A semi-synthetic red light photoswitch for optogenetic control of protein


 Red light is useful for optogenetic control of various protein activities in mammalian deep tissues due to its high tissue penetration, low invasiveness, and low light scattering. However, technology that enables for the optogenetic manipulation using red light remains elusive. Here we develop a red light-activatable, semi-synthetic photoswitch, named MagRed. MagRed is composed of a red light-absorbing bacterial phytochrome incorporating a mammalian endogenous chromophore, biliverdin, and its photo-state-specific de novo synthetic binder. MagRed allows us to reassemble split-proteins using red light and thereby develop a red light-activatable Cre recombinase, which is applicable for mammalian deep tissues. Additionally, we take advantage of MagRed to develop red light-inducible transcription system based on the CRISPR-Cas9 system, enabling for high induction (up to 378-fold) of multiple user-defined endogenous target genes. With high versatility and regulatability, MagRed provides a powerful technology that easily facilitates optogenetics applications for a variety of biological research areas.

RpBphP1 upon red light illumination but also to its apo-protein irrespective of red light 85 illumination. The light-independent binding of PpsR2 with the apo-protein of RpBphP1 causes 86 spontaneous activation of the optogenetic tools in the dark condition and thereby greatly 87 hampers their regulatability. These benchmarking studies reveal that core technology with high 88 versatility and regulatability that enables for the optogenetic manipulation using red light 89 remains elusive. 90 To better facilitate optogenetics using red light, we developed a red light-activatable,   108 To develop a red light photoswitch, we applied synthetic biological approaches to generate a 109 photo-state specific de novo binding partner of DrBphP, a bacterial phytochrome derived from 110 Deinococcus radiodurans (Fig. 1a). DrBphP incorporates BV as a mammalian endogenous 111 chromophore and reversibly photoconverts between a Pr dark-state (λmax = 701 nm) and a Pfr 112 photo-state (λmax = 752 nm) with two-wavelength light illuminations (Supplementary Fig. 1a). 113 In addition to the chromophore availability in mammalian cells and the controllability using  Fig. 1b). The large conformational change of DrBphP is an advantage for developing its de 118 novo synthetic binder that selectively binds to the Pfr photo-state but not to the Pr dark-state. 119 Importantly, previous studies reported that DrBphP exhibits biexponential slow dark reversion 120 kinetics with decay amplitude of 24% for 7 min and with that of 76% for the following 1,291 121 min 27,28 , which is much slower than that of RpBphP1 (t1/2=2.83 min 24 ) (Supplementary Fig.   122 2). The slow dark reversion kinetics of DrBphP is beneficial for keeping it associated with the 123 photo-state specific binder even after turning off red light illumination, thereby enabling for a 124 sustained activation of optogenetic tools even by single or pulsed illumination. In contrast, 125 RpBphP1 does not have such a controllability due to its fast dark reversion kinetics 24 . 126 To generate a binding partner of DrBphP, we applied Affibody, the Z domain of 127 immunoglobulin-binding staphylococcal protein A. Thirteen residues in the first and second 128 helices of Affibody were randomized to generate its ribosome-displayed library 29 (Fig. 1b). 129 We purified the DrBphP-PSM protein and immobilized it on magnetic beads, and then  Table 1). 136 We assessed whether the top 10 candidates of the prioritized Affibody clones could 137 interact with DrBphP-PSM in mammalian cells by the tetR-tetO-based firefly luciferase (fluc) 138 expression system with VP16, a transcription activation domain (Fig. 1c). Of the tested clones, 139 one Affibody clone named Aff6 displayed a high bioluminescence intensity upon red light 140 6 illumination at 660 nm, which was comparable with that induced by a direct fusion of tetR and 141 VP16 (Fig. 1d, e, Supplementary Fig. 3). This result indicates that Aff6 binds to the Pfr photo-142 state of DrBphP-PSM with a high affinity. However, its Light/Dark contrast was considerably 143 low (1.2 fold-induction) because it also showed a substantially high leakiness in the dark. We 144 found that the leak in the dark was significantly decreased by using the full-length DrBphP 145 instead of DrBphP-PSM (Fig. 1e) Fig. 4). We conducted saturation mutagenesis at the V18 residue in Affi6 and found that V18F, 153 V18W and V18H mutations significantly improved the Light/Dark contrast compared to the 154 original Aff6 (Fig. 1d, e and Supplementary Fig. 5). Truncation of the N-terminal 155 unstructured three residues from Aff6_V18F further enhanced the Light/Dark contrast (8.3-156 fold induction with 67% activity of tetR-VP16) (Fig. 1d, e). We named the pair of DrBphP and 157 Aff6_V18FΔN "MagRed".

