRI gamma-chains with polyclonal rabbit anti-FceRIc antibodies (Upstate).

ELISA for the detection of sFceRI in cell culture supernatants and serum
To improve sensitivity, wells were first incubated with a goatanti-mouse coating antibody (5 mg/ml, Sigma), then with antialpha chain mAb (CRA1 0.5 mg/ml, clone AER-37; eBioscience). After a blocking step with 10% FCS in PBS, wells were incubated with sera (1:2 dilution) overnight. After repetitive washing, platebound sFceRI was loaded with IgE (Bioreclamation) and detected with a goat anti-human IgE-HRP conjugated second step (Caltag). Rather than using IgE purified from different patient sera, IgE was purchased to ensure the quality and consistency of this reagent in our ELISA. Conversion of 3,39,5,59-tetramethyl-benzidine liquid substrate (TMB, Sigma) was measured at 450 nm. Results are given as optical density (OD). To control for intra-assay variability, we included on each plate a positive and a negative control sample consisting of a pool of three positive or three negative patients respectively. The levels of circulating IgE-sFceRI complexes were determined by omitting the IgE-loading step of the protocol. Levels of free sFceRI were then calculated as follows: OD total sFceRI 2OD IgE-sFceRI complexes = OD free sFceRI .
In a subset of patients, sFceRI was measured in plasma and serum in parallel. For conversion of plasma samples into serum, BD Serum Separation Tubes (Becton Dickinson) were used according to the manufacturer's guidelines.

Production of sFceRI by MelJuso-ac cells and detection in cell culture supernatants
MelJuso-ac cells were grown to confluence and incubated with cIgE overnight. Excess cIgE was washed away and ligand-bound receptor was activated with haptenized antigen (BSA-or OVA-, 1 mg/ml, both from Biosearch Technologies, Novato, CA). Cell culture supernatants were collected after the indicated time periods and analyzed for the presence of soluble alpha chain by ELISA or by immunoprecipitation.

Exosome removal
To remove exosomes from cell culture supernatants, MelJusoac supernatants were treated with a sequence of ultracentrifugation steps following the protocol published by Thery et al. [30]. Briefly, exosome-free supernatants were obtained by the following consecutive centrifugations: 300 g for 5 minutes, 1200 g for 20 min, and 10000 g for 30 min, followed by a final centrifugation step at 110000 g for 1 h.

Flow cytometry analysis
Surface expression of FceRI on MelJuso-ac cells was determined by staining with the anti-human FceRI alpha mAb CRA1. IgE binding was tested by culturing MelJuso-ac cells in the presence or absence of NP-specific cIgE and staining with phycoerythrin (PE)-conjugated NP (NP-PE; Biosearch Technologies). For the detection of sFceRI, cells were loaded with either a mix of cIgE (100 ng/ml) in culture supernatants that contained sFceRI or supernatants from unstimulated cell cultures for 30 min on ice. A number of different ratios of cIgE to culture supernatants was analyzed. FceRI-bound cIgE was stained with NP-PE. Analysis was performed on a BD FACScan TM flow cytometer using CellQuest software for acquisition and data analysis (both from Becton Dickinson).

Immunofluoresces Microscopy
For FceRI alpha internalization, MelJuso-ac cells were grown on coverslips (No 1.5), stained first with purified mouse anti-FceRI alpha CRA1 antibody for 20 min at 37uC and subsequently with an anti-mouse Alexa Fluor 568 for 45 min at 37uC to induce receptor crosslinking. Cells were fixed with 4% paraformaldehyde for 20 min and mounted using Prolong Antifade reagent (Invitrogen). Both antibodies were diluted in HBSS supplemented with 10 mM HEPES (Invitrogen) and 5% NuSerum (Invitrogen). CRA1 was diluted at 1:100 and anti-mouse Alexa Fluor 568 was diluted at 1:400. Plasma membranes of fixed cells were stained with Alexa Fluor 647-conjugated wheat germ agglutinin (WGA, diluted at 1:1000) for 10 min. For time point t = 0 cells were fixed before incubation with anti-mouse Alexa Fluor 568 and WGA. All incubation steps were carried out in a humidified chamber. Confocal images were acquired on a Nikon TE2000 inverted microscope coupled to a Yokogawa spinning disk confocal unit (Perkin-Elmer Inc.) and an Orca AG scientific-grade cooled CCD camera (Hamamatsu Photonics K.K.). Slidebook software (Intelligent Imaging Innovations Inc.) was used for image capture, processing, and analysis.

Statistical Analysis
Correlations between serum IgE and serum sFceRI were calculated by Spearman's rank correlation test using SPSS for Windows (version 16.0, SPSS Inc., Chicago, IL). Spearman's rank correlation coefficients are displayed as 'rho', a p-value of .0.05 was considered significant.

Supporting Information
Table S1 Serum levels of sFceRI in atopic patients. (DOC)