Protochromic absorption changes in the two-cysteine photocycle of a blue/orange cyanobacteriochrome

Abstract

Cyanobacteriochromes (CBCRs) are phytochrome-related photosensors with diverse spectral sensitivities spanning the entire visible spectrum. They covalently bind bilin chromophores via conserved cysteine residues and undergo 15Z/15E bilin photoisomerization upon light illumination. CBCR subfamilies absorbing violet-blue light use an additional cysteine residue to form a second bilin–thiol adduct in a two-Cys photocycle. However, the process of second thiol adduct formation is incompletely understood, especially the involvement of the bilin protonation state. Here, we focused on the Oscil6304_2705 protein from the cyanobacterium Oscillatoria acuminata PCC 6304, which photoconverts between a blue-absorbing 15Z state (^15ZPb) and orange-absorbing 15E state (^15EPo). pH titration analysis revealed that ^15ZPb was stable over a wide pH range, suggesting that bilin protonation is stabilized by a second thiol adduct. As revealed by resonance Raman spectroscopy, ^15EPo harbored protonated bilin at both acidic and neutral pH, but readily converted to a deprotonated green-absorbing 15Z state (^15ZPg) at alkaline pH. Site-directed mutagenesis revealed that the conserved Asp-71 and His-102 residues are required for second thiol adduct formation in ^15ZPb and bilin protonation in ^15EPo, respectively. An Oscil6304_2705 variant lacking the second cysteine residue, Cys-73, photoconverted between deprotonated ^15ZPg and protonated ^15EPr, similarly to the protochromic photocycle of the green/red CBCR subfamily. Time-resolved spectroscopy revealed ^15ZPg formation as an intermediate in the ^15EPr–to–^15ZPg conversion with a significant solvent-isotope effect, suggesting the sequential occurrence of 15EP–to–15Z photoisomerization, deprotonation, and second thiol adduct formation. Our findings uncover the details of protochromic absorption changes underlying the two-Cys photocycle of violet-blue–absorbing CBCR subfamilies.

Introduction

Phytochromes are photosensors that bind a linear tetrapyrrole (bilin) chromophore and are found in higher plant, algae, bacteria, and fungi (1–3). They undergo reversible photoconversion typically between the red-absorbing state (Pr) and the far-red light-absorbing state (Pfr),² which is triggered by the 15Z/15E photoisomerization of the bilin (Fig. S1) (4). In the dark, phytochromes convert to thermally stable Pr or Pfr via a process called dark reversion. The bilin is bound to the pocket of a GAF (cGMP phosphodiesterase/adenylyl cyclase/FhlA) domain via a conserved Cys residue that is found in the GAF domain for phycocyanobilin (PCB) and phytochromobilin (Fig. 1A, canonical Cys) or in the N terminus for biliverdin (5). The GAF domain interacts with the phytochrome domain via tongue structure and, in many cases, with the Per/Arnt/Sim (PAS) domain with the knot structure (6–8). These domains are required to complete the photocycle of phytochromes.

Figure 1.

Sequence and structural information of CBCR GAF domains. A, multiple alignment of the bilin-binding pocket of the GAF domain of CBCR subfamilies undergoing blue/orange (B/O), blue/green (B/G), green/blue (G/B), blue/blue (B/B), green/red (G/R), and red/green (R/G) photocycles. Sequences of the GAF domain of phytochromes (R/FR) were also included. Key residues interacting with the bilin were colored accordingly. Residue numbers of TePixJ and (cyan) corresponding Oscil6304_2705 (green) are indicated. B, PVB chromophore and its associated hydrogen bond network in the structures of TePixJ of ^15ZPb (upper) and ^15EPg (lower).

Cyanobacteriochromes (CBCRs) are unique phytochrome-related photosensors that are confined to but widely distributed among cyanobacteria (2, 9–14). Like phytochromes, CBCRs utilize the 15Z/15E bilin photoisomerization to trigger their photocycle, but they require only the GAF domain and show substantial variation in the absorbing wavelength spanning near-UV to far-red parts of the spectrum (15–33). Recent genome analyses show that some cyanobacteria harbor up to several dozen CBCR GAF domains (34–37), suggesting that they employ one of the most complex photosensing systems among all organisms. Although the physiological roles of most CBCRs have yet to be revealed, some were shown to regulate photosynthetic antenna composition (37–45), cell morphology (46), phototaxis (47–50), and cell aggregation (28, 51).

CBCRs are conventionally classified into subfamilies based on their peak absorption in the 15Z and 15E configurations of the bilin, and several chemical processes are known to regulate spectral tuning. CBCR subfamilies such as blue/green (B/G), green/blue (G/B), and blue/orange (B/O) utilize “two-Cys photocycles” that are responsible for their absorption of light in the near-UV to blue region (19, 22, 27, 52). These photocycles employ a second Cys residue to form an additional thiol adduct to the bilin C10 atom, which disconnects the conjugated system between the B- and C-rings and leads to the absorption blue shift (15, 53, 54). In these cases, the second Cys is a part of the highly-conserved DXCF motif, as revealed by alignment of the GAF domains of these CBCR subfamilies (Fig. 1A, second Cys). The B/G and G/B subfamilies incorporate PCB. In these CBCR subfamilies, PCB is autocatalytically isomerized it to phycoviolobilin (PVB), which disconnects the conjugated system between A- and B-rings and yields green absorption in the thiol-free state (19, 53). Crystal and NMR structures have been determined for TePixJ of the B/G subfamily in both blue-absorbing 15Z (^15ZPb) and green-absorbing 15E (^15EPg) photostates (Fig. 1B) (54–56), where all four pyrrole rings are fully protonated in both states.

The red/green (R/G) CBCR subfamily also has an advantage in that crystal and/or NMR structures of both red-absorbing 15Z state (^15ZPr) and green-absorbing 15E state (^15EPg) are available (55, 57) (unpublished structures 5DFX and 5M82 in the PDB). In this subfamily, structural changes occur in its A- and D-rings upon light illumination, whereas the coplanarity of B- and C-rings remains almost unchanged (57). Recent quantum mechanics/molecular mechanics (QM/MM) analysis suggested that distortion of the D-ring in the photoproduct state plays more crucial roles than the A-ring in the absorption blue shift of ^15EPg (58). Photocycle of the R/G subfamily is thus called the “trapped-twist photocycle” (24). NMR and resonance Raman (RR) spectroscopy revealed that the four pyrrole nitrogens are fully protonated in both ^15ZPr and ^15EPg (57, 59), indicating that a change of bilin protonation state is not involved in the spectral tuning of the R/G subfamily, although deprotonated intermediates had previously been suggested in their forward and reverse photocycles (60). Additionally, some members of this lineage acquired a second Cys residue in an insertion loop distinct from the common DXCF motif of the B/G, G/B, and B/O subfamilies, as well as blue-violet–sensing capacity (18, 25, 32); these are referred to as the insert-Cys subfamily.

