An Intermediate pH Unfolding Transition Abrogates the Ability of IgE to Interact with Its High Affinity Receptor FcϵRIα*

The interaction between IgE-Fc (Fcϵ) and its high affinity receptor FcϵRI on the surface of mast cells and basophils is a key event in allergen-induced allergic inflammation. Recently, several therapeutic strategies have been developed based on this interaction, and some include Fcϵ-containing moieties. Unlike well characterized IgG therapeutics, the stability and folding properties of IgE are not well understood. Here, we present comparative biophysical analyses of the pH stability and thermostability of Fcϵ and IgG1-Fc (Fcγ). Fcϵ was found to be significantly less stable than Fcγ under all pH and NaCl conditions tested. Additionally, the Cϵ3Cϵ4 domains of Fcϵ were shown to become intrinsically unfolded at pH values below 5.0. The interaction between Fcϵ and an Fcγ-FcϵRIα fusion protein was studied between pH 4.5 and 7.4 using circular dichroism and a combination of differential scanning calorimetry and isothermal titration calorimetry. Under neutral pH conditions, the apparent affinity of Fcϵ for the dimeric fusion protein was extremely high compared with published values for the monomeric receptor (KD < 10-12 m). Titration to pH 6.0 did not significantly change the binding affinity, and titration to pH 5.5 only modestly attenuated affinity. At pH values below 5.0, the receptor binding domains of Fcϵ unfolded, and interaction of Fcϵ with the Fcγ-FcϵRIα fusion protein was abrogated. The unusual pH sensitivity of Fcϵ may play a role in antigen-dependent regulation of receptor-bound, non-circulating IgE.

Immunoglobulin E (IgE) is important for host defense against parasites and for protective inflammation. Yet, IgE-mediated signaling through its receptors is also a focal point of inflammatory allergic disease (1). The constant domain of IgE (Fc⑀) is responsible for binding to its two receptors, Fc⑀RI and CD23 (also known as Fc⑀RII). Fc⑀RI is expressed on the surface of mast cells, basophils, and some antigen presenting cells in humans (2,3), whereas Fc⑀RII is expressed primarily on antigen presenting cells, including B-cells (4). Binding of IgE to Fc⑀RI, in particular, leads to increased levels of Fc⑀RI on the cell surface, up-regulation of proteins involved in cell survival pathways, and cellular sensitization for an allergic event (2,(5)(6)(7).
Cross-linking of cell surface IgE⅐Fc⑀RI complexes with multivalent antigens or anti-IgE antibodies leads to receptor signaling and cellular degranulation, the release of preformed secretory granules containing histamine, serotonin, various lipids, proteases, and other acute inflammatory agents, and the expression and release of inflammatory cytokines (1). Receptor crosslinking also leads to endocytosis of aggregated IgE⅐Fc⑀RI complexes and exposure to endosomes and lysosomes. Endocytosis is a potentially important pathway for IgE regulation and Fc⑀RI turnover during an antigen-mediated allergic response (8 -10).
Fc⑀RI expressed on mast cells and basophils is a heterotetramer consisting of one ␣, one ␤, and two ␥ chains (1). The ␣ chain, an immunoglobulin family member, is the subunit responsible for high affinity binding of IgE (K D ϳ 10 Ϫ10 M) (11). It has been shown that mice lacking B cells and passively sensitized with anti-trinitrophenyl IgE respond to antigen treatment or anti-IgE for more than 1 month following IgE administration even with the absence of detectable serum IgE (12). This implies that high affinity between IgE and Fc⑀RI translates into an incredibly long in vivo half-life of the complex on the surface of mast cells and basophils.
The Fc⑀/Fc⑀RI␣ interaction occurs in a structurally homologous manner to what has been observed for IgG 1 constant domain (Fc␥) binding to Fc␥RI-III (13)(14)(15). Domain 2 of the Fc⑀RI␣ subunit interacts with the homodimeric interface of the C⑀3 domains of Fc⑀. C⑀4 of Fc⑀ does not participate directly in the binding but is crucial for maintaining high affinity (16 -18). 2 The C⑀2 domain, which is unique to IgE and virtually replaces the hinge found in IgG, is much less involved in binding but plays an important role in controlling the proper IgE/ Fc⑀RI␣ stoichiometry and binding kinetics (17,19).
The production and purification of IgGs, particularly human IgG1, for diagnostic or therapeutic applications is now fairly routine. IgE-based therapeutics, however, have only recently begun to build some momentum (20 -22). Unlike IgG therapeutics, standard methods for producing, handling, and formulating IgE or Fc⑀, particularly at an industrial scale, have not been established. In particular, it is not known whether Fc⑀ is similar to Fc␥ in terms of its thermostability and pH sensitivity. Exposure to destabilizing conditions could lead to denaturation or aggregation. Even low levels of Fc⑀ dimer or higher order aggregates could have profound effects on cellular or in vivo functional assays by inadvertently cross-linking Fc⑀RI in the absence of antigen (5,7). Our initial studies revealed an unusual sensitivity of Fc⑀ to various buffer and pH conditions. This led us to characterize in detail the biophysical properties of Fc⑀ important for maintaining IgE-Fc⑀RI interactions and for IgE half-life.
Here we describe detailed biophysical analyses of Fc⑀ focused on deriving domain-specific unfolding information for Fc⑀ and how buffer conditions, particularly pH, affect Fc⑀RI␣ binding. Using circular dichroism (CD) 3 and differential scanning calorimetry (DSC), we demonstrate that Fc⑀ is much less thermostable than Fc␥. The native structure of Fc⑀ is also highly sensitive to intermediate pH levels (5.0 and below) and high ionic strength. Binding of Fc⑀ to an Fc␥-Fc⑀RI␣ fusion protein is extremely high affinity at all pH values where Fc⑀ remains folded. Unfolding of the C⑀3 and C⑀4 domains below pH 5.0 abolishes Fc⑀ binding to the Fc␥-Fc⑀RI␣ fusion. This pH-dependent unfolding property of Fc⑀ may be important for providing a mechanism of IgE release from its high affinity receptor inside the cell and may help define a unique pathway for the regulation of receptor-bound IgE in vivo.

EXPERIMENTAL PROCEDURES
Subcloning of Fc⑀, Fc␥, Fc␥-C⑀2, Fc␥-C⑀2C⑀3, and Fc␥-Fc⑀RI␣-Fc⑀ was subcloned from mRNA extracted from human blood B cells (CD20ϩ and CD20ϩ IgMϪ). Complementary DNA primers, GTCACTATGCCACCATCAG-CTTG and TTTACCGGGATTTACAGACACCGC, which amplify from the C terminus of the C⑀1 domain to the C terminus of the C⑀4 domain, were used to create an Fc⑀ vector insert via PCR. The resulting insert was cloned into a TA vector (TOPO TA cloning kit, Invitrogen). The modified TA vector was used as template for PCR using primers GCGGCCGCCT-CACCATGGGCTGGAGCCTGATCCTGCTGTTCCTGGT-GGCCGTGGCCACCCGCGTGCTGAGCTTCACCCCGCC-CACCGTGAA and CGTCGTTTAATTAATCATTTACCG-GGATTTACAG. The resulting PCR product was also introduced into the TA vector. After sequencing the 2nd generation TA vector, a fragment was digested from the vector using the enzymes NotI and PacI and ligated into the expression vector PV-90 (Biogen Idec). The theoretical subcloned sequence following signal processing was (in Kabat numbering) 248 FTPP . . . PGK 610 (23). The plasmid was introduced into CHO cells by electroporation. Clones were selected with geneticin (Invitrogen). An N-terminal signal sequence, MGWSLILLFLVAVATRVLS, was built-in to the expression vector for secretion into the media and was included for the majority of the protein constructs described here (24). Cells were cultured for 14 days at 28°C, 5% CO 2 humidified air, in BCM16 serum-free media. A secreted, dimeric protein of the expected molecular mass, ϳ71.3 kDa (as per SDS-PAGE and static light scattering analysis), was found in the media following culture. The Fc⑀ protein identity was also confirmed by Edman degradation and intact mass analysis.
