Introduction of reversible cysteine ligation ability to the biliverdin‐binding cyanobacteriochrome photoreceptor

Cyanobacteriochrome (CBCR) photoreceptors are distantly related to the canonical red/far‐red reversible phytochrome photoreceptors. In the case of the CBCRs, only the GAF domain is required for chromophore incorporation and photoconversion. The GAF domains of CBCR are highly diversified into many lineages to sense various colors of light. These CBCR GAF domains are divided into two types: those possessing only the canonical Cys residue and those with both canonical and second Cys residues. The canonical Cys residue stably ligates to the chromophore in both cases. The second Cys residue mostly shows reversible adduct formation with the chromophore during photoconversion for spectral tuning. In this study, we focused on the CBCR GAF domain AnPixJg2_BV4, which possesses only the canonical Cys residue. AnPixJg2_BV4 covalently ligates to the biliverdin (BV) chromophore and shows far‐red/orange reversible photoconversion. Because BV is a mammalian intrinsic chromophore, BV‐binding molecules are advantageous for in vivo optogenetic and bioimaging tool development. To obtain a better developmental platform molecule, we performed site‐saturation random mutagenesis and serendipitously obtained a unique variant molecule that showed far‐red/blue reversible photoconversion, in which the Cys residue was introduced near the chromophore. This introduced Cys residue functioned as the second Cys residue that reversibly ligated with the chromophore. Because the position of the introduced Cys residue is distinct from the known second Cys residues, the variant molecule obtained in this study would expand our knowledge about the spectral tuning mechanism of CBCRs and contribute to tool development.


Introduction
Cyanobacteriochrome (CBCR) photoreceptors are distant relatives of the phytochromes showing red/far-red reversible photoconversion [1].The photosensory module of the phytochromes is composed of three domains: Per/Arnt/Sim, cGMP-phosphodiesterase/ adenylate cyclase/FhlA (GAF), and phytochrome-specific (PHY) domains [2].By contrast, CBCRs require only the GAF domain for chromophore incorporation and proper photoconversion.CBCRs covalently bind the linear-tetrapyrrole chromophore, also called bilin pigment, and show photoconversion between two bistable light-absorbing states in most cases.Rotation of the double bond between rings C and D, Z/E isomerization, occurs as a primary photoreaction.The Z-isomer is photochemically the dark-adapted state, whereas the E-isomer is the photoproduct state.A vast repertoire of CBCR GAF domains showing various photoconversions, such as red/green, red/blue, blue/green, and farred/orange, has been characterized.In the context of molecular compactness and spectral diversity, CBCR GAF domains are advantageous for developing optogenetic and bioimaging tools [3].
Initial studies have identified that CBCRs covalently bind phycocyanobilin (PCB) or phycoviolobilin (PVB) [4][5][6][7][8].Because PVB has no double bond between rings A and B, the PVB-binding GAF domains sense the relatively shorter wavelength UV-to-green region, whereas the PCB-binding ones sense the UV-to-red region.In PVB-binding GAF domains, PCB is initially incorporated into the CBCR GAF domain, followed by the isomerization of PCB to PVB [6,9].Among the PCB-binding CBCR GAF domains, many GAF domains form a huge expanded red/green (XRG) lineage [7,[10][11][12][13][14].Typical XRG GAF domains covalently bind PCB and show red/green reversible photoconversion, but many atypical ones have been identified to show violet/blue, violet/orange, red/blue, or green/blue photoconversion.Furthermore, we have identified that the CBCR GAF domains from a unique cyanobacterium Acaryochloris marina covalently bind biliverdin (BV) and show far-red/orange reversible photoconversion [15,16].Because BV has the most extended p-conjugated system among natural linear-tetrapyrrole chromophores, the sensing light window is expanded to the far-red region, which penetrates into the mammal deep tissues escaping from absorption by hemoglobin.Furthermore, BV is a mammalian intrinsic chromophore; therefore, BV-binding CBCR GAF domains are promising platforms for in vivo optogenetic and bioimaging tools [17].In this context, to understand the molecular basis of BV incorporation into CBCR GAF domains, we compared BVacceptable molecules with BV-rejective molecules and succeeded in identifying four residues crucial for BV incorporation [18].The crystal structure of the BVbinding CBCR variant introducing these four residues, AnPixJg2_BV4, elucidated a detailed molecular basis to accept BV.
Cys residues within the CBCR GAF domains play crucial roles in chromophore ligation and spectral tuning.Most CBCR GAF domains possess highly conserved Cys residue (canonical Cys) for stable chromophore ligation [1].In addition, many CBCR GAF domains have a second conserved Cys residue (second Cys) [4][5][6]10,12,14,[19][20][21][22][23][24][25][26].The typical role of the second Cys residue is reversible or stable attachment to C10 of the chromophore, which vastly shortens the p-conjugated system of the chromophore, leading to UV-to-blue absorption.Notably, the second Cys residues-from structurally different positions-have been reported to ligate to C10 of the chromophore.Four distinct structural positions have been identified for the second Cys residue [10,12,14,19].Of these, the second Cys residue within the highly conserved DXCF motif has an additional function in that it is crucial for isomerization from PCB to PVB upon chromophore incorporation [6,9,21].
In this study, we performed site-saturation mutagenesis on AnPixJg2_BV4, which does not possess a second Cys residue, and serendipitously obtained a variant molecule that acquired a Cys residue to show far-red/blue reversible photoconversion.The acquired Cys residue seemed to reversibly ligate to the chromophore to function as a "second Cys" at a unique position.

