Molecular characterization of DXCF cyanobacteriochromes from the cyanobacterium Acaryochloris marina identifies a blue-light power sensor

Cyanobacteriochromes (CBCRs) are linear tetrapyrrole-binding photoreceptors that sense a wide range of wavelengths from ultraviolet to far-red. The primary photoreaction in these reactions is a Z/E isomerization of the double bond between rings C and D. After this isomerization, various color-tuning events establish distinct spectral properties of the CBCRs. Among the various CBCRs, the DXCF CBCR lineage is widely distributed among cyanobacteria. Because the DXCF CBCRs from the cyanobacterium Acaryochloris marina vary widely in sequence, we focused on these CBCRs in this study. We identified seven DXCF CBCRs in A. marina and analyzed them after isolation from Escherichia coli that produces phycocyanobilin, a main chromophore for the CBCRs. We found that six of these CBCRs covalently bound a chromophore and exhibited variable properties, including blue/green, blue/teal, green/teal, and blue/orange reversible photoconversions. Notably, one CBCR, AM1_1870g4, displayed unidirectional photoconversion in response to blue-light illumination, with a rapid dark reversion that was temperature-dependent. Furthermore, the photoconversion took place without Z/E isomerization. This observation indicated that AM1_1870g4 likely functions as a blue-light power sensor, whereas typical CBCRs reversibly sense two light qualities. We also found that AM1_1870g4 possesses a GDCF motif in which the Asp residue is swapped with the next Gly residue within the DXCF motif. Site-directed mutagenesis revealed that this swap is essential for the light power–sensing function of AM1_1870g4. This is the first report of a blue-light power sensor from the CBCR superfamily and of photoperception without Z/E isomerization among the bilin-based photoreceptors.

Photosynthetic organisms utilize light not only as an energy source but also as a signal. Linear tetrapyrrole (bilin) chromophores are involved in both of these functions, which involve the photosynthetic antenna complex, phycobilisomes, and photoreceptor families of phytochromes and cyanobacteriochromes (CBCRs). 2 Phycobiliproteins covalently bind phycocyanobilin (PCB), phycoerythrobilin, phycoviolobilin (PVB), and phycourobilin to transfer the light energy to chlorophyll (1,2). They stably absorb specific light qualities, although phycoerythrocyanin shows unexpected photoconversion in vitro (3). Conversely, phytochromes and CBCRs covalently bind a linear tetrapyrrole chromophore and show reversible photoconversion that is triggered by Z/E isomerization of the double bond between the C and D rings (4,5).
The phytochromes and CBCRs commonly possess a cGMP phosphodiesterase/adenylate cyclase/FhlA (GAF) domain that plays a central role in chromophore incorporation. Only the GAF domain is needed for proper photoconversion of the CBCRs, whereas N-terminal Per/Arnt/Sim (PAS) and/or phytochrome-specific (PHY) domains are also needed for proper photoconversion of the phytochromes (5). The light qualities sensed by the phytochromes are mostly restricted to the redto-far-red region, although algal phytochromes show diverse spectral properties (6). Conversely, the light qualities sensed by CBCR GAF domains vary widely, ranging from ultraviolet to far-red. To date, biliverdin, PCB, and PVB are known to bind to the CBCR apoprotein as a chromophore via a conserved canonical Cys residue (7)(8)(9). The CBCRs are categorized into various subfamilies and their color-tuning mechanisms are correspondingly diverse. Among them, two-Cys CBCRs have been detected from several distinct subfamilies (10 -13). For most of these, the second Cys residue transiently ligates to C10 of the chromophore during the photoconversion cycle (10 -16). Ligation of the second Cys residue to C10 results in a large blueshift due to shortening of the conjugated system of the chromophore. Among the two-Cys CBCRs, DXCF CBCRs are widely distributed among cyanobacteria and are the most intensively analyzed so far (7,10,14,(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32).
We focused on the GAF domains of the DXCF lineage because of their sequence variation (Fig. 1). AM1_5704g1 and AM1_1499g1 lack the second Cys residue within the DXCF motif. To address the two-Cys photocycle in this study, we analyzed the other seven CBCR GAF domains (AM1_1378g, AM1_0048g1, AM1_0048g2, AM1_6305g1, AM1_0829g, AM1_ 1870g4, and AM1_5997g4) ( Fig. 2A, domains highlighted by dotted squares). Among them, AM1_1870g4 possesses an arranged GDCF motif, in which the Asp residue is swapped with the next residue. These seven CBCR GAF domains also retain residues that are important for chromophore incorporation and proper photoconversion, including the canonical Cys residue (Fig. 2B).
In previous studies, by co-expressing heme oxygenase and PcyA to produce PCB in E. coli, the known GAF domains of the DXCF lineage were revealed to bind PCB and catalyze isomerization from PCB to PVB in some cases (14,18,28,32). Based on these findings, we also expressed the seven GAF domains as His-tagged proteins in PCB-producing E. coli and purified them by nickel-affinity chromatography. All GAF domains except AM1_5997g4 are likely to covalently bind a linear tetrapyrrole chromophore, as judged by zinc-induced fluorescence detection (Fig. 3A). Furthermore, denatured spectra corresponded well to the covalently bonded PCB or PVB, as described below (see Figs. 5 and 6B). We could not detect any mutations critical for chromophore binding in AM1_5997g4 based on sequence comparison (Fig. 2B). In the case of AM1_1378g, a major band in the CBB-stained gel did not correspond to a zinc-induced fluorescent band (Fig. 3A). Instead, a minor band in the CBB-stained gel (Fig. 3A, triangle) corresponded well to the fluorescent band. Together with a very low specific absorbance ratio (SAR) value (Table 1), this minor band in the CBB-stained gel was concluded to be AM1_1378g (Fig. 3A). In total, intensities of CBB-stained bands did not correspond well to those of zinc-induced fluorescent bands (Fig.  3A). This is due to two parameters: binding chromophore species and binding efficiency. By additionally taking the SAR values into consideration (Table 1), we concluded that the binding efficiencies of AM1_0048g1 and AM1_6305g1 were better than seen with the other CBCR GAF domains. Next, we performed spectral analyses using native and acid-denatured proteins. We describe these results for each photocycle type.