An in-house vector containing the constant region of human IgG 1 was used as template for PCR amplification of human Fc␥ using primers GCGGCCGCCTCACCATGGGCTGGAGCC-TGATCCTGCTGTTCCTGGTGGCCGTGGCCACCCGCG-TGCTGAGCGAGCCCAAATCTTGTGACAA and TTAAT-TAATCATTTACCCGGAGACAGG. The PCR product was cloned into the TA vector and sequenced in-house. The remaining plasmid manipulations, transfections, and cell culture were performed as described above for Fc⑀. The mature sequence of the Fc␥ protein was predicted to be 226 EPK . . . PGK 478 . The secreted, dimeric protein was found to have an average mass ϳ56.4 kDa by SDS-PAGE and static light scattering analyses. Mass analysis with the purified protein reduced with dithiothreitol and treated with PNGase F yielded a dominant ion at 25,964 kDa corresponding to mature Fc␥ with the C-terminal lysine cleaved at Ͼ90%. Cys 230 (residue 5 in the mature protein) forms a disulfide bond with light chain in fulllength IgG 1 . No nonnative inter-or intra-molecular disulfides were detected following long term (i.e. several months) incubation at 4°C.
The chimeric protein, Fc␥-Fc⑀RI␣, containing human Fc␥ connected at its C terminus (Nterm-226 EPK . . . SPG 477 -Cterm) by a 9-amino acid linker, SRENLYFQG, to the N terminus of human Fc⑀RI␣ (starting with -VPQ . . . ), was created for enhanced receptor expression. The human Fc⑀RI␣ DNA insert encoding the extracellular domain without its signal peptide (Val 26 to Tyr 202 , residue numbering includes signal peptide) was obtained by PCR from IMAGE clone 4294467 (ATCC) using the primers GCGGCGTCTAGAGAGAACCTGTACT-TCCAGGGCGTCCCTCAGAAACCTAAGGTC and GCGG-CGGTCGACGAATTCTTAGTACTTCTCACGCGGAG-CTT. The PCR product was subsequently digested with XbaI and SalI and subcloned into a modified INPEP4-Fc vector containing a signal peptide of the mouse -light chain fused to the human immunoglobulin heavy chain IgG 1 constant region (Fc) (Kabat number 226-477). The Fc␥ portion contained a T318A (EU299) mutation that abolishes glycosylation at Asn 314 (EU297). The C-terminal Lys on Fc␥ was not included because this residue was cleaved at Ͼ90% within the cell culture for Fc␥ (see above). The plasmid was transfected into dihydrofolate reductase-CHO DG44 cells using electroporation (25) and cloned in medium containing CHO-S-SFMII (Invitrogen) supplemented with HT (Invitrogen) in the presence of 400 g/ml G418 (Invitrogen) using the limiting dilution method. Cell culture was performed using the same protocols as used for the Fc⑀ producing CHO cells.
The chimeric proteins, Fc␥-C⑀2 and Fc␥-C⑀2C⑀3, were prepared by linking the C⑀2 or C⑀2C⑀3 domains of IgE to the C terminus of Fc␥ via a 15-amino acid linker GSGGS(GGGGS) 2 . The C-terminal Lys on Fc␥ was also not included here. The sequences of the Fc␥-C⑀2 and Fc␥-C⑀2C⑀3 protein domains began at the same N terminus as the above Fc⑀ construct and 3 The abbreviations used are: CD, circular dichroism; Fc␥, IgG constant domain fragment; IgG, immunoglobulin G; C⑀2, C⑀3, and C⑀4, second, third, and fourth constant region domains of the IgE heavy chain, respectively; DSC, differential scanning calorimetry; C␥2 and C␥3, second and third heavy chain constant domains of IgG, respectively; ANS, 1-anilino-8naphthalene sulfonate; ITC, isothermal titration calorimetry; CHO, Chinese hamster ovary; SEC, size exclusion chromatography; LC/MS, liquid chromatography/mass spectrometry; HPLC, high pressure liquid chromatography; SPR, surface plasmon resonance; PBS, phosphate-buffered saline; ⌬H A 0 (T), molar enthalpy of association; ⌬C P 0 , molar heat capacity of association; T M , midpoint of thermal unfolding transition; ⌬S A 0 (T), molar entropy of association.

pH Unfolding of IgE-Fc Abrogates Binding to Fc⑀RI␣
contained the following C termini: . . . SNP 364 for Fc␥-C⑀2 and . . . GPR 499 for Fc␥-C⑀2C⑀3. CHO cell lines producing high titers of Fc␥-C⑀2 and Fc␥-C⑀2C⑀3 were subcloned from the bulk transformants using the method of Brezinsky et al. (26) and cultured using the same protocol as described for the Fc⑀ CHO clone.
Protein Purification-Purifications were performed using an AKTA Explorer (GE-Healthcare). CHO cell media containing Fc⑀ was titrated to pH 9.0 using 1 M Tris base, diluted 10 times with distilled deionized H 2 O, and loaded on a Q Sepharose column equilibrated in 20 mM, Tris, pH 9.0, at 10 ml/min. The column was washed with running buffer, and Fc⑀ was eluted using a linear gradient to 200 mM NaCl. Fc⑀ eluted between 110 and 170 mM. To concentrate the protein, the eluant was diluted 10 times, passed over a 60-ml Q Sepharose column, washed, and eluted in a single step using 20 mM Tris, 1 M NaCl, pH 9.0. The concentrated eluant was titrated to pH 8.0 using 2 M NaCH 3 COO, pH 5.0, and further concentrated using an Amicon stir unit with a 10,000 M r cut-off YM10 membrane, and passed over a 350-ml Superdex 200 column equilibrated with PBS to remove aggregates, multimers, and remaining contaminants. Purified Fc⑀ was concentrated using an Amicon Ultra-15 centrifugal filter device, 10,000 M r (Millipore). Protein concentration was measured by AU 280 nm using an extinction coefficient of 1.4 ml mg Ϫ1 cm Ϫ1 .
CHO supernatants containing recombinant Fc␥, Fc␥-Fc⑀RI␣, or Fc␥-C⑀2 were loaded onto a Protein A-Sepharose FF (Amersham Biosciences) column equilibrated with 50 mM Tris, 500 mM NaCl buffer at pH 7.5. The column was washed and protein was eluted using a step gradient to 100% 100 mM glycine, pH 3.0. Protein-containing fractions were titrated to about pH 7.0 using 1 M Tris base, pooled, and dialyzed against PBS. The Fc␥-Fc⑀RI␣ fusion protein was further purified by size exclusion chromatography (SEC) as described above for Fc⑀. Purified Fc␥, Fc␥-Fc⑀RI␣, and Fc␥-C⑀2 were concentrated in the same manner as Fc⑀. Fc␥ and Fc␥-Fc⑀RI␣ concentrations were determined by AU 280 nm using extinction coefficients of 1.4 and 1.9 ml mg Ϫ1 cm Ϫ1 , respectively. Fc␥-C⑀2 concentration was determined using the modified Lowry method and bovine IgG as the standard.