Site-saturation random mutagenesis on AnPixJg2_BV4
To obtain BV-binding variant molecules with better chromophore-binding efficiency and better expression yield than the parent AnPixJg2_BV4 molecule, we first computationally performed site-saturation mutagenesis at the residues near the chromophore using the Rosetta protein design software suite (see Materials and Methods section and Figs S1 and S2 for details) [27].The calculations suggest that the replacement of Thr292 and Phe319 with several other residues improves the BV-binding affinity of AnPixJg2_BV4 (Fig. S3).Thr292 is positioned just after the Asp residue, which forms hydrogen bonds with the pyrrole nitrogen, whereas Phe319 is two residues before the canonical Cys residue, which covalently ligates to the ring A carbon (Fig. 1A, left).Because constructing various site-directed mutants individually is laborious, we established a method for site-saturation mutagenesis and colony colorization by using BV-producing Escherichia coli (see Materials and Methods section for details; Fig. S4A).Saturation mutagenesis at position Phe319 produced colonies with color variations (Fig. S4B).The details of these variants will be reported in the near future.In this study, we have focused on the serendipitous discovery of a variant molecule with unexpected spectral features targeting Thr292.

Introduction of Cys at Thr292
Saturation mutagenesis at Thr292 produced many pale brown colonies (Fig. 1B), which is the standard color for E. coli colonies, indicating that replacement at this position primarily affected chromophore-binding efficiency.Conversely, some colonies were green, which indicates that the chromophore-binding ability was retained (Fig. 1B).We picked up one of these green colonies (highlighted by the red circle, Fig. 1C), extracted the plasmid DNA, and determined the mutation by DNA sequencing.We found a Thr292 to Cys mutation in this plasmid (AnPixJg2_BV4_TC).We next expressed this protein in BV-producing Escherichia coli in a large-scale liquid culture and compared it with the parent molecule.The cell pellet expressing the AnPixJg2_BV4_TC variant showed a clearly different color from that expressing the parent molecule (Fig. 1D).The purified AnPixJg2_BV4_TC variant is covalently ligated to a chromophore judging from fluorescence detection using SDS/PAGE (Fig. S5).Unexpectedly, the variant protein showed reversible photoconversion between a far-red lightabsorbing (Pfr, black line in Fig. 1F) state, peaking at 702 nm, and a blue light-absorbing state (Pb, blue line in Fig. 1F), peaking at 473 nm (Fig. 1E,F,H), which is in contrast to the far-red/orange reversible photoconversion of the parent molecule AnPixJg2_BV4 (Fig. 1E,G,H).The AnPixJg2_BV4_TC variant showed heterologous behavior, including major farred/blue and minor far-red/orange reversible photoconversions, which were described in detail in the next paragraph.Judging from the absorption ratio of the Pfr peak to the protein UV peak, the BV-binding affinity of AnPixJg2_BV4_TC is comparable to that of the AnPixJg2_BV4 parent molecule (Fig. 1F,G).Although the Pfr states of these molecules are almost identical to each other, the Pb state of the AnPixJg2_BV4_TC variant protein was ~147-nm blueshifted in comparison with the Po state of the parent molecule.Based on the blue-absorbing property, the Pb state was considered to be derived from the covalent bond formation between C10 of the chromophore and the introduced Cys residue as well as known CBCRs, such as DXCF-type and insert-Cystype, which have the second Cys residue (Fig. 1J) [9,10,19,28,29].Acid denaturation analysis identified that states Pfr and Pb bound Z-configurated and Econfigurated BV, respectively (Fig. 1I, Fig. S6A-D), showing that the Cys-adducted Pb state is the photoproduct state (Fig. 1J).This result also indicates that the introduced second Cys residue does not have the ability to isomerize BV to the other chromophore, which is in contrast with the DXCF-type second Cys residue [6].
After prolonged irradiation with far-red light, the spectrum showed that a small component absorbed the orange-light region, peaking at 620 nm, which is similar to the Po state of the parent molecule (Fig. 1F, blue line).Orange-light illumination resulted in a decrease in orange light absorption and a concomitant increase in far-red light absorption (Fig. 1F, dark red line).The difference spectrum (dark red line minus blue line) is almost identical to that of the far-red/ orange reversible component of the parent molecule AnPixJg2_BV4 (Fig. 1H, gray solid line and gray dashed line).These results indicate that a small portion of the preparation does not show reversible Cys ligation activity but shows far-red/orange reversible photoconversion that is comparable with that of the parent molecule.
Next, we measured the thermal relaxation kinetics (also called dark reversion) of the AnPixJg2_BV4_TC variant molecule to monitor the Pb, Po, and Pfr peak absorbances at 25 °C (Fig. 2A-D).The Po peak absorbance decreased in the initial phase but increased after that, indicating at least two components (Fig. 2C).Because we detected major far-red/ blue and minor far-red/orange reversible components in the purified sample of AnPixJg2_BV4_TC, the initial decrease would reflect the decay of the Po state, and the later increase would correspond to the Pfr accumulation originated from the Pfr/Pb reversible component.Generation of a large amount of the Pfr state from the Pb state resulted in an increase of the Po peak absorbance because of the band-tail absorption, even in the orange region.In this context, we fitted two exponential components to the three kinetic curves monitored at the Pb, Po, and Pfr peak absorbances with high R-values (> 0.99) (Fig. 2B-F, Table 1).The fitting parameters are summarized in Table 1.The rate constants of the faster and slower components calculated from the three kinetic curves are similar to each other (Fig. 2E).The average half-lives of the faster and slower components were 319 s and 1662 s, respectively.Because a change in the Pb peak absorbance of the Pfr/ Po reversible component during the conversion process is prominently slight compared with that of the Pfr/Pb reversible component, the absorbance change of the faster component is much smaller than that of the slower component (Fig. 2F).Even single exponential component could be fitted to the thermal relaxation kinetics of the Pb peak absorbance with high R-value (> 0.99), and rate constant of this component is almost identical to the slower one of the two exponential components (Table 1).In conclusion, these results consistently showed that the Pb photoproduct state slowly converts to the Pfr dark-adapted state under the dark condition, whereas the Po photoproduct state readily converts to the Pfr dark-adapted state.Because the Po and Pb photoproduct states showed faster and slower thermal relaxation than the Po photoproduct state of the parent AnPixJg2_BV4 molecule (1199 s at 25 °C), respectively [18], the Cys introduction at the T292C position may affect the stability of both the Po and Pb photoproduct states.Cys ligation would stabilize the photoproduct state, whereas the presence of the Cys residue without ligation would destabilize the photoproduct state.
Because the release of the introduced Cys residue is fast without blue light illumination, we examined whether blue light illumination affected the Pb-to-Pfr conversion kinetics.The results showed that blue light illumination largely promoted Pb-to-Pfr conversion at 25 °C (Fig. 3).This fact indicates that the Pb state is light-responsive.Blue light illumination did not result in the full recovery of the Pfr state at 25 °C (Fig. 3), which is indicative of light-dependent bleaching.

