Light-responsive monobodies for dynamic control of customizable protein binding

Customizable, high affinity protein-protein interactions, such as those mediated by antibodies and antibody-like molecules, are invaluable to basic and applied research and have become pillars for modern therapeutics. The ability to reversibly control the binding activity of these proteins to their targets on demand would significantly expand their applications in biotechnology, medicine, and research. Here we present, as proof-of-principle, a light-controlled monobody (OptoMB) that works in vitro and in vivo, whose affinity for its SH2-domain target exhibits a 300-fold shift in binding affinity upon illumination. We demonstrate that our αSH2-OptoMB can be used to purify SH2-tagged proteins directly from crude E. coli extract, achieving 99.8% purity and over 40% yield in a single purification step. This OptoMB belongs to a new class of light-sensitive protein binders we call OptoBinders (OptoBNDRs) which, by virtue of their ability to be designed to bind any protein of interest, have the potential to find new powerful applications as light-switchable binders of untagged proteins with high affinity and selectivity, and with the temporal and spatial precision afforded by light.

domain 46 is a valuable resource for our rational protein engineering approach. Our strategy to 126 develop a light-sensitive HA4 was to design various chimeras of this monobody with the light-127 oxygen-voltage-(LOV) sensing domain of Phototropin1 from Avena sativa, AsLOV2, and test 128 their ability to bind and release the SH2 domain depending on light conditions. 129

130
To build our chimeras, we used a truncated version of AsLOV2, which induces light-dependent 131 conformational changes in engineered nanobody chimeras more efficiently than its full-length 132 counterpart 45 . Our strategy was to insert the shortened AsLOV2 domain in all seven structurally-133 conserved, solvent-exposed loops of HA4 (Fig. 1b). Given the large conformational change of 134 AsLOV2 triggered by light (Fig. 1a), our hypothesis was that the native conformation of the 135 monobody domain in some chimeras would be preserved in the dark, allowing it to bind to SH2,136 but disrupted in the light, causing it to release its target. Guided by the crystal structure of HA4 137 bound to SH2 (PDB ID: 3k2m) 46 , we explored potential sites within the seven solvent-exposed 138 loops in HA4 where we could insert AsLOV2. We selected as many positions as possible in 139 each loop, avoiding those were we have reasons to believe the dark state conformation of 140 AsLOV2 would disrupt the core b-sheets of the monobody or interfere with the monobody-SH2 141 binding interaction. We also excluded positions where the light-triggered conformational change 142 of the AsLOV2 Ja helix might be impeded by clashes with the monobody core. After this 143 6 structural analysis, we selected 17 AsLOV2 insertion sites across all solvent exposed loops of 144 HA4, as well as N-or C-terminal fusions ( Fig. 1b and   sheets, bSh1(black) and bSh2 (light blue), and seven structurally conserved loops (red), where 152 AsLOV2 was inserted in our chimeras. Loop L4, where AsLOV2 is inserted in OptoMB, is 153 shown with an arrow. The cartoon below shows the relative size and location of loops (red), 154 including positions of AsLOV2 insertion (red lines) and insertion sites of identified light-155 responsive chimeras: MS29 and SS58 (in OptoMB). c, Schematic diagram of pull-down screens 156 used to identify light-responsive chimeras. Co 2+ agarose beads (pink) were used to immobilize 157 His6-YFP-SH2 (yellow-gray), which were then incubated with HA4-AsLOV2 chimeras (blue-158 orange) in either dark or light. After washing and eluting with imidazole, the eluents were 159 resolved on SDS-PAGE, were differences in protein band intensity between samples exposed to 160 different light conditions reflect differences in chimera SH2 binding. To find chimeras that can bind the SH2 domain in the dark but not in the light, we screened our 170 constructs using an in vitro pull-down assay (Fig. 1c). First, we produced an N-terminally His-171 tagged fusion of yellow fluorescent protein (YFP) and SH2 domain (His6-YFP-SH2) in E. coli, 172 and immobilized it onto cobalt-charged agarose beads. We then incubated the beads with crude 173 extracts of E. coli expressing each of the different AsLOV2-HA4 chimeras, in either blue light or 174 darkness. After washing the beads under the same light conditions (see Methods), we eluted with 175 imidazole and resolved the products with denaturing polyacrylamide gel electrophoresis  to analyze the binding of each chimera in different light conditions (Fig. 1c,d). We 177 anticipated that chimeras that bind to SH2 preferentially in the dark would show a more intense 178 band on SDS-PAGE for samples that were incubated and washed in the dark, relative to the 179 samples treated in the light (Fig. 1c,d and Supplementary Fig. 1). 180