Analytical Size Exclusion Chromatography with In-line Static Light Scattering-Various pH samples of Fc⑀ and Fc␥ were prepared by dilution of 1.8 and 4.0 mg/ml stocks, respectively, into 25 mM phosphate, 25 mM citrate, 150 mM NaCl buffers at pH 2, 3, 4, 4.5, 5.0, 5.3, 5.7, 6.0, 6.5, 7.0, 8.0, and 9.0. Final protein concentrations were 320 g/ml. Additional samples were prepared by diluting the Fc⑀ and Fc␥ stocks into 25 mM phosphate, 25 mM citrate, 750 mM NaCl at pH 4, 5, and 6. The stock solutions were in PBS (10 mM phosphate, pH 7.4, 140 mM salt). All samples were incubated for 3 or more hours at room temperature before analysis. Eighty l of each sample was injected onto a TSKgel G3000SW XL, 5 mm, 250-Å Analytical SEC column (Tosoh Biosciences) equilibrated in 10 mM phosphate, 150 mM NaCl, 0.02% sodium azide at pH 7.2 using an Agilent 1100 HPLC system. Whereas injection of the various pH samples onto the SEC column using a similar running buffer may result in some protein refolding or re-equilibration, the running buffer was amenable for the column and allowed for uniform comparison of elution times via the use of a single running buffer. Light scattering data for material eluting from the SEC column were collected using a miniDAWN static light scattering detector coupled to an in-line refractive index meter (Wyatt Technologies). Light scattering data were analyzed using the ASTRA V software provided by the manufacturer.
ANS Binding-The fluorescence of 1-anilino-8-naphthalene sulfonate (ANS, Sigma) was measured using a Victor3 multilabel fluorescence plate reader (PerkinElmer Life Sciences) with a 360 Ϯ 20 nm excitation cutoff filter and a 460 Ϯ 20 nm emission cutoff filter. All fluorescence measurements were made using 96-well fluorescence plates (NUNC MaxisorbF) with 150 l/well and a 0.1-s sample averaging period per well. ANS (20 M) was incubated with 20, 40, 80, 160, and 320 g/ml bovine ␣-lactalbumin (Sigma) at pH 2.0 and 7.0 to establish appropriate protein concentrations to obtain adequate signal to noise. At 80 g/ml ␣-lactalbumin at pH 2.0, the ANS fluorescence signal was ϳ20-fold higher than the noise; therefore, 80 g/ml protein concentrations were utilized for the Fc⑀ and Fc␥ measurements. To study ANS binding to Fc⑀ or Fc␥, each protein was diluted into a 25 mM phosphate, 25 mM citrate, 150 mM NaCl buffer series at pH 2, 3, 4, 4.5, 5.0, 5.3, 5.7, 6.0, 6.5, 7.0, 8.0, and 9.0 and incubated with the fluorescent dye. Additional Fc⑀ and Fc␥ samples were prepared in 25 mM phosphate, 25 mM citrate, 750 mM NaCl at pH 4.0, 5.0, and 6.0.
CD Spectroscopy-CD measurements were performed using a Jasco J-810 spectropolarimeter equipped with a thermoelectric Peltier device for temperature control and an external water bath as a heat sink. Fc⑀, Fc␥, and Fc␥-Fc⑀RI␣ spectra (195-260 nm) were taken using the continuous scan mode at 100 nm/min. Bandwidth was set to 1 nm and data pitch to 0.2 nm. The response time of the instrument was set to 1 s. Five spectra were averaged for increased signal to noise. All spectra were taken at 25°C. Fc⑀, Fc␥, and Fc␥-Fc⑀RI␣ samples were dialyzed against 10 mM sodium citrate, 10 mM NaCl at pH 4.5, 5.0, 5.5, and 6.0. Additional Fc⑀, Fc␥, and Fc␥-Fc⑀RI␣ samples at pH 7.4 were prepared by dialysis against PBS. To map the apparent unfolding transition of Fc⑀ in detail, additional Fc⑀ samples were prepared by dialysis against the same citrate buffer at pH 4.8 and 5.2. Background buffer scans at pH 5.0, 6.0, and 7.4 were virtually identical; therefore, all background subtractions were performed using the pH 5.0 background spectrum. All CD spectra were obtained using 3.75 M protein with the dimer molecular mass (71.8 kDa) used for Fc⑀ and the monomer molecular mass used for Fc␥ (24.7 kDa) and the divalent Fc␥-Fc⑀RI␣ fusion protein (49.8 kDa). Thermal melts of Fc⑀ were performed in the cuvette by heating the samples at 5°C intervals at a rate of 1°C/min. Between each 5°C heating interval, three replicate scans were performed using the parameters described above.
Limited Proteolysis and LC/MS-Fc⑀ was digested by trypsin and Glu-C at pH 4.5, where significant protein structural changes away from the natively folded material occurred. Five samples were prepared by diluting 1 mg of Fc⑀ into 0.4 ml of 70 mM sodium acetate solutions at pH 4.5. The pH of the acetate solution was unperturbed subsequent to the addition of Fc⑀ in PBS. Various concentrations of trypsin (30, 300, and 3000 units, Sigma T-1426) and endoproteinase Glu-C (10 and 100 units, Sigma P-2922) were added to individual pH 4.5 Fc⑀ samples. Digestions were incubated at 37°C. Time points (100 l) were taken for LC/MS analysis after 0, 2, 10, 60, and 300 min and after overnight incubation. Proteolysis was halted by the addition of 50 mM EDTA and 10 mM phenylmethylsulfonyl fluoride. Dithiothreitol was then added to Ͼ10 mM for disulfide reduction, and each time point was subsequently frozen at Ϫ80°C. Reagent blanks containing trypsin or Glu-C but excluding Fc⑀ were also prepared for identifying non-Fc⑀ peaks in the LC chromatograms. Based on the level of proteolysis observed by SDS-PAGE analysis, the 30-unit trypsin and 100-unit Glu-C samples were chosen for LC/MS peptide mass mapping. All time points were analyzed by HPLC MS as described previously (27). Total ion chromatograms were analyzed using the Agilent Deconvolution and Peptide Tools software programs.
DSC-DSC scans were performed using an automated capillary DSC (capDSC, MicroCal, LLC). Protein and reference solutions were sampled automatically from 96-well plates using the robotic attachment. Prior to each protein scan, two buffer scans were performed to define the baseline for subtraction. All 96-well plates containing protein were stored within the instrument at 6°C.
