Bivalent Ligands with Rigid Double-Stranded DNA Spacers Reveal Structural Constraints on Signaling by FcεRI¹

Allergic reactions are initiated by Ag-induced cross-linking of IgE Abs that are bound to high-affinity receptors, FcεRI, on the surface of mast cells and basophils. FcεRI belongs to a family of multisubunit immune recognition receptors that include TCRs, B cell receptors, and FcRs. Many members of this family share similar signal transduction and cellular activation pathways (1, 2). Although clustering of these receptors by their appropriate ligands appears necessary for cell activation, a detailed understanding of structural constraints in this process is lacking (3).

The earliest detectable signaling event upon FcεRI cross-linking is phosphorylation of FcεRI β and γ subunit immunoreceptor tyrosine-based activation motif regions by the Src family tyrosine kinase, Lyn (4). Phosphorylation of the γ subunit leads to recruitment and activation of Syk kinase via its tandem Src homology domain 2 (5, 6). This results in tyrosine phosphorylation of adapter proteins that participate in the activation of phospholipase C (PLC)⁴γ1 and PLCγ2 (7, 8). These key enzymes hydrolyze phosphatidylinositol-4,5-bisphosphate to generate inositol-1,4,5-trisphosphate and 2,3-diacylglycerol, activators of Ca²⁺ mobilization and protein kinase C, respectively. Activation of this and other signaling pathways culminates in the release of histamine and other mediators of the allergic response.

Previous studies on FcεRI (9) and the TCR (10) indicated that relatively subtle structural differences in ligand binding and receptor cross-linking can affect the efficacy of signal transmission by members of this receptor family. In other studies, well-defined bivalent ligands have been used to study the relationship between binding and cross-linking of IgE-FcεRI complexes and subsequent cell activation. In some of these studies, the theoretically predicted extent of equilibrium cross-linking was found satisfactory to account for the biological response (11, 12, 13), whereas, in others, analysis of the data using this theory led to the conclusion that cyclic cross-links formed by some bivalent ligands and bivalent IgE bound to FcεRI can limit effective signaling (14, 15). Consistent with these latter findings, Schweitzer-Stenner et al. (16) showed that flexible bivalent ligands ≥130 Å in length form cyclic monomers with IgE in solution, whereas bivalent ligands <45 Å in length efficiently form cyclic dimers.

To extend these studies and gain additional insights into structural requirements for effective cross-linking, we have embarked on a systematic study using rigid bivalent ligands. For this purpose we used dsDNA as a versatile spacer. Largely because of base pairing, dsDNA forms an α helix that is rigid, with a persistence length of at least 500 Å (17), and thus provides a practical means to create bivalent ligands with rigid spacers of differing lengths. For specific binding, dinitrophenyl (DNP) DNA polymers were synthesized as a series of ligands whose lengths and flexibilities were varied. Both monovalent and bivalent forms of dsDNA ligands were constructed and characterized in binding and functional studies. We found that bivalent ligands with shorter rigid lengths of 44–51 Å stimulate degranulation responses, whereas ligands of longer rigid lengths (68–102 Å) do not. Tyrosine phosphorylation of FcεRI β, Syk, and linker for activation of T cells (LAT) is stimulated strongly by all of these ligands, but Ca²⁺ responses are small, indicating that structural constraints limit coupling between stimulated tyrosine phosphorylation and more downstream signaling events.

Oligonucleotide sequences in the range of 13–30 mer were selected to achieve high melting temperatures while minimizing self-annealing and misalignment between complementary strands (18). For this purpose, the DNAstar software package (DNAstar, Madison, WI) was used to select the sequences shown in Table I. These oligonucleotide sequences were synthesized by Synthegen (Houston, TX) or Sigma-Genosys (The Woodlands, TX) with a six-carbon, amine-modified 5′ linker and purified by reversed-phase HPLC according to the manufacturer’s instructions.

