key: cord-0260662-nt0zhr6u authors: Rathinaswamy, Manoj K; Fleming, Kaelin D; Dalwadi, Udit; Pardon, Els; Harris, Noah J; Yip, Calvin K; Steyaert, Jan; Burke, John E title: HDX-MS optimized approach to characterize nanobodies as tools for biochemical and structural studies of class IB phosphoinositide 3-kinases date: 2021-06-01 journal: bioRxiv DOI: 10.1101/2021.06.01.446614 sha: 7aa3f15386b1c1aae3300b15cf300961ff98a790 doc_id: 260662 cord_uid: nt0zhr6u There is considerable interest in developing antibodies as modulators of signaling pathways. One of the most important signaling pathways in higher eukaryotes is the phosphoinositide 3-kinase (PI3K) pathway, which plays fundamental roles in growth, metabolism and immunity. The class IB PI3K, PI3Kγ, is a heterodimeric complex composed of a catalytic p110γ subunit bound to a p101 or p84 regulatory subunit. PI3Kγ is a critical component in multiple immune signaling processes and is dependent on activation by Ras and GPCRs to mediate its cellular roles. Here we describe the rapid and efficient characterization of multiple PI3Kγ single chain camelid nanobodies using hydrogen deuterium exchange mass spectrometry (HDX-MS) for structural and biochemical studies. This allowed us to identify nanobodies that stimulated lipid kinase activity, blocked Ras activation and specifically inhibited p101-mediated GPCR activation. Overall, this reveals novel insight into PI3Kγ regulation and identifies sites that may be exploited for therapeutic development. Highlights – HDX-MS rapidly identifies epitopes of camelid single-chain nanobodies raised against Class IB PI3K complexes, p110γ/p101 and p110γ/p84 – A nanobody targeting p101 improves local resolution in EM studies with p110γ/p101 facilitating structural characterization of the complex – Nanobodies that bind at the interfaces with the lipidated activators Ras and Gβγ can prevent activation of p110γ/p101 and p110γ/p84 which normally packs against the VL in conventional dual chain antibodies (Desmyter et 98 al., 1996) . This coupled with their small size (~15 kDa versus 150 kDa for conventional 99 antibodies) provide nanobodies the potential to bind specifically to epitopes that are 100 inaccessible for conventional antibodies, with high affinity. These advantages have 101 resulted in their widespread use in research, testing and therapy (Uchański et al., 2020) . 102 Nanobodies have proven to be exceptional tools for optimizing structural biology 103 approaches of protein assemblies. In X-ray crystallography, they can stabilize flexible 104 protein regions, prevent aggregation/oligomerization, and offer novel crystal contact sites 105 into distinct signaling pathways will be powerful in dissecting the molecular mechanisms 132 of signaling and in developing novel therapeutics. Critical to determining the usefulness 133 of nanobodies as structural chaperones and signaling modulators is the ability to rapidly 134 determine their binding sites and how they might alter protein conformational dynamics. 135 Here we report the characterization of multiple PI3Kg binding nanobodies, and 136 describe their application for both structural biology approaches and modulation/inhibition 137 of the activation and kinase activity of PI3Kg. We utilized hydrogen deuterium exchange 138 mass spectrometry (HDX-MS) to rapidly identify binding epitopes in p110g, p101, and 139 p84. This allowed us to identify and characterize p110g binding nanobodies that activate 140 lipid kinase activity, block activation by Ras, and a p101 binding nanobody that disrupts 141 GPCR activation. Nanobodies were identified that stabilized flexible regions/domains, 142 allowing for high resolution cryo electron microscopy (cryo-EM) studies (details described 143 in a separate study). Overall, this work provides an HDX-MS enabled flow-path for the 144 rapid characterization of nanobodies for structural and biochemical studies (Figure 1) . 