Class IB phosphoinositide 3-kinases γ (PI3Kγ) are second-messenger-generating enzymes downstream of signalling cascades triggered by G-protein-coupled receptors (GPCRs). PI3Kγ variants have one catalytic p110γ subunit that can form two different heterodimers by binding to one of a pair of non-catalytic subunits, p87 or p101. Growing experimental data argue for a different regulation of p87–p110γ and p101–p110γ allowing integration into distinct signalling pathways. Pharmacological tools enabling distinct modulation of the two variants are missing. The ability of an anti-p110γ monoclonal antibody [mAb(A)p110γ] to block PI3Kγ enzymatic activity attracted us to characterize this tool in detail using purified proteins. In order to get insight into the antibody–p110γ interface, hydrogen–deuterium exchange coupled to MS (HDX-MS) measurements were performed demonstrating binding of the monoclonal antibody to the C2 domain in p110γ, which was accompanied by conformational changes in the helical domain harbouring the Gβγ-binding site. We then studied the modulation of phospholipid vesicles association of PI3Kγ by the antibody. p87–p110γ showed a significantly reduced Gβγ-mediated phospholipid recruitment as compared with p101–p110γ. Concomitantly, in the presence of mAb(A)p110γ, Gβγ did not bind to p87–p110γ. These data correlated with the ability of the antibody to block Gβγ-stimulated lipid kinase activity of p87–p110γ 30-fold more potently than p101–p110γ. Our data argue for differential regulatory functions of the non-catalytic subunits and a specific Gβγ-dependent regulation of p101 in PI3Kγ activation. In this scenario, we consider the antibody as a valuable tool to dissect the distinct roles of the two PI3Kγ variants downstream of GPCRs.
Class I phosphoinositide 3-kinases (PI3Ks) are lipid kinases that transduce extracellular signals to trigger PtdIns(3,4,5)P3 synthesis, an essential second messenger at the plasma membrane. PtdIns(3,4,5)P3, together with its metabolites, PtdIns(3,4)P2 and PtdIns(3,5)P2, play fundamental roles in the regulation of basic cellular processes, such as proliferation, differentiation, growth and chemotaxis [1–8]. Class I PI3Ks are heterodimers composed of a catalytic (p110) and a non-catalytic subunit of the p85- or p101-type. Based on their interaction with non-catalytic subunits and their specific modes of regulation, class I PI3Ks can be further subdivided into class IA and class IB [2,3,9–12]. Class IA is characterized by heterodimers consisting of a catalytic p110α, p110β or p110δ subunit associated with a p85-type non-catalytic subunit, which has dual roles acting as an adaptor and a regulator [11,13–16]. Although the p85-type subunit is indispensable for class IA PI3K stability and regulation, the p110 catalytic subunit determines the signalling specificity [17–24].
The class IB PI3Ks are represented by two enzymes consisting of one catalytic p110γ subunit associated with either a p101 or a p87 (also known as p87PIKAP or p84) non-catalytic subunit [25–29]. Both PI3Kγ variants, i.e. p87–p110γ and p101–p110γ, are stimulated by Gβγ heterodimers released upon G-protein-coupled receptor (GPCR) activation and by active Ras proteins [25–39]. The former view of p87 and p101 being redundant adapters in Gβγ-mediated recruitment of PI3Kγ variants to the membrane compartment [27–29] has been challenged by previous data showing a different contribution of Gβγ and Ras on the two PI3Kγ variants . In particular, distinct Gβγ-binding affinities of the non-catalytic subunits for p110γ are intriguing [38,40,41]. These findings support data showing that PI3Kγ variants integrate into different and independent signalling cascades [39,42–44]. We have previously reported specific features for p87 and p101, such as diverse spatial and temporal distribution in human tissues and a different regulatory impact on p110γ activity, which may contribute to the differential regulation of the PI3Kγ variants [40,41]. These findings, in combination with the fact that only a single class IB catalytic subunit is present in cells, led us to postulate that p87 and p101 serve as signal-discriminating regulatory subunits defining specific functions for both p87–p110γ and p101–p110γ variants . However, the exact molecular mechanisms that maintain the specificity and selectivity of the two PI3Kγ variants are still unknown.