Coupling of DNP-succinimidyl ester to the 5′ amine-modified termini of the oligonucleotides was performed using a FluoReporter DNP-X Amine Labeling kit (Molecular Probes, Eugene, OR) with minor modifications. In brief, ethanol precipitation of the oligonucleotides was omitted, and the DNP modification reaction proceeded overnight at room temperature. The reaction was terminated by removal of the unreacted DNP substrates by repetitive Microcon filtration (Millipore, Bedford, MA) or spin column filtration (Amersham Pharmacia Biotech, Piscataway, NJ). DNP-labeled DNA and unlabeled DNA were recovered as the retentate of the Microcon filtration or eluted from the spin column, then loaded on a reversed-phase C-18 HPLC column (Waters, Milford, MA) for further purification. Well-separated peaks were resolved by a linear gradient of 200 mM triethylammonium acetate (pH 6.05) and a 60/40 mixture of acetonitrile/dH₂0, and peak assignments were verified using matrix-assisted laser desorption ionization/time-of-flight mass spectrometry (BioResource Center, Cornell University, Ithaca, NY). These ssDNA ligands were spin vacuum dried, resuspended in 10 mM Tris, 1 mM EDTA, and 135 mM NaCl (pH 8.2), dried again, redissolved to remove traces of triethylammonium buffer, and stored at 4°C. Single-stranded ligand preparations were characterized with UV/visible absorbance spectrometry and denaturing PAGE. Ligand preparations that demonstrated a single distinct oligonucleotide band with appropriate DNP:DNA ratios (18) were used for annealing to form dsDNA ligands.

Bivalent double-stranded ligands were formed using equimolar concentrations of 5′ DNP-labeled DNA and its complementary 5′ DNP-labeled DNA strand. To form the monovalent double-stranded ligands, a 10% excess of unlabeled oligonucleotide molecules was added to the DNP-labeled complementary strands to enhance the likelihood that all of the DNP-labeled molecules were incorporated into double-stranded ligands. DNA mixtures were heated to 100°C for 5 min and allowed to cool to ambient temperature (19).

Spectroscopic measurements, HPLC analyses, and native gel electrophoresis were used to characterize the ligands qualitatively and quantitatively. DNP:DNA ratios were determined spectroscopically for the monovalent and bivalent ligands using the molar extinction coefficient of 1.7 × 10⁴ M⁻¹cm⁻¹ for ε_360(DNP). A_260(DNA) = 1 corresponds to 33 μg/ml DNA for ssDNA, and A_260(DNA) = 1 corresponds to 50 μg/ml DNA for dsDNA, where A_260(DNA) = A_260(DNP+DNA) − (0.53 × A_360(DNP)). The average monovalent ligand DNP:DNA ratio was calculated to be 0.9 ± 0.2 and the average bivalent dsDNP:DNA ratio was determined to be 1.7 ± 0.3 (Table II). These values are within experimental uncertainty of the expected ratios of 1 and 2, respectively.

Monovalent and bivalent DNP-dsDNA were electrophoresed on 20% native polyacrylamide gels in 89 mM Tris base, 89 mM boric acid, 2 mM EDTA (pH 8). Fig. 1,A shows a representative polyacrylamide gel run under these conditions and stained with SYBR Green (Molecular Probes), then imaged with a PhosphoImager (Molecular Dynamics, Sunnyvale, CA). The monovalent DNP forms of the ligands migrate slightly faster than the bivalent DNP forms. The lighter, faster-migrating bands in each lane are the ssDNA oligomers, which are in excess in the monovalent DNP ligand preparations, as expected. Relative densities of bands in Fig. 1,A were further quantified using the program ImageQuant (Molecular Dynamics). Preparations with <80% of their DNA in appropriate double-stranded bands (e.g., the monovalent 20-mer sample in Fig. 1,A) were not used for further studies. Bivalent and monovalent dsDNA ligands were also evaluated by gel permeation chromatography using a Pharmacia Superose 12 column (Amersham Pharmacia Biotech). Results consistent with those in Fig. 1 A confirmed that the ligands used were of expected size and purity (18).

Mouse monoclonal anti-DNP IgE was affinity purified from hybridoma H1 26.82 (20) as previously described (21). Bispecific IgE was generated in our laboratory as described by Subramanian (22). Briefly, quadroma cells were formed from the fusion of anti-DNP hybridoma H1 26.82 cells and anti-dansyl hybridoma 27.74 cells (23). The bispecific anti-DNP, anti-dansyl IgE secreted from the quadroma was purified by sequential affinity chromatography. Other experiments that characterize our bispecific IgE have been reported previously (24).