159
Photoswitching property of MagRed. 160 To further characterize MagRed in mammalian cells, we performed a bioluminescence assay 161 using a split-firefly luciferase (split-fluc) 30 (Fig. 2a, b). In HEK 293T cells expressing 162 MagRed-fused split-fluc, repeated induction of bioluminescence was feasible using the 660 nm 163 and the 800 nm pulsed illuminations (Fig. 2c). This result demonstrates that the two-colored 164 pulsed illuminations can independently and repeatedly control the association (switch-ON) and 165 the dissociation (switch-OFF) of MagRed. Next, we measured the dissociation kinetics of 166 MagRed. After the 660 nm illumination was turned off, MagRed maintained its association 167 form with approximately 70% efficiency, following 21 % decrease for the first 10 min (Fig.   168 2d). This biphasic slow dissociation kinetics of MagRed is correlated with the biexponential 169 slow dark reversion kinetics of DrBphP (Supplementary Fig. 6) RNA aptamer and MS2 coat protein (Fig. 3a). We designed all configurations for CPTS and 179 examined their transcription activities using a luciferase reporter (Fig. 3b). In addition, we also 180 tested the existing red light photoswitch RpBphP1-PpsR2/QPAS1 in CPTS as benchmark 181 experiments to compare with MagRed. Most of the configurations for CPTS using MagRed 182 showed significantly high Light/Dark contrasts, which were up to 70-fold induction (Fig. 3b-183 d, and Supplementary Fig. 7a). On the other hand, all the configurations for CPTS using 184 RpBphP1-PpsR2/QPAS1 showed much lower Light/Dark contrasts, which were up to 4.2-fold 185 induction ( Fig. 3b-d). We found that the low Light/Dark contrasts of RpBphP1-PpsR2/QPAS1-186 based CPTS were attributable to their high leak activities in the dark (Supplementary Fig. 7b, 187 c). These results demonstrate that MagRed can control the domain recruitment more precisely 188 and dynamically than RpBphP1-PpsR2/QPAS1 in CPTS.

189
To examine why RpBphP1-PpsR2/QPAS1-based CPTS shows high leak activity in 190 the dark, we accessed the effect of BV concentration on its transcription activity. We found 191 that the addition of excess amount of BV decreased the dark leak activity of CPTS based on 192 RpBphP1-PpsR2/QPAS1 (P=0.00781, Supplementary Fig. 8a, b). Following this result, we 193 hypothesized that the apo-protein of RpBphP1, which is not yet incorporated with BV, could 194 cause the high leak activation of RpBphP1-PpsR2/QPAS1-based CPTS in the dark. To examine 195 the effect of the apo-protein of RpBphP1 on the transcription activity, we generated mutants of 196 RpBphP1, in which the cysteine residues covalently bound to BV were substituted to alanine 197 (C20A) and serine (C20S), respectively. These mutants exhibited high transcription activities 198 regardless of red light illumination (Fig. 3e). The results indicate that PpsR2 binds not only to 199 the holo-protein of RpBphP1 upon red light illumination but also to its apo-protein irrespective  Fig. 3f and Supplementary Fig. 9). 205 The results demonstrate that Aff6_V18FΔN does not bind to the apo-protein of DrBphP, 206 leading to the minimum leak activation of MagRed-based optogenetic tools in the dark. Among the tested configurations for MagRed-based CPTS (Fig. 3c), we focused on 208 the configurations #1 and #3. Additionally, taking advantage of the small size of 209 Aff6_V18FΔN (6.2 kDa), we designed tandem fusions of Aff6_V18FΔN to enhance its 210 apparent affinity with DrBphP. The tandem dimer of Aff6_V18FΔN substantially enhanced 211 the transcription activity of CPTS upon red light illumination without increasing the leak 212 activity in the dark in both configurations #1 and #3 (Fig. 4a). Especially, the configuration #3 213 with the tandem dimer of Aff6_V18FΔN resulted in an enhanced Light/Dark contrast (619-214 fold induction). Although we also tested the configuration #3 with the tandem trimer and 215 tetramer of Aff6_V18FΔN, the trimer and tetramer constructs did not exhibit further 216 enhancement of the transcription activity (Fig. 4a). Therefore, we concluded the configuration 217 #3 with the tandem dimer of Aff6_V18FΔN as the best version of MagRed-based CPTS called 218 "Red-CPTS". We confirmed that Red-CPTS works robustly in various cell lines from different 219 origins (Supplementary Fig. 10). To examine whether Red-CPTS can be applicable to 220 endogenous gene targets, we next applied Red-CPTS to multiple endogenous genes, delivering 221 a group of four gRNAs separately targeting human ASCL1, HBG1, IL1R2, and MYOD1 222 promoters. We found that the targeted endogenous genes were simultaneously activated in 223 HEK 293T cells by red light. Red-CPTS significantly increased all the targeted gene 224 transcriptions with high Light/Dark contrasts, which were up to 378-fold induction (Fig. 4b). 225 Importantly, in all the targeted endogenous genes, the mRNA levels of Red-CPTS-transfected 226 cells in the dark were comparable to those of mock-transfected cells, demonstrating that Red-227 CPTS has no obvious leak activity in the dark and thereby enables for robust regulation of 228 multiplexed user-defined endogenous gene activation using red light illumination.