Other CBCR subfamilies, such as the green/red (G/R) subfamily, are less understood at the structural level. The G/R subfamily has been extensively studied for its physiological roles in the regulation of photosynthetic antenna composition (39–41, 45). pH titration analysis demonstrated that members of this subfamily photoconvert between deprotonated ^15ZPg and protonated ^15EPr (41), with a clear change of the pK_a for the bilin aromatic π system following the 15Z/15E photoisomerization. Because the absorption of green or red light is strongly dependent on the bilin protonation state rather than the configuration of the 15,16-double bond, the associated photocycle is called the “protochromic photocycle” (41). Although no structural information for the G/R subfamily is currently available, comprehensive mutagenesis analysis demonstrated that a protochromic triad of three key residues was essential for the bilin pK_a shift, Lys, Leu, and Glu, which are called the protochromic triad residues (Fig. 1A) (41). Ultrafast transient spectroscopy of the G/R subfamily revealed that the ^15ZPg state was heterogeneous (61, 62), which these researchers interpreted as tautomerization between B- and C-ring deprotonated species. A recent study using RR spectroscopy and QM/MM analysis strongly implicated the presence of the B-ring deprotonated component in ^15ZPg of the G/R subfamily (63). The protochromic triad residues of the G/R subfamily are partially conserved in the B/B CBCR subfamily (Fig. 1A) (16, 19, 21), in which two-Cys linkage is retained in both stable photostates.

The aim of this study is to investigate the role of the bilin protonation state in the two-Cys photocycle. In some of the two-Cys CBCR subfamilies, the incorporated PCB is autocatalytically isomerized to PVB, whose conversion is typically less efficient in Escherichia coli cells than in cyanobacterial cells (65). The resultant PCB/PVB heterogeneity makes the interpretation of results more complicated in the component analyses of pH titration and time-resolved spectroscopy (41, 66). In addition, the change of protonation state of PVB causes a relatively small absorption shift compared with that of PCB (52), which makes it difficult using spectroscopy to distinguish the titration of the bilin from that of charged residues interacting with the bilin.

To overcome these difficulties, we focused on Oscil6304_2705 of the B/O CBCR. This protein has previously been shown to photoconvert between blue-absorbing ^15ZPb and orange-absorbing ^15EPo states in the two-Cys photocycle using PCB as the only chromophore (27). The Oscil6304_2705 protein consists of only a single GAF domain and may transduce light signals via protein–protein interaction, but its physiological roles or interacting protein(s) are not yet known. In addition to the stationary ^15ZPb and ^15EPo states of Oscil6304_2705, deprotonated green-absorbing ^15ZPg and/or ^15EPg states were observed at alkaline pH, in a variant protein lacking the second Cys residue, and also in the reverse reaction in time-resolved spectroscopy. From these results, we propose three sequential processes for the reverse photocycle of the two-Cys photocycle of the B/O subfamily: initial 15E–to–15Z photoisomerization, subsequent deprotonation, and slow formation of second thiol adduct. Therefore, the B/O subfamily combines protochromic and two-Cys photocycles in separate timescales. These results provide insight into the molecular mechanisms by which CBCRs absorb light in the violet-blue region using intrinsically red-absorbing bilin chromophores.

Results

Absorption and RR spectra of the blue/orange cyanobacteriochrome

We expressed and purified the full length of Oscil6304_2705 protein from E. coli cells harboring PCB-producing plasmid (67). The dependence of the protein expression on temperature was investigated based on maximum blue color density of E. coli cells after illumination of blue light (Fig. S2). Overnight induction at 25 °C showed the highest expression level for all conditions examined, and the cells grown under this condition were used throughout this study (Fig. S2). Purified Oscil6304_2705 holo-protein showed reversible photoconversion between the blue-absorbing state (Pb) peaking at 423 nm and the orange-absorbing state (Po) peaking at 604 nm (Fig. 2, A and C), in rough agreement with previous characterizations of this protein (27). Acid-denaturation spectra of Pb and Po showed absorption maxima at 662 and 586 nm, which correspond to 15Z and 15E configurations of covalently bound PCB, respectively (Fig. 2D) (27, 68). Therefore, these photostates were designated as ^15ZPb and ^15EPo, respectively. The ^15EPo gradually converted to ^15ZPb in the dark, indicating that the ^15ZPb is the thermally stable dark state (Fig. S3). The acid-denaturation spectra also confirmed that Oscil6304_2705 incorporated PCB without isomerization to PVB, as expected (27). The covalent binding of the PCB was confirmed by fluorescence from acrylamide gel after SDS-PAGE (Fig. S4) (69). We did not experimentally confirm the binding site of PCB in Oscil6304_2705 protein, but the sequence alignment suggests that PCB is covalently bound at Cys-101 as shown in structures of phytochromes and CBCRs that covalently incorporate PCB (Fig. 1) (7, 54–56).

Figure 2.

Spectral characterization of WT and single amino acid variants. Photographs of protein solution of WT undergoing blue/orange photocycle (A) and C73A variant undergoing green/orange photocycle (B) are shown. C–J, native absorption spectra of WT (C), C73A (E), D71A (G), and H102L (I) and acid-denaturation absorption spectra of WT (D), C73A (F), D71A (H), and H102L (J). For H102L, absorption after blue light illumination was also included as a green dot (I). Spectra corresponding to the 15Z state (red) or 15E state (blue) of the bilin were indicated.

To obtain insights into the structure and protonation state of the bilin, we utilized RR spectroscopy (70). The RR spectra of ^15EPo was measured with 785 nm excitation (Fig. 3, middle traces). The observed RR spectrum exhibits prominent bands at 1633, 1329, 820, and 658 cm⁻¹ (Fig. 3, middle black trace). The overall spectral pattern is similar to that for the ^15EPr state of RcaE of the G/R subfamily (Fig. 3, upper traces), whose PCB chromophore is protonated (63). We also measured the RR spectra of ^15EPo in D₂O, where exchangeable protons such as NHs and OHs were deuterated (Fig. 3, middle red trace). The H/D exchange changes the frequencies and intensities of the bands in the 1633 and 1500–1250 cm⁻¹ regions, and new bands appear at 1086 and 936 cm⁻¹. Similar D₂O-induced changes are also seen in the protonated ^15EPr state of RcaE (Fig. 3, upper traces), suggesting that the PCB is protonated in ^15EPo for Oscil6304_2705. The feature at 1586 cm⁻¹ in ^15EPo also disappears upon deuteration, which was not clearly observed in the RR of ^15EPr state of RcaE (Fig. 3, upper traces) (63). This band could correspond to the C=N stretching mode coupled with a N–H bending as observed in the RR spectra of protonated bilin in phytochromes (71, 72).

Figure 3.