In the present study, we have identified and characterized a functional monoclonal anti-p110γ antibody that specifically inhibits the Gβγ-induced p87–p110γ enzymatic activity via contacting the C2 domain of p110γ. Our results point to a differential impact of the non-catalytic subunits thereby revealing a specific Gβγ-dependent regulatory role of p101 in PI3Kγ activation.
Cell cultures and expression plasmids
Human embryonic kidney (HEK)-293 cells (German Resource Centre for Biological Materials) were cultured and transfected with expression plasmids encoding p101 and p110γ as described previously [27,37,38]. For preparation of whole cell lysates, cells were directly lysed by adding 1× Laemmli sample buffer .
Expression and purification of recombinant proteins
Sf9 cells (fall armyworm ovary; Invitrogen) were cultured and infected as described previously . Recombinant baculoviruses for expression of Gβ1γ2, PI3Kγ and PI3Kβ subunits as well as their expression in Sf9 cells and purification of (His)6-tagged recombinant Gβ1(His)6γ2, (His)6p110γ, p87–(His)6p110γ, p101–(His)6p110γ and p85–(His)6p110β have been described elsewhere [38,40,41,46–48]. The pFastBac™ HTb baculovirus transfer vector (Invitrogen) was used to generate human full-length N-terminally (His)6-tagged H-Ras using BamHI/XhoI cloning site. H-Ras was produced in Sf9 insect cells and isolated using the Triton X-114 partition method as described previously [48,49]. The post-translational processing and lipidation of the protein was verified by MS analysis. Purified proteins were quantified by Coomassie Brilliant Blue staining after SDS/PAGE (10% acrylamide) with BSA as the standard. The proteins were stored at −80°C.
Hydrogen–deuterium exchange coupled to MS measurements
Hydrogen–deuterium exchange coupled to MS (HDX-MS) analyses of PI3Kγ in the presence and absence of an anti-p110γ monoclonal antibody [mAb(A)p110γ] were performed following a similar protocol as described previously [21,48]. The rate of exchange of full-length p110γ(His)6 alone and in the presence of a 3-fold molar excess of mAb(A)p110γ were compared. Reactions were initiated by mixing 10 μl of protein solution with 40 μl of deuterated buffer containing 20 mM Hepes, pH 7.2, 50 mM NaCl and 0.5 mM EGTA. Deuteration reactions were run for 3, 30, 300 and 3000 s of on-exchange at 23°C, before being quenched by addition of 20 μl of a 2 M guanidinium chloride and 1.2% formic acid solution. The final deuterium concentration during the reaction was 78%. Every time point and state was a unique experiment and every HDX-MS experiment was repeated twice. Samples were immediately frozen in liquid nitrogen and stored at −80°C for less than 1 week.
Analysis of the p110γ deuteration level was done as described previously , by sequentially digesting the protein with pepsin, separating the fragments on a C18 column and measuring the masses of peptides on a LTQ Orbitrap XL mass spectrometer. Manually selected peptides were then examined for deuterium incorporation by the HD-examiner software (Sierra Analytics). Results are presented as relative levels of deuteration with no correction for back exchange.