PCB incorporation into the T292C variant molecule
We next expressed the AnPixJg2_BV4_TC variant molecule in PCB-producing E. coli.The purified PCBbinding AnPixJg2_BV4_TC variant is covalently ligated to a chromophore, judging from the fluorescence detected on the SDS/PAGE gel (Fig. S5D).The purified protein showed two absorption peaks in thered-to-far-red region (Fig. 4A, black line).Irradiation of this sample with red and far-red light sources resulted in a decrease of these absorbances, and a concomitant increase of absorption in the blue-to-green region (Fig. 4A, blue line, Fig. 4C), which contrasts with the red/green reversible photoconversion of the AnPixJg2_BV4 parent molecule derived from the PCB-producing E. coli (Fig. 4B).Using various light sources, we succeeded in extracting two independent photoconversion components: red/green reversible component (Fig. 4A, black line VS green line, Fig. 4D) and far-red/blue reversible component (Fig. 4A, black line VS red line, Fig. 4E).Acid denaturation analysis revealed that the red/green and farred/blue reversible components bound PCB and BV, respectively (Fig. 4F, Fig. S6A,B,E-I).AnPixJg2_BV4_TC, but not the parent molecule, could bind BV even in the PCB-producing E. coli.This indicates that the introduction of Cys residue at the T292C position results in a relative decrease of the PCB-binding affinity to the BV-binding affinity.Because the BV-binding affinity of AnPixJg2_BV4_TC is comparable to that of the parent molecule as described above, we can conclude that the binding affinity of the AnPixJg2_BV4_TC to PCB decreased.
The red/green reversible component is almost identical to that of the AnPixJg2_BV4 parent molecule derived from the PCB-producing E. coli (Fig. 4D), while the far-red/blue reversible component is almost identical to that of AnPixJg2_BV4_TC derived from the BV-producing E. coli.It is of note that the difference spectrum of the red/green reversible component has no negative peaks in the blue light region (Fig. 4D), indicating that the introduced Cys residue, T292C, does not bind to the chromophore in the photoproduct state of the PCB-binding AnPixJg2_BV4_TC.

Introduction of Cys at Tyr293
We introduced the Cys residue at Tyr293, which corresponds to the second Cys residue within the DXCF motif.Expression of this variant protein, AnPixJg2_B-V4_YC, in BV-producing E. coli resulted in very low chromophore-binding efficiency (Fig. 5A), whereas PCB was efficiently incorporated into this variant  protein (Fig. 5B).The BV-and PCB-binding ones showed far-red/orange and red/green reversible photoconversion that is comparable with that of the parent AnPixJg2_BV4 molecule (Fig. 5C).Notably, we observed absorption in the blue light region at ~405 nm in both states of the PCB-binding one.
Because the absorption in this region did not change upon photoconversion, the absorption may correspond to the Cys-adducted non-photoconvertible component.This suggests that although the Cys residue at Tyr293, just after Thr292, could attach to the PCB, this ligation is not reversible.Conversely, the same Cys residue did not ligate to the BV chromophore in contrast to that seen for Thr292.

Introduction of the Cys residue for AM1_C0023g2_SG
To expand our analysis, we focused on another BVbinding scaffold, AM1_C0023g2_SG [16].We replaced Thr309 (Thr292 in AnPixJg2_BV4) with Cys and purified this variant protein, AM1_C0023g2_SG_TC, from both BV-and PCB-producing E. coli.Both far-red/ blue and red/green reversible photoconversions were observed for BV-and PCB-binding ones, respectively (Fig. 6A-C).The Pfr and Pb states from BVproducing E. coli bound Z-and E-configurated BV, respectively (Fig. 6D, Fig. S7A-D), whereas the Pr and Pg states from PCB-producing E. coli bound Z-and E-configurated PCB, respectively (Fig. 6E, Fig. S7E-H).These observations were basically similar to those for AnPixJg2_BV4, namely, the introduced Cys residue could reversibly form a covalent bond with the BV but not with the PCB as well as the  at Thr292 position and identified two molecules (NEP87939g1 and HFS08092g1).We expressed these CBCR GAF domains in the BV-and PCB-producing E. coli and purified the chromophore-binding proteins for both GAF domains.However, we could not obtain chromophore-binding proteins from BV-producing E. coli.The PCB-binding NEP87939g1 and HFS08092g1 showed red/green and orange/green reversible photoconversion, respectively (Fig. 7), indicating that the Cys residue at T292C did not form a covalent bond with the chromophore during photoconversion.