230
Red light-activatable DNA recombination system based on split-Cre reassembly. 231 In addition to the domain recruitment strategy applied for CPTS, next we examined whether 232 MagRed can be applied for the split-protein reassembly with red light illumination. Among the 233 existing optogenetic tools based on the split-protein reassembly system with the blue light-234 activatable photoswitches, site-specific DNA recombinase is one of the most attractive 235 targets 11,12,20,31-33 . Especially, Cre recombinase is the most widely used DNA recombinase in 236 biology, biotechnology and biomedical studies 34-37 . We fused MagRed to split-Cre fragments 237 to develop a red light-activatable Cre recombinase applicable in mammalian systems (Fig. 5a). 238 To test all configurations, we fused either DrBphP or Aff6_V18FΔN to the newly created N-239 and C-terminal ends of split-Cre fragments (CreN and CreC) as well as the original N-terminal 240 end of CreN (Fig. 5b). We also tested two split positions (CreN59/CreC60 and 241 CreN104/CreC106) for the split-Cre fragments. The DNA recombination activities were 242 examined using a Floxed-STOP fluc reporter in HEK 293T cells (Supplementary Fig. 11a). 243 Most of the configurations using MagRed and split-Cre exhibited red light-dependent DNA 244 recombination with significant Light/Dark contrasts ( Fig. 5b-d, and Supplementary Fig. 12a). 245 Of the tested configurations, NLS-CreN104-Aff6_V18FΔN and NLS-DrBphP-CreC106 gave 246 the highest Light/Dark contrast (31-fold induction). This MagRed-based red light-activatable 247 Cre recombinase was named "RedPA-Cre". We confirmed that additional BV supplementation 248 did not have significant effect on the DNA recombination activity of RedPA-Cre 249 ( Supplementary Fig. 13), revealing that RedPA-Cre works robustly at the endogenous BV 250 concentration of living mammalian cells as Red-CPTS does. 251 We also tested RpBphP1-PpsR2/QPAS1 for the split-protein reassembly strategy with  Fig. 11b). 277 Upon red light illumination for 24h, RedPA-Cre induced mCherry fluorescence with 55% 278 efficiency of that induced by iCre (Fig. 5e). 279 Next, we examined whether RedPA-Cre could be applied to bicistronic designs.     Emulsiflex C5 high-pressure homogenizer at 12,000 psi (Avestin Inc., Ottawa, Canada).

392
Homogenates were centrifuged at 165,000g for 30 min, and supernatants were filtered through 393 a 0.8-µm cellulose acetate membrane. Filtrated samples were loaded onto a nickel-affinity His-394 trap column (GE Healthcare, Little Chalfont, UK) using an ÄKTAprime plus (GE Healthcare) 395 System. The column was washed using the lysis buffer containing 50 mM and 100 mM 396 imidazole, followed by elution using a linear gradient of the lysis buffer containing 100 to 400 397 mM imidazole (1 mL/min, total 15 min). After incubation with 1 mM EDTA for 1 h on ice, the 398 purified proteins were dialyzed against the lysis buffer to remove EDTA and imidazole.

399
Purified proteins concentration was determined by the Bradford method.     bioluminescence signals were integrated over 1 s at room temperature and plotted every 1 s.

527
To illuminate 660 nm light to the cells, we stopped the bioluminescence measurement and 528 removed the sample dish from the luminometer. We then irradiated the sample dish with 660 529 nm light (10 W m -2 ) for 1 min using the 660 nm LED array and immediately placed the sample 530 dish back into the luminometer to resume the bioluminescence measurement. The procedures 531 for illuminating 800 nm light to the cells were identical to those described above.

548
The procedures for plating and transfection were identical to those described above, except for 549 using a mixture of sgRNAs separately targeting four endogenous genes and omitting the