Gel electrophoresis, immunoblotting and antibodies
Generation and characterization of the anti-serum against the Gβ1 subunit are detailed elsewhere [31,50]. Specific antibodies against p87 and p101 were gifts from Michael Schaefer (Rudolf-Boehm-Institut für Pharmakologie und Toxikologie, Leipzig, Germany) and Len Stephens (Babraham Institute, Cambridge, U.K.) respectively. mAb(A)p110γ and mAb(B)p110γ were raised against full-length human p110γ using mouse hybridoma cells and were characterized earlier . Large-scale preparations of mAb(A)p110γ were generated in co-operation with BioGenes. mAb(B)p110γ was as described earlier [31,40,41]. Generation and characterization of mAb(C)p110γ, raised against the N-terminal 210 amino acids of catalytic p110γ, was as detailed earlier . Anti-Ras antibody was purchased from BD Biosciences. Anti-p110β antibody was purchased from Cell Signaling Technology. Proteins were fractionated by SDS/PAGE (10% acrylamide) and transferred onto nitrocellulose membranes (Hybond™-C Extra, GE Healthcare). Visualization of specific antisera was performed using the ECL system (GE Healthcare) or the SuperSignal® West Pico Chemiluminescent Substrate (Pierce) according to the manufacturers’ instructions. Chemiluminescence signals were estimated using the VersaDoc™ 4000 MP imaging system (Bio-Rad Laboratories).
Immunoprecipitation of PI3K
Purified recombinant p110γ, p87–p110γ and p101–p110γ and p85α–p110β variants were subjected to immunoprecipitation (IP) using mAb(A)p110γ, mAb(B)p110γ or mAb(C)p110γ. IP experiments were performed as detailed previously  with some modifications. In brief, Protein A–Sepharose CL-4B beads (GE Healthcare) were pre-incubated with or without antibody, washed, incubated overnight with cleared cell lysates or purified proteins and washed again. Proteins bound to beads were either tested for their lipid kinase activity or eluted by adding 1× Laemmli sample buffer  and subjected to SDS/PAGE.
Analysis of PI3K enzymatic activity
Analytical ultracentrifugation analyses
Molecular mass and complex stability of purified p87–p110γ and p110–p110γ heterodimers were analysed by sedimentation equilibrium analysis using a Beckman Optima XL-I centrifuge using the AN-60Ti rotor with the absorption optics set to 280 nm. Analyses were conducted in a buffer containing 20 mM Tris/HCl, pH 8.0, 150 mM NaCl, 2 mM DTT and 0.033% deca(ethylene glycol) dodecyl ether (C12E10) at 10°C. Sample and buffer (120 μl each) were loaded into six-channel cell assemblies. Replicate scans were taken following a 24 h equilibration at 6000 rev/min and then following a second 24 h equilibration at 11000 rev/min. Scans were also taken at 22 h at each speed so that equilibration could be confirmed. The equilibrium protein concentration distributions were globally analysed using the program HeteroAnalysis version 1.1.58 [51,52]. Sednterp version 20120828 Beta (http://sednterp.unh.edu) was used to calculate the partial specific volume of the proteins from their sequence and the density of the buffer from its composition neglecting the contribution of the detergent. The sedimentation parameters were corrected to standard conditions (20, w) using these values. The 280-nm molar absorption coefficients calculated from each protein's sequence were used to calculate the concentrations of the protein complexes (http://web.expasy.org/protparam/).
Results (means±S.E.M.) were analysed using Student's t test (*P≤0.05; **P≤0.01).
Inhibition of monomeric p110γ by mAb(A)p110γ
A monoclonal anti-p110γ antibody [mAb(A)p110γ] raised against full-length human catalytic p110γ subunit used in earlier IP experiments  displayed interesting features attracting our attention. mAb(A)p110γ failed to visualize p110γ in immunoblots (Figure 1A); however, it was able to interact with the intact protein in solution enabling IP experiments (Figure 1B). The feature of recognizing native p110γ made it worthwhile to test whether mAb(A)p110γ interferes with p110γ activity. As shown in Figure 1(C), incubation with mAb(A)p110γ led to a drastic reduction in p110γ lipid kinase activity stimulated by Gβ1γ2, defining mAb(A)p110γ as a putative PI3Kγ inhibitor.