Discussion
In this study, we focused on site-saturation mutagenesis at the Thr292 position.Although mutation at this position was mostly predicted to result in better BV incorporation by the computational method, we experimentally obtained many variants with worse BV incorporation than the parent molecule (Fig. 1B).We consider that such prediction is quite difficult and not fruitful in some cases.Conversely, in the case of site-saturation mutagenesis at the Phe319 position, we obtained many variants retaining the BV-binding capability (Fig. S4B).Further analyses of these mutants revealed that some variant molecules showed higher BV-binding capability and/or molecular coefficient (Suzuki et al. in preparation).We will discuss the potential applicability of computational prediction in the following publication.Instead, we have serendipitously obtained a singly mutated variant molecule AnPixJg2_BV4_TC (Thr292 was replaced with Cys) based on the parent AnPixJg2_BV4, which shows far-red/blue reversible photoconversion upon BV incorporation.Because the blue-absorbing Pb has an absorption peak at 473 nm, which is highly atypical among known CBCR GAF domains, this molecule provides a color variation to the CBCR palette.The Cys residue at T292C reversibly ligates to BV but not to PCB.Notably, Cys introduced at Tyr293 next to Thr292 did not ligate to BV.Taken together, the specific combination of Cys at T292C and BV is important for the reversible Cys ligation in the AnPixJg2_BV4 scaffold.
Although the AnPixJg2_BV4_TC variant molecule showed efficient reversible Cys-adduct formation, a small portion of the sample did not show adduct formation (Fig. 1F).To estimate the ratio of the far-red/ blue reversible component to total photoconvertible proteins, prolonged far-red illumination to the AnPixJg2_BV4_TC sample was performed to generate Pb and Po (Fig. 1F, Pb + Po, blue line).Then, we irradiated the Po state with orange light to prepare the sample in which only the far-red/orange reversible components are present as the Pfr state, whereas thefar-red/blue reversible components are present as the Pb state (Fig. 1F, Pb + Pfr, red line).After the absorbance of the "Pb + Po" sample at 702 nm was subtracted as background, we compared the Pfr absorbance of the "Pb + Pfr" sample (0.0486) with that of the "Pfr + Pfr" sample (0.301) in which both components are present as the Pfr state (Fig. 1F, Pfr + Pfr, black line) to calculate the ratio of the far-red/blue components to the far-red/ orange components.The results revealed that 83.9% of the preparation showed far-red/blue reversible photoconversion (Fig. 1J).Based on the absorbance changes of the fast (Po-to-Pfr) and slow (Pb-to-Pfr) thermal relaxation kinetics monitored at the Pfr peak absorption (Fig. 2E), the ratio of the far-red/blue components to the far-red/orange reversible components was calculated.As a result, 84.5% of the preparation showed thermal relaxation from the Pb state to the Pfr state (Fig. 1J).These two calculated values (83.9% and 84.5%) are almost identical to each other.This means that the artificially introduced Cys residue efficiently ligated to the chromophore.Furthermore, this artificially introduced Cys residue could reversibly ligate to the chromophore.
In the case of DXCF-type CBCRs, the structures of both Cys-non-adducted and Cys-adducted state have been revealed [28,29].In the Cys-non-adducted states structure, the Cys side chain is not oriented to C10 of the chromophore, indicating that the distance and orientation of the Cys side chain to the chromophore are crucial for covalent bond formation.In this context, the introduced Cys residue in the AnPixJg2_BV4_TC variant would show structural change upon photoconversion, in which the Cys side chain of not the darkadapted state but the photoproduct would be oriented to C10 of the chromophore.We have determined the Pfr structure of the AnPixJg2_BV4 parent molecule.In this structure, the side chain of Thr292, corresponding to T292C Cys residue, is located under ring A and not directed to C10 of the chromophore [18].Although the structure of the Po photoproduct state of the parent molecule remains unknown, the structures of the PCBbinding red/green reversible CBCR GAF domains (homologous to AnPixJg2_BV4) are determined in both states [30,31].Both proteins show a structural change in the side chains of the corresponding Thr residues, but the side chain is far from and not oriented to C10 of the chromophore in the Pg photoproduct state.Because the PCB-binding AnPixJg2_BV4_TC did not show reversible Cys ligation to the chromophore, the structure of the PCB-binding one in the photoproduct state should be vastly different from that of the BV-binding one.The introduced Cys residue in the BV-binding one might show a significant structural change to approach C10 of the chromophore in the Pb photoproduct state.By contrast, the Pb photoproduct state of AnPixJg2_BV4_TC has an absorption peak at 473 nm (Fig. 1F), which is 30-70 nm redshifted compared with those of the typical Pb states, forming a covalent bond between the second Cys and C10 of the chromophore [4,5,[9][10][11][12]14,19,20,[23][24][25][26].Although the BV chromophore has an additional double bond in ring D compared with PCB and PVB, this difference would not fully explain such a large redshift.Taken together, we alternatively propose that covalent bond formation occurs not at C10 but at C4 or other carbons near rings A and B, because C4 is known to be susceptible to nucleophilic attack [2].This may explain the redshifted feature of the Pb photoproduct of AnPixJg2_BV4_TC, albeit speculative.In the near future, we would like to perform detailed biophysical studies to clarify the covalent bond position of this protein.It is especially crucial to determine the three-dimensional structure of the Pb photoproduct of AnPixJg2_BV4_TC.