To assess chromatographically the binding and cross-linking of DNP-dsDNA ligands to anti-DNP IgE, a 2- to 3-fold molar excess of monovalent or bivalent ligands were incubated with 1 μM IgE for 1 h at room temperature. Samples (200 μl) were eluted through a 10 × 300 mm HPLC size-exclusion Superose 6 gel permeation column (Amersham Pharmacia Biotech) in 20 mM sodium phosphate, 100 mM sodium sulfate, and 0.02% sodium azide (pH 7.4) at a flow rate of 0.75 ml/min; protein absorbance was monitored at 280 nm. Molecular weight standard proteins from Bio-Rad (Hercules, CA) were used for calibration.

Equilibrium binding experiments with FITC-labeled IgE were conducted as described previously by Erickson et al. (25). Briefly, fluorescence measurements were made on an SLM 8000 fluorometer in time-based acquisition mode. Excitation and emission wavelengths were 490 and 520 nm, respectively. Equilibrium titrations with soluble and cell-bound FITC-IgE were conducted at 37°C with continual stirring in the cuvette. Experiments with cells were conducted in balanced salt solution (BSS; 20 mM HEPES (pH 7.4), 135 mM NaCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 5.6 mM glucose containing 0.1% gelatin) in the presence of 2 μM cytochalasin D to prevent receptor internalization (26).

RBL-2H3 cells (27) were maintained in Eagle’s MEM supplemented with 20% FCS, sensitized with excess anti-DNP IgE, and plated at a density of 2.5 × 10⁵ cells/well in 48-well plates for at least 90 min at 37°C. Adherent cells were washed and ligands were added in BSS. Unless otherwise indicated, 2 μM cytochalasin D was included during stimulation to enhance degranulation stimulated by bivalent ligands (15). After 1 h at 37°C, aliquots of supernatant were taken from each well to assay the extent of β-hexosaminidase release from the cells as previously described (15). Stimulated release of enzyme activity is expressed as a percentage of the total cellular β-hexosaminidase activity present in cell lysates after solubilization in 0.5% Triton X-100.

Cells sensitized with saturating concentrations of IgE were plated at 2 × 10⁶ cells/well in six-well tissue culture plates for at least 2 h at 37°C. Adherent cells were washed and incubated with BSS containing ligand for 10 min at 37°C. In most experiments, 2 μM cytochalasin D was present during stimulation similar to the degranulation experiments. Cell activation was terminated by rapid removal of the supernatant and addition of ice-cold lysis buffer (20 mM Tris (pH 8), 100 mM NaCl, 60 mM sodium pyrophosphate, 0.04 U/ml aprotinin, 0.02% sodium azide, 1 mM PMSF, or 4-((2-aminoethyl)benzenesulfonylfluoride)) containing 0.5% Triton X-100. Wells were scraped with the pipette tip and lysates were centrifuged for 5 min at 13,000 × g to pellet insoluble debris. Anti-Syk immunoprecipitations were conducted as previously described (15). Aliquots of lysates were mixed with 5× SDS sample buffer (50% glycerol, 0.25 M Tris base (pH 6.8), 5% SDS, 0.5% bromophenol blue) then boiled and centrifuged for 5 min at 13,000 × g. Supernatants were either loaded on a polyacrylamide gel or stored at −20°C for analysis at a later time. For some experiments, cells were lysed directly in SDS sample buffer following removal of the stimulation solutions, and these SDS lysates were immediately boiled and centrifuged, then loaded onto a polyacrylamide gel.

Samples from whole cell lysate preparations and immunoprecipitations were analyzed by Western blotting as previously described (15), except that blots were probed overnight at 4°C with a 1/10,000 dilution of HRP-conjugated anti-phosphotyrosine mAb 4G-10 (Upstate Biotechnology, Lake Placid, NY). For quantitative analysis, nonsaturating blots were scanned using Un-Scan-It (Silk Scientific, Orem, UT) in its linear optical range.

RBL-2H3 cells were suspended in BSS containing 1 mg/ml BSA and 0.25 mM sulfinpyrazone at a concentration of 1 × 10⁷ cells/ml, then incubated with 3 μg/ml indo-1 AM (Molecular Probes) for 1 min at 37°C, diluted, and sensitized with excess IgE for 1.5 h at 37°C. These cells were washed in BSS containing BSA and sulfinpyrazone and then resuspended to a final concentration of 2 × 10⁶ cells/ml. Measurements of indo-1 fluorescence were made on an SLM 8000 fluorometer in time-based acquisition mode as previously described (28). Excitation and emission wavelengths were 330 and 399 nm, respectively, for detection of intracellular indo-1 Ca²⁺ fluorescence changes. Cells were stirred at 37°C in the presence or absence of 2 μM cytochalasin D, and ligands were added as indicated.