For the Fc⑀ and Fc␥ pH titrations (Figs. 3, B and C, and 4, A and B), three separate pH panels were created with 15, 150, or 750 mM NaCl. The buffers included 70 mM glycine at pH 2.8, 3.3, and 3.5; 70 mM sodium acetate at pH 4.0, 4.5, 4.8, 5.0, and 5.2; 35 mM sodium citrate at pH 5.5, 6.0, and 6.5; and 35 mM Tris at pH 7.0, 7.5, and 8.0. Stock solutions of Fc⑀ and Fc␥ at 1.8 and 4.0 mg/ml, respectively, in PBS were diluted to 0.5 mg/ml in each buffer solution. Identical PBS dilutions into all members of the buffer panel were made for baseline measurements, and it was confirmed that the pH of each sample was not altered by the residual PBS. Scans were performed from 10 to 90°C at 1°C/min using the medium feedback mode for enhanced peak resolution. Additionally, the Fc⑀ and Fc␥-C⑀2 proteins were co-dialyzed against 10 mM sodium phosphate, 15 mM NaCl, pH 2.5, diluted to 1 mg/ml with dialysate, and scanned under identical conditions as described for the buffer panel.
For measurement of the thermodynamic interaction parameters between Fc⑀ and Fc␥-Fc⑀RI␣, each protein was dialyzed against a panel of buffer solutions at pH 4.5, 5.0, 5.5, 6.0, and 7.4 as described above for the CD experiments. Using the dimeric molecular weight for apo-and holo-Fc⑀ and the monomeric molecular weight for apo-and holo-Fc␥-Fc⑀RI␣, 2.5 M protein solutions were prepared using matched dialysates for dilution and incubated for at least 2 h at 4°C prior to DSC analysis. Dialysates were used within the reference cell of the calorimeter to define the baseline of each protein scan. Additional baseline corrections were performed by scanning with dialysate in both the reference and sample cells. For the binding experiments, scans were performed from 20 to 95°C using a scan rate of 4.0°C/min and the low feedback mode.
Scans were analyzed using the Origin software supplied by the manufacturer. Subsequent to the subtraction of reference baseline scans, non-zero protein scan baselines were corrected using a third order polynomial. The unfolding parameters for the multidomain unfolding profiles of Fc⑀, Fc␥, Fc␥-Fc⑀RI␣, and Fc⑀/Fc␥-Fc⑀RI␣ were deconvoluted using the multipeak fitting routine within the software.
Isothermal Titration Calorimetry (ITC)-Fc⑀ and Fc␥-Fc⑀RI␣ were dialyzed against 10 mM citrate, 10 mM NaCl at pH 5.5 and 6.0 as described above for the CD experiments. Additional samples were also prepared by dialyzing Fc⑀ and Fc␥-Fc⑀RI␣ against 1ϫ PBS. Fc⑀ stock solutions were concentrated to ϳ100 M using 6 ml of Vivaspin MW5000 centrifugal concentration units (VivaSciences). ITC experiments were performed on a VP-ITC microcalorimeter (MicroCal, LLC). Aliquots of 70 M Fc⑀ (15 l) were injected into the reaction cell containing 5 M solutions of Fc␥-Fc⑀RI␣ to obtain a final Fc⑀/Fc␥-Fc⑀RI␣ ratio of ϳ2:1. Dialysates were used in the reference cell. A 4-min equilibration period was used between all Fc⑀ injections with an initial delay of 60 s. All samples were degassed for ϳ10 min prior to each experiment. The ITC internal sample jacket was set 10°C below the constant temperature within the reaction and reference cells, except for the 10°C titrations where the bath was set to 1°C. Numerical integration of the data were performed using the ITC data analysis software supplied by MicroCal (Origin). ⌬H A 0 (T) values were calculated based on the difference between the average heat liberated/absorbed during the binding phase of the injections and the average heat of dilution found once the receptor, Fc-Fc⑀RI␣, was saturated with Fc⑀. The titration midpoints that defined the stoichiometry, n, were determined without data fitting. None of the ITC curves were utilized further for determination of binding constants due to the lack of titration points in the transition region. The absence of multiple titration points at the concentration where Fc⑀ saturated the receptor was an indication that the affinity was too high to be measured by ITC alone.
Evaluation of the apparent binding thermodynamics for Fc⑀/ Fc␥-Fc⑀RI␣ using DSC and ITC-K U (T M ) values for the isolated C⑀3C⑀4 domains of Fc⑀ and the apodomains of Fc⑀RI␣ were extrapolated from the measured calorimetric enthalpies and T M values of the protein domains (i.e. with Fc⑀ alone or Fc␥-Fc⑀RI␣ alone in the calorimeter) using theoretical heat capacities, 7.2 and 3.1 kcal mol Ϫ1 K Ϫ1 , respectively, based on the molecular weight of their cooperatively folded domains (28). ⌬C P 0 and ⌬H A 0 (25°C) values for the Fc⑀/Fc␥-Fc⑀RI␣ interaction were determined by ITC and estimated at pH 5.0 by fitting the experimentally derived values at pH 5.5, 6.0, and 7.4 to a second order polynomial. All other thermodynamic parameters were measured directly by fitting the DSC peaks within the Origin Software. The parameters were input into expressions derived by Brandts and Lin (29) for two proteins that demonstrate independent unfolding transitions in isolation but whose unfolding becomes thermodynamically coupled upon formation of a complex.

RESULTS
pH-dependent Unfolding of Fc⑀-Concentrated Fc⑀ was diluted into a set of buffers ranging from pH 2 to 9 and chromatographed on an SEC column (Fig. 1A). The molecular weight(s) of the eluted Fc⑀ peak(s) was determined by static light scattering and refractive index analyses. Following incubations between pH 5.3 and 9.0, Fc⑀ eluted as a single peak with pH Unfolding of IgE-Fc Abrogates Binding to Fc⑀RI␣ the expected molecular mass of 71 kDa. At pH 5.0 and below, multiple peaks were observed within the chromatogram (Fig.  1A). Between pH 3.0 and 5.0, the additional peaks were limited to small multimers (i.e. dimers and tetramers) and apparent monomeric material, which eluted later than the folded Fc⑀ protein did. At pH 2.0, higher order aggregates were observed.
Using the same buffer set, the fluorescent hydrophobic dye ANS was added to solutions containing either Fc␥ or Fc⑀. ANS is a hydrophobic dye used to probe for non-native or partially folded protein species. Its fluorescence is significantly quenched in aqueous solution; however, upon incorporation into the "fluid" hydrophobic interior of unfolding or molten globule-like proteins, the dye exhibits a sharp increase in fluorescence. Under neutral pH conditions, neither Fc␥ nor Fc⑀ associated with ANS as judged by fluorescence (Fig. 1B). In samples containing Fc␥, titration to pH 3.0 and below led to significant increases in ANS fluorescence corresponding to the loss of tertiary structure due to unfolding (Fig. 1B). For Fc⑀ samples, sharp increases in ANS fluorescence were observed at pH values below 5.0 (Fig.  1B), 2 pH units above the pH where ANS bound Fc␥. In 750 mM NaCl and pH values below 6, the dye bound to Fc⑀ (Fig. 1B,  inset), suggesting that high ionic strengths may facilitate the unfolding event.
To more directly evaluate secondary structure, CD spectra of Fc⑀ were taken under buffer conditions ranging from pH 4.5 to 7.4. Between pH 5.2 and 7.4, the spectra of Fc⑀ were found to be identical and contained a single minimum between 216 and 217 nm indicative of significant ␤-sheet and typical of Ig domains (Fig. 1C). At pH 5, the Fc⑀ spectrum shifted in a random coil direction (the minimum shifted toward 200 nm), and by pH 4.5, the spectrum was predominantly random coil. Based on the remaining negative signal at wavelengths between 210 and 220 nm and the overall spectral minimum, which did not shift entirely to 197 nm as would be expected for a random coil polypeptide, residual secondary structure was present at pH 4.5 (Fig. 1C).