Notably, the T292C Cys residue could not form a covalent bond with the PCB chromophore, which is in contrast with the case for BV (Fig. 4A,C).In addition, Cys introduction at Thr292 resulted in a decrease in PCB-binding efficiency but not in BV-binding efficiency (Figs 3C and 4A).Our structural studies on the dark-adapted states of the PCB-binding AnPixJg2 and the BV-binding AnPixJg2_BV4 have revealed that the canonical Cys residue forms a covalent bond with C3 1 for PCB and C3 2 for BV [18,29].This difference would lead to differential chromophore positioning within the chromophore-binding pocket.The shallower insertion of the PCB into the chromophore-binding pocket would cause steric hindrance with the side chain of the  introduced Cys residue and result in the unsuitable positioning of C10 (or other carbons) for covalent bond formation.
Although we have introduced the Cys residue at the position corresponding to the DXCF Cys, the variant molecule AnPixJg2_BV4_YC showed lower binding affinity to both PCB and BV than the parent molecule (Fig. 5A,B).Further, we could not detect any reversible Cys-adduct formation for both PCB-and BVbinding proteins.Because Tyr293 is positioned just after Thr292 (Fig. 7E), a slight difference in Cys positioning would be critical for covalent bond formation with C10 (or other carbons) of the chromophore.Taken together, the exquisite positioning of both the chromophore and the Cys residue is essential for reversible Cys-adduct formation.
In this study, we could give the additional colortuning mechanism to the CBCR GAF domain by introducing a single Cys residue for ligation to the chromophore in a reversible manner.In nature, the second Cys has been acquired in several lineages of diverse CBCR GAF domains [4,10,12,14,20,21,23,24,26].Notably, the DXCF Cys residue has been acquired and lost several times during evolution [25].In addition, insert-Cys-type, RB1-type, and RB2-type second Cys residues derived from distinctive positions have been identified to ligate to C10 of the chromophore [10,12,14] (Fig. 7E).In this context, this study has identified the fifth position of the second Cys for reversible ligation.Because the artificial introduction of a single Cys residue enables drastic color tuning, from the orange-absorbing Po state to the blueabsorbing Pb state, such a mutation tends to be fixed during the evolutionary process in multiple lineages.
The BV-binding AM1_C0023g2_SG_TC showed reversible photoconversion between the far-redabsorbing Pfr dark-adapted state and the blueabsorbing Pb photoproduct state, which is similar to that for AnPixJg2_BV4_TC (Fig. 6).However, the Pb photoproduct state of AM1_C0023g2_SG_TC has the absorption peak at 392 nm, which is ~80 nm blueshifted in comparison with that of AnPixJg2_BV4_TC.Although the same Cys residue would ligate to the chromophore, the absorption peaks are significantly distinct from each other.In the case of AnPixJg2_BV4_TC, we speculate that the introduced Cys residue may ligate to C4 or other carbons of rings A and B instead of C10.In this context, in the case of AM1_C0023g2_SG_TC, the introduced Cys would ligate to C10, similar to the canonical dual Cys-type CBCR GAF domains.
We have previously obtained a BV-binding CBCR variant showing far-red/blue reversible photoconversion based on the AM1_1186g2 scaffold [32].The AM1_1186g2 wild-type molecule covalently incorporates PCB and shows reversible photoconversion between the red-absorbing Cys-non-adducted darkadapted state and the blue-absorbing Cys-adducted photoproduct [12].The second Cys residue is located at a unique position distinct from the DXCF Cys.The variant molecule containing four replacements (AM1_1186g2_KCAP_QV) efficiently incorporates BV and shows reversible photoconversion between the Cysnon-adducted far-red-absorbing dark-adapted state (Pfr) and the Cys-adducted blue-absorbing photoproduct state (Pb) [32].This photoconversion is similar to that of the variant molecule AnPixJg2_BV4_TC, which was developed in this study.However, AM1_1186g2_K-CAP_QV has the drawback that the photoconversion speed from the Pfr state to the Pb state is very slow, taking ~1 h for full photoconversion.By contrast, AnPixJg2_BV4_TC shows faster Pfr-to-Pb photoconversion, taking ~15 min for full photoconversion (Fig. 2A).Furthermore, AnPixJg2_BV4_TC showed rapid Pb-to-Pfr thermal relaxation, with a half-life of 1662 s for the slower component at 25 °C (Fig. 2B).These characteristics of AnPixJg2_BV4_TC would be advantageous for optogenetic tool development compared with AM1_1186g2_KCAP_QV for at least two reasons.First, fast Pfr-to-Pb photoconversion enables quick regulation of the target molecule by far-red light illumination.Second, fast Pb-to-Pfr thermal relaxation enables regulation only by turning the far-red light on or off.As the final goal for optogenetics tool development based on the CBCR molecules, it is necessary to transduce the conformational change within the CBCR molecules upon photoconversion to the target output molecule.In this context, one of the solutions may be a circular permutation of the CBCR molecules in which the N-and C-terminal ends are near the chromophorebinding pocket.By fusing the circular permutated CBCR molecule with the split target molecule, photoconversion occurring within the CBCR molecule may affect the maturation of the split molecule.
In this study, we did not discover CBCR molecules showing Cys-adduct formation using the T292C Cys residue among the natural repertoires.CBCR molecules possessing this Cys residue showed red/green and orange/green reversible photoconversion without Cysadduct formation (Fig. 7).However, we further detected other CBCR molecules possessing this Cys residue, and there would be novel CBCR molecules in nature.Future studies may detect natural CBCR molecules showing Cys-adduct formation using this Cys residue.