RR spectroscopy of RcaE and Osil6304_2705. The observed Raman spectra of ^15EPr of RcaE (upper) in the previous study (63), and those of ^15EPo (middle) and ^15ZPb (lower) of Oscil6304_2705 in this study. Spectra were measured at room temperature in the buffer containing H₂O (black) or D₂O (red). Important Raman bands are assigned as follows: νC=C, the C=C stretching mode; νC=NH, the C=N stretching coupled with an in-plane NH bending vibration; νCC (ring), the CC stretching mode of a pyrrole ring; νCN, the CN stretching mode, δNH/δCH, the in-plane NH and CN bending mode; δND, the in-plane ND bending mode; γCH, the out-of-plane CH bending mode of a methine bridge; δCCC, the C–C–C bending deformation mode.

The RR spectrum of ^15ZPb was measured with 406-nm excitation under continuous illumination of red light (Fig. 3, lower traces), as the excitation caused ^15ZPb to ^15EPo conversion. The RR spectrum of ^15ZPb was obtained by subtraction of the RR spectrum of ^15EPo from the observed RR spectrum. The obtained RR spectrum of ^15ZPb differs significantly from that of ^15EPo. The C=C stretching band for ^15ZPb is broad (Fig. 3, lower black trace), and at least two peaks are seen near 1625 and 1641 cm⁻¹. The RR spectra in the 1250–1550 cm⁻¹ region also differs between ^15ZPb and ^15EPo. The normal modes in this region involve C–H and/or N–H bending motions of the bilin (73), and these are expected to be sensitive to the bilin structure. These results indicate a large structural difference between the two photostates. The effects of H/D exchange on the RR spectra were also examined for ^15ZPb (Fig. 3, lower red trace). The RR spectrum of ^15ZPb shows a small but detectable band at 1367 cm⁻¹, which exhibits a large downshift to 1042 cm⁻¹ upon deuteration. Thus, the 1367 cm⁻¹ band can be assigned to a normal mode involving the N–H bending coordinate of the bilin in the protonated ^15ZPb state.

Absorption spectroscopy of single amino acid variants

To elucidate the absorption tuning mechanism of the two-Cys photocycle in Oscil6304_2705, we performed site-directed mutagenesis using structures of ^15ZPb (PDB code 4GLQ) and ^15EPg (PDB code 2M7V) of TePixJ as a reference (Fig. 1B) (54, 56). Multiple sequence alignment shows that residues in the bilin-binding pocket are highly conserved between the B/O and B/G subfamilies (Fig. 1A). Cys-73 of Oscil6304_2705 corresponds to the Cys-494 of TePixJ (Fig. 1A) that is responsible for the second linkage formation and the absorption blue shift (15). Asp-71 and His-102 of Oscil6304_2705 correspond to Asp-492 and His-523 of TePixJ, respectively, which form a hydrogen bond network with pyrrole nitrogens of B- and C-rings (Fig. 1B). The two residues are also conserved in phytochromes and are implicated in tuning the bilin pK_a in both red- and far-red–absorbing states of phytochromes (55, 57, 74). These points imply that Asp-71 and His-102 could be responsible for bilin protonation in Oscil6304_2705. Among a total of 211 single GAF domain proteins orthologous to Oscil6304_2705, residues corresponding to Cys-73, Asp-71, and His-102 are conserved in 210, 193, and 210 proteins, respectively (supporting information). We introduced single amino acid substitutions for these three residues in Oscil6304_2705 and examined their effects on the absorption spectra.

Single amino acid variants of these three residues incorporated the bilin covalently (Fig. S4). The C73A variant photoconverted between a green-absorbing state (Pg) peaking at 557 nm and Po peaking at 584 nm (Fig. 2, B and E). Acid-denaturation spectra showed that the C73A variant photoconverts between ^15ZPg and ^15EPo states. Notably, acid-denatured spectrum of the ^15EPo in the C73A variant showed a red-shifted absorption peak at 595 nm with a peak shoulder at 661 nm (Fig. 2F, blue line), indicating that forward 15Z to 15E photoisomerization is not complete. Thus, the absence of Cys-73 resulted in formation of thiol-free ^15ZPg and reduced forward photoisomerization, but did not substantially affect formation of protonated ^15EPo. The D71A variant showed double absorption peaks of Pb at 414 nm and Pg at 565 nm after red illumination, whereas it showed a single peak of Po at 594 nm after blue illumination (Fig. 2G). Acid-denaturation spectra of 15Z and 15E states of D71A were similar to those of WT (Fig. 2, D and H), confirming the full 15Z/15E photoisomerization. Thus, the absence of Asp-71 resulted in a mixture of ^15ZPb and ^15ZPg due to the inhibition of second thiol adduct formation, but it did not affect the ^15EPo formation.

For His-102, we created H102A, H102Q, and H102L variants, but these variants showed photoinert Pb peaking at ∼420 nm (Fig. 2I and Fig. S5). As expression levels of H102A and H102Q variants were low (Fig. S4), we utilized the H102L variant for further analysis. Acid-denaturation spectra of H102L revealed that the Pb consists mainly of the 15E state (Fig. 2J). The mixture of 15Z and 15E states in the H102L variant suggests the photoequilibrium of forward and reverse reactions due to spectral overlap of ^15ZPb and ^15EPb, which reached the equilibrium during our purification procedure under illumination of white light. Thus, the loss of the imidazole group of His-102 caused efficient thiol adduct formation not only in the 15Z state but also in the 15E state of the bilin.

In summary, our site-directed mutagenesis experiments suggest that Asp-71 stabilizes thiol adduct formation between Cys-73 and C10 of the 15Z state of the bilin and that His-102 is important for cleavage of the thioether linkage in the 15E state.

pH-induced spectral changes and pK_a fitting

We next investigated the protonation state of the bilin in WT and C73A, D71A, and H102L variants by monitoring their pH-dependent changes in absorption (41, 74). Aliquots of the protein were converted to the 15Z or 15E state by light illumination, supplemented with an excess amount of buffer at each pH point, and immediately measured for their absorption spectra in the dark (Figs. 4 and 5). Changes of the absorption peaks were fitted with one or two titrating groups and the Henderson-Hasselbalch equation, as reported previously (41). The pK_a values of the absorption changes were estimated and summarized (Table 1). For pK_a calculation, we excluded the points of pH 5.0–6.0 that showed increased scattering due to partial protein aggregation. We favored fitting with one titrating group, because secondary pK_a values for minor transitions were not robust by changes in initial values of the fitting equation.

Figure 4.

pH titration of WT and single amino acid variants. A, pH-dependent absorbance spectra of the 15Z and 15E states of WT (A and B), C73A (C and D), D71A (E and F), and H102L (G) are shown. pH values are shown between 5.0 (red) and 11.0 (purple) in 0.5 pH unit increments.

Figure 5.

pK_a estimation of WT and single amino acid variants. A, pH-dependent absorption changes of 15Z and 15E states of WT (A and B), C73A (C and D), D71A (E and F), and H102L (G) are shown. The peak wavelengths and bilin configuration are shown as insets. The absorption changes were fitted with one or two titrating groups of the Henderson-Hasselbalch equation to estimate pK_a (solid line) (41) (B–F).