DXCF cyanobacteriochromes from Acaryochloris marina Blue/green photocycle, AM1_1378g
To assess the chromophore structures binding to the CBCR GAF domains, we used an acid denaturation assay. Because chemical denaturation removes interactions with the protein, the absorbance spectra and photochemistry of denatured biliproteins can be used to identify their bilin chromophore species and to assign C15 double-bond configurations (3,40). The 15E chromophores of acid-denatured biliproteins that absorb a shorter wavelength region can photoisomerize to 15Z and absorb a longer wavelength region, but 15Z chromophores are inert. Covalently bonded 15Z-and 15E-PCB showed absorption maxima around 660 and 600 nm, respectively, while covalently bonded 15Z-and 15E-PVB showed absorption maxima around 600 and 530 nm, respectively. In the case of a mixture of PCB and PVB, two absorbance peaks around 660 and 600 nm were detected for the 15Z chromophores.
Based on the spectral properties of the native and acid-denatured proteins, AM1_1378g covalently bound PVB and showed

DXCF cyanobacteriochromes from Acaryochloris marina
reversible photoconversion between a Pb form with a 15Z-isomer peaking at 414 nm and a Pg form with a 15E-isomer peaking at 525 nm (Figs. 4A, 5A, and 6 and Table 1). Because PVB but not PCB was observed in the acid-denatured spectra (Figs. 5A and 6B), isomerization from PCB to PVB was quite efficient even in E. coli. AM1_1378 has an MA domain for signal output and is orthologous to SyPixJ1 and TePixJ, the CBCR GAF domains that also show blue/green reversible photoconversion ( Fig. 2A) (17,23).