145 146 Llamas were immunized with either the p110g/p101 or p110g/p84 complex and 149 putative binders were identified using phage display from the B-cells as previously 150 described (Pardon et al., 2014) (Fig. 1A) . We identified 88 potential binders, which were 151 classified according to the sequence of the third complementarity determining region 152 (CDR3), into 49 families. Representative nanobodies from all families were recombinantly 153 expressed in WK6 E.coli cells and purified using Nickel affinity chromatography. Members 154 of four families could not be expressed, leading to 45 purified nanobodies. Streptavidin 155 pulldown assays were performed using strep-tagged p110g/p101 and p110g/p84 156 complexes with the purified nanobodies to determine binding. To identify nanobodies that 157 bound to p110g, p101, or p84 we carried out pull downs on both p110g/p101 and 158 p110g/p84. Nanobodies that bound both complexes were assumed to bind p110g, while 159 ones that bound specifically to either p110g/p101 or p110g/p84 were assumed to be p101 160 or p84 binders. Twenty-six nanobodies were identified as positive hits, of which twenty-161 two bound to p110g, three bound to p101, and one to p84 (Fig. 1B, Table S1 , and Source 162 data). We used differential scanning fluorimetry (DSF) with nanobodies bound to PI3Kg 163 to identify possible stabilizing effects. DSF measures the unfolding of proteins as a 164 function of temperature, which allows for the identification of stabilizing binding partners 165 by observing increases in the melting temperature (Tm) (Fig. 1C) . Fifteen nanobodies 166 showed higher Tm values, including eight nanobodies which induced differences of 0.5°C 167 or greater, indicating significant stabilizing effects (Table S1) . The changes in HDX as a result, allow for the identification of potential binding interfaces 195 or conformational changes induced upon complex formation with nanobodies. The full 196 details of HDX data collection and analysis are shown in Table S2 , with the full raw 197 deuterium incorporation data, and differences in exchange with each nanobody included 198 in the source data excel file. Many nanobodies bound to regions of either p110g or the p101 and p84 subunits 240 that have not been characterized structurally up to this point. We first used utilized 241 negative stain electron microscopy to analyze three nanobodies that bind novel regions 242 of the p110g-p101 complex: NB1-PIK3R5 (p101 C-terminus), NB2-PIK3R5 (p101) and 243 NB5-PIK3CG (p110g ABD) ( Fig 3A) . 2D analysis revealed that additional densities along 244 the periphery of the complex corresponding to unique binding sites of these nanobodies. 245 We were able to model the approximate location of these binding sites by integrating data 246 from HDX-MS and negative stain EM (Fig. 3D) . We also conducted cryo-EM analysis of 247 the p110g-p101 complex, generating a map at an overall resolution of 3.4 Å. However, 248 this 3D reconstruction revealed that a large portion of p101 was poorly resolved in the 249 EM density map, with this region matching the area where NB1-PIK3R5 bound based on 250 the negative stain analysis and HDX-MS. Hence, we reconstituted a ternary complex of 251 p110g-p101 with NB1-PIK3R5 and vitrified this sample for cryo-EM analysis. We were 252 able to obtain a 3D reconstruction at an overall resolution of 2.9 Å, with greatly improved 253 local resolution compared to the apo complex at the NB1-PIK3R5 binding site (Fig. 3B+C ) 254 (the full details of this structural reconstruction are described in a separate manuscript). 255 The improved quality of the EM density map resulting from the nanobody serves as a 256 proof of principle for the HDX-MS guided approach to rapidly optimize nanobodies as led to investigate the potential role of nanobodies binding this region. We selected three 280 nanobodies for full biochemical characterization (the RBD binding nanobody NB5-281 PIK3CG, the p101 binding nanobody NB1-PIK3R5, and the ABD binding nanobody NB7-282 PIK3CG) using in vitro lipid kinase assays of its basal activity and activation by lipidated 283 Gbg and Ras ( Fig 1F, Fig 4) . Assays were carried out with both the p110g-p101 and 284 p110g-p84 complexes, to determine any complex specific modulatory effects. The 285 development of biomolecules that can specifically target unique p110g complexes would 286 be an important tool to decipher PI3Kg signaling, as ATP competitive inhibitors will equally 287 inhibit both complexes. 