Table 1.

Summary of the pK_a estimation

Calculated pK_a values for absorption changes in the pH titration in Fig. 5. Protein variant, 15Z/15E configuration, and absorption peak wavelength utilized for the fitting are shown. Errors are reported as standard deviations of the mean for the fitted values. We utilized the pK_a value with smaller errors for the transition. For example, pK_a of ^15ZPg/^15ZPo conversion of C73A 15Z state can be calculated as 5.8 ± 0.8 from the decrease of ^15ZPg at 550 nm and 6.4 ± 0.2 from the increase of ^15ZPo peak at 600 nm. We described 6.4 ± 0.2 as reliable pK_a values for the ^15ZPg/^15ZPo conversion in the main text.

Protein variant and D-ring configuration	pKa	Absorption peaks for pKa calculation
WT 15E	9.0 ± 0.1	15EPo (600 nm)
	9.4 ± 0.4, 9.6 ± 0.4	15EPb (420 nm)
	10.0 ± 0. 1	15EPg (530 nm)
C73A 15Z	5.8 ± 0.8	15ZPg (550 nm)
	6.4 ± 0.2	15ZPo (600 nm)
C73A 15E (with minor 15Z)	6.7 ± 0.7, 9.9 ± 0.3	15EPg (530 nm)
	6.2 ± 0.7, 9.4 ± 0.1	15EPo (600 nm)
D71A 15Z	6.0 ± 0.2	15ZPo (610 nm)
	9.7 ± 0.1	15ZPb (420 nm)
	9.7 ± 0.2	15ZPg (550 nm)
D71A 15E	10.6 ± 0.9	15EPg (550 nm)
	8.8 ± 0.1	15EPo (610 nm)

The spectrum of WT ^15ZPb was maintained over a pH range from 5 to 11 with only small absorption changes (Figs. 4A and 5A). Similar results were obtained for the H102L variant, which contains mainly ^15EPb (Figs. 4G and 5G). The robustness of the ^15ZPb and ^15EPb peaks suggests the generation of rubinoid species upon a second thiol adduct formation (Fig. 5, A and G) (15). Rubinoid has fully-protonated and neutral π systems with pK_a values outside the pH range of water (75). The small absorption changes in the ^15ZPb of WT and ^15EPb of H102L seem to represent equilibrium constants for the dissociation of second thioether linkage but not pK_a of the pyrrole nitrogen (Fig. 5, A and G). Therefore, we did not apply the pK_a calculation for these absorption changes.

In contrast to ^15ZPb, ^15EPo of the WT was stable at low and neutral pH, but readily converted to ^15EPg at high pH (Figs. 4B and 5B). These results suggest that ^15EPo converted to ^15EPg by bilin deprotonation, which probably occurred at the B- or C-ring pyrrole nitrogen atoms (61, 63). Notably, a decrease of ^15EPo and increase of ^15EPg showed distinct pK_a values, pK_a 9.0 ± 0.1 and pK_a 10.0 ± 0.1, respectively (Table 1). In addition, increase and decrease of ^15EPb peaks were observed with pK_a 9.4 ± 0.4 and pK_a 9.6 ± 0.6, respectively (Figs. 4B, blue arrow, and 5B and Table 1). These results suggest protonation equilibrium between ^15EPo, ^15EPb, and ^15EPg with two close pK_a values, which is caused by titration of the bilin pyrrole nitrogen and a residue with a high pK_a value.

We examined the photosensitivity of WT protein at extreme pH conditions (Fig. 6). At pH 5.0, WT protein showed efficient ^15ZPb to ^15EPo forward reaction and ^15EPo to ^15ZPb reverse reaction (Fig. 6A). The slight increase of peak at 615 nm at pH 5.0 can be assigned as formation of a native thiol-free 15Z state, because it differs from the denatured spectrum of the 15Z state peaked at ∼660 nm at pH 5.0 (41). At pH 11.0, WT protein showed efficient ^15ZPb to ^15EPg forward reaction but reduced ^15EPg to ^15ZPb reverse reaction (Fig. 6B). The slight increase of peak at ∼550 nm at pH 11.0 can be assigned as formation of a native thiol-free 15E state, because it differs from the denatured spectrum of 15E state peaked at ∼600 nm at high pH (41). Thus, these results suggest that Oscil6304_2705 protein was not denatured at extreme pH conditions within our experimental timescale, albeit with slight dissociation of the second thiol adduct.

Figure 6.

Photocycle of WT protein at extreme pH conditions. WT protein was photoconverted to the ^15ZPb state with red light illumination at pH 7.5. Then, an excess amount of buffer of pH 5.0 (A) or pH 11.0 (B) was added. Absorption spectra were measured after pH change or each illumination of LED light shown as insets.

The absorption of the C73A variant was sensitive to the pH in both 15Z and 15E states: ^15ZPo and ^15EPo were formed at low pH values, whereas ^15ZPg and ^15EPg were formed at high pH values (Figs. 4, C and D, and 5, C and D). The absorption change between ^15ZPg and ^15ZPo showed pK_a 6.4 ± 0.2, whereas change between ^15EPg and ^15EPo showed pK_a 9.4 ± 0.1 (Fig. 5, C and D, and Table 1), indicating that the proton affinity of the bilin was ∼1000-fold lower in the 15Z state than in the 15E state. Interestingly, this variant was proved to photoconvert between deprotonated ^15ZPg and protonated ^15EPo at neutral pH values (Fig. 2, B and E), which is similar to the protochromic photocycle of the G/R subfamily (41). C73A variant also showed a secondary pK_a 6.2 ± 0.7 in the 15E state (Fig. 5D and Table 1), which is derived from titration of the residual 15Z states due to the incomplete photoisomerization capacity of this variant.

The D71A variant showed ^15ZPo and ^15EPo formation at low pH values, whereas it showed ^15ZPg and ^15EPg formation at high pH values (Figs. 4, E and F, and 5, E and F). The absorption changes of ^15ZPo and ^15EPo showed pK_a 6.0 ± 0.2 and 8.8 ± 0.1, respectively (Table 1). Thus, the D71A variant also showed a significant pK_a shift upon 15Z/15E photoisomerization, suggesting that Asp-71 is not necessary for the drastic pK_a shift of the bilin. The 15Z state of the D71A variant showed additional pK_a 9.7 ± 0.1 that reflects the conversion between ^15ZPb and ^15ZPg (Figs. 4E and 5E and Table 1). This pH-dependent formation of ^15ZPb is similar but more drastic than the titration of the 15E state of the WT (Figs. 4B, blue arrow, and 5B).