In order to test the selectivity of the mAb(A)p110γ antibody, we measured its effect on the activity of the class IA PI3Kβ, another Gβγ-sensitive PI3K. Recombinant and functionally active Gβγ-sensitive p85α-p110β was purified following heterologous expression in Sf9 cells (Figures 2A and 2B). IP experiments (Figure 2C) as well as analysis of the immunoprecipitates in the lipid kinase assays (Figure 2D) showed complete lack of interaction between mAb(A)p110γ and p85α–p110β. Correspondingly, mAb(A)p110γ did not inhibit lipid kinase activity of purified p85α–p110β (Figure 2E).
Mapping of p110γ regions affected by interaction with mAb(A)p110γ
Since mAb(A)p110γ was generated by an immunization and selection protocol using full-length human catalytic p110γ subunit, the epitope of p110γ targeted by mAb(A)p110γ was unknown. To determine the p110γ epitope recognized by mAb(A)p110γ, we used HDX-MS. HDX-MS is a powerful technique that can map protein–protein and protein–lipid interactions, as well as provide useful information on the dynamics of proteins [53,54]. The technique is based on the differences in exchange rate of amide protons from a protein with solvent, a reaction that is influenced by secondary structure and solvent exposure.
To map the regions in p110γ that are affected by the interaction with mAb(A)p110γ, we compared the HDX rates of p110γ in solution and when in a complex with mAb(A)p110γ. A large proportion of the C2 domain shows a reduced HDX rate in the p110γ–mAb(A)p110γ complex, suggesting that the antibody binds this region of p110γ (Figures 3A and 3B). More precisely, the most solvent-exposed part of the C2 domain, spanning residues 382–413, has a strongly reduced dynamics, probably stabilizing the β-strand underneath (residues 414–428). Interestingly, binding of mAb(A)p110γ seems to induce allosteric changes in p110γ, as increased HDX rates are observed in two distinct domains of p110γ: the helical and kinase domains (Figure 3B). The increased dynamics in the p110γ helical domain (551–607, 622–630, 636–650) overlaps with the previously identified Gβγ-binding site (546–607) . The two helices within the kinase domain that show increased dynamics (1035–1050) correspond to a region essential for inhibition of p110α activity by its regulatory subunit .
In summary, HDX-MS experiments revealed that mAb(A)p110γ associates with the C2 domain of p110γ and induces conformational changes in the helical and kinase domains. Since both domains are important for PI3Kγ regulation, binding of mAb(A)p110γ to p110γ might affect kinase enzymatic activity.
Effect of mAb(A)p110γ on p87–p110γ and p101–p110γ heterodimer activity
Class IB PI3Kγ is present as two distinct functional p87–p110γ and p101–p110γ heterodimers in vivo [26,38,41,42]. We tested how mAb(A)p110γ affects the enzymatic activities of these two PI3Kγ variants stimulated by Gβ1γ2. Two additional monoclonal antibodies raised against full-length human catalytic p110γ subunit [mAb(B)p110γ] and N-terminal amino acids 1–210 of p110γ [mAb(C)p110γ] were also included in order to validate the specificity of interactions. As depicted in Figure 4(A), significant differences in the ability of the antibodies to affect lipid kinase activities of the two PI3Kγ variants became apparent. Although incubation of p87–p110γ with mAb(A)p110γ resulted in drastic reduction in Gβ1γ2-stimulated lipid kinase activity, inhibition of p101–p110γ activity by this antibody, at the concentrations tested, was weak. In contrast, mAb(B)p110γ and mAb(C)p110γ were ineffective in inhibiting enzymatic activity of either PI3Kγ variant under the identical experimental conditions (Figure 4A). The intriguing finding of the differential mAb(A)p110γ-mediated effect on the two PI3Kγ variants showing only weak inhibition of p101–p110γ as compared with strong inhibition of p87–p110γ prompted us to check whether mAb(A)p110γ was able to interact with p110γ when associated with p101. Comparable to monomeric p110γ (Figure 1A), immunoblotting (IB) analysis revealed that mAb(A)p110γ does not recognize denatured p101–p110γ complex (Figure 4B). In contrast, mAb(B)p110γ and mAb(C)p110γ recognize p110γ in immunoblots (Figure 4B). Nonetheless, the capability of mAb(A)p110γ to directly bind to p110γ when complexed with p101 could be verified by IP (Figure 4C).