288 The RBD-binding nanobody NB7-PIK3CG had no effect on basal lipid kinase 289 activity of either p110g-p101 or p110g-p84 (Fig. 4A+D ). For both complexes it completely 290 blocked activation by lipidated Ras (Fig. 4A+D ). This nanobody appeared to have a 291 limited effect on Gbg activation of the p110g-p101 complex, as it was still robustly 292 activated, however, for the p110g-p84 complex it caused complete disruption of both Ras 293 and Gbg activation (Fig. 4A+D ). This potentially could be utilized as a biased inhibitor that 294 would preferentially inhibit p110g-p84 over p110g-p101. The contact site of the p101 295 binding nanobody NB1-PIK3R5 partially overlapped with the one identified for lipidated 296 Gbg on membranes (Vadas et al., 2013) . In light of this, we hypothesized that this 297 nanobody would be able to specifically inhibit Gbg-mediated activation of p110g/p101. 298 Gbg activation of p110g/p101 was almost completely inhibited in the presence of NB1-299 PIK3R5 ( Fig 4B) , with no effect on Ras activation. This nanobody caused no significant 300 differences in lipid kinase activity under any conditions for the p110g/p84 complex (Fig 301 4E ). Due to its ability to potently inhibit only GPCR activation of the p110g-p101 complex, 302 NB1-PIK3R5 will be a powerful tool to selectively inhibit only p110g-p101 over p110g-p84 303 to decipher their specific roles in PI3Kg signaling. 304 We tested the effect of the p110g ABD binding nanobody (NB5-PIK3CG) on lipid 305 kinase activity. The ABD of p110g is structurally uncharacterized, but the ABD domain of 306 class IA PI3Ks is known to be a critical regulator of lipid kinase activity (Vadas et al., 307 2011) . The NB5-PIK3CG nanobody activated lipid kinase activity under all conditions 308 tested for both p110g-p101 and p110g-p84 (Fig. 3C+F ). This reveals an unexpected and 309 previously undescribed role of the ABD as a regulator of p110g signaling, with molecules 310 targeting this region able to modulate lipid kinase activity. of the NB1-PIK3R5 nanobody, which stabilized the dynamic C-terminus of the p101 365 subunit, allowed us to obtain a high resolution map of the p110g-p101 complex by cryo-366 EM, which is described in depth in another manuscript. Overall, our HDX-MS based 367 approach allows for a repeatable method to rapidly identify the most suitable nanobodies 368 to optimize X-ray crystallography and cryo-EM approaches. 369 Our combined HDX-MS and EM structural studies revealed multiple nanobodies 370 that bound at critical regulatory interfaces involved in the binding of Ras and Gbg in both 371 p110g and p101. We identified two nanobodies (NB6-PIK3CG + NB7-PIK3CG) that bound 372 to the RBD in p110g, which contains the Ras binding interface (Pacold et al., 2000) . 373 Membrane reconstitution assays of Ras activation showed that NB7-PIK3CG potently 374 inhibited Ras activation for both the p110g-p101 and p110g-p84 complexes. Intriguingly, 375 for p110g-p84, this nanobody also disrupted activation by Gbg. The p110g-p84 complex this inhibitory effect and still be activated by Gbg. This difference between the two PI3Kg 385 complexes will likely enable this nanobody to be used as a biased p110g-p84 inhibitor. A 386 similar effect was seen with a C2 binding antibody that was able to selectively inhibit 387 GPCR activation of p110g-p84 over p110g-p101 (Shymanets et al., 2015) . Together these 388 biomolecules will be useful as tools to study the specific roles of p110g-p84 in cell 389 The p110g-p101 complex contains Gbg binding sites in both p110g and p101 391 The protein was concentrated to 1 mg/mL using a 10,000 kDa MWCO Amicon concentrator, aliquoted, 853 snap-frozen in liquid nitrogen and stored at -80°C. The samples were immediately frozen in liquid nitrogen at -80˚C. This was then used to calculate the specific activity of the enzyme. The number of deuteron difference for all peptides analyzed over the entire deuterium exchange time 1127 course is shown for p110g. The Molecular Mechanism of Transport by the 643 Gβγ is a 647 direct regulator of endogenous p101/p110γ and p84/p110γ PI3Kγ complexes in 648 mouse neutrophils Allosteric nanobodies uncover a role of hippocampal mGlu2 653 receptor homodimers in contextual fear consolidation Structure of PINK1 in