Time-resolved spectroscopy of the reverse photocycle

We showed that the bilin can be deprotonated by alkalization of buffer pH or by mutations in residues of the bilin-binding pocket in Oscil6304_2705 (Figs. 2 and 4). However, RR spectroscopy and pH titration analysis suggested that stationary states of ^15ZPb and ^15EPo are both protonated (Figs. 3 and 4). Therefore, we investigated the existence of the deprotonated species as intermediates in the photocycle of Oscil6304_2705 using time-resolved spectroscopy.

We focused on the reverse reaction from ^15EPo to ^15ZPb. After excitation of ^15EPo with a flash of green light, absorption changes at 300–750 nm were monitored at 10-nm intervals over a timescale of microseconds to seconds (Fig. 7A). Datasets were analyzed with a sequential irreversible model (for details, see under “Experimental procedures”). Time-resolved absorption changes from ^15EPo to ^15ZPb were fitted with three distinct time constants: 0.33 ms (τ₁), 4.7 ms (τ₂), and 2.0s (τ₃) (Fig. 7A). Three kinetically-defined intermediates with time constants of τ₁, τ₂, and τ₃ were designated as P₆₁₀, P₅₃₀-I, and P₅₃₀-II, respectively, according to their determined spectra (Fig. 7B). Difference absorption spectra showed the appearance of a positive 350-nm band within 36 μs after the excitation (Fig. S6A). The absorption spectrum of P₆₁₀ is similar to that of ^15EPo but shows a red-shifted Po peak and a higher Soret peak (Fig. 7B). From 36 μs to 3.3 ms, a positive 520-nm band and a negative 610-nm band were observed (Fig. S6B). The absorption spectrum of P₅₃₀-I was very similar to that of the deprotonated ^15ZPg identified in the WT and D71A variants at high pH values and also in the C73A variant at a neutral pH value (Figs. 2E, 4, B–F, and 7B). This strongly suggests that the bilin is deprotonated in P₅₃₀-I. From 3.3 to 31 ms, small but detectable changes were observed at positive 520- and 610-nm bands and a negative 350-nm band (Fig. S6B). The absorption spectrum of P₅₃₀-II is similar to that of P₅₃₀-I (Fig. 7B), suggesting that the bilin is still deprotonated in P₅₃₀-II. The final P₅₃₀-II to ^15ZPb transition occurred at a substantially slower time range from 31 ms to 11 s, which gave a positive 430-nm band and negative 350- and 530-nm bands (Fig. S6C). The coupling of the negative green band and the positive blue band in the P₅₃₀-II to ^15ZPb transition suggests that the thiol adduct formation occurs in the deprotonated bilin.

Figure 7.

Time-resolved spectroscopy of the reverse photocycle. A, time trace of the absorption changes at selected wavelengths from ^15EPo to ^15ZPb conversion. Absorption changes were monitored for the samples in the buffer containing 100% H₂O (black) or 100% D₂O (blue). These traces were fitted with three exponentials (solid line). Time constants τ₁, τ₂, and τ₃ are shown as vertical dotted lines. B, absorption spectra of three intermediates P₆₁₀ (black), P₅₃₀-I (red), and P₅₃₀-II (blue) and the final product (green), which was calculated by the subtraction of absorption changes from the initial ^15EPo spectrum (gray, solid line). The good agreement between the final product (green) and ^15ZPb spectra (gray, dashed line) demonstrated that the final product corresponds to ^15ZPb. C, effect of the H₂O/D₂O ratio on the time constants τ₁, τ₂, and τ₃.

To obtain the evidence of bilin deprotonation in the two-Cys photocycle, we substituted H₂O with D₂O and investigated the effect of this in the transient absorption spectra. The τ₁ clearly increased by 2.4-fold upon the D₂O substitution (Fig. 7C). The amplitude of this increase was correlated to the ratio of H₂O/D₂O (Fig. 7C), suggesting that the increase of τ₁ was caused by the isotope effect of deuterium. The increase of τ₁ value is similar to the H/D substitution of the transient deprotonation in the phytochrome (76). In contrast, τ₂ and τ₃ were not significantly affected by the H/D substitution (Fig. 7C). These results are consistent with our conclusion that the P₆₁₀ to P₅₃₀-I transition is caused by bilin deprotonation. Considering that bilin photoisomerization generally occurs within the order of picoseconds (66, 77, 78), our transient spectroscopy experiments suggest that the reverse photocycle from ^15EPo to ^15ZPb of Oscil6304_2705 consists of the three sequential chemical reactions on the bilin: 1) 15E to 15Z photoisomerization to produce P₆₁₀; 2) the deprotonation of pyrrole nitrogen producing P₅₃₀-I, with subsequent conformational changes producing P₅₃₀-II; and 3) second thiol adduct formation to produce ^15ZPb as the final photoproduct. Thus, Oscil6304_2705 of the B/O subfamily combined the protochromic photocycle and the two-Cys photocycle in separate timescales.

Discussion

In this study, we investigated the bilin protonation state and its associated absorption changes in the two-Cys photocycle of Oscil6304_2705 of the B/O subfamily of CBCRs. pH titration and RR spectroscopy suggested that the bilin chromophore is protonated in the stationary states of ^15ZPb and ^15EPo. pH titration of single amino acid variants showed that bilin deprotonation caused ^15ZPg and ^15EPg formation in the absence of a second thiol adduct. The pK_a of ^15ZPg was lower than the pK_a of ^15EPg by almost 3 orders of magnitude, suggesting the changes in the bilin protonation state upon photoisomerization. Consistently, the deprotonated ^15ZPg was also detected as intermediates with a significant solvent isotope effect in the reverse photocycle from ^15EPo to ^15ZPb using time-resolved spectroscopy. To our knowledge, this is the first detailed characterization of protochromic absorption changes underlying the two-Cys photocycle.

Although no structural information about the B/O subfamily is currently available, structural information for the closely related DXCF CBCR of TePixJ of the B/G subfamily suggested that Asp-71, Cys-73, and His-102 residues of Oscil6304_2705 directly interact with the bilin and tune its absorption (Fig. 1B) (54–56). C73A and D71A variants showed almost no effects on the formation of protonated ^15EPo (Fig. 2, E and G), indicating that Cys-73 and Asp-71 are not responsible for the bilin protonation in the 15E state. In contrast, H102A, H102Q, and H102L variants were deficient in the ^15EPo formation. H102L variant was capable of bilin photoisomerization and enhanced thiol adduct formation in the 15E state (Fig. 2I and Fig. S5). These results suggest that the imidazole group of His-102 directly interacts with the 15E state of the bilin and is necessary for the proton transfer. pH titration analysis of phytochrome Agp1 showed that the His residue corresponding to His-102 is responsible for maintaining the bilin pK_a (74). Changes in the protonation state in the imidazole group corresponding to His-102 were reported in the stationary ^15ZPr of phytochrome Cph1 (79–82). Therefore, we speculate that the protonated imidazole group of His-102 donates its proton to the bilin. Alternatively, the neutral imidazole group of His-102 can receive proton from other charged residue(s) or B- or C-ring propionate via the hydrogen bond network and transfer the proton to the bilin.