To evaluate structural features of IgE-FcεRI cross-linking leading to cell activation, we prepared and characterized a series of dsDNA bivalent ligands in which the dsDNA acts as a linear, rigid spacer between DNP haptens attached via flexible linkers (Tables I and II), as well as their respective monovalent dsDNA analogs. As described in Materials and Methods, the monovalent and bivalent dsDNA ligands were characterized with gel electrophoretic, spectroscopic, and chromatographic analyses (Fig. 1,A). We then assessed the efficacy of these ligands to form stable cross-links with IgE in solution. For this purpose, DNP-dsDNA ligands were mixed with anti-DNP IgE and isocratically eluted through a size-exclusion gel permeation column. Fig. 1,B illustrates these results with chromatographs of IgE alone and combined with monovalent 15-mer, bivalent 15-mer, and bivalent 20-mer ligands. As monitored by protein absorbance at 280 nm, IgE with a monovalent 15-mer ligand elutes slightly ahead of IgE alone, probably due to the small contribution of bound ligand to the IgE mass (Fig. 1,Ba). In contrast, IgE incubated with bivalent ligands elutes in broader bands that are shifted to substantially shorter retention times indicative of oligomeric complexes of IgE. For the complexes with a bivalent 15-mer ligand, the peak is slightly ahead of the 670-kDa standard but is skewed toward a distribution of smaller species, indicative of a mixture of small oligomers dominated by trimers and dimers of IgE (Fig. 1,Bb). The complexes formed with the bivalent 20-mer ligand appear somewhat smaller on average than those with the bivalent 15-mer ligand (i.e., primarily dimers), with a shoulder in the region of IgE monomers (Fig. 1 Bc). In similar experiments, the bivalent rigid 30-mer and bivalent flexible 30-mer ligands also cause the formation of oligomeric IgE complexes of IgE of similar sizes and amounts (29). In multiple experiments, no consistent correlation between the length of the rigid spacer and the size distributions of oligomeric complexes was observed. Because gel permeation can separate dissociated ligands from IgE during the course of chromatography, the extent of oligomerization detected does not represent the equilibrium distribution of these complexes. Nevertheless, the results indicate that all of the bivalent dsDNA ligands cause the formation of stable IgE oligomers with a limited distribution of sizes. This latter feature is consistent with the efficient formation of small cyclic oligomers.

To evaluate affinities of the DNP-dsDNA ligands binding to anti-DNP IgE, a series of equilibrium titrations was performed with IgE in solution and bound to FcεRI on the RBL cell surface. We observed that they all bound with similar rates that typically reached a steady state of binding after a few seconds (data not shown). The equilibrium titration data were fit to determine an apparent K_d, using methods previously described (25). As anticipated from the structural identity near the DNP groups of all of our ligands, the representative data in Fig. 2,A show that the binding of monovalent 15- and 20-mer ligands to IgE in solution are nearly identical. Likewise, the monovalent 13- and 30-mer ligands bind with indistinguishable high affinity to solution IgE as summarized in Table II. Also shown in Fig. 2,A are representative data for monovalent DNP-dsDNA binding to cell-bound IgE. These and other data summarized in Table II confirm that the affinity for binding to IgE-FcεRI on cells (K_d = 10–20 nM) is not significantly different from that for IgE in solution, consistent with previous studies on monovalent ligands (25). Fig. 2,B shows the binding of the bivalent 15- and 20-mer ligands to IgE in solution compared with IgE on cells. For IgE in solution, the average K_d values are 13 ± 4 and 12 ± 6 nM for the bivalent 15- and 20-mer ligands, respectively. The bivalent 13- and 30-mer ligands have a similar apparent affinity, 16 ± 2 and 14 ± 2, respectively, as summarized in Table II. The similarity of all of these values to those for the monovalent DNP-DNA ligands indicates that the two ends of these bivalent ligands bind independently to IgE in solution. In contrast, substantial affinity enhancement is observed for bivalent ligands and IgE bound to FcεRI on cells, as illustrated for bivalent 15- and 20-mer ligands in Fig. 2 B. The average apparent K_d for the bivalent 15-mer ligand is 3 ± 3 nM and the average apparent K_d for the bivalent 20-mer ligand is 4 ± 3 nM, compared with the values of 12–13 nM for these bivalent ligands binding to IgE in solution. Because this difference is observed for bivalent but not monovalent ligands, we conclude that intermolecular cross-linking of IgE-FcεRI locally concentrated on cells is substantial and is the cause of the increased apparent affinity for the bivalent ligands.