Blue/teal photocycle, AM1_0048g1 and AM1_0048g2
Based on the spectral properties of the native and acid-denatured proteins, AM1_0048g1 covalently bound both PCB and PVB and showed reversible photoconversion between a Pb form with a 15Z-isomer peaking at 418 nm and a Pt form with a 15E-isomer peaking at 498 nm (Figs. 4B, 5B, and 6 and Table 1). Similarly, AM1_0048g2 covalently bound both PCB and PVB and showed reversible photoconversion between a Pb form with a 15Z-isomer peaking at 409 nm and a Pt form with a 15E-isomer peaking at 506 nm (Figs. 4C, 5C, and 6 and Table 1). A very low level of PCB corresponding to an absorbance of around 662 nm for the denatured spectrum of the 15Z-isomer was detected for AM1_0048g1, whereas AM1_0048g2 bound a larger amount of PCB corresponding to the same absorbance region (Fig. 5, B and C). Difference spectra of the denatured proteins also support this interpretation (Fig. 6B). The absorption intensity around 605 nm was higher than that of around 669 nm for AM1_0048g1 but lower for AM1_0048g2 (Fig. 6B). Because the conjugated system of PCB is extended to the ring A, PCB chromophorylation should result in a red shift of the native 15E-isomer whose chromophore is free from the second Cys residue. Notably, both Pt forms contained red-shifted absorption shoulders of around 570 nm (Fig. 4, B and C). Irradiation of the red-shifted components of AM1_0048g1 and AM1_0048g2 with orange and red light sources peaking at 590 and 650 nm, respectively, resulted in photoconversion to the Pb forms (Fig. 7, orange lines). The red-shifted shoulder absorption of AM1_0048g2 was more prominent than that of AM1_ 0048g1 (Fig. 4, B and C), which is consistent with the finding that the binding ratio of PCB to PVB in AM1_0048g2 was higher than that in AM1_0048g1 (Figs. 5, B and C, and 6B). All these results demonstrate that the red-shifted components correspond to those that bind PCB. In both cases, the absorption peaks of the Pb forms of the PCB-binding components (Fig. 7, orange lines) are almost same as those that bind PVB (Fig. 7, blue lines). A mixture of PCB and PVB was also observed for the

DXCF cyanobacteriochromes from Acaryochloris marina
other CBCRs, when reconstituted in the PCB-producing E. coli (14, 18 -20, 22, 28, 32). However, when the same CBCR GAF domains were expressed in the cyanobacterial cells, the binding chromophore consisted of only the PVB chromophore (7, 18 -20, 22). Thus, AM1_0048g1 and AM1_0048g2 are also expected to bind only PVB in the cyanobacterial cells. Unknown factors may be needed for proper isomerization activity from PCB to PVB.    Figure 4. Photoconversion of DXCF CBCRs from A. marina. A, absorption spectra of the blue-absorbing form (Pb, blue) and the green-absorbing form (Pg, orange) of AM1_1378g. B, absorption spectra of the blue-absorbing form (Pb, blue) and the teal-absorbing form (Pt, orange) of AM1_0048g1. C, absorption spectra of the blue-absorbing form (Pb, blue) and the teal-absorbing form (Pt, orange) of AM1_0048g2. D, absorption spectra of the green-absorbing form (Pg, blue) and the teal-absorbing form (Pt, orange) of AM1_6305g1. E, absorption spectra of the blue-absorbing form (Pb, blue) and the orangeabsorbing form (Po, orange) of AM1_0829g.

DXCF cyanobacteriochromes from Acaryochloris marina
Some blue/teal reversible CBCRs retain two Phe residues that are predicted to be located close to ring D, and these Phe residues have been shown to play a role in twisting ring D from the plane of the B-C rings to form blue-shifted Pt forms (27). As AM1_0048g1 also retains these two Phe residues, its Pt formation should be established by a similar mechanism; however, AM1_ 0048g2 does not possess these Phe residues, indicating involvement of other residues in Pt formation. Tlr1999 derived from Thermosynechococcus elongatus BP-1 also showed a similar blue/ teal reversible photoconversion without these Phe residues (22).

DXCF cyanobacteriochromes from Acaryochloris marina
AM1_0048 has two blue/teal reversible CBCR GAF domains but no known signal-output domains. At its C-terminal region, however, there is a region consisting of ϳ300 amino acid residues with no known functional domains ( Fig. 2A), indicating that this region may function in some sort of signaling role.