Taken together, mAb(A)p110γ inhibits Gβγ-stimulated lipid kinase activity of p87–p110γ more potently than of p101–p110γ.
Interaction of p87–p110γ or p101–p110γ heterodimers with phospholipid vesicles
The HDX-MS data demonstrate binding of mAb(A)p110γ to the C2 domain of p110γ (Figure 3B). The C2 domain of p110γ, similarly to other C2 domains, is considered to mediate protein–lipid interactions [56–58]. This encouraged us to check whether mAb(A)p110γ interferes with Gβ1γ2-mediated association of p87–p110γ or p101–p110γ to phosholipid vesicles in the absence and presence of another known PI3Kγ regulator, i.e. H-Ras. Strikingly, mAb(A)p110γ differently affected Gβ1γ2-mediated phospholipid vesicle association of PI3Kγ variants. Whereas mAb(A)p110γ strongly reduced Gβ1γ2-mediated vesicle association of p87-p110γ in a concentration-dependent manner, association of p101–p110γ remained unchanged (Figure 5A). mAb(A)p110γ did not change binding of p101–p110γ to phospholipid vesicles upon exposure to both regulators, Gβ1γ2 and H-Ras (Figure 5B). However, concomitant incubation with Gβ1γ2 and prenylated H-Ras partially rescued phospholipid vesicle association of p87–p110γ in the presence of mAb(A)p110γ. Nonetheless, membrane association was impaired by high concentrations of mAb(A)p110γ (Figure 5B). It should be pointed out that in these experiments p87, p101 and p110γ were found in ratios corresponding the starting condition suggesting that the stoichiometry of the PI3Kγ variants bound to phospholipid vesicles was not affected by mAb(A)p110γ (Figure 5, grey or white bars compared with black bars). Control experiments excluded that the association of Gβγ or H-Ras to phospholipid vesicles was significantly affected by mAb(A)p110γ (Table 1). High complex stability was supported by equilibrium analytical ultracentrifugation showing Kd values of ≤0.2 μM for p87–p110γ and ≤0.1 μM for p101–p110γ (Figure 6).
The interference of mAb(A)p110γ with Gβγ-binding was tested by co-IP of p87–p110γ or p101–p110γ with Gβ1γ2 and H-Ras (Figure 7). In the case of p87–p110γ, a reduction in Gβ1γ2 monitored by Gβ1-immunoreactivity was evident, whereas H-Ras-levels remained unaffected (Figure 7). Taken together, the data show a mAb(A)p110γ-dependent inhibition of Gβ1γ2-induced recruitment of p87–p110γ to the lipid compartment. Next, we investigated the consequences for enzymatic activity.
Concentration-dependent inhibition of PI3Kγ variants by mAb(A)p110γ
We studied concentration-dependent inhibition of variously stimulated lipid kinase activities of p87–p110γ and p101–p110γ in the presence of increasing concentrations either of the pan-PI3K inhibitor wortmannin (Figures 8A–8D) or mAb(A)p110γ (Figures 8E–8H). Wortmannin, which blocks all class I PI3Ks by covalent binding to a lysine residue in the ATP-binding pocket of p110 isoforms , inhibited both PI3Kγ variants at similar IC50 concentrations under all conditions tested and failed to differentiate between the two PI3Kγ variants.