The Receptor-binding Domains of Fc⑀, C⑀3C⑀4, Unfold Below pH 5-To determine whether Fc⑀ unfolding observed at pH values below 5.0 was limited to particular domains or occurred over the entire Fc⑀ region, limited proteolysis with either Glu-C or trypsin was performed at pH 4.5. Digestion time courses were analyzed by reverse phase HPLC MS. The first peaks appeared within 10 min of Glu-C or trypsin digestion and corresponded predominantly to peptides from the C⑀4 domain and to a lesser extent the C⑀3 domain. Several peptides from the C⑀3 domain appeared at a moderately slower rate. After 5 h, peptides covering Ͼ80% of the sequences of the C⑀3 and C⑀4 domains were observed ( Fig. 2A). After 24 h of proteolysis, several small peaks corresponding to initial C⑀2 proteolytic events appeared (not shown).
To confirm that the C⑀2 domain was stable at low pH values relative to the C⑀3 and C⑀4 domains, a recombinant C⑀2 domain without C⑀3 and C⑀4 was created by fusing the domain to the C terminus of Fc␥. Both Fc⑀ and the Fc␥-C⑀2 fusion protein were subjected to DSC analysis. DSC is capable of deconvoluting individual protein unfolding events within multidomain proteins and was used to investigate the stability of C⑀2 within the context of the multidomain proteins Fc⑀ and Fc␥-C⑀2. Both Fc⑀ and Fc␥-C⑀2 were analyzed by DSC at multiple pH values between 2.5 and 8.0 (data not shown). Under all pH conditions, a single transition with a midpoint of unfolding (T M ) between 60 and 70°C corresponding to the denaturation   OCTOBER 13, 2006 • VOLUME 281 • NUMBER 41 of C⑀2 within both Fc⑀ and Fc␥-C⑀2 was observed. Fc⑀ and the Fc␥-C⑀2 construct were additionally dialyzed against a sodium phosphate buffer at pH 2.5, and the matched solutions were analyzed by DSC (Fig. 2B). Both proteins displayed a C⑀2 unfolding transition near 65°C indicating that the C⑀2 domain was stably folded even at very low pH values.
pH-dependent Stability of Fc⑀-Based on the pH-dependent unfolding described above, we investigated whether Fc⑀ may have an attenuated stability between pH 7.0 and 5.0 (i.e. above the pH where unfolding is observed). Thermal denaturation was chosen as the method for measuring Fc⑀ stability under the different conditions. Thermal denaturation of Fc⑀ at various pH values was first monitored by far-UV CD (Fig. 3A). At pH 7.0, there was one transition for the unfolding of all three domains (C⑀2-4). A similar transition was observed at pH 6.0, although the apparent T M decreased by 1°C. Thermal unfolding of Fc⑀ at pH 5.2 resulted in a much broader transition(s) that began 6°C lower than at neutral pH. At pH 4.8, two clear transitions were evident, with the lower temperature transition at ϳ42°C corresponding to the unfolding of the C⑀3C⑀4 domains and the higher temperature transition at ϳ70°C corresponding to the unfolding of C⑀2. The C⑀2 transition was identified based on the Fc␥-C⑀2 DSC analyses and the Fc⑀ limited proteolysis studies at pH 4.5.
The pH-dependent stabilities of both Fc⑀ and Fc␥ were compared using DSC. The unfolding transitions of both Fc⑀ and Fc␥ were found to be irreversible and scan rate dependent (data not shown), suggesting that irreversible aggregation affects the apparent T M values of both proteins (30,31). Therefore, we compared the relative stabilities of the Fc⑀ and Fc␥ domains by measuring T M values under identical instrument conditions. The C⑀2 domain of Fc⑀ was pH stable and exhibited an unfolding transition very similar to C⑀2 measured within the Fc␥-C⑀2 fusion protein at pH values between 2.5 and 8.0 (compare Figs. 2B and 3B). At pH 8.0, the C⑀2 transition occurred at a similar T M as the C⑀3C⑀4 domains suggesting possible unfolding cooperativity; however, its T M was similar when measured in the context of Fc⑀ or Fc␥-C⑀2. Because the C⑀3C⑀4 domains did not raise the thermostability of C⑀2, the apparent overlap of the unfolding transitions of C⑀2 and C⑀3C⑀4 was unlikely due to FIGURE 2. A, schematic (rectangle) diagram of Fc⑀ with peptides detected by LC/MS subsequent to digestion at pH 4. 5, 37°C labeled above (Glu-C ) and below (trypsin). Peptides/domains with masses Ͻ800 or Ͼ10,000 Da were not included in the diagram as the small peptides could correspond to multiple sequences, and the large peptides were too large for precise intact mass determination. Peptide sequences that appeared slowly (after 1 day) upon Glu-C or trypsin digestion at pH 4.5 are underlined. All non-underlined peptide sequences were detectable within 1 h of digestion. B, DSC traces of Fc⑀ (black line) and Fc␥-C⑀2 (gray line) performed using samples dialyzed against the same pH 2.5 phosphate buffer. Schematic diagrams of the Fc⑀ and Fc␥-C⑀2 proteins are shown above the DSC curves.

pH Unfolding of IgE-Fc Abrogates Binding to Fc⑀RI␣
cooperativity between these domains at pH 8.0. The C⑀3C⑀4 domains of Fc⑀ were pH labile (Fig. 3B). The C⑀3C⑀4 domains unfolded cooperatively with a maximum T M of 57°C at pH 8. As the pH was lowered, the single unfolding transition of the C⑀3C⑀4 domains shifted to lower temperatures, whereas the unfolding transition of the C⑀2 domain remained relatively static. As the pH was lowered to pH 4.8, the C⑀3C⑀4 unfolding transition decreased to 44°C, and at pH 4.5 the unfolding transition disappeared entirely, presumably because the domains became intrinsically unfolded at all temperatures.
The C␥2 and C␥3 domains of Fc␥ were generally much more stable than the homologous C⑀3C⑀4 domains of Fc⑀ (Fig. 3C). Unlike the cooperative unfolding observed for the C⑀3C⑀4 domains of Fc⑀, the C␥2 and C␥3 domains unfolded in two separate transitions. The C␥2 domain was significantly less thermostable than the C␥3 domain. The C␥2 transition was identified by the effect that deglycosylation had on its thermostability. 4 The C␥3 domain was identified by its high thermostability, a characteristic of the isolated domain as investigated previously (32). The stability of both domains decreased with decreasing pH. Unlike the C⑀3C⑀4 domains, however, the sharp transition toward an intrinsically unfolded state at all temperatures only occurred below pH 3.