Green/teal photocycle, AM1_6305g1
Based on the spectral properties of the native and aciddenatured proteins, AM1_6305g1 covalently bound PVB and showed reversible photoconversion between the Pg form with a 15Z-isomer peaking at 557 nm and the Pt form with a 15Eisomer peaking at 491 nm (Figs. 4D, 5D, and Fig. 6 and Table 1). Because the Pt form of AM1_6305g1 is spectrally similar to those of blue/teal reversible CBCRs, the Pt form of AM1_ 6305g1 is likely to be free from the second Cys residue as well as the other Pt forms (22,25,28). The chromophore of the Pg form is apparently free from the second Cys residue because of its red-shifted absorbance. Thus, the second Cys residue should not be covalently bound to the chromophore in either form. To verify this speculation, we added iodoacetamide (IAM) to both the Pg and Pt forms of AM1_6305g1 at a final concentration of 50 mM (Fig. 8), which is an excessive amount relative to the protein concentration and is sufficient to block Cys residues of the other CBCR GAF domains (13,22). Protein samples were incubated for 30 min after IAM addition. In both cases, IAM addition had little effect on their spectral properties. In the presence of IAM, both Pt and Pg forms showed teal/green reversible photoconversion. These results demonstrate that the second Cys residue is not covalently bound to the chromophore in either form. Conversely, the acid-denatured spectra clearly showed that only the PVB chromophore attached to AM1_ 6305g1 (Fig. 5D). In AM1_6305g1, the second Cys residue did not contribute to covalent bond formation but rather to efficient isomerization from PCB to PVB.
In previous studies, two CBCRs, NpR5113g1 and DpxA from Nostoc punctiforme and Fremyella diplosiphon, respectively, were reported to show green/teal reversible photoconversion (28,29). Interestingly, although NpR5113g1 and DpxA are definitely orthologous to each other, AM1_6305g1 is not a closely related ortholog of these two GAF domains (Fig. 1). NpR1597g1 and NpR5113g3, which show blue/teal reversible photoconversion, are phylogenetically closer to NpR5113g1 and DpxA than AM1_6305g1. Complicated evolutionary event(s) have likely occurred within this cluster. Although we tried to identify residues specifically conserved among these three green/teal reversible CBCRs in comparison with other blue/teal DXCF CBCRs (22,28,29,32), we could not identify such residues near the chromophore based on the structure. Thus, multiple residues may indirectly affect positioning of the second Cys to interfere with covalent bond formation in their Pg forms.

Blue/orange photocycle, AM1_0829g
Based on the spectral properties of the native and acid-denatured proteins, AM1_0829g covalently bound PCB and showed reversible photoconversion between the Pb form with a 15Zisomer peaking at 417 nm and the Po form with a 15E-isomer peaking at 579 nm (Figs. 4E, 5E, and 6 and Table 1). No PVB chromophore was detected from the acid-denatured spectra (Fig. 5E), indicating no isomerization activity from PCB to PVB. The Po form is red-shifted in comparison with the Pg forms of the typical DXCF CBCRs. This red-shifted spectrum is due to PCB attachment in which the second Cys residue is free from the chromophore and so the conjugated system is extended to ring A to absorb orange light.
Some CBCRs that show blue/orange reversible photoconversion have previously been identified. These CBCRs are categorized into two subfamilies (Fig. 1, blue/orange and CikA subfamilies). One subfamily, which includes NpF4973, is characterized by a whole-protein architecture being composed of only a GAF domain (26). Another subfamily contains CikA homologs, including SyCikA and NpF1000 (21,28). Interestingly, AM1_0829g showed low sequence similarity with both subfamily CBCRs. In this context, AM1_0829g independently lost isomerization activity from PCB to PVB to sense red-shifted orange light.

Blue/blue photocycle, AM1_1870g4
Purified AM1_1870g4 absorbs blue light with an absorbance maximum at 416 nm ( Fig. 9A and Table 1). We could not detect any photoconversion after blue-light illumination. Therefore, we next measured the absorption spectrum during irradiation with blue light, taking the possibility of rapid dark reversion into consideration. As a result, irradiation of AM1_1870g4 with blue light resulted in a slight absorption increase (Fig. 9A). These measurements showed that the photoproduct state of AM1_1870g4 also absorbs blue light but is slightly red-shifted relative to the dark-adapted form, with a maximum at 424 nm ( Table 1). The spectral overlap of the dark-adapted form and