These binding results can be used to estimate the ligand concentration for maximal cross-linking according to the theory of Dembo and Goldstein (30): C_Tmax = (K_d/2) + IgE_Total (equation 1), where C_Tmax is the total ligand concentration that corresponds to maximal cross-linking and IgE_Total is the total IgE concentration (14). Accordingly, the total ligand concentration for maximal cross-linking on cells is ∼6 nM for the bivalent DNP-dsDNA ligands under our experimental conditions. This value provides a reference for the interpretation of the degranulation results that are described below.

Responses were compared over a wide range of dsDNA ligand concentrations. Fig. 3 A summarizes normalized results for three or more experiments with each ligand, including >10 experiments each for the bivalent 15-mer and bivalent 20-mer ligands. None of the monovalent ligands induces cellular degranulation significantly above baseline unstimulated levels, even at very high ligand concentrations. The shorter bivalent ligands (13 mer and 15 mer) stimulated similar amounts of degranulation in RBL cells with optimal responses at ∼100 nM. In contrast, the magnitude of the degranulation response is substantially less for the bivalent 20-mer ligand, and no stimulated degranulation is observed for the bivalent 30-mer ligand, except for a small amount at the very highest concentration tested. In these experiments, optimal doses of multivalent Ag stimulate an average of 67% β-hexosaminidase release, whereas the bivalent 15-mer ligand stimulates an average response of 19 ± 10% at an optimal dose of 100 nM, and spontaneous (unstimulated) release was usually ≤2%. Thus, maximal degranulation stimulated by the most effective bivalent dsDNA ligands is generally about one-third of that stimulated by multivalent Ag.

To test whether the lack of stimulated degranulation with the bivalent dsDNA 30-mer ligand is due to the rigid separation length of ∼100 Å for that ligand, we prepared and characterized a ligand of the same extended length that contains a flexible spacer in the middle of the rigid dsDNA (Table I). As shown in Fig. 3 B, this flexible 30-mer ligand stimulates degranulation similarly to the rigid bivalent 15-mer ligand when tested in the same experiment, contrasting sharply with that observed for the rigid 30-mer ligand. This indicates that a flexible hinge in the middle of an ∼100 Å rigid spacer permits signaling for degranulation not observed for the inflexible rod.

The concentrations of the bivalent 13- and 15-mer ligands that stimulate maximal degranulation (∼100 nM) are substantially greater than the concentration predicted for maximal cross-linking of ∼6 nM, according to equation 1 and our equilibrium binding results. Little or no stimulated degranulation is observed for any bivalent dsDNA ligand at 6 nM, indicating that the cross-linked species of IgE-FcεRI predominating at maximal cross-linking by these ligands does not trigger a degranulation response. Similar results were observed previously with bivalent N,N′-bis(DNP-caproyl-l-tyrosyl)-l-cysteine, and a detailed analysis of these data revealed that cyclic dimers containing two IgE-FcεRI and two ligands are the dominant, nonstimulatory cross-linked species at maximal cross-linking (14). Therefore, to restrict cross-linked complexes to linear dimers we used a bispecific IgE that recognizes DNP groups in one Fab binding site and dansyl groups in the other Fab binding site (22, 24). This bispecific IgE was prepared and purified from a quadroma cell line derived from the same anti-DNP hybridoma that secretes the bivalent anti-DNP IgE used in our other experiments, and the DNP binding affinity for the bispecific IgE was previously confirmed to be identical to the DNP binding affinity of the parental bivalent IgE (22). Thus, bispecific IgE functions as a monovalent receptor for the DNP-dsDNA ligands, such that bispecific IgE cross-linked with bivalent ligand can form only linear dimers.

