Trapping IgE in a closed conformation by mimicking CD23 binding prevents and disrupts FcεRI interaction

Anti-IgE therapeutics interfere with the ability of IgE to bind to its receptors on effector cells. Here we report the crystal structure of an anti-IgE single-domain antibody in complex with an IgE Fc fragment, revealing how the antibody inhibits interactions between IgE and the two receptors FcεRI and CD23. The epitope overlaps only slightly with the FcεRI-binding site but significantly with the CD23-binding site. Solution scattering studies of the IgE Fc reveal that antibody binding induces a half-bent conformation in between the well-known bent and extended IgE Fc conformations. The antibody acts as functional homolog of CD23 and induces a closed conformation of IgE Fc incompatible with FcεRI binding. Notably the antibody displaces IgE from both CD23 and FcεRI, and abrogates allergen-mediated basophil activation and facilitated allergen binding. The inhibitory mechanism might facilitate strategies for the future development of anti-IgE therapeutics for treatment of allergic diseases.

outsider. For example in "The human FcεRI is expressed as a tetramer or a trimer lacking the signal amplifying β-subunit." It is not clear whether only the trimer or both the trimer and the tetramer lack the subunit.
Page 5, 1st paragraph, analysis of the surface plasmon resonance experiment: The combination of a KD value with a χ2 is nonsensical. The KD should have a margin of confidence, whereas χ2 is a property of the fit (and often meaningless).
Page 9, SAXS analysis: It is not possible to draw any conclusion of the ensemble of solution structure from C ORAL modeling. C lustering of the results only imply that the reconstruction is stabl e, but C ORAL by design will always find the SINGLE model that describes the data best. Less clustered solutions imply that the reconstruction is more ambiguous. To address the presence of several configurations, one should apply ensemble based tools such as MES or EOM. One could for example imagine a coexistence of strongly bent and extended confirmations of IgE Fc which could give the same SAXS signal as a moderately bent confirmation. Further, there is absolutely no point in discussing differences in chi-squared values. This can be easily understood that the same sample at lower concentration generally gives a lower chi -squared for the same model. A somewhat more meaningful estimator of fit-quality would be the p-value from the C ORMAP test (Franke et al., 2014). Page 11, 2nd paragraph, discussion of SAXS data: Following the above, it would be more appropriate to state that the data can be described by such models, as you cannot exclude the existence of other solutions. Page 13, Acknowledgements: I would strongly recommend to follow the acknowledgment guidelines for the beamlines used for this study: http://www.diamond.ac.uk/Beamlines/Mx/C ommon/C ommon-Manual/Acknow ledging-MXbeamlines.html http://www.esrf.eu/UsersAndScience/UserGuide/Preparing/Guidelines ToUsers Page 16, 2nd paragraph (Affinity measurements): The model underlying the KD calculation should be specified.
Page 18, 2nd paragraph (Small angle X-ray scattering data collection and rigid body modeling): In the first line, 026 probably refers to sdab 026? The structure of the data analysis section is very confusing. If the authors really used C RYSOL3 to calculate the scattering curves of the models generated by C ORAL, this step is not only unnecessary but possibly also harmful, as C RYSOL does not necessarily treat regions added by C ORAL correctly. C hi-squared values and predicted curves should be taken directly from C ORAL. It is also not clear whether parts of the proteins missing in the crystal structures (Linker, termini) were explicitly modeled. This is the strong point of C ORAL, otherwise SASREF could have been used. Generally, for the publication of SAXS data, it is recommended to include a table listing "Datacollection and scattering-derived parameters" (http://scripts.iucr.org/cgibin/paper?S0907444912012073). The authors should include such a table. In addition, it is always a good idea to show Krakty plots (in the Supplement). Response: The reviewer is correct, the basophil analysis has been performed using exemplary one patient sensitized to birch pollen. We consider the functional data shown in Figure 4 mainly as initial data that support the structural data on the IgE Fc conformation and suggest that the sdab exhibit effects on free IgE and IgE on effector cells. Much broader studies on efficacy of the sdab will be needed to assess its potential as anti-IgE biological. Such broad analyses however need to be addressed in future studies.
We agree however with the reviewer and appreciated the suggestion that analysing a range of basophil donors is highly interesting and could provide additional, more quantitative insights into the effects. To address this point we carefully recruited a panel of allergic patients without prior specific SIT. In addition to serological data we determined changes of surface IgE and basophil activation upon incubation with the sdab. For control purposes we also expressed, purified and included an inactive sdab mutant.  (Fig. S7). Next we analysed the impact on basophil activation for a panel of 6 patients with birch pollen allergy (Table  S1). The high prevalence of birch pollen allergy and the fact that birch pollen major allergen Bet v 1 is the predominant allergen for the vast majority of birch pollen allergic patients in Northern Europe allows for a good comparability of obtained results. Therefore we analysed changes in basophil allergen threshold sensitivity, CDsens, to Bet v 1 in basophil activation tests. CDsens is a well-established measure to quantify effector cell sensitivity to an allergen and reflects the amount of allergen needed for efficient activation. In accordance with the reduced surface IgE, the 026 sdab significantly reduced CDsens for patient´s basophils by 50-95%. (Fig. 4C-E). Levels of specific IgE (sIgE) to Bet v 1 and total IgE (tIgE) in serum suggest that lower concentrations of sIgE ans as a consequence thereof smaller ratios of sIgE to tIgE translate into a more efficient reduction of CDsens (Fig. S8). These findings are in line with the observation that polysensitisation is often accompanied with a less severe phenotype of the allergic response."