DXCF cyanobacteriochromes from Acaryochloris marina
the photoproduct results in a dark-minus-light difference spectrum with a maximum at 374 nm and a minimum at 437 nm (Fig. 9B), shifted relative to the observed peak wavelengths. The rapid dark reversion observed with AM1_1870g4 raised the possibility that it might function as a sensor for light intensity, as reported for some red/green CBCRs and DXCIP CBCRs (33)(34)(35). To behave as a sensor for light intensity, dark reversion should be rapid enough that photoconversion of the photoproduct does not substantially contribute to photoequilibrium as well as the known light-power sensors. Therefore, we monitored changes in absorbance at 424 nm during photoconversion and dark-reversion cycles with different light intensities at different temperatures (Fig. 10A). At all temperatures tested (15,20,25, and 30°C), the light-dependent absorbance changes decreased with lower light intensities. Plotting the change in absorbance at 424 nm (⌬A 424 ) versus light intensity demonstrated that ⌬A 424 increased and was almost, but not completely, saturated at 15°C, probably due to quite fast dark reversion (Fig. 10B). These results suggest that suppression of dark reversion at lower temperatures resulted in a shift of the equilibrium in favor of the light-adapted form.
We therefore measured dark-reversion kinetics at different temperatures. At all temperatures tested (15,20,25, and 30°C), semi-logarithmic plots of the dark-reversion kinetics were linear, indicating that dark reversion is a first-order reaction (Fig.  10C). Slower dark reversion was observed at lower temperatures. These results are consistent with the results of the light intensity dependence described above. The half-lives at 15, 20, 25, and 30°C were 24.6, 12.0, 5.2, and 2.6 s, respectively. An Arrhenius plot of the corresponding rate constants was constructed (Fig. 10D), and AM1_1870g4 exhibited a linear Arrhenius relationship. From this plot, we calculated the activation energy to be 26 kcal/mol. This value is higher than those (14 -16 kcal/mol) of the light-intensity sensors previously reported for the other CBCR lineage (33,35). The higher activation energy

DXCF cyanobacteriochromes from Acaryochloris marina
means that AM1_1870g4 possesses higher thermosensitivity than the other light-intensity sensors. Because the optimal temperature for growth of A. marina is around 25°C, AM1_1870g4 is likely to integrate light intensity and temperature signals in natural environments. A dark-reversion property generally enables photoreceptors to sense light intensity signals in a temperature-dependent manner. In fact, other photoreceptors belonging to different families such as phytochrome, light, oxygen, or voltage (LOV) and blue-light using flavin (BLUF) proteins have also been revealed to show similar behaviors (41)(42)(43)(44)(45). In the case of the phytochrome proteins, Cph1 and Agp1 showed unexpected temperature-dependent His kinase activities, which affected conjugation in Agrobacterium. In this context, we should monitor output activity of the C-terminal His kinase activity under various light and temperature conditions to understand the detailed mechanisms integrating light intensity and temperature signals (46 -48).
We next measured the acid-denatured spectra of both the dark-adapted form and the photoproduct (Fig. 11). To get as much photoproduct as possible, the sample was irradiated with strong blue light (1320 mol m Ϫ2 s Ϫ1 ) on ice prior to denaturation. As a result, both the dark-adapted form and photoproduct showed similar spectra peaking at 580 nm with a significant shoulder around 520 nm. In both cases, white light illumination resulted in a red shift, and the final products corresponded well to covalently bound 15Z-PVB (Fig. 11, A and B). These results mean that both forms contained 15Z-and 15E-PVB (Fig. 11). The photoproduct contained a slightly larger amount of the 15E-isomer than the dark-adapted form (Fig. 11C). The photoproduct was produced by blue-light illumination with 1320 mol m Ϫ2 s Ϫ1 light intensity on ice. Under these conditions, photoconversion should be almost saturated, judging by the lightdependent absorbance change (Fig. 10, A and B). In this context, the slightly larger amount of the 15E-isomer in the photoproduct is not likely to be derived from photoconversion of the native protein. Alternatively, this situation may be derived from photochemical equilibrium irrespective of photoconversion of the native protein. This hypothesis was also supported by a site-directed mutagenesis study, as described below.
If this hypothesis is correct, this protein may be the first example among the bilin-based photoreceptors that shows photoconversion without Z/E isomerization. Because AM1_1870g4 is present within the tandemly arranged GAF domain cluster ( Fig.  2A), one may assume that AM1_1870g4 acts not as a photoreceptor but as an antenna to transfer light energy to the other GAF domains. To verify such a possibility, we determined the fluorescence quantum yield of AM1_1870g4, which was calculated to be only 0.6%. This result clearly rejects the antenna role of AM1_1870g4. To further prove this hypothesis, it will be necessary to perform other experiments such as observation of light-dependent output activity. However, this study may provide a clue for novel photoreceptors that bind phycoerythrobilin and phycourobilin, which contain no double bonds between the C and D rings.