In the presence of high salt, the C␥2 and C␥3 domains of Fc␥ and the C⑀3 and C⑀4 domains of Fc⑀ were slightly destabilized. This was seen as a small shift in their T M values at 150 and 750 mM NaCl relative to 15 mM NaCl in the intermediate pH range between 5.0 and 7.0 (Fig. 4A). These small stability differences are unlikely to have a major affect on the in vitro half-life of Fc␥ within this pH range because the T M of both the C␥2 and C␥3 domains remained above 60°C. Interestingly, in high salt, the C⑀2 domain of Fc⑀ appeared to be slightly more stable. C⑀2 was especially stabilized at neutral pH and 750 mM NaCl with a T M more than 7°C higher than the T M measured using 15 mM NaCl (Fig. 4B). In contrast, NaCl significantly destabilized the C⑀3C⑀4 domains between pH 5 and 6 ( Fig. 4B). C⑀3C⑀4 began to unfold (i.e. a measurable fraction of intrinsically unfolded material at all temperatures) at pH 5.0 in low salt. In high salt, the unfolding transition was shifted 0.5 pH units (to pH 5.5).
pH Dependence of the Fc⑀/Fc⑀RI␣ Interactions-Many protein-protein interactions are highly pH-dependent. Such pH dependencies often have significant biological implications. For example, the pH dependence of the interaction between IgG1-Fc␥ and FcRn is important for serum IgG recirculation and half-life (33). Because marked decreases in Fc⑀ stability were observed as the pH was decreased, we investigated whether pH-dependent stability changes attenuated the ability of IgE to interact with its high affinity receptor, Fc⑀RI␣. To facilitate the expression of soluble Fc⑀RI␣, the receptor subunit was fused to Fc␥ (26). The resulting protein was a dimeric chimera of Fc␥ and Fc⑀RI␣. A mutation within Fc␥ (T299A) that abrogated glycosylation was utilized to prevent potential nonspecific binding of Fc␥ to Fc⑀RI␣.
We first investigated whether Fc⑀RI␣ alone exhibited pHdependent unfolding similar to the C⑀3C⑀4 domains of IgE.
CD spectra of Fc␥-Fc⑀RI␣ in buffers ranging from pH 4.5 to 7.4 indicated no major structural changes under these conditions. Fig. 5A shows the differences between the spectra of Fc␥-Fc⑀RI␣ and Fc␥ and is representative of the CD spectra of Fc⑀RI␣. The resulting difference spectra closely resembled the Fc⑀RI␣ CD spectrum published at pH 7.2 (34). Thus, Fc⑀RI␣ did not appear to undergo a pH-dependent unfolding at intermediate pH levels.
Combining Fc⑀ and Fc␥-Fc⑀RI␣ in solution led to changes in the CD spectra of the complexed proteins (Fig. 5, B-IE). Between pH 5 and 7.4, the spectra of the complex were compared with added spectra of the isolated Fc⑀ and Fc␥-Fc⑀RI␣ proteins. Across these pH values, the absolute value of the spectral intensity of the complex decreased slightly, and the minimum shifted from 216 to 217 nm indicating potential ␤-sheet stabilization and possibly the loss of some dynamic flexibility. The spectral changes upon complex formation were highly similar to the spectral changes observed by Sechi and co-workers (34) for the Fc⑀⅐Fc⑀RI␣ complex at pH 7.2. While reproducible, the changes in the CD spectra upon Fc⑀⅐Fc⑀RI␣ complex formation were fairly minimal and would certainly not reflect   large structural changes such as the folding/unfolding of an entire Ig domain. Due to the unfolded nature of Fc⑀ at pH 4.5, the summed spectra of the apoproteins changed significantly from that observed at higher pH values (Fig. 5F). The spectrum of the physical mixture of Fc⑀ and Fc␥-Fc⑀RI␣ was identical to the summed spectra of the apoproteins indicating that mixing the proteins together at pH 4.5 produced no changes in structure.
pH Dependence of the Fc⑀/Fc⑀RI␣ Binding Thermodynamics-Quantification of Fc⑀-Fc⑀RI␣ binding affinities was attempted by surface plasmon resonance (SPR). Although we were able to reproduce results in the literature (11), the sensorchip surfaces were poorly behaved at and below pH 6.0 and precluded obtaining interpretable results. Additionally, the kinetics obtained at pH 7.4 did not fit to a single exponential and were highly mass transfer limited, suggesting that the measured K D (2.6 ϫ 10 Ϫ10 M) represented an upper limit of the K D value (i.e. the limit of detection for SPR kinetic measurements) between the two proteins.
We therefore investigated the relative binding strengths of Fc⑀ to Fc␥-Fc⑀RI␣ at various pH values using DSC. The DSC curves of the individual Fc⑀ and Fc␥-Fc⑀RI␣ proteins at pH 7.4 are shown in Fig. 6. An Fc␥(T299A) construct was also analyzed to differentiate the unfolding transitions of Fc␥ from the single transition observed for Fc⑀RI␣ in the Fc␥-Fc⑀RI␣ fusion protein, 5 as the two transitions could not be separated by deconvolution (not shown). In particular, the two domains of Fc⑀RI␣ were found to unfold at the same temperature as the Fc␥(T299A) C␥2 domain. The DSC curve of complexed Fc⑀/ Fc␥-Fc⑀RI␣ was significantly different from the simple addition of the isolated Fc⑀ and Fc␥-Fc⑀RI␣ curves (dotted lines in Fig. 6) indicative of a strong interaction between the two proteins. The transitions corresponding to the C⑀3C⑀4 domains of Fc⑀ and domains 1 and 2 of Fc⑀RI␣ coalesced into a single transition that occurred at a higher temperature (ϳ8°C higher) than the unfolding transitions of C⑀3C⑀4 and Fc⑀RI␣ in isolation. Additionally, there was a significant increase in the calorimetric enthalpy of the transition (Fig. 6).
A mathematical treatment of DSC changes resulting from protein complex formation has been developed to obtain approximate binding affinities for high affinity interactions (29). The equations used by Brandts and Lin (29) were applied to Fc⑀ and Fc␥-Fc⑀RI␣ interactions with the following assumptions: 1) unfolding in all cases was two-state and reversible; and 2) the temperature-dependent equilibrium of the interaction could be described by, ] are the concentrations of unfolded Fc⑀RI␣ and C⑀3C⑀4, respectively, and [F Fc⑀RI␣-C⑀3C⑀4 ] is the concentration of the complex that was folded at all temperatures. Although the unfolding was not two-state for Fc⑀ or Fc⑀RI␣ due to aggregation at temperatures where unfolded protein exists, the scans were performed at maximal rates (4°C/min) to reduce aggregation and approximate two-state behavior.
Most of the necessary parameters for determining the affinity were obtained from the DSC experiment; however, the heat capacity of the interaction (⌬C P 0 ) and the enthalpy of the interaction at 25°C (⌬H A 0 (25°C)), the temperature where we report the affinity, needed to be determined by ITC. The ⌬C P 0 (Ϫ760 cal mol Ϫ1 K Ϫ1 ) and ⌬H A 0 (25°C, Ϫ20.4 kcal mol Ϫ1 ) values for binding of Fc⑀ to Fc␥-Fc⑀RI␣ at pH 7.4 were measured by ITC 5 S. J. Demarest and F. Taylor, manuscript in preparation.

pH Unfolding of IgE-Fc Abrogates Binding to Fc⑀RI␣
and found to be similar to previously determined values (35). The remaining thermodynamic parameters derived by DSC for determination of the apparent binding affinity of Fc⑀ to Fc␥-Fc⑀RI␣ are shown in Table 1. An extremely high affinity was calculated using the DSC method (K D Ͻ10 Ϫ12 M) that was significantly stronger than the value measured by SPR at pH 7.4 (K D ϭ 2.6 ϫ 10 Ϫ10 M, Table 2). However, for reasons described above, the SPR experiments likely reflect the lower limit (i.e. weakest possible K D ) of the interaction.