Now, analyses of surface IgE were performed using basophils from patients sensitized to birch pollen, yellow jacket venom, and honeybee venom and the experiments were done in a concentration and time-dependent manner. For analysis of basophil activation we focused on birch pollen allergic patients since tree pollen allergy is an excellent model and the vast majority of patients react to the dominant component
The discussion section now reads as follows: "Apparently, the sdab removes IgE from the surface of effector cells as shown for basophils obtained from patients with different types of allergies. The reduction of IgE resulting correlates within a decreased sensitivity of the effector cells to allergen-dependent activation as shown in a cohort of birch pollen sensitized patients. It is imperative to consider that sensitivity of effector cells is shaped by the characteristics of surface IgE, e.g. ratio of specific and total IgE, affinity and repertoire complexity 39 . Although we used the dominant birch pollen allergen, Bet v 1, patient-specific sensitization profiles influence the reduction of cellular sensitivity, an effect likely to be even more pronounced in multisensitized individuals." Minor point: Page 11, line 8: "CD23 contacts residues in…." Change to "contact residues"? Response: We reworded the sentence to "CD23 is contacting residues in …"

Reviewer #2 (Remarks to the Author): Specific comments: p. 4, in the introduction please describe sdab 026-was it from a camel or llama, or is it a human VHH? How was it isolated?
Response: We agree that more information about the sdab is needed and introduced additional information on the sdab into the introduction section. The section now reads as follows: "Here we report the crystal structure and structure in solution of the human IgE Fc in complex with the sdab 026, a llama-derived, humanized sdab selected against IgE." p. 6, please give a better description of the 'hepta-saccharide'-was this a Nag2Man4? Response: We added information about the heptasaccharide to the text: "In our structure we observe this as a (N-acetylglucosamine)2(mannose)5 (NAG2MAN5) hepta-saccharide, which …" p. 13, in the materials and methods, hopefully your crystals were 'cryocooled', not frozen. Response: The wording has been corrected.
In Table 1, please include Rmerge and Rpim. Under the refinement statistics, what are the ligand atoms? I would specify as water or carbohydrate or whatever they are to be more clear. Response: The missing information has been added and the wording has been specified.

In the Figure 3 legend, please include the PDB code for the Fc-FceRI complex shown in panel D.
Response: The PDB code has been added as requested.