181
We found that AsLOV2 insertions in two different HA4 loops produce chimeras with the 182 expected behavior in our pull-down assays. One promising chimera has AsLOV2 inserted 183 8 between residues Met29 and Ser30 (a site we call MS29, following a naming system for sites 184 used in this study), located in loop L2 of HA4 (Fig. 1b,d and Supplementary Fig. 1). We only 185 saw an effect involving this loop when AsLOV2 was inserted at position MS29 and loop L2 was 186 shortened by removing the three surrounding amino acids (Ser30, Ser31 and Ser32, see 187 Supplementary Sequences). Another chimera with positive results has AsLOV2 inserted between 188 Ser58 and Ser59 (site SS58) located in loop L4 of HA4 (Fig. 1b,d). Insertions at other positions 189 within loop L4 ( 57 YSSS 60 ) show smaller degrees of variation in band intensity between beads 190 treated in the light versus in the dark ( Fig. 1d and Supplementary Fig. 1). This suggests that loop 191 L4 is a "hot spot" for favorable orientations between AsLOV2 and the monobody to produce 192 light-responsive chimeras that switch between a conformation that allows target binding (in the 193 dark) and one that promotes target dissociation (in the light). Compared to AsLOV2 insertions at 194 other positions in loop L4, the insertion at SS58 shows the largest qualitative difference between 195 the chimera bound to beads in different light conditions ( Fig. 1d and Supplementary Fig. 1  The light dependency of OptoMB interactions with YFP-SH2 can also be analyzed in solution by 241 size exclusion chromatography (SEC). We prepared mixtures of purified YFP-SH2 and OptoMB 242 (or HA4 monobody, as control), in which OptoMB was added in excess (see Methods). We then 243 loaded each sample to a gel filtration column under continuous darkness or blue light 244 ( Supplementary Fig. 2c). We found that for both the dark-incubated OptoMB sample (Fig. 2c) 245 and the HA4 monobody in either light condition ( Supplementary Fig. 2d), the YFP-SH2 elutes 246 primarily as a monobody-bound complex. In contrast, the illuminated OptoMB sample shows a 247 higher retention time for the YFP-SH2-OptoMB complex and a larger proportion of the YFP-248 SH2 eluting as a monomer on its own (Fig. 2c). Taken together, our bead-imaging and SEC data 249 are both consistent with a blue light-triggered reduction in the OptoMB-SH2 binding affinity, 250 both in solution and on protein-coated surfaces. 251