Site-directed mutagenesis of AM1_1870g4
AM1_1870g4 possesses a GDCF motif in which the Asp residue is swapped with the next residue in comparison with the canonical DXCF motif. This unique arrangement of the Asp position may be responsible for the blue-light power-sensing function. To test this, we swapped this motif back to the canonical DGCF motif. The mutant protein, AM1_1870g4-DGCF, efficiently incorporated PVB as well as the wildtype protein and absorbed light in the blue region (Figs. 3B and 12A). AM1_1870g4-DGCF showed reversible photoconversion that is clearly distinct from that of the wildtype protein (Fig. 12). The  Figure 11. Acid-denatured spectra of AM1_1870g4. A, acid-denatured spectra of the dark-adapted form before (blue) and after (orange) light illumination. B, acid-denatured spectra of the photoproduct before (blue) and after (orange) light illumination. C, acid-denatured spectra of the dark-adapted form (blue) and the photoproduct (orange) before light illumination. D, Z-E difference spectrum.

DXCF cyanobacteriochromes from Acaryochloris marina
ground state with 15Z-PVB peaking at 420 nm (Figs. 12A, blue line, and 13A) partially converted to the photoproduct with 15E-PVB peaking at 425 nm (Figs. 12A, orange line, and 13B). The ground state was generated upon irradiation with light around 470 nm, whereas the photoproduct was generated upon irradiation with light around 410 nm. Highly overlapped absorption spectra of these two states resulted in partial photoconversion. Reversible photoconversion was repeated without noticeable deterioration. The difference spectrum had a positive peak at 405 nm and a negative peak at 457 nm (Fig.  12B). These results indicate that the power-sensing function without photoisomerization, as observed for the wildtype protein, largely depends on the arranged "GDCF" motif. Blue-light absorption of both forms suggests that the second Cys residue stably binds to the C10 of the chromophore. We introduced further mutations into AM1_1870g4-DGCF to try to obtain proteins showing typical blue/green reversible photoconversion based on sequence comparison, but we have failed to do so to date (data not shown).
We next performed site-directed mutagenesis focused on the canonical Cys (Cys-782) and second Cys (Cys-754) residues (Figs. 14 and 15). Singly (C754S and C782S) and doubly (C754S/ C782S) mutated proteins were constructed. All variant proteins bound a chromophore, although C782S and C754S/C782S variants exhibited reduced chromophore binding judging by the SAR values (Fig. 14 and Table 1). Furthermore, these two variants (C782S and C754S/C782S) showed no covalent bond formation as assessed by fluorescence detection (Fig. 3C). The canonical Cys residue, Cys-782, was essential for covalent bonding and efficient chromophore incorporation, whereas the C754S variant efficiently incorporated the chromophore comparable with the wildtype protein (Fig. 3C). None of these variants showed any detectable photoconversion or dark reversion (Fig. 14), indicating that these Cys residues are essential for the light power-sensing function of AM1_1870g4. Both C754S and C754S/C782S variants, which lack the second Cys residue, absorbed light in the red region (Fig. 14, A and C), probably due to no isomerization from PCB to PVB and no covalent bond formation between the second Cys residue and C10 of the chromophore. However, the C782S variant absorbed light in the blue region (Fig. 14B), probably due to covalent bond formation between the second Cys residue and C10 of the chromophore. Acid denaturation revealed that no variants bound any PVB chromophore species (Fig. 15). Not only the second Cys residue but also the canonical Cys residue are essential for isomerization activity from PCB to PVB. Denatured C782S had a spectrum peaking around 600 nm with a significant shoulder ΔAbsorbance Wavelength (nm) Figure 13. Acid-denatured spectra of AM1_1870g4-DGCF. A, acid-denatured spectra under the dark condition before (blue) and after (orange) light illumination. B, acid-denatured spectra of AM1_1870g4-DGCF under blue-light illumination before (blue) and after (orange) light illumination. C, difference spectra of acid-denatured AM1_1870g4-DGCF before and after white light illumination.