Reviewer #3 (Remarks to the Author):
Parts of the introduction read rather awkwardly, it might profit from thorough proofreading by an outsider. For example in "The human FcεRI is expressed as a tetramer or a trimer lacking the signal amplifying β-subunit." It is not clear whether only the trimer or both the trimer and the tetramer lack the subunit. Response: In order to improve readability we reworded sentences, phrases and wording in the introduction. We hope to have improved readibility here.
Page 5, 1st paragraph, analysis of the surface plasmon resonance experiment: The combination of a KD value with a χ2 is nonsensical. The KD should have a margin of confidence, whereas χ2 is a property of the fit (and often meaningless). We agree with the reviewer and removed χ2. A short statement and a reference to available information have been added.
Page 9, SAXS analysis: It is not possible to draw any conclusion of the ensemble of solution structure from CORAL modeling. Clustering of the results only imply that the reconstruction is stable, but CORAL by design will always find the SINGLE model that describes the data best. Less clustered solutions imply that the reconstruction is more ambiguous. To address the presence of several configurations, one should apply ensemble based tools such as MES or EOM. One could for example imagine a coexistence of strongly bent and extended confirmations of IgE Fc which could give the same SAXS signal as a moderately bent confirmation. Further, there is absolutely no point in discussing differences in chi-squared values. This can be easily understood that the same sample at lower concentration generally gives a lower chi-squared for the same model. A somewhat more meaningful estimator of fit-quality would be the p-value from the CORMAP test (Franke et al., 2014). Response: We agree on the comments regarding the interpretation of clustering within rigid body models, and we have removed the prior statements hinting at a biological meaning of tight/loose clustering in the CORAL output models. Also we have removed comparisons of the chi2 values after CORAL modelling, we now merely provide the values. The reviewer's suggestion of attempting ensemble optimization is highly relevant considering the known flexibility of IgE Fc and we now describe EOM analysis for both unbound IgE Fc and the Fc-sdab complex. In both cases two representative models were output with the major conformation closely resembling the models from CORAL rigid body modelling while a second minor conformation is somewhat different. Overall the conclusions from CORAL and EOM are the same, unbound IgE Fc appears to adopt mainly the bend conformation in solution whereas in the sdab complex a novel less bend conformation appears to be the dominating conformation. The entire SAXS paragraph has been extensively rewritten, the presentation of the rigid body modelling has been truncated, the EOM analysis has been added, and a new SAXS data table requested by the reviewer has been added.
Page 11, 2nd paragraph, discussion of SAXS data: Following the above, it would be more appropriate to state that the data can be described by such models, as you cannot exclude the existence of other solutions. Response: We agree and now write in the discussion "Being unable to crystallize sdab in complex with the full IgE Fc we used SAXS to analyze the complex and free IgE Fc in solution. This allowed for the first time insights into the positioning of the Cε2 domain in solution for both the free Fc as well as in the complex with the sdab. Two different approaches to SAXS based modelling suggested that with respect to the sdab complex, the IgE Fc Cε2 domains appear to adopt a major conformation in between the two well established bent and extended IgE Fc conformations. In unbound IgE Fc, the dominating conformation appears to be the bend conformation. In addition, EOM analysis suggested that one minor conformation somewhat different from the major conformation may be present in both cases." Page 13, Acknowledgements: I would strongly recommend to follow the acknowledgment guidelines for the beamlines used for this study: http://www.diamond.ac.uk/Beamlines/Mx/Common/Common-Manual/Acknowledging-MXbeamlines.html http://www.esrf.eu/UsersAndScience/UserGuide/Preparing/GuidelinesToUsers Response: We appreciate the suggestion and modified the acknowledgements accordingly.
Page 16, 2nd paragraph (Affinity measurements): The model underlying the KD calculation should be specified. The information was added as suggested.
Page 18, 2nd paragraph (Small angle X-ray scattering data collection and rigid body modeling): In the first line, 026 probably refers to sdab 026? The structure of the data analysis section is very confusing. If the authors really used CRYSOL3 to calculate the scattering curves of the models generated by CORAL, this step is not only unnecessary but possibly also harmful, as CRYSOL does not necessarily treat regions added by CORAL correctly. Chi-squared values and predicted curves should be taken directly from CORAL. It is also not clear whether parts of the proteins missing in the crystal structures (Linker, termini) were explicitly modeled. This is the strong point of CORAL, otherwise SASREF could have been used. Response: We now give the chi2 values output from directly from CORAL rather than with CRYSOL3 and we now also give the comparison between experimental and calculated data as output from CORAL in Fig 5B  and 5D. To comfort the reviewer there are no notable differences to the chi2 values and predicted scattering curves from CRYSOL3, but we agree that using the CORAL output is a more stringent approach.
I would like to thank the authors for the efforts they invested into improving the manuscript. It was a very pleasant read. I only have a few very minor comments that should be easily addressed: Page 20 ("Small angle scattering data collection and modeling") "C alculation of theoretical scattering profiles of atomistic structures from C ORAL modeling and their fits to the experimental data (as measured by the chi2 value) were done using C RYSOL3." In your response you state "We now give the chi2 values output from directly from C ORAL rather than with C RYSOL3 and we now also give the comparison between experimental and calcula ted data as output from C ORAL in Fig  5B  and 5D. To comfort the reviewer there are no notable differences to the chi2 values and predicted scattering curves from C RYSOL3, but we agree that using the C ORAL output is a more stringent approach." But this is not how I would understand the above sentence -I would naively assume that all chi2 values come from C RYSOL3. Perhaps a formulation along these lines would be less misinterpretable: "Fits of atomistic models to the experimental data (as measured by the chi2 value) were done using C RYSOL3 and atomistic models to fit the data were build using C ORAL." Figure 5B,D; Supplement Figure S10A,B,C : The y-axis deserves some units. If your data comes from BM29, I suspect these might be kDa mg/mL. If your data was scaled, the units are of course arbitrary, but that could explicitly stated. Supplement Table S2: If BSA is used as a standard, the exact product and supplier should be specified, as well as how the sample was prepared (buffer, concentration,…). 72 kDa are not the molecular weight of BSA, does this value take dimers and/or contaminants into account? Additionally, for data collected at ESRF BM29 you should have a water measurement. At higher angles, water scatters at 0.0163/cm, which can be used to calculate the scaling. Supplement Figure S10B,C : it would be nice if you could show how the data continues outside of the Guinier range.