In the pH titration experiments, we observed that second thiol adduct was stable for pH 5–11 in the 15Z state of WT and also 15E state of H102L variants. This robustness shows that the chromophore is not deprotonated in this pH range as expected for a rubinoid pigment (75) and that the linkage is not affected by titration of other residues or bilin propionates. In contrast, we observed pH-dependent thiol adduct formation in the 15E state of the WT (Fig. 4B, blue arrows). There is a proton equilibrium between ^15EPo, ^15EPb, and ^15EPg with two close pK_a values around pH 9 and 10. A possible explanation for this equilibrium is the “neutral model” that, around pH 9, the bilin deprotonation caused loss of its positive charge and the neutralized bilin enhanced its interaction with the neutral thiol group of Cys-73. Around pH 10, deprotonation of the thiol yields negatively charged thiolate, which destabilizes the interaction with the neutral bilin. An alternative explanation is that in the “charge model” deprotonation of Cys-73 precedes the bilin deprotonation within the protein environment, and the negatively-charged thiolate efficiently attacks the positively-charged protonated bilin, causing the decrease of ^15EPo at around pH 9. Notably, we observed that pH-dependent thiol adduct formation in the 15Z state of the D71A variant is around pH 7–9, where the bilin is deprotonated (Fig. 4E, blue arrows). This indicates the occurrence of second thiol adduct between the deprotonated bilin and protonated thiol group at pH 7–9, which is destabilized by deprotonation of the thiol over pH 9. Therefore, we favor the neutral model for the thiol group attachment to the bilin. For another factor other than charge, overall conformation of the bilin could affect the thiol–adduct formation. The deprotonated B-ring nitrogen could interact with the hydrogen of the C-ring nitrogen, and the coplanarity of the B/C-rings could promote the accessibility and reactivity of the thiol group to the bilin.

Based on this structural information and experimental results obtained in this study, we propose a hypothetical model for the molecular process of the reverse photocycle of Oscil6304_2705 protein, as shown in Fig. 8. It is commonly accepted that the first process of the photoconversion of the phytochrome superfamily is 15Z/15E photoisomerization of the bilin (4), which typically occurs on the order of picoseconds (66, 77, 78). Our monitoring system in transient absorption spectroscopy covers a slower timescale from the order of microseconds to seconds. Therefore, the orange-absorbing P₆₁₀ was assigned to protonated and thiol-free ^15ZPo (Fig. 7). We propose that the transition from P₆₁₀ to P₅₃₀-I reflects bilin deprotonation, which is based on the following: 1) the spectral similarity of P₅₃₀-I and deprotonated ^15ZPg species in the pH titration analysis; and 2) the sensitivity of τ₁ of this transition to D₂O and the insensitivity of other steps. Because site-directed mutagenesis suggested that His-102 interacts with ^15EPo, but not Asp-71 and Cys-73, we speculate that His-102 also acts as an acceptor during the ^15ZPo to ^15ZPg transition (Fig. 8, green arrow). The bilin deprotonation could occur in the B- or C-ring (61, 63), but we currently do not have information about which pyrrole ring is deprotonated. The P₅₃₀-I to P₅₃₀-II transition was insensitive to H/D substitution and therefore may reflect a conformational change of the bilin or of residues such as Cys-73 or Asp-71. The final P₅₃₀-II to ^15ZPb transition was also insensitive to H/D substitution, suggesting that proton transfer is not rate-limiting of the nucleophilic addition reaction of thiol (15, 83). We assume that the unshared electron pair of the sulfur atom of Cys-73 attacks the C10 atom (Fig. 8, blue arrows), with formation of a covalent bond between the B-ring (or C-ring) nitrogen and the proton derived from the thiol group of Cys-73.

Figure 8.

Hypothetical model of the reverse reaction of Oscil6304_2705. Stationary ^15EPo converted to ^15ZPo with the bilin 15Z/15E photoisomerization. Then, ^15ZPo converted to ^15ZPg with the bilin deprotonation. Finally, ^15ZPg converted to ^15ZPb with the formation of a second thiol adduct. Intermediates and their time constants identified by time-resolved spectroscopy are shown. Hydrogen bond networks with the bilin and chemical reactions suggested by the site-directed mutagenesis are shown.

The molecular process of the two-Cys photocycle has been extensively studied in TePixJ and SesA (Tlr0924), which belong to the B/G subfamily. They photoconvert between ^15ZPb and ^15EPg, where the PVB chromophore is protonated in both stationary states (15, 65, 84). Notably, the residues in the bilin-binding pocket are highly conserved between B/O and B/G subfamilies (Fig. 1A), suggesting the possibility that the B/G subfamily also undergoes bilin deprotonation during the photocycle. The D492S variant of TePixJ was partially deficient in but still performed Pb/Pg conversion (54), which is similar to the D71A variant of Oscil6304_2705. H523Q and H523L variants of TePixJ also showed absorption changes of ^15ZPb/^15EPb (54), which is similar to the H102L variant of Oscil6304_2705. These results imply that the protonation state of PVB is maintained by the imidazole group of His-523 in TePixJ and that the replacement of His-523 with Leu caused the bilin deprotonation, as shown in H102L variants of Oscil6304_2705. In addition, members of the B/B subfamily harbor a Leu residue in the position of His-102 and photoconvert between ^15ZPb and ^15EPb (Fig. 1A) (16, 19, 21), suggesting that they intrinsically utilize the loss of the imidazole group to enhance the thiol adduct formation for blue light-sensing. Some CBCRs harbor a Tyr residue in the position corresponding to His-102, implying that the phenol group of Tyr can act as a proton donor/acceptor or that they do not undergo bilin deprotonation because of the high pK_a value of Tyr compared with His.

Time-resolved spectroscopy of the ^15EPg to ^15ZPb photoconversion in SesA identified two ^15ZPg as intermediates (78). Because both protonated and deprotonated PVBs show green light absorption (52), the protonation state of these ^15ZPg intermediates has to be examined by methods other than absorption spectroscopy such as NMR or RR. However, green-absorbing states were not identified from ^15ZPb to ^15EPg conversion (85), suggesting that the bilin protonation state did not change in the forward reaction of the two-Cys photocycle of SesA. Notably, ^15ZPb formation of Oscil6304_2705 shows a time constant of 2.0 s, which is ∼100-fold slower than the 23.6 ms of SesA (78), demonstrating substantial diversity in the lifetime of the intermediates of the CBCR photocycle.