Computational saturation mutagenesis using Rosetta
Computational mutation analysis was performed using the crystal structure of AnPixJg2_BV4 in complex with BV (PDB ID: 5ZOH).Solvent molecules and ions other than two water molecules interacting with BV were removed.The PYMOL software (PyMOL Molecular Graphics System, Version 2.4.0 Schr€ odinger, LLC, New York, NY, USA) was used to add hydrogen atoms to BV and create a structure data file (sdf)-formatted file of BV.The sdf file was converted to a params file using the molfile_to_params.pyscript in the Rosetta software suite [27].The information on the covalent bond between the sulfur atom of Cys321 of AnPixJg2_BV4 and the carbon atom of BV was additionally described in the params file as well as in the FoldTree file for the Rosetta calculation [33].
The complex structure of AnPixJg2_BV4 and BV thus prepared was used as the initial structure the mutational analysis of AnPixJg2_BV4 by the Rosetta protein design software suite (version 3.13) [27].All calculations were performed with RosettaScripts [34] using the REF2015 score function.First, optimization of the initial structure was performed with the PackRotamersMover (prepack) and FastRelax Mover (relax) modules [35,36].Prepack was used to select optimal side chains from rotamer libraries with the main chain fixed.Subsequently, relax was performed to fine-tune the complex structure and minimize the total energy of the system.To optimize the complex structure by sampling various conformations, the following six types of relax calculations were performed, with and without prepack calculations, and with and without constraints on the main chain and BV: Optimization 1: without prepack, without BV and mainchain constraints.Optimization 2: without prepack, without BV constraints, with main-chain constraints.Optimization 3: without prepack, with BV and mainchain constraints.Optimization 4: with prepack, without BV and mainchain constraints.Optimization 5: with prepack, without BV constraints, with main-chain constraints.Optimization 6: with prepack, with BV and main-chain constraints.
Each optimization was performed 5000 times independently.For each run, the total_score, which indicates the stability of the overall complex structure, was obtained using ScoreMover, and the interface_delta, which indicates the binding energy between the protein and BV, was obtained using InterfacecScoreCalculator (Fig. S1) [37].Both scores are expressed in Rosetta energy units (REUs), and smaller scores indicate higher stability and tighter binding.In each of the six optimization methods, structures with both the lowest total_score and with the lowest inter-face_delta were selected, and a total of 12 structures were obtained as optimized structures for the AnPixJg2_BV4-BV complex.These structures were named, for example, Opt11 and Opt12 for those selected by total_score and interface_delta in Optimization 1, respectively.
Next, using these 12 optimized structures as starting structures, we computationally performed comprehensive saturation mutagenesis, in which each of the following 28 residues of AnPixJg2_BV4 close to BV was replaced, one at a time, with one of 19 different amino acids other than the wild-type residue [a total of 532 mutants (= 28 sites 9 19 types)]: residues 256, 258, 260, 268, 289, 290, 291, 292, 293, 294, 301, 302, 308, 310, 316, 317, 318, 319, 320, 322, 325, 326, 331, 334, 336, 348, 350, and 352.The amino acid substitution was introduced using the PackRo-tamersMover (fixbb) module, and subsequent energy minimization of the complex structure with main-chain constraints was performed using the relax module with and without constraints on BV in the same way as for optimization of initial structures.These calculations were independently performed five times for each mutant.The average of five total_score or interface_delta values for each mutant is shown in Fig. S2.The sum of the average total_score or average interface_delta of all 12 calculations was calculated for each mutant, and the difference in that value between the mutant and wild type is shown in Fig. S3.