For these experiments, we compared the 15- and 20 mer ligands, as they differ in length by only 17 Å yet show a marked difference in their capacity to stimulate degranulation with bivalent IgE. Fig. 3,Ca is representative of three separate degranulation dose-response curves for the bivalent 15- and 20-mer ligands on RBL cells sensitized with saturating concentrations of bispecific IgE. These can be compared with dose-response curves for these ligands with the same cells sensitized with the parental bivalent anti-DNP IgE (Fig. 3,Cb). Most notably, the concentration of ligand that produces maximal cross-linking with bispecific IgE is significantly shifted to lower ligand concentrations compared with results observed using bivalent IgE. Furthermore, the maximum for degranulation with bispecific IgE is observed at the concentration that is predicted to produce maximal cross-linking with these ligands (6 nM) according to equation 1. The magnitude of the degranulation response with bispecific IgE for both the 15- and 20-mer ligands is small and virtually identical (Fig. 3,Ca), unlike the case with bivalent IgE (Fig. 3,Cb). Thus, the degranulation response to linear dimers formed by these ligands and bispecific IgE is weak but measurable, and it is not as sensitive to the rigid spacing lengths of these ligands as it is for the functionally active cross-linked complexes formed with bivalent IgE. These results imply that the stronger degranulation responses seen with bivalent anti-DNP IgE and bivalent 13- and 15-mer ligands at higher concentrations (Fig. 3 A) are due to cross-linked IgE-FcεRI complexes larger than dimers, and these larger complexes show a stronger dependence on rigid spacer length. Small, cyclic complexes (i.e., cyclic dimers) are likely to be the predominant species at ligand concentrations yielding maximal cross-linking (6 nM), and they are apparently ineffective in stimulating degranulation for all dsDNA spacer lengths, as they were for N,N′-bis(DNP-caproyl-l-tyrosyl)-l-cysteine (14, 15).

To determine whether ligand length-dependent differences in degranulation are due to differences in the earliest signaling events, we investigated stimulated tyrosine phosphorylation. Fig. 4,A shows the time course for this activity in whole cell lysates induced by multivalent Ag (100 ng/ml), the bivalent 15-mer ligand (100 nM), and the monovalent 15-mer ligand (100 nM), at optimal doses for stimulated degranulation. Under these conditions, the amount of tyrosine phosphorylation is maximal 10 min after addition of the bivalent ligand, and multivalent Ag stimulates a similar amount of tyrosine phosphorylation that is maximal between 2 and 10 min of incubation. Several proteins are tyrosine phosphorylated in response to FcεRI cross-linking, including the β subunit of FcεRI and the adapter protein LAT, as indicated in Fig. 4 A. The monovalent 15-mer ligand does not cause detectable phosphorylation of LAT or FcεRI β, and increased phosphorylation of higher m.w. proteins apparent with this ligand at longer times was not consistently observed in other experiments. In contrast, 100 nM bivalent 15-mer ligand consistently stimulates tyrosine phosphorylation of FcεRI β, which is phosphorylated by Lyn (4), and LAT, which is phosphorylated by Syk (31), similarly to multivalent Ag.

For subsequent experiments, we chose the 10-min time point corresponding to maximum cellular phosphorylation to compare amounts of tyrosine phosphorylation of FcεRI β, LAT, and Syk stimulated by the bivalent ligands of different lengths at 100 nM. As shown in a representative Western blot in Fig. 4,B, all bivalent ligands stimulate similar phosphorylation of FcεRI β subunit and LAT. Furthermore, in multiple experiments this stimulated tyrosine phosphorylation is similar in magnitude to that for multivalent Ag at an optimal dose and time. Fig. 4 C shows the results of quantitative analysis of LAT phosphorylation for five experiments and confirms that all of the DNP-dsDNA ligands stimulate amounts of tyrosine phosphorylation similar to multivalent Ag.

To compare further the tyrosine phosphorylation stimulated by these different ligands, we immunoprecipitated Syk tyrosine kinase and examined its autophosphorylation following receptor aggregation by either DNP-BSA (100 ng/ml) or the bivalent DNA ligands. As shown in Fig. 5,A, stimulated Syk phosphorylation is substantial for all of the bivalent ligands relative to the Ag-mediated response. Scanning and quantitation from six independent experiments showed that tyrosine phosphorylation stimulated by bivalent 13- and 15-mer ligands is not significantly different from that for multivalent Ag, although that for the bivalent 20- and 30-mer ligands is somewhat less (Fig. 5 B). These small differences in Syk phosphorylation follow trends similar to the degranulation results with these ligands. However, the most striking observation is that Syk tyrosine phosphorylation caused by these bivalent ligands (and that of FcεRI β and LAT described above) are much more similar to that for multivalent Ag than are the degranulation responses. In particular, the 20- and 30-mer ligands did not stimulate significant amounts of degranulation, yet they were quite effective at stimulating early tyrosine phosphorylation events.