In the presence of mAb(A)p110γ, basal lipid kinase activities of the two PI3Kγ variants were inhibited in a concentration-dependent manner with IC50 values of 7.2±1.3 nM and 17.8±5.2 nM for p87–p110γ and p101–p110γ respectively (Figure 8E). Strikingly, the Gβ1γ2-stimulated activity of p87–p110γ was inhibited ~30-fold more potently as compared with the p101–p110γ counterpart (IC50 of 1.6±0.5 nM compared with 46.5±12.6 nM; Figure 8F). In contrast, mAb(A)p110γ inhibition of H-Ras-stimulated variants was indistinguishable (Figure 8G). When the enzymes were co-stimulated by Gβ1γ2 and H-Ras, p87-p110γ was 10-fold more potently inhibited as compared with p101–p110γ by mAb(A)p110γ (IC50 of 4.3±0.4 nM compared with 49.5±4.9 nM; Figure 8H). Thus, mAb(A)p110γ not only represents a valuable experimental tool to understand the different regulation of PI3Kγ variants but also serves to selectively intervene into Gβγ-induced p87–p110γ lipid kinase activity.
We recently described p87–p110γ as a constitutively and ubiquitously expressed class IB PI3Kγ variant . In contrast, p101–p110γ appeared as an inducible counterpart which is up-regulated upon activation and expressed in various tissues side-by-side with p87–p110γ. In line with this view, growing experimental evidence indicates a divergent function and regulation of the two class IB PI3Kγ variants [38,39,42–44]. Unfortunately, pharmacological tools discriminating between the two variants are not available . In the present study, we identified a monoclonal antibody mAb(A)p110γ as a potent inhibitor of PI3Kγ isoforms acting at low nanomolar concentrations. mAb(A)p110γ blocked basal lipid kinase activities of either p87–p110γ or p101–p110γ with potencies comparable to that of wortmannin, an inhibitor acting at the ATP-binding site. Interestingly, enzymatic activities were differentially inhibited with a significant preference for p87–p110γ following stimulation by Gβγ. This preferential inhibition of p87–p110γ activity by mAb(A)p110γ persisted even in experiments stimulating the PI3Kγ variants simultaneously with Ras and Gβγ.
The mAb(A)p110γ was generated using full-length human p110γ protein for immunization and selection procedure and, therefore, the exact antibody–p110γ interaction site was unknown [37,61]. HDX-MS, an approach that has provided insight into PI3K regulation at the membrane and by regulatory partners [21,48,62], identified dynamic changes within three domains of p110γ upon association with mAb(A)p110γ. Residues 382–428 in the C2 domain of p110γ were protected from HDX, most probably due to binding of the antibody to this region. In addition, antibody–p110γ interaction induced increased dynamics in both the helical and the kinase domain of p110γ, probably as a result of allosteric modifications.
Generally, C2 domains have been associated with membrane interactions. The C2 domain of p110γ was also proposed to be involved in the interaction of p110γ with the plasma membrane . However, recent data looking at lipid-binding sites of class I PI3Ks have identified the C-terminal helix of the kinase domain rather than the C2 domain to be involved in binding to lipids [21,48,63]. Our data obtained in phospholipid pull-down assays are in agreement with these recent data. The necessity of the C2 domain of p110γ to act as the membrane interaction module in the regulation of PI3Kγ was not hitherto experimentally validated. Although Kirsch et al.  have shown that the phospholipid binding of a p110γ fragment comprising amino acids 740–1068 was significantly lower than the binding of full-length p110γ, this truncation construct lacked more than just the C2 domain (comprising residues 357–522). In addition to phospholipid binding, C2 domains have been reported to exhibit additional functions. In p110α, the C2 domain seems to be crucial for the inhibitory function of p85 on p110, whereas the C2 domain of p110β harbours a nuclear localization signal motif mediating translocation into the nucleus [11,15,65].