The interaction between Fc⑀ and Fc␥-Fc⑀RI␣ remained extremely strong at pH 6.0 and 5.5 ( Table 2). Measurements of ⌬C P 0 (Ϫ670 cal mol Ϫ1 K Ϫ1 ) and ⌬H A 0 (25°C, Ϫ8.9 kcal mol Ϫ1 ) at pH 6.0 and ⌬C P 0 (Ϫ520 cal mol Ϫ1 K Ϫ1 ) and ⌬H A 0 (25°C, Ϫ6.9 kcal mol Ϫ1 ) at pH 5.5 by ITC were utilized in the affinity calculations in Table 2 (Fig. 7, A-C). No changes in binding stoichiometry were observed. At pH 7.4, 6.0, and 5.5, the binding isotherms at all temperatures were too sharp (i.e. no points in the transition region) for accurate determination of the bind-ing affinity by ITC (Table 2) and indicated a K D Ͻ 10 Ϫ9 M, consistent with the values obtained using DSC.
DSC experiments with Fc⑀, Fc␥-Fc⑀RI␣, and Fc⑀⅐Fc␥-Fc⑀RI␣ complexes were performed at pH 6.0 and 5.5 to obtain parameters for determining the affinity at these pH values (Tables 1  and 2, and Fig. 8, A and B). Dropping the pH from 7.4 to 6.0 did not attenuate the affinity within the error of the experiment ( Table 2; compare Figs. 6 and 8A). At pH 5.5, however, the affinity appeared to be attenuated significantly (K D Ͻ 10 Ϫ10 M; Table 2 and Fig. 8B). Complex formation between Fc⑀ and Fc␥-Fc⑀RI␣ at both pH 6.0 and 5.5 resulted in large increases in the apparent T M values of the C⑀3C⑀4 domains of Fc⑀ and domains 1 and 2 of Fc⑀RI␣ similar to what was observed at pH 7.4 (compare Figs. 6 and 8, A and B), indicating the persistence of a strong interaction.
Titration to pH 5.0 and 4.5 profoundly weakened the interaction between Fc⑀ and Fc␥-Fc⑀RI␣. DSC experiments clearly indicated that the interaction between the two proteins was highly attenuated at pH 5.0 and completely abrogated at pH 4.5 (Fig. 8, C and D). At pH 5.0, a structural change within Fc⑀ (Figs. 1C and 5E) appeared coincidentally with a reduced stability of the C⑀3C⑀4 domains (Fig. 3B). Formation of the complex between Fc⑀ and Fc␥-Fc⑀RI␣ at pH 5.0 resulted in a marginal increase in the T M of the C⑀3C⑀4 domains and the apodomains of Fc⑀RI␣ over what was observed for the two proteins alone (Fig. 8C). The integrated enthalpy under the DSC curve of the Fc⑀⅐Fc␥-Fc⑀RI␣ complex at pH 5.0 was also greatly decreased compared with that observed at higher pH values. At pH 4.5, the mathematically combined DSC traces of apo-Fc⑀ and apo-Fc␥-Fc⑀RI␣ were nearly identical to the DSC trace measured for the mixture of the two proteins indicating that the interaction was completely abolished at pH 4.5 (Fig. 8D).
We speculated that Fc⑀ bound to Fc⑀RI (as if on the surface of a cell) may be afforded some protection to pH-dependent denaturation via its strong interaction with the receptor. To test this idea, Fc⑀ and Fc␥-Fc⑀RI␣ were mixed at pH 7.4 prior to dialysis at pH 4.5. The potential complex was removed from dialysis tubing and analyzed by DSC. Fc⑀ was found to be completely

DISCUSSION
In this study, we showed that the IgE-Fc demonstrated an unusual pH-dependent instability unlike that found for the IgG 1 -Fc. The pH-dependent unfolding of Fc⑀ was localized to the C⑀3C⑀4 domains and abolished the ability of IgE to interact with its high affinity receptor, Fc⑀RI, below pH 5.0. The thermostability of C⑀3C⑀4 decreased significantly as the pH was reduced from pH 7.4 to 5.0. Attenuated Fc⑀ thermostability did not significantly attenuate binding to Fc⑀RI until intrinsically unfolded Fc⑀ became present at pH 5.0 and below.
Unfolding Cooperativity and Stability Differences between Fc⑀ and Fc␥-The C␥2 and C␥3 domains of Fc␥ were found to unfold independently of one another, whereas the C⑀3 and C⑀4 domains of Fc⑀ unfolded cooperatively under all conditions tested. Analyses of the structures of Fc⑀ (14) and Fc␥ (36) reveal a similar amount of surface area buried at the C⑀3/C⑀4 interface compared with the C␥2/C␥3 interface (500 Å 2 for the C⑀3/C⑀4 interface versus 460 Å 2 for the C␥2/C␥3 interface). There is also a strong homology between the C␥2/C␥3 and C⑀3/C⑀4 interfaces, with near identical Fc residue positions burying surface area between the domains (in terms of the Kabat numbering system). However, the amino acid identity at the two interfaces is moderate and similar to the overall sequence identity between Fc␥ and Fc⑀ (34% identity, 64% homology). It is likely that differences between the cooperativity of unfolding for the IgG and IgE Fc regions are a consequence of specific amino acid interactions at the interfaces of the Fc domains because the contact areas and structural compositions are similar. It is possible that the IgG1-Fc in particular may not unfold cooperatively due to the extremely high apparent stability of the C␥3 domain. However, similar thermal unfolding experiments with human IgG2, IgG3, and IgG4 also demonstrated separate unfolding transitions for the C␥2 and C␥3 domains of these isotypes that contained relatively less stable C␥3 domains than    (16,18). We found that portions of the C⑀3 domain proteolyzed more slowly than C⑀4 at pH 4.5. It is possible that incomplete unfolding of the C⑀3 domain hinders the rapid proteolysis of C⑀3. When we removed C⑀4 from Fc⑀ using a recombinant fusion construct Fc␥-linker-C⑀2C⑀3, this protein demonstrated a strong tendency to aggregate and poor overall biophysical behavior. Although Fc⑀RI␣ has been shown to bind the C⑀3 domain, removal of C⑀4 from Fc⑀ abrogates the ability of Fc⑀ to interact with cell surface Fc⑀RI (37). Our results and those of others suggest C⑀3 folding is dependent on C⑀4 and explain why the presence of C⑀4 is critical for high affinity Fc⑀RI binding (16,18,37).
A cooperatively folded structure for C⑀3 in the presence of C⑀4 does not discount potential flexibility of various loops or flexibility at the C⑀2/C⑀3 interface important for the observed biphasic binding kinetics to Fc⑀RI␣ (11) and the existence of closed versus open forms of the receptor-binding region (15,38). In fact, we did observe changes in the CD spectrum of Fc⑀/Fc␥-Fc⑀RI (Fig. 5) upon formation of the complex. This is indicative of stabilization of ␤-structures or limited structural changes resulting from the "locking down" of flexible loops similar to what was observed by Sechi and co-workers (34).