The degranulation results described above reveal that spatial parameters of the bivalent DNP-dsDNA ligands determine the magnitude of the degranulation response, yet early signaling events as exemplified by tyrosine phosphorylation of FcεRI β, LAT, and Syk are robust for all of these ligands at the concentrations for maximal degranulation (100 nM). Thus, we investigated whether these ligands could stimulate Ca²⁺ mobilization commensurate with their tyrosine phosphorylation responses. Fig. 6 shows representative time courses for the Ca²⁺ response of IgE-sensitized cells stimulated by multivalent Ag or by the bivalent 13-mer ligand (100 nM) in the absence or presence of cytochalasin D. As previously shown (28), cytochalasin D increases the Ag-stimulated Ca²⁺ response in these cells. In the presence of cytochalasin D, only a small, transient Ca²⁺ response is seen with the 13-mer ligand (Fig. 6,D), and this is not detectable in the absence of cytochalasin D (Fig. 6 B). No detectable Ca²⁺ mobilization was seen with 100 nM bivalent 20- or 30-mer ligands, even in the presence of cytochalasin D (data not shown). These results indicate that robust stimulation of tyrosine phosphorylation by these bivalent dsDNA ligands fails to cause even transient Ca²⁺ responses that are detectable. Thus, reduced degranulation responses observed with these ligands are likely to be the result of very limited, asynchronous Ca²⁺ mobilization, suggesting that the structural constraints imposed by these rigid bivalent ligands prevents effective coupling of stimulated tyrosine phosphorylation to Ca²⁺ mobilization.

This study took advantage of the rigid, linear structure of dsDNA in the oligonucleotide length range of 13–30 bp to create bivalent ligands with rigid spacers of 44–102 Å. We showed that all of these ligands bind and cross-link IgE on cells and in solution with similar efficiency (Figs. 1 and 2 and Table II). They trigger RBL cell degranulation in a length-dependent manner, with the shortest ligands stimulating about one-third as much as multivalent DNP-BSA (Fig. 3). The incapacity of the longest ligand to stimulate degranulation over a wide range of concentrations is overcome by introducing a flexible spacer in the middle of the dsDNA structure (Fig. 3 B), and this argues that the rigid 30-mer ligand does not allow sufficient proximity of cross-linked IgE-FcεRI complexes for significant stimulation of the degranulation response. For all of these ligands, a large portion of the flexible spacers that separate DNP groups from the rigid dsDNA segments are expected to be buried in the Ab-combining sites (25). Surprisingly, all of these ligands, including the rigid 30-mer ligand, stimulate robust tyrosine phosphorylation responses, including Syk activation. This indicates that the structural constraint on stimulated degranulation occurs at a step that is downstream of these early tyrosine phosphorylation events. Only limited Ca²⁺ mobilization is observed even with the shorter ligands that trigger significant degranulation responses, indicating that signaling by these ligands is limited at a step upstream of this essential process.

Based on these results, we propose a structural hypothesis in which the spacing between directly cross-linked receptor complexes is an important determinant for efficient activation of a step proximal to Ca²⁺ mobilization. Regulation of PLCγ activity is a likely candidate. Because the products of PLCγ-mediated hydrolysis of phosphatidylinositol-4,5-bisphosphate, i.e., inositol-1,4,5-trisphosphate and 2,3-diacyl glycerol, are diffusable mediators, these second messengers are not expected to be restricted by the spacing of cross-linked receptors. However, activation of PLCγ by IgE receptors is a complex process that requires tyrosine phosphorylation by Syk (32) and possibly Bruton’s tyrosine kinase (2) and depends on adapter proteins for this process, including LAT (33) and SLP-76 (34). Thus, it is possible that PLCγ bound to one receptor complex is transphosphorylated and thereby activated by Syk associated with an adjacent, cross-linked receptor. This process might well be limited if these cross-linked receptors are held too far apart. As another possibility, recent experiments have implicated activation of the Rho family GTPase Cdc42 in FcεRI-mediated Ca²⁺ mobilization (35), and this might also be critically affected by the spacing between cross-linked FcεRI.