DXCF cyanobacteriochromes from Acaryochloris marina
around 700 nm, and white light illumination after denaturation resulted in a red shift peaking around 700 nm (Fig. 15B). These results suggest that the chromophores of the C782S variant are a mixture of the 15Z-isomer and the 15E-isomer, which is similar to the wildtype protein. However, white light illumination after denaturation did not affect the absorption spectra of the C754S and C754S/C782S variants (Fig. 15, A and C), indicating that their bound chromophores consisted of only 15Z-isomers. Together with the DGCF mutation analysis, covalent bond formation between the second Cys residue (Cys-754) and C10 and the arranged "GDCF" motif may result in accumulation of the 15Eisomer to some extent, irrespective of photoconversion. The denatured spectrum of the C754S variant peaking at 665 nm corresponds to that of covalently bound 15Z-PCB (Fig. 15A). Conversely, the 15Z-isomers of the C782S and C754S/C782S variants peaking at 676 and 689 nm, respectively, were red-shifted in comparison with the covalently bound 15Z-PCB (Fig. 15, B and C).
Because these variants lack the canonical Cys residue, these red-shifted components should correspond to non-covalently bound 15Z-PCB. In line with this assumption, the photochemical difference spectrum of the denatured C782S variant was red-shifted in comparison with that of the covalently bound PCB (Fig. 15D).

DXCF cyanobacteriochromes from Acaryochloris marina Unique properties of AM1_1870
AM1_1870 possesses a quite complicated domain architecture consisting of one phytochrome-type GAF domain and two CBCR GAF domains for light sensory input and two-component hybrid signaling systems for output ( Fig. 2A) (36). To characterize the phytochrome-type GAF domain of AM1_1870, we expressed AM1_1870g1-g2 covering the GAF-PHY region ( Fig.  2A) in the PCB-producing E. coli. Although the purified fraction included many contaminating proteins, probably due to low expression yield, zinc-induced fluorescence was detected from the protein band at ϳ60 kDa (Fig. 16, inset, arrowheads), which corresponds well to the theoretical molecular weight. The purified AM1_1870g1-g2 could bind PCB and showed red/far-red reversible photoconversion (Fig. 16). Together with a previous study (36), AM1_1870 is concluded to possess three light sensory systems: red/far-red-, blue/blue-, and red/green-sensing systems.
There have been no reports of CBCRs that sense blue-light power. To date, red and green light power sensors have been identified (33)(34)(35). Most cyanobacteria possess flavin-binding photoreceptors such as LOV and BLUF proteins. Because flavin absorbs in the UV-to-blue region, they function as blue-light sensors. Because some LOV and BLUF proteins derived from cyanobacteria show rapid dark reversion, they are also likely to function as a blue-light power sensor (49,50). Notably, BLUF proteins were not detected in A. marina genomic information. Although one LOV protein was detected in A. marina, this protein possesses an arranged SCHFL motif instead of the highly conserved NCRFL motif that is important for photoconversion. Although there is no experimental evidence, AM1_1870g4 may function as a blue-light power sensor to compensate for a lack of flavin-based power sensors.
Recently, plant photoreceptors of phototropin and phytochrome B have been reported to integrate light and temperature signals in vivo (42)(43)(44). Because several CBCRs, including AM1_1870g4, integrate light and temperature signals in vitro, these photoreceptors are also likely to physiologically perceive both light and temperature signals in vivo. Future physiological studies in this regard are required.

Concluding remarks
In this study, we identified DXCF CBCR GAF domains showing blue/green, blue/teal, green/teal, blue/orange, or blue/blue reversible photoconversion from A. marina. Together with previous studies of XRG CBCR GAF domains and a phytochrome protein from A. marina (9,13,36,37,51), A. marina can sense various light qualities covering the blue-to-far-red wavelength range. It is noteworthy that expression of only the GAF domain may not reflect the original spectral property of its full-length protein. In fact, phytochromes need the PAS and/or PHY domains in addition to the GAF domain for proper chromophore incorporation and photoconversion (52)(53)(54). Further studies such as spectral analyses using full-length proteins and physiological analyses using photoreceptor disruptants are needed to reveal detailed photoperception mechanisms.