complex with its substrate 657 ubiquitin 659 2013. p87 and p101 subunits are distinct regulators determining class IB PI3K 660 specificity Different inhibition of Gβγ-stimulated class IB phosphoinositide 3-kinase 664 (PI3K) variants by a monoclonal antibody. Specific function of p101 as a Gβγ-665 dependent regulator of PI3Kγ enzymatic activity Transient conformers of LacY are trapped by nanobodies Allosteric 673 nanobodies reveal the dynamic range and diverse mechanisms of G-protein-674 coupled receptor activation Nanobodies as therapeutics: big 676 opportunities for small antibodies Gbg sensitivity of a PI3K is dependent upon a tightly associated adaptor, p101 Nanobodies to study protein 683 conformational states Structural basis for 686 activation and inhibition of class I phosphoinositide 3-kinases Molecular determinants of PI3Kγ-mediated activation downstream of G-691 protein-coupled receptors (GPCRs) Structural flexibility of the G alpha s alpha-helical domain in the beta2-adrenoceptor Distinctive 700 Activation Mechanism for Angiotensin Receptor Revealed by a Synthetic Nanobody VIB-CMB COVID-19 Response Team Structural Basis for Potent Neutralization of Betacoronaviruses by 706 Single-Domain Camelid Antibodies PI3Kg catalytic and regulatory subunits were co-expressed in Spodoptera frugiperda (Sf9) cells using the 755 baculovirus expression system The resuspended pellets were sonicated for 2.5 minutes at level 4.0 with cycles consisting of 759 15 seconds ON/OFF using the Misonix Sonicator 3000. Triton X-100 was added to the cell lysate at a final 760 concentration of 0.1% (v/v), and the lysates were centrifuged at 14,000 rpm at 4 C for 45 minutes in a JA-761 20 rotor. The supernatant was loaded onto a HisTrap™ FF column glycerol) and a high salt wash was performed using NiNTA A High Salt Buffer 10 mM imidazole pH 8.0, 5% (v/v) glycerol, 2mM bME). The protein was washed on an AKTA Start 765 FPLC (GE) system with 4 column volumes (CV) of NiNTA A, 4 CV of 94% NiNTA A/6% NiNTA B Tris pH 8.0, 100 mM NaCl, 200 mM imidazole pH 8.0, 2 mM bME, 5% (v/v) glycerol), and eluted in 2 CV The eluted protein was loaded onto a StrepTrap™ column (GE) equilibrated with gel 768 filtration buffer (GFB For pulldowns, the column was washed with 2 CV GFB and the tagged protein was 770 eluted in 2 CV GFB containing 2.5 mM Desthiobiotin. For studies using untagged protein, the washed 771 column was incubated on ice overnight with 100 uL Tobacco Etch Virus Protease (1 mg/mL) diluted in GFB 772 and eluted the following day Gel filtration chromatography was performed on an AKTA Pure (GE) using a Superdex™ 200 /300 Increase column (GE) equilibrated in GFB. The fractions containing the protein were pooled and 775 concentrated, flash frozen in liquid nitrogen, and stored at -80 C. All purification steps were analyzed using Nanobody generation and small scale nanobody purification Nanobody discovery was carried out as previously described by the Steyaert lab at the VIB-VUB Briefly, one llama was immunized 6 times with in total 900µg of final pixel size of 4.67 Å/pixel. For the p110g-p101-NB5-PIK3CG complex, 25 micrographs were acquired 970 at a nominal magnification of 49,000x at a defocus of -1 80 micrographs of the p110g-p101-NB2-PIK3R5 complex were acquired at a nominal 972 magnification of 45,000x at a defocus of -1.2µm and binned twice to obtain a final pixel 50 micrographs of the p110g-p101-NB1-PIK3R5 complex were acquired at a nominal magnification 974 of 45,000x at a defocus of -1.2µm and binned twice to obtain a final pixel size of 4.53 Å/pixel. For each 975 dataset, the contrast transfer function (CTF) of each micrograph was estimated using CTFFind4 These templates were then used to autopick 35629, 6425, 46848, and 23112 particles for the 978 apo-, NB5-PIK3CG, NB3-PIK3R5, and NB1-PIK3R5-bound datasets, respectively, and extracted with a box 979 size of 320Å. Particles were then subjected to 2D classification and the classes showing clear additional 980 density for Sample Preparation and Data Collection Flat 2/2-T 300 mesh grids were glow discharged for 25s at 15mA using a Pelco easiGlow glow-984 3μL of purified p110g-p101 complex with or without bound nanobody was then applied to the 985 grids at