252
To quantify the changes in OptoMB-SH2 binding, we determined the kinetic rate constants and 253 binding affinity in different light conditions. In these assays we took advantage of three classes 254 of mutations in AsLOV2 to vary properties of the OptoMB binding switch. Bio-layer 255 interferometry uses visible light for measuring changes in binding, so we first prepared a light-256 insensitive OptoMB by introducing the well-characterized C450V mutation in AsLOV2 that 257 renders it light-insensitive 36,51 (OptoMBC450V). Conversely, we ensured that illumination could 258 drive efficient conversion to the lit state by generating additional OptoMB variants with 259 mutations that extend the lifetime of the lit state after illumination from ~80 to 821 or up to 260 4300 sec in vitro (AsLOV2 V416I 49 or V416L 52 respectively). Finally, we added the AsLOV2 261 G528A and N538E mutations (to make the triple mutant OptoMBV416I_G528A_N538E), which have 262 been reported to stabilize the dark state conformation and potentially decrease leakiness in the 263 absence of illumination 53,54 . We fit the resulting binding and dissociation data to a mass-action 264 kinetic binding model ( purifying a protein of interest simply by shifting illumination conditions (Fig. 3a), a procedure 287 that we termed "light-controlled affinity chromatography" (LCAC). We immobilized two 288 variants of His-tagged OptoMBs harboring either the wild-type AsLOV2 or the triple-mutant 289 (OptoMBV416I_G528A_N538E) described above, onto Co-charged Agarose beads to make an αSH2-290 OptoMB affinity resin. We then incubated OptoMB-coated beads with crude lysate from E. coli 291 overexpressing YFP-SH2. After washing thoroughly in the dark (see Methods), we eluted with 292 blue light either in batch (Fig. 3b) or in a column ( Supplementary Fig. 3). After elution, beads 293 were washed with imidazole to recover any remaining protein bound to the beads in order to 294 estimate the capacity and yields of the resin. With these initial LCAC purification trials, we 295 achieved 95-98% purity in a single step, with yields ranging from 18-30% and binding capacities 296 from 112-145 nmol (4.5-6 mg) of SH2-tagged YFP per mL of OptoMB resin, depending on the 297 OptoMB variant used (Supplementary Table 3). 298   299 To test whether LCAC could be applied to larger and more complex proteins, we also used it to 300 purify the main pyruvate decarboxylase from Saccharomyces cerevisiae, Pdc1p. This enzyme 301 catalyzes the decarboxylation of pyruvate to acetaldehyde for ethanol fermentation, and is 302 composed of a homotetramer of 61 kDa monomers 55 , significantly larger than YFP. We fused the 303 SH2 domain to the N-terminus of Pdc1p (SH2-PDC1) and performed LCAC to purify it from 304 crude E. coli lysate, as described above, using a resin coated with OptoMBV416I_G528A_N538E. This 305 procedure enabled purification of Pdc1p to 96% purity with a 39% yield ( Fig. 3c and  306 Supplementary Table 3). It is noteworthy that this purification works considerably well, despite 307 the potential binding avidity of Pdc1p tetramers that would be predicted to increase the protein's 308 apparent affinity in both the light and dark. These results demonstrate that LCAC can be applied 309 to purify relatively large proteins with quaternary structures of up to at least 300 kDa (including 310 the fused SH2 domain), achieving a high degree of purity and an acceptable yield. Our results 311 with YFP-SH2 and SH2-PDC1 further demonstrate that OptoMB-assisted purification is 312 compatible with both N-and C-terminal SH2 tags. to determine whether an alternative resin could be used for LCAC, we immobilized our 338 OptoMBV416I_G528A_N538E onto cyanogen bromide-activated sepharose beads (CNBr-beads), which 339 immobilizes proteins by making covalent bonds with its primary amines (see Methods). 340 Following the same purification protocol as above, we found that CNBr-beads are also effective 341 at purifying both YFP and Pdc1p ( Supplementary Fig. 4). A single step of CNBr-based 342 purification achieved yields above 40% and purity of 96.7-99.8%, surpassing any other LCAC 343 method tested (Supplementary Table 3). These gains are likely related to the reduced non-344 specific binding of E. coli proteins to CNBr-sepharose and covalent OptoMB attachment which 345 allowed for more extensive washing at higher salt concentrations. We observed that the total 346 loading capacity is not as high as that of OptoMB-coated Co-agarose (Supplementary Table 3 OptoMB binding would cause the OptoMB to redistribute between the cytosol and plasma 361 membrane (PM) (Fig. 4a), as has been observed for conventional optogenetic protein-protein 362 interactions in previous studies [56][57][58] . As a control, we expressed irFP-labeled HA4 monobody 363 instead of OptoMB, which would be expected to bind to the membrane-localized SH2-mCherry-364 CAAX regardless of illumination conditions. 365 366 Fluorescence imaging confirmed the PM localization of SH2-mCherry-CAAX (Fig. 4b,c left 367 panels) as well as constitutively PM-bound HA4-irFP (Fig. 4b). In contrast, we observe a light-368 dependent shift in OptoMB localization (Fig. 4c) Here we show that by taking a rational protein engineering approach it is possible to develop a 398 light-switchable monobody (OptoMB). Our OptoMB design strategy was based on the fusion of 399 a light-switchable AsLOV2 domain to a structurally conserved loop of the H4A monobody 400 which binds with high affinity and selectivity to the SH2 domain of the human Abl kinase. We 401 find that this chimera binds to SH2 in the dark with similar affinity as its parental monobody 402 H4A; but in blue light its binding affinity drops 150-to 300-fold relative to the dark state 403 (Supplementary Table 2). Furthermore, the light responsiveness of OptoMB is effective to 404 control binding to proteins fused to SH2 at either their N-or C-terminus. OptoMBs, along with 405 the light-dependent nanobodies (OptoNBs) we describe in an accompanying study 45 , belong to a 406 new class of light-dependent protein binders we call OptoBinders (OptoBNDRs), which offer 407 promising new in vivo and in vitro applications. 408 409 A close inspection of the structural model of OptoMB and binding measurements suggest a 410 possible mechanism for the light-dependent binding affinity of OptoMB. The monobody protein 411 fold consists of two antiparallel b-sheets (bSh1 and bSh2) that interact with each other to form 412 the protein core (Fig. 1b). In our original chimera screens, we inserted AsLOV2 in all 413 intervening loops within (L1, L3, L6 and L7) and between (L2, L4 and L5) these b-sheets. OptoMB-SH2 off-rate (koff). While we do observe as much as a 10-fold increase in koff in the 436 light, consistent with this hypothesis, we found that most of the change in affinity comes from as 437 much as a 46-fold change in the binding rate (kon) between dark and light conditions 438 (Supplementary Table 2). This suggests that the binding site is disrupted enough to inhibit 439 OptoMB-SH2 association in the first place. Yet despite what is likely to be a substantial 440 conformational change in the OptoMB, we find that productive binding is fully reversible in 441  Table 2). In principle, these changes could increase the overall 468 energy beyond the 3.8 kcal/mol measured for the wild-type AsLOV2 domain. The structural 469 targets set for optimization and the expected results will of course vary depending on the 470 particular application objectives and the individual OptoMB/ligand pair in question. 471