Site-saturation random mutagenesis and screening on the LB plate
Phe319 and Thr292 of AnPixJg2_BV4 were independently mutated by site-saturation random mutagenesis using the primers listed in Table S1 (No. 1 and No. 2 for Phe319, No. 3 and No. 6 for Thr292).All PCR and vector construction analyses were performed using KOD One (TOYOBO, Osaka, Japan) and the In-Fusion HD Cloning Kit (Clontech, Mountain View, CA, USA).The mutant plasmids were introduced into C41 (DE3) pKT270 [38].To separate the growth phase from the protein induction phase on the LB plate, we set the nitrocellulose membrane on the LB plate (Merck).After colonies had formed on the membrane, the membrane was transferred to the LB plate containing 100 lM Isopropyl-b-D-thiogalactopyranoside (IPTG) and incubated at room temperature for 2-5 days.

Site-directed mutagenesis
Site-directed mutagenesis was performed by the PCR-based method using primers and KOD One.Thr292 and Tyr293 of AnPixJg2_BV4 were replaced with Cys using the primer sets (Table S1

Protein expression and purification
The wild-type and variant molecules were expressed in PCBand BV-producing E. coli.The cells were grown at 37 °C in 2.5 mL LB medium overnight and diluted 100-fold with 250-mL LB medium until the optical density at 600 nm was 0.4-0.6.Protein expression was induced by the addition of IPTG to a final concentration of 0.1 mM, and the cells were cultured at 18 °C overnight.The cells were resuspended in cell lysis buffer [20 mM HEPES-NaOH (pH 7.5), 100 mM NaCl, 10% w/v glycerol] containing 500 lM Tris(2-carboxyethyl) phosphine and disrupted by three passages through the Emulsiflex C5 high-pressure homogenizer at 83 MPa (Avestin, Ottawa, Canada).After the cell extracts were centrifuged at 10 000 g and 15 000 g for 30 min, 30 mM imidazole (final concentration) was added to the supernatant, which was filtered with a 0.8-lm cellulose ether membrane.The Histagged proteins were purified by using a HisTrap HP column (1 mL) (Cytiva, Tokyo, Japan) and € AKTA Pure system (Cytiva).The column was washed with the lysis buffer containing 100 mM imidazole, and then the His-tagged proteins were eluted with a linear gradient of lysis buffer containing 100-400 mM imidazole.After the purified protein was incubated with 1 mM EDTA for 1 h, the purified protein was dialyzed with the lysis buffer containing 1 mM dithiothreitol.

SDS/PAGE and zinc-induced fluorescence assay
The purified proteins dissolved in the electrophoresis buffer [60 mM DTT, 2% (w/v) SDS, 60 mM Tris-HCl (pH 8.0)] with the purple loading dye (New England Biolabs, Ipswich, MA, USA) were heat denatured (95 °C, 3 min).The samples were subjected to SDS/PAGE with 12% (w/v) acrylamide gel.The gels were stained with Coomassie brilliant blue R-250.For the Zn-induced fluorescence assay, electrophoresed gels were soaked in 20 mM zinc acetate at room temperature for 30 min.Fluorescence was visualized through a 600-nm longpath filter upon excitation with blue light (k max = 470 nm) and green light (k max = 527 nm) through a 562-nm short-path filter and imaged using the WSE-1096100 LuminoGraph (ATTO, Tokyo, Japan) and WSE-5500 VariRays (ATTO).

Spectral analysis
Ultraviolet and visible absorption spectra of the proteins were measured using the UV-2600 spectrophotometer (SHIMADZU, Kyoto, Japan) at room temperature.To measure the thermal relaxation kinetics of the BV-binding AnPixJg2_BV4_TC, the absorbances at 702 nm, 620 nm, and 473 nm were monitored at 25 °C.In addition, to measure the photoconversion kinetics of the BV-binding AnPixJg2_BV4_TC, the absorbance at 702 nm was monitored while irradiated with or without blue light [473 nm, 7.933 lmolÁ(m 2 s) À1 ] at 25 °C.The Opto-Spectrum Generator (Hamamatsu Photonics, Inc., Shizuoka, Japan) was used to generate monochromic light of various wavelengths for photoconversion.For the denaturation assay, both the darkadapted state (15Z-isomer) and photoproduct state (15Eisomer) of the native proteins were 5-fold diluted into 7 M guanidinium chloride (GdmCl)/1% (vol/vol) HCl and their absorption spectra were measured at room temperature before and after 3 min of irradiation with white light.

Fig. 1 .
Fig. 1.Serendipitous discovery and spectral analysis of BV-binding AnPixJg2_BV4_TC.(A) 3D structures of the BV-binding AnPixJg2_BV4 (left) and the PCB-binding AnPixJg2 (right).(B) E. coli colony colorization of the AnPixJg2_BV4-T292X mutant on LB plate.(C) An enlarged view of (B).E. coli colony expressing AnPixJg2_BV4_TC is highlighted by the red circle.(D) E. coli cell pellet expressing the AnPixJg2_BV4 parent molecule (left) and the AnPixJg2_BV4_TC variant molecule (right).(E) Color change of solutions of purified BV-binding AnPixJg2_BV4_TC variant molecule (Upper) and AnPixJg2_BV4 parent molecule (Lower).(F) Photoconversion of AnPixJg2_BV4_TC.The black line is the absorption spectrum derived from preparation, including only the Pfr dark-adapted state.The blue line is the absorption spectrum derived from preparation, including the far-red/blue reversible Pb photoproduct state and the far-red/orange reversible Po photoproduct state.The dark red line is the absorption spectrum derived from preparation, including the far-red/blue reversible Pb photoproduct state and the far-red/orange reversible Pfr dark-adapted state.(G) Absorption spectra of the far-red-absorbing Pfr dark-adapted state (dark red line) and the orange-absorbing Po photoproduct state (orange line) of AnPixJg2_BV4.(H) Normalized Pfr-minus-Pb (black solid line) and Pfr-minus-Po (gray solid line) difference spectra of the AnPixJg2_BV4_TC variant molecule in comparison with the Pfr-minus-Po difference spectra (gray dashed line) of the AnPixJg2_BV4 parent molecule.(I) Normalized acid-denatured Z-minus-E difference spectra of AnPixJg2_BV4_TC (black line) and AnPixJg2_BV4 (gray dashed line).(J) Possible structural change of the chromophore during photoconversion.