Our data argue for a different effect of mAb(A)p110γ on Gβγ-mediated stimulation of p87–p110γ and p101–p110γ. HDX-MS analyses indicate that binding of mAb(A)p110γ to the p110γ C2 domain induces allosteric changes in the helical domain. Since the helical domain is responsible for Gβγ binding , it is possible that the conformational changes directly affect the affinity of Gβγ for p110γ. Additionally, the different potencies by which mAb(A)p110γ inhibits Gβγ stimulation of PI3Kγ variants may be a consequence of a distinct effect of the two non-catalytic subunits, i.e. p87 and p101, on PI3Kγ activity (Figure 9). Alternatively, since the p110γ helical domain is stabilized by the associated p87 or p101 regulatory subunits [48,66], one possibility of discriminative inhibition of PI3Kγ variants is that p101 protects from allosteric changes induced by mAb(A)p110γ more than p87 does. This would explain the reduced inhibitory effect of the antibody for the p101–p110γ heterodimer compared with p87–p110γ and to p110γ.
Ample evidence suggests that p101 acts as a Gβγ adaptor [26,32,37,38]. Since p101 is able to rescue the stimulatory effect of Gβ1 mutants deficient in stimulating p110γ  and enhance Gβγ-induced stimulation of lipid-associated p110γ , we characterize p101 as a Gβγ-dependent regulator of PI3Kγ enzymatic activity. HDX-MS analysis on the p101–p110γ complex has identified two regions within the C-terminal part of p101 to mediate PI3Kγ activation by Gβγ . In contrast, whether p87 functionally interacts with Gβγ remains an open question. Although p87 exhibits a significant degree of homology with p101 at the C-terminal region [27–29], up to now we could not find any evidence that it displays a Gβγ-adapter function or serves as a Gβγ-dependent regulator [38,40,41]. Therefore, we suppose that in the presence of Gβγ, mAb(A)p110γ induces structural alterations in the helical domain that result in more drastic consequences for p87–p110γ than for p101–p110γ on phospholipid vesicle recruitment and enzymatic activation.
Taken together, we have characterized the inhibitory action of the monoclonal anti-p110γ antibody mAb(A)p110γ, mapped the antibody–p110γ interface and present new structure–function insights into PI3Kγ activity. Specific features of mAb(A)p110γ to differentially block Gβγ-mediated association of p87–p110γ and p101–p110γ, and hence their enzymatic activities, provide the basis for a selective inhibition of Gβγ-initiated hormonal pathways of PI3Kγ variants and argues for a specific Gβγ-dependent regulatory role for p101 in PI3Kγ activation. This supports the idea of a differential regulatory impact of p87 and p101 on PI3Kγ activation.
The expert technical assistance of Renate Riehle and Rosi Maier is greatly appreciated. We thank all members of the Nürnberg laboratory previously located in Düsseldorf and in Tübingen. We thank John Burke, Olga Perisic, Mark Skehel, Sarah Maslen, Farida Begum and Sew-Yeu Peak-Chew for help with the HDX-MS setup.
Aliaksei Shymanets, Christian Harteneck and Bernd Nürnberg designed the study. Aliaksei Shymanets, Prajwal, Oscar Vadas, Cornelia Czupalla, Jaclyn LoPiccolo, Alessandra Ghigo and Eberhard Krause performed the experiments. Aliaksei Shymanets, Oscar Vadas, Michael Brenowitz, Eberhard Krause, Emilio Hirsch, Reinhard Wetzker, Roger Williams, Christian Harteneck and Bernd Nürnberg analysed and interpreted the data and wrote the paper.
This work was supported by the Deutsche Forschungsgemeinschaft [grant numbers IRTG 1302 (to B.N.) and RTG 1715 (to R.W.)]; Telethon [grant number GGP14106 (to E.H.)]; the Swiss National Science Foundation fellowship [grant number PA00P3_134202 (to O.V.)]; the European Commission fellowship [grant numbers FP7-PEOPLE-2010-IEF and N°275880 (to O.V.)]; and the Medical Research Council [grant number U10518430 (to R.W.)].
Present address: Department of Pharmaceutical Sciences, University of Geneva, CH-1211 Geneva 4, Switzerland.