472
The performance of our light-dependent OptoMB/SH2 binding pair is comparable to that of 473 other optical dimerization systems previously used to control transcription, protein-protein 474 interactions, or protein localization 60-62 . However, these previous systems have been developed 475 by fusing proteins of interest to specific light-dependent interaction partners (such as PhyB/PIF3 476 or Cry2/CIB) evolved in photosensitive organisms. While the demonstrations we used in this 477 study rely on fusing SH2 to different proteins, the true potential of this technology is not so much 478 to replace the light-responsive tags used in previous optogenetic systems with SH2 and OptoMB,479 but in the possibility of engineering other monobodies against different targets of interest to 480 make their specific interaction light-dependent. The relative simplicity of the monobody fold and 481 the likely structural mechanism that confers light-dependent binding makes it reasonable to 482 expect that other monobodies could be engineered to be light-dependent; starting from inserting 483 AsLOV2 in loop L4 and then optimizing the precise position of the insertion site and the linkers. 484 Monobodies could in principle be designed and selected to bind any protein of interest 6,31 or to 485 different epitopes on a single a target 31 . Our approach may thus be useful to design reversible 486 interactions to a protein of interest without the need for a binding tag, which for some proteins 487 may be impractical, not possible, or interfere with their natural activity.  activates phototropin kinase activity. Biochemistry 43, 16184-16192 (2004 Med. Chem. 6, 1909-1926(2014.

Plasmid construction of chimeras for bacterial expression 650
One-step isothermal assembly reactions (Gibson assembly) were performed using previously 651 described methods 61 . The monobody HA4 and the SH2 domain (codon optimized for E. coli 652 expression) were ordered as gBlocks from Integrated DNA Technologies (IDT) containing 653 homology arms. The following vectors from the pCri system 63 were used: pCri-7b for constructs 654 without a 6x-histidine tag; pCri-8b for constructs with a 6x-histidine tag (N-terminus), and pCri-655 11b for constructs with both SUMO and 6x-histidine tags (N-terminus). As the pCri vectors 656 contain YFP, the synthetized SH2 domain was inserted into pCri-7b and 8b (previously 657