Bacterial strains and growth media
The E. coli strain JM109 was used for plasmid construction, and the E. coli strain C41 harboring pKT271 that encodes heme oxygenase and PcyA to produce PCB was used for protein expression (55). Bacterial cells were grown in Luria-Bertani (LB) medium containing kanamycin with or without chloramphenicol at 20 g ml Ϫ1 . For protein expression, the cells were grown in LB medium at 37°C until the optical density at 600 nm was 0.4 -0.8, and then isopropyl ␤-D-1-thiogalactopyranoside was added at a final concentration of 0.1 mM. Subsequently, the cells were cultured at 18°C overnight followed by collection of the cell pellets by centrifugation.

Bioinformatics
Nucleotide and amino acid sequences of DXCF CBCR proteins from A. marina were obtained from CyanoBase (56). Motif analyses were performed by SMART searches run on the internet (57). Multiple sequence alignments were constructed using Clustal_X and manually edited based on the structural information of TePixJ (58). The phylogenetic tree was con-

DXCF cyanobacteriochromes from Acaryochloris marina
structed by the neighbor-joining method and visualized by using iTOL on the internet (59).

Plasmid construction
DXCF CBCR GAF domains were amplified from A. marina genomic DNA using PrimeSTAR Max DNA polymerase and the appropriate nucleotide primers ( Table 2). The amplified DNA fragments were cloned into the plasmid pET28a, and sitedirected mutagenesis was performed as described previously using the appropriate primers (Table 2) (35). All expression constructs were verified by nucleotide sequencing.

Protein expression and purification
The His-tagged proteins were expressed in E. coli C41 pKT271. Cells were disrupted in buffer A (20 mM HEPES-NaOH, pH 7.5, 0.1 M NaCl, and 10% (w/v) glycerol) containing 0.5 mM tris(2-carboxyethyl)phosphine by three passages through an Emulsiflex C5 high-pressure homogenizer at 83 MPa (Avestin). The mixture was centrifuged at 165,000 ϫ g for 30 min, and then the supernatants were loaded onto a nickel-affinity His-trap column (GE Healthcare) using the ÄKTAprime plus chromatography system (GE Healthcare) after filtration with a 0.2 m cellulose ether membrane. The column was washed with buffer A containing 100 mM imidazole, and then the His-tagged proteins were eluted with a linear gradient of buffer A containing 100 -400 mM imidazole. Following incubation with 1 mM EDTA for 1 h, the proteins were dialyzed against buffer A containing 1 mM dithiothreitol (DTT) without EDTA and imidazole.

Electrophoresis and zinc-induced fluorescence assay
Purified proteins in 2% (w/v) SDS, 60 mM DTT, and 60 mM Tris-HCl, pH 8.0, were separated using SDS-PAGE with a 12% (w/v) acrylamide gel, followed by staining with Coomassie Bril-liant Blue R-250 (CBB). For the zinc-induced fluorescence assay after SDS-PAGE, the gel was soaked in 20 mM zinc acetate at room temperature for 30 min. Fluorescence was then visualized through a 600-nm long-path filter upon excitation with blue ( max ϭ 470 nm) and green ( max ϭ 527 nm) light through a 562-nm short-path filter using WSE-6100 LuminoGraph (ATTO) and WSE-5500 VariRays (ATTO) machines.

Spectroscopy and dark-reversion kinetics
Ultraviolet and visible absorption spectra of the proteins were recorded with a UV-2600 spectrophotometer (Shimadzu) at 15, 20, 25, and 30°C using a temperature controller. Monochromatic light of various wavelengths for photoconversion was generated using an Opto-Spectrum Generator (Hamamatsu Photonics, Inc.). Acid denaturation used 8 M urea, pH 2.0, followed by recording an absorption spectrum, white light illumination for 3 min, and recording of a second absorption spectrum.
To monitor the photoconversion and dark-reversion processes, absorbance at 424 nm of the proteins against various intensities of blue light (430 nm: 1320, 1130, 1020, 890, 750, 620, 490, 360, 230, 90, 30, 15, and 5 mol m Ϫ2 s Ϫ1 ) was measured for 15 s with dark intervals of 90 s. The half-lives and the Arrhenius parameters were estimated from the dark-reversion kinetics at the different temperatures. The fluorescence quantum yield was measured with Quantaurus-QY (Hamamatsu Photonics, Inc.).  Fw