Structure and Function Analysis of an Antibody Recognizing All Influenza A Subtypes

Kallewaard, Nicole L.; Corti, Davide; Collins, Patrick J.; Neu, Ursula; McAuliffe, Josephine M.; Benjamin, Ebony; Wachter-Rosati, Leslie; Palmer-Hill, Frances J.; Yuan, Andy Q.; Walker, Philip A.; Vorlaender, Matthias K.; Bianchi, Siro; Guarino, Barbara; De Marco, Anna; Vanzetta, Fabrizia; Agatic, Gloria; Foglierini, Mathilde; Pinna, Debora; Fernandez-Rodriguez, Blanca; Fruehwirth, Alexander; Silacci, Chiara; Ogrodowicz, Roksana W.; Martin, Stephen R.; Sallusto, Federica; Suzich, JoAnn A.; Lanzavecchia, Antonio; Zhu, Qing; Gamblin, Steven J.; Skehel, John J.

Crystals were frozen by direct immersion in liquid nitrogen and diffraction datasets were collected at 100K at the IO2 and IO4 beamlines at the diamond light source (Harwell, UK). Diffraction datasets were indexed and integrated with XDS (Kabsch, 2010) and scaled with XSCALE. . For solving the MEDI8852 Fab fragment, a search model was generated by blasting the sequences of its heavy and light chains and using the structures with the highest sequence homology. The search models were generated in chainsaw from CCP4 (Collaborative Computational Project, Number 4, 1994). Molecular replacement was carried out with Phaser (McCoy et al., 2007) , which found two constant domains and two variable domains corresponding to two Fab fragments. The structure was refined by rigid body and simulated annealing refinement in Phenix (Adams et al., 2010) , then completely rebuilt into the electron density, and then refined by alternating cycles of model building in Coot (Emsley and Cowtan, 2004) and TLS, restrained coordinate and B factor refinement in Refmac (Murshudov et al., 2011) . The complex structures were solved by molecular replacement using the cognate HA and MEDI8852 Fab structures as search models. For the H5 complex, Phaser found solutions for two HA trimers as well as for two variable domains of MEDI8852. Electron density was visible for the other 4 variable domains as well and atomic models were placed in the appropriate positions. After rigid body refinement in Phenix, electron density appeared for the constant domains, which were manually placed in the structure. There was clear difference density for HCDR3 and LCDR1 of MEDI8852 in the complex. To build the model of the complex, an unbiased omit map was calculated for each domain (HA1 fusion, HA1 esterase, HA1 head, HA2, MEDI8852 heavy chain variable, MEDI8852 heavy chain constant, MEDI8852 light chain variable, MEDI8852 light chain constant) in Phenix. Figures of merit were calculated for these Fourier maps using SigmaA (Collaborative Computational Project, Number 4, 1994) and the maps were subjected to twofold and six fold noncrystallographic symmetry averaging in DM. A sharpening B factor of 75 A 2 was applied to the structure factors before averaging. HCDR3 and LCDR1 were completely rebuilt into the averaged maps, and other residues were rebuilt according to the averaged maps as well as difference Fourier maps. The H7 structure was solved by molecular replacement in Phaser which found the one HA monomer as well as one variable and one constant domain of MEDI8852 in the asymmetric unit. The structure was refined by restrained coordinate refinement with global NCS restraints and TLS refinement in Refmac, and TLS and group B factor refinement in Phenix with one TLS group per domain (see above). Crystallographic statistics are summarized in Table S4 .