Fig. 2 .
Fig. 2. Measurement of the thermal relaxation kinetics of the BV-binding AnPixJg2_BV4_TC.(A) Normalized absorption changes of the BVbinding AnPixJg2_BV4_TC at 25 °C during photoconversion and thermal relaxation.Absorbances at 702 nm (dark red line), 620 nm (orange line), and 473 nm (blue line) were monitored.Far-red light (FRL) was illuminated during the time period highlighted by the pink background.(B-D) Experimental (black line) and fitting curves (red dotted line) of the thermal relaxation kinetics.The experimental curves were fitted with KALEIDAGRAPH 4.1 (Synergy Software, Reading, PA, USA) to the following equation with two exponentially decaying components: A(t) = A(∞) + A fast exp(Àk fast t) + A slow exp(Àk slow t), where A fast and A slow are the absorbance changes of the fast and slow components, respectively, and k fast and k slow are the rate constants of the fast and slow components, respectively.The fitting parameters are shown in Table 1.(E) Rate constants of the fast and slow components derived from two-exponential curve fit.(F) Absorbance changes of the fast and slow components derived from two-exponential curve fit.

Fig. 4 .
Fig.4.AnPixJg2_BV4_TC purified from PCB-producing E. coli.(A) Absorption spectra of the AnPixJg2_BV4_TC derived from the PCBproducing E. coli.This sample heterologously included the BV-binding and PCB-binding components.Using various monochromic light sources, four kinds of combinations of the absorbing states were prepared; the BV-binding Pfr dark-adapted state and the PCB-binding Pr dark-adapted state (black), the BV-binding Pb photoproduct state and the PCB-binding Pr dark-adapted state (red), the BV-binding Pfr darkadapted state and the PCB-binding Pg photoproduct state (green), and the BV-binding Pb photoproduct state and the PCB-binding Pg photoproduct state (blue).(B) Absorption spectra of the AnPixJg2_BV4 parent molecule derived from the PCB-producing E. coli.Red and green lines indicate the red-absorbing Pr dark-adapted state and the green-absorbing Pg photoproduct state, respectively.(C) Normalized (BV Dark + PCB Dark )-minus-(BV Photo + PCB Photo ) difference spectrum of AnPixJg2_BV4_TC, in which both the Pfr/Pb and Pr/Pg reversible photoconversions were observed (Black solid line) in comparison with normalized Pr-minus-Pg difference spectrum of AnPixJg2_BV4 (Gray dashed line).(D) Normalized (BV Dark + PCB Dark )-minus-(BV Dark + PCB Photo ) difference spectrum of AnPixJg2_BV4_TC, in which only the Pr/Pg reversible photoconversion was observed (Black solid line) in comparison with normalized Pr-minus-Pg difference spectrum of AnPixJg2_BV4 (Gray dashed line).(E) Normalized (BV Dark + PCB Dark )-minus-(BV Photo + PCB Dark ) difference spectrum of AnPixJg2_BV4_TC, in which only the Pfr/Pb reversible photoconversion was observed (Black solid line) in comparison with normalized Pfr-minus-Pb difference spectrum of AnPixJg2_BV4_TC derived from the BV-producing E. coli (Gray dashed line).(F) Normalized acid-denatured Z-minus-E difference spectra of the AnPixJg2_BV4_TC variant molecule derived from the PCB-producing E. coli (Green and red solid lines) in comparison with those of the AnPixJg2_BV4 parent molecule derived from the BV-(Gray dashed line) and PCB-producing E. coli (Black dashed line).Green and red solid lines correspond to the acid-denatured Z-minus-E difference spectra of the PCB-and BV-binding components, respectively.

Fig. 5 .
Fig. 5. Spectral analysis of the BV-and PCB-binding AnPixJg2_B-V4_YC.(A) Absorption spectra of the far-red-absorbing Pfr darkadapted state (dark red line) and the orange-absorbing Po photoproduct state (orange line) of the BV-binding AnPixJg2_BV4_YC.(B) Absorption spectra of the red-absorbing Pr dark-adapted state (red line) and the green-absorbing Pg photoproduct state (green line) of the PCB-binding AnPixJg2_BV4_YC.(C) Normalized difference spectra of the BV-binding AnPixJg2_BV4_YC (black line) and the PCBbinding AnPixJg2_BV4_YC (gray line).

Fig. 6 .
Fig. 6.Spectral analysis of the BV-and PCB-binding AM1_C0023g2_SG_TC.(A) Absorption spectra of the far-red-absorbing Pfr dark-adapted state (dark red line) and the blue-absorbing Pb photoproduct state (blue line) of the BV-binding AM1_C0023g2_SG_TC.(B) Absorption spectra of the red-absorbing Pr dark-adapted state (red line) and the green-absorbing Pg photoproduct state (green line) of the PCB-binding AM1_C0023g2_SG_TC.(C) Normalized difference spectra of the BV-(black line) and PCB-binding (gray line) AM1_C0023g2_SG_TC.(D, E) Normalized acid-denatured Z-minus-E difference spectra of AM1_C0023g2_SG_TC (black line) and AM1_C0023g2_SG (gray dashed line) for the BV-(D) and PCB-binding ones (E).

Fig. 7 .
Fig. 7. Native CBCRs possessing the Cys residue corresponding to the T292C position.(A) Absorption spectra of the red-absorbing Pr state (red line) and the green-absorbing Pg state (green line) of the PCB-binding NEP87939g1.(B) Pr-minus-Pg difference spectra of the PCBbinding NEP87939g1.(C) Absorption spectra of the orange-absorbing Po state (orange line) and the green-absorbing Pg state (green line) of the PCB-binding HFS08092g1.(D) Po-minus-Pg difference spectra of the PCB-binding HFS08092g1.(E) Multiple sequence alignment of the CBCRs.

Table 1 .
Fitting parameters for measurement of the thermal relaxation kinetics.The errors are fitting errors.