NpF2164g3 of the insert-Cys subfamily photoconverts between ^15ZPv and ^15EPo using PCB, whose spectra are slightly blue-shifted but very similar to those of the B/O subfamily (18). Time-resolved spectroscopy of NpF2164g3 identified three intermediates in the reverse reaction from ^15EPg to ^15ZPv: ^15ZLumi-Or peaked at 630 nm (τ ∼9 ns), ^15ZMeta-Og peaked at 550 nm (τ ∼750 ns), and ^15ZMeta-Or peaked at 650 nm (τ ∼1 ms) (86). Although additional intermediates with slower time constants could be possible, these data imply that the thiol adduct formation occurs in protonated ^15ZMeta-Or. Therefore, we speculate that photoconversion mechanisms could differ between insert-Cys subfamily and the B/O subfamily, especially as regards bilin protonation. These differences might derive from the complex evolutionary history of insert-Cys CBCRs, which evolved from the B/G subfamily via the loss of the Cys residue in the DXCF motif yielding R/G CBCRs and followed by acquisition of a large loop containing the insert-Cys residue (32). The involvement of the protonation state of diverse CBCR subfamilies and their evolution are interesting issues for further exploration.

Experimental procedures

Plasmid construction

The sequences of the synthetic oligonucleotides used in this study are shown in Table S1. The ORF of Oscil6304_2705 from the cyanobacterium Oscillatoria acuminata PCC 6304 (Oscil6304_2705) was synthesized using Invitrogen GeneArt DNA synthetic service (Thermo Fisher Scientific) with codon optimization for E. coli. The synthesized gene fragment was amplified by PCR using the primer set of Oscil6304_2705_1 and Oscil6304_2705_2. The pET28a expression vector (Merck) was amplified using the primer set of pET28a-InFusionF2 and pET28a-InFusionR2. The two DNA fragments were assembled using the InFusion cloning kit (Takara Bio, Japan), resulting in the expression plasmid of Oscil6304_2705 as an N-terminal His-tagged protein. Site-directed mutagenesis was performed using KOD plus mutagenesis kit (Toyobo, Japan) with the primer sets of O6304_2705_C73A_F and O6304_2705_C73A_R for the C73A variant, O6304_2705_D71A_F and O6304_2705_D71A_R for the D71A variant, O6304_2705_H102L_F and O6304_2705_H102 for the H102L variant, O6304_2705_H102Q_F and O6304_2705_H102_R for the H102Q variant, and O6304_2705_H102A_F and O6304_2705_H102_R for the H102A variant. Sequences of the constructs were confirmed by Sanger sequencing.

Alignment and protein structures

Alignments of amino acid sequences of the GAF domains were prepared using MAFFT version 7.429 (87). The PDB file was visualized using PyMOL. Sequences of CBCR GAF domains were derived from the following organisms: TePixJ and SesA from Thermosynechococcus elongatus BP-1; SyCcaS, SyPixJ1, SyCph1, SyCph2, SyCikA, and Slr1393 from Synechocystis sp. PCC 6803; IflA and FdRcaE from Fremyella diplosiphon; Oscil6304_2705, Oscil6304_4336, and Oscil6304_4203 from O. acuminata PCC 6304; NpCcaS, NpF4973, NpR5313, NpF1000, and NpR6012 from Nostoc punctiforme ATCC 29133; Alr3356, All1688, and AnPixJ from Anabaena sp. PCC 7120; S7335_RcaE from Synechococcus sp. PCC 7335; Lep6406_CcaS from Leptolyngbya sp. PCC 6406; AM1_1378 from Acaryochloris marina MBIC11017; RfpA from Leptolyngbya sp. JSC-1; PaBphP from Pseudomonas aeruginosa ATCC 15692; and DrBphP from Deinococcus radiodurans. Sequences of 211 GAF domains of B/O CBCR were obtained by BLASTP search in the nr database of NCBI using Oscil6304_2705 as a query with threshold of sequence identity 41% (88).

Light sources

For illumination of purified proteins and E. coli cells, LEDs emitting at 460 nm with 260 milliwatts (LXML-PB01-0023; Lumileds), 515 nm with 100 milliwatts (SHD-HBGX1; Shimarisudo), and 620 nm with 120 milliwatts (LXHL-MD1D; Lumileds) were used as blue-, green-, and red-light sources, respectively.

Protein expression and purification

The expression vector for His-tagged Oscil6304_2705 of WT or single amino acid variant was transformed to E. coli C41(DE3) harboring PCB-producing plasmid pKT271 (67). The transformants were precultured overnight in 10 ml of LB medium, and then transferred to 1 liter of LB containing 0.05 mm 5-aminolevulinic acid and cultured for 2 h at 37 °C with shaking at 200 rpm. 20 μg/ml kanamycin and 20 μg/ml chloramphenicol were added to LB throughout this study. After the addition of 1 mm isopropyl β-d-thiogalactopyranoside, the culture was further incubated for 2 h at 37 °C, 16.5 h at 25 °C, or 16 h at 16 °C with shaking at 100 rpm. Then, the E. coli cells were harvested by centrifugation for 10 min at 8273 × g at 4 °C. The expression level of the protein was estimated based on color changes of the harvested cells by illumination of blue and red light (Fig. S2). The cells were resuspended in 50 ml of disruption buffer (0.1 m NaCl, 20 mm HEPES-NaOH, pH 7.5, 1 mm DTT) and broken by a French press (no. 5501-M; Ohtake) twice at 1500 kg/cm². The homogenate was centrifuged at 164,700 × g for 30 min at 4 °C, and its supernatant was supplemented with 20 mm imidazole. His-tagged proteins were purified with a nickel-affinity column (HisTrap 5 ml; GE Healthcare UK Ltd.) using the Akta start system (GE Healthcare UK Ltd.) at a late flow at 5 ml/min in a cold chamber maintained at 4 °C. Proteins were eluted by 45 ml of a linear gradient of imidazole from 20 to 420 mm with 1.5 ml fractionation. The purified fractions were combined, supplemented with 5 mm EDTA for the removal of any leaked Ni, and dialyzed at 4 °C with the disruption buffer to remove the imidazole and EDTA. Proteins of single amino acid variants were concentrated using centrifugal filters (Amicon Ultra-0.5 ml, 10 kDa; Merck), as their expression levels were lower compared with WT. For dark reversion assay, RR, and time-resolved spectroscopies, the concentration of NaCl in the buffer was increased to 1 m, which efficiently suppressed the protein aggregation at high concentration.

SDS-PAGE and fluorescence detection of PCB

Purified samples were denatured in 62.5 mm Tris-HCl, pH 6.8, 2% (w/v) sodium lauryl sulfate, 5% (v/v) β-mercaptoethanol, 5% (w/v) sucrose, and 0.005% (w/v) bromphenol blue. The samples were incubated at 65 °C for 3 min and then separated by SDS-PAGE on a 15% acrylamide gel. After separation, the acrylamide gel was immersed in 1 mm zinc acetate for 10 min and scanned with a Typhoon FLA 9000 imager (GE Healthcare) with green laser excitation for detection of the fluorescence of the covalently bound bilin. The gel was then stained with a Coomassie Brilliant Blue-based stainer, EzStain AQUA (ATTO, Tokyo, Japan).