a concentration of 0.45 mg/ml . Grids were then prepared using a Vitrobot Mark IV Scientific) by blotting for 1.5s at 4°C and 100% humidity with a blot force of -5 followed by plunge freezing 987 in liquid ethane. Grids were screened for particle and ice quality at the UBC High Resolution 988 Macromolecular Cryo-Electron Microscopy (HRMEM) facility using a 200kV Glacios TEM Scientific) equipped with a Falcon 3EC DED. All datasets were then collected at the Pacific Northwest Cryo-990 For the apo p110g-p101 complex, 6153 super-resolution movies were collected 992 using SerialEM with a total dose of 50e -/Å 2 over 50 frames at a physical pixel size of 1.079Å/pix, using a 993 defocus range of -0.8 to -2μm. For the nanobody-bound p110g-p101 complex, 6808 super-resolution 994 movies were collected using SerialEM with a total dose of 36.4e -/Å 2 over 50 frames at a physical pixel size 995 of 1 Cryo-EM image analysis For the 999 nanobody-bound p110g-p101 complex dataset, patch motion correction using default settings was first 1000 applied to all movies to align the frames and Fourier crop the outputs by a factor of 2. The contrast transfer 1001 function (CTF) of the resulting micrographs was estimated using the patch CTF estimation job with default 1002 settings. 2D class averages from a previous dataset were low-pass filtered to 15Å and used as templates 1003 to auto-pick 3,762,631 particles, which were then extracted with a box size of 300 pixels 2D class averages that were ice contamination or showed no features were discarded. The remaining 1006 952,705 particles were next used for ab initio reconstruction and heterogenous refinement using 2 classes Per-particle local 1009 motion correction was then carried out on the remaining 662,855 particles. The particles were then used 1010 for ab intio reconstruction and heterogeneous refinement using 4 classes. 320,179 particles from the most 1011 complete class were used to carry out homogenous refinement using the 3D reconstruction for that class 1012 as a starting model, yielding a reconstruction with an overall resolution of 2.99Å based on the Fourier shell 1013 correlation (FSC) 0.143 criterion. The particles were further refined using local CTF refinement before being 1014 used for non-uniform refinement with simultaneous global CTF refinement, yielding a map with an overall 1015 resolution of 2.90Å. Finally, the map was subjected to a final non-uniform refinement using a mask 1016 enveloping the entire volume with the rotation fulcrum centered at the low-resolution nanobody The contrast transfer function (CTF) of the resulting 1020 micrographs was estimated using CTFFIND4 with default settings. 2D class averages from a previous 1021 dataset were low-pass filtered to 15Å and used as a template to auto-pick 4,792,176 particles, which were 1022 then down-sampled by 2 (resulting pixel size of 1.079 Å/pix) and extracted with a box size of 320 pixels. The particles were subjected to multiple rounds of 2D classification with the 2D class re-center threshold 1024 set to 0.05, and a circular mask of 200Å. 2D class averages that did were ice contamination or did not align 1025 to high-resolution were then discarded classes were then used for ab initio reconstruction and 1028 heterogenous refinement using 2 classes twice iteratively. 196,390 particles from the better 3D 1029 reconstruction were used to carry out homogenous refinement using the 3D reconstruction for that class as 1030 a starting model, yielding a reconstruction with an overall resolution of 3.49Å based on the Fourier shell 1031 correlation (FSC) 0.143 criterion. The map was further refined non-uniform refinement, yielding a map with 1032 an overall resolution of 3.36Å. The full details of the p110g Supplemental Figures and Tables for An HDX-MS optimised approach to characterise nanobodies as tools for 1056 biochemical and structural studies Protein samples were rapidly thawed and injected onto an integrated fluidics system containing a HDx-3 890 PAL liquid handling robot and climate-controlled chromatography system (LEAP Technologies), a Dionex 891 Ultimate 3000 UHPLC system, as well as an Impact HD QTOF Mass spectrometer (Bruker). The protein 892 was run over either one (at 10°C)