An alternative hypothesis that could explain the poor downstream signaling observed, especially with the longer dsDNA ligands, is kinetic proofreading, in which early signaling by these ligands is too transient to permit adequate activation of more downstream events (24, 36). This explanation does not appear to be applicable in the present case, because the binding of all of these dsDNA ligands is similarly tight (Table II) and the time course of tyrosine phosphorylation stimulated by these ligands is at least as sustained as that with multivalent Ag (Fig. 4 and data not shown). Furthermore, bivalent dsDNA ligands of different spacer lengths do not show obvious differences in their binding kinetics (data not shown). Instead, we favor the structural hypothesis that implies spatial constraints for efficient coupling between early activation events (i.e., stimulated tyrosine phosphorylation) and more downstream events (i.e., activation of PLCγ and/or Ca²⁺ mobilization). Consistent with this structural hypothesis are the length-dependent differences in degranulation that we observe and the activity of the flexible 30-mer ligand compared with the rigid 30-mer ligand.

Our results in Figs. 4 and 5 further indicate that, unlike Ca²⁺ mobilization and degranulation, the initial cross-link-dependent tyrosine phosphorylation of FcεRI, LAT, and Syk does not require close proximity of directly cross-linked IgE-receptor complexes. For example, activation of Syk (indicated by LAT phosphorylation, Fig. 4) is only slightly reduced with the longest DNP-dsDNA ligand, which separates cross-linked IgE-combining sites by as much as 100 Å and fails to stimulate degranulation. Although the actual receptor-receptor separation distances are not known, these results do not appear to support a transphosphorylation model for signal initiation, which predicts that Lyn bound to one receptor in a cross-linked aggregate phosphorylates an adjacent, tethered receptor (37). Rather, the results are more consistent with the lipid raft model (38), in which the local proximity of cross-linked FcεRI and active Lyn in liquid-ordered membranes promotes receptor phosphorylation without the structural constraints implicit in the transphosphorylation model.

Using a bispecific IgE that is effectively univalent for DNP ligands, we could evaluate signaling by linear dimers formed with the dsDNA bivalent ligands (Fig. 3,C). Only small amounts of degranulation were observed in this situation, but the peak of the degranulation response was clearly maximal at ∼6 nM, the concentration predicted to cause maximal cross-linking from our binding results and the theory of Dembo and Goldstein (30). Interestingly, the bivalent 15- and 20-mer ligands yielded indistinguishable degranulation dose-response curves in this situation (Fig. 3,Ca), suggesting that stimulation differences observed for these ligands with bivalent IgE (Fig. 3 Cb) are related to structural constraints imposed by the formation of cyclic dimers that are relieved with larger aggregates at the higher concentrations (∼100 nM). Previous results with bivalent IgE and bivalent linear polymers of avidin (39) did not show the same length-dependent restriction on degranulation responses that we observe with the DNP-dsDNA ligands, and it is likely that those avidin polymers did not form cyclic complexes efficiently because of the protein mass in the spacer. Our results with the DNP-dsDNA ligands are more similar to those obtained with one of three monoclonal anti-FcεRI α characterized by Ortega and colleagues (9, 40). This particular mAb, designated H10, bound and cross-linked FcεRI as well as the other two, and it stimulated substantial FcεRI tyrosine phosphorylation but weaker PLC activation, Ca²⁺ mobilization, and degranulation compared with the other two mAb, or to multivalent Ag (40). H10 is an IgG2b mAb, whereas the other two mAb are both of the IgG1 subclass, and it is possible that segmental flexibility of the these different subclasses (41) influences geometrical constraints of the FcεRI dimers formed.

In summary, our data show that structural constraints in cross-linked IgE receptor complexes can strongly affect signaling, leading to stimulated degranulation. More specifically, the bivalent dsDNA ligands we prepared and characterized reveal a critical dependence on the interreceptor spacing for productive downstream signaling but only limited dependence on this spacing for the earliest signaling events activated. Future experiments will be aimed at understanding the particular signaling step that is most strongly influenced by these structural constraints.

We are grateful for helpful discussions with Byron Goldstein (Los Alamos National Laboratory).