The pH stability profiles of Fc␥ and Fc⑀ were also very different. The pH-dependent unfolding of C⑀3C⑀4 occurred at pH 5.0 likely due to the protonation of one or more basic amino acids that subsequently destabilizes the native fold. pHdependent protein stability is often observed for proteins whose folds either stabilize the burial of charged amino acids or are dependent on strong charge-charge interactions. In the pH range where Fc⑀ appears to undergo its unfolding transition (pH 4.4 -5.5), the most likely amino acid to undergo protonation would be histidine. Four of 9 histidines within Fc⑀ (His 422 , His 480 , His 490 , and His 528 ) were found to be Ն90% buried (14). One is in the C⑀3 domain and the remaining 3 are in the C⑀4 domain. Of the four Fc⑀ histidines, only His 528 is conserved at the homologous position of Fc␥ (36). The two histidines, His 528 in IgE and His 429 in IgG, also occupy nearly identical chemical environments suggesting that His 528 of IgE is unlikely to play a role in the intermediate pH unfolding transition of Fc⑀. Alternately, His 422 , His 480 , and His 490 of Fc⑀ are all occupied by polar residues at their homologous positions within IgG (Asn 325 , Ser 383 , and Tyr 391 , respectively) and more likely involved in the pH sensitivity of Fc⑀.
The stability of the C⑀2 domain was relatively insensitive to pH. Our DSC experiments demonstrate that the domain is natively folded between pH 2.5 and 9.0. C⑀2 likely remains folded at even lower pH values considering that the acidic groups within the domain should all be protonated by pH 2.5 and incapable of inducing large electrostatic changes in domain stability. C⑀2 contains significantly fewer charged groups than the C⑀3 and C⑀4 domains of Fc⑀. There are only 16 charged amino acids within C⑀2 and only one, Glu 270 , is Ͼ80% buried within the protein fold (15,19). In contrast, C⑀3 and C⑀4 contain 25 and 28 charged amino acids, respectively. In these two domains, 12 of the charged amino acids are Ͼ85% buried within the fold, and several additional charged amino acids within C⑀4 are Ͼ80% buried (15). The different charged residue compositions of C⑀2 and C⑀3C⑀4 as well as the extent to which they are buried within their folds very likely lead to the different pH stability behaviors.
pH Effects on the Thermodynamics of the Fc⑀/Fc␥-Fc⑀RI␣ Interaction-The affinity between Fc⑀ and the Fc␥-Fc⑀RI␣ fusion protein was extremely high under all pH conditions where the C⑀3C⑀4 domains of Fc⑀ remained folded. DSC is uniquely suited to the study of very strong binding interactions (29). There are a multitude of parameters that must be determined with reasonable accuracy to obtain the K D by DSC including some that are estimated (i.e. ⌬C P 0 unfolding for both the isolated C⑀3C⑀4 domains and domains 1 and 2 of Fc⑀RI␣). Due to the combination of experimental errors from the many variables, the magnitude of the K D represents the best that can be expected from the DSC measurements and likely includes an error in the range of 10 -100-fold. Encouragingly, the change in the Fc⑀/Fc␥-Fc⑀RI␣ DSC curves upon complex formation looked very similar to DSC simulations of a theoretical K D ϭ 10 Ϫ12 M interaction described by Brandts and Lin (29). The experiments described here were performed at protein concentrations 2 orders of magnitude lower than their simulated curves; therefore, the simulations would predict that the Fc⑀/Fc␥-Fc⑀RI␣ K D was close to 10 Ϫ14 M. The existence of two binding sites per single Fc␥-Fc⑀RI␣ fusion protein may improve the affinity of the construct for IgE over what has previously been observed for the monomeric soluble receptor. However, thermal unfolding of both Fc⑀ and Fc␥-Fc⑀RI␣ is irreversible and complicates the interpretation of the thermodynamic parameters as additional avidity could be induced during unfolding due to aggregation. Therefore, we simply state that the K D Ͻ 10 Ϫ12 M.
Whereas the magnitude of the affinity between Fc⑀ and Fc␥-Fc⑀RI␣ at pH 7.4 and 6.0 was relatively unchanged, the thermodynamic parameters of the interaction changed remarkably. The unfavorable ϪT⌬S A 0 (25°C) entropy term at pH 7.4 (ϩ0.7 kcal mol Ϫ1 ) changed by Ϫ10 kcal mol Ϫ1 upon titration to lower pH values (ϪT⌬S A 0 (25°C) ϭ Ϫ10 kcal mol Ϫ1 at both pH 6.0 and 5.5). These results suggest a number of possibilities that all relate to the presence of charged groups directly at the binding interface. First, apparent changes in enthalpy are likely to arise from differences in buffer ionization. The buffer was changed from phosphate at pH 7.4 to acetate at pH 6.0 and below. It is possible that uptake of one or more protons at the binding interface may account for much of the enthalpy change because phosphate has an intrinsically low enthalpy of ionization (39).
The titration of one or more residues between pH 7.4 and 6.0 may also affect the overall thermodynamics of binding and potentially the binding mode between the two proteins as well. No secondary structure changes were observed for apo-Fc⑀ or apo-Fc␥-Fc⑀RI␣ by CD upon titration from 7.4 to 5.2; therefore, the significantly altered binding thermodynamics could be the result of changes in charge density at or near the binding interface, local changes in the loop structures, or perhaps dynamic changes in quarternary structure. Consistent with the latter possibility, "open" and "closed" states have been shown to exist for Fc⑀ in crystal structures with Fc⑀RI␣ bound and absent (40). Titration to pH 6.0 and below also resulted in anomalous Fc⑀ binding to the carboxy dextran-coated surface of the SPR chips (not shown) suggesting a change in polarity on the surface of Fc⑀. Large changes in the heat capacity of the interaction also suggest a change in how the proteins are associating at the different pH values. Quarternary structure rearrangements at low pH that modify the surface exposure of apo-FcE or apo-Fc⑀RI␣ could explain the experimentally determined ⌬C P 0 changes. One could speculate that this is due to stabilization of the closed form at lower pH.
Biological Importance of the pH-dependent Fc⑀ Unfolding Event-It seems likely that the unusual pH sensitivity of Fc⑀, which differs greatly from what was observed for Fc␥, plays a role in its biological function or regulation. We have found the same pH sensitivity in an independent construct containing mouse Fc⑀, 7 suggesting that this pH sensitivity is a general property of IgE. Mildly acidic solutions of sodium acetate have been used to strip bound IgE from the surface of basophils suggesting that the pH unfolding/unbinding mechanism of Fc⑀ abolishes the ability of IgE to remain bound to Fc⑀RI (8,41).
As the half-life of Fc⑀RI-associated IgE is extremely long and differs from the short half-life of IgE in the serum (12), the regulation of mast cell surface-bound IgE is a key factor determining the duration of allergic sensitivity. High levels of monomeric IgE increase the amount of Fc⑀RI found on the surface of mast cells and basophils and induce enhanced survival of these inflammatory cells (6,10). Cross-linking of receptor bound IgE and subsequent endocytosis is one recognized method of antigen-specific cell surface IgE clearance (8). It has been shown that cross-linking leads to shuttling of endocytic vesicles containing IgE/Fc⑀RI from prelysosomal to lysosomal compartments (8). Additionally, IgE/Fc⑀RI cross-linking leads to the accumulation of IgE fragments in the lysosomal-like secretory granules (9). These mechanisms of antigen-specific IgE turnover appear to involve the introduction of IgE to moderately acidic lysosomal compartments. Under these conditions, we would predict that the Fc⑀ portion of IgE unfolds and dissociates from its receptor. This release may lead to IgE proteolysis within lysosomal compartments, whereas the less pH-sensitive Fc⑀RI␣ may recycle back to the cell surface. Thus, the pH sensitivity of Fc⑀ may play a key role in the turnover of receptor-bound IgE.