Spectroscopy, pH titration, and pK_a estimation

For pH titration, dialyzed samples were diluted with milliQ water to 10 mm NaOH, 50 mm NaCl, and 0.5 mm DTT. 400 μl of each sample was converted to the 15Z or 15E state by light illumination. After full conversion, 100 μl of 1 m of the following buffer (final 200 mm) was added in the dark: MES-NaOH for pH 5.0, 5.5, 6.0, and 6.5; HEPES-NaOH for pH 7.0, 7.5, 8.0, and 8.5; and glycine-NaOH for pH 9.0, 9.5, 10.0, 10.5, and 11.0. We used an aliquot of the sample for one pH point measurement to minimize the protein denaturation due to exposure to extreme pH for a long time. For the acid-denaturation absorption spectra, 100 μl of each dialyzed sample was converted to the 15Z or 15E state by light illumination and then supplemented with 400 μl of 10 m urea, pH 2.0. After mixing the sample by pipetting in the dark, the samples were immediately measured using a UV-visible spectrophotometer (model V-650; JASCO). The pK_a fitting of the absorption changes with one or two titrating groups was performed as described previously (41). Peaks of blue absorption at pH 5.0–6.0 and peaks of green and orange absorption at pH 5.0–5.5 were excluded from the calculation of pK_a.

To assess the photoconversion capacity under extreme pH, 400 μl of WT protein was photoisomerized to 15Z or 15E states in 10 mm HEPES-NaOH, 500 mm NaCl, and 0.5 mm DTT, and then the pH was changed by addition of the excess amount of buffer of pH 5.0 or pH 11.0 as described above. LED light of each color was sequentially illuminated for 20 s. To assess the dark reversion activity, WT protein was photoisomerized to 15E states in 20 mm HEPES-NaOH, 1 m NaCl, and 1 mm DTT, and absorption spectra were measured at 6-h intervals in the dark at room temperature. For these analyses, we used transmission integrating sphere (model ISV-722; JASCO) for UV-visible spectroscopy to mitigate the scattering due to partial protein aggregation.

RR spectroscopy

To obtain the RR spectrum of ^15EPo, purified protein solution was illuminated with blue light for 5 min before the measurement. RR spectra were measured for 60 min at room temperature with 785-nm excitation using a 0.3-m imaging spectrograph (SpectraPro 2300i; Princeton Instruments) equipped with a charge-coupled device (CCD, Pixis256E; Roper Scientific). The excitation light was obtained from a diode laser (ibeam-smart-785; Toptica), and back-scattered light was collected by collection optics. The laser power at the sample was ∼60 milliwatts. A long wave path edge filter (LP03-785RS-25; Semrock) was used to eliminate the intense Rayleigh light. RR spectrum of ^15ZPb was measured under continuous red illumination at room temperature with 406-nm excitation. The spectrometer system for ^15ZPb was composed of a laser diode (LM406; Ondax Inc.), a 0.5-m single spectrograph (Spex 500M; HORIBA Jobin Yvon), and a liquid nitrogen-cooled UV-coated CCD detector (Spec-10:400B; Roper Scientific Inc.). A 90° scattering geometry was employed, and a 0.19-m spectrometer (Triax190; HORIBA Jobin Yvon) was used to remove the excitation light. The laser power at the sample was ∼5 milliwatts. All of the spectra were taken at room temperature, and custom-made software was used to eliminate the noise spikes in the spectra caused by cosmic rays. RR spectra of ^15EPr RcaE with 785 nm excitation was obtained from our previous work (63).

Time-resolved absorption spectroscopy

The reverse reaction from ^15EPo to ^15ZPb was examined by the flash-induced absorbance changes. Here, we used the second harmonic (532 nm, 7 ns, and 7 mJ) of a Q-switched Nd:YAG laser (Minilite I, Continuum, Santa Clara, CA) for the excitation pulse. ^15ZPb has no absorption at 532 nm, and so ^15EPo was selectively photoexcited. The optical layout of the apparatus was essentially the same with that described previously (89). The monitoring light source was a 150-watt Xenon lamp, and its beam was perpendicular to that of the excitation pulse. The photomultiplier was used to detect the monitoring light passing through the sample in 10-mm quartz cuvette. Two monochromators were placed in the rear of the monitoring light source and in front of the photomultiplier to select the measuring wavelength and exclude the scattered pulse from the sample, respectively. At each selected wavelength, 20 laser pulses were used to improve the signal–to–noise ratio. Before each flash excitation, the sample was completely converted to ^15EPo by 20-s illumination of blue light, which was made by a 150-watt halogen lamp and an interference filter KL-430 (Toshiba). The protein concentration was adjusted so that the absorbance values of ^15EPo peak became 0.5. The sample temperature was kept at 20 °C. We independently measured the baseline shift “without” excitation pulses and then subtracted it from the corresponding data measured “with” excitation pulses. At respective wavelengths, the absorbance changes were recorded between −44 ms to 10 s with two AD converters at 0.5- and 50-μs intervals, respectively. The data points were then selected by choosing a logarithmic timescale to reduce the number of points.

To examine the kinetic isotope effect, we prepared four samples containing 0, 33.3, 66.7, and 100% D₂O. For 0 and 100% D₂O samples, the transient-absorbance changes were measured at 300–750 nm with a 10-nm interval. They were analyzed based on an irreversible sequential model (90). Here, we supposed the following model: ^15EPo → P₁ → P₂ → ··· → P_n → ^15ZPb, where P_n represents nth kinetically-defined state. This model contains only forward reactions, and so the P_i states (i = 1 − n) may contain a few physically defined intermediates when the reverse reactions exist between them. The detailed analysis procedures were essentially the same with that described previously (64). Briefly, the respective data set of all wavelengths was fitted simultaneously by the following multiexponential functions shown in Equation 1,

$$\Delta A\left(\lambda ,t\right)={\displaystyle \sum _{i=1}^n{B}_i\left(\lambda \right)·\exp \left(-t/{\tau }_i\right)+B_0\left(\lambda \right)}$$ (Eq. 1)

where ΔA(λ, t) means the measured time-dependent absorbance changes at λ nm, τ_i indicates the decay time constant of the ith kinetically-defined state, and B_i(λ) and B₀(λ) are the parameters depending on λ. Here, we used two to five exponential functions (n = 2–5) and found the reductions in the standard deviation of weighted residuals were saturated at a three-exponential function, indicating the existence of three kinetically-defined states (P₁–P₃). Using the determined fitting parameters, the absorption differences (Δϵ_i) between P_i and the initial ^15EPo were calculated. Finally, the absolute spectra of P_i states were obtained by adding the spectrum of ^15EPo to Δϵ_i. For 33.3 and 66.7% D₂O samples, the transient-absorbance changes were measured at typical five wavelengths (350, 430, 520, 570, and 620 nm). Respective datasets were fitted simultaneously by three-exponential functions to determine the decay time constants, τ_i.

Author contributions

T. S., T. K., R. M., K. K., C. Y., T. F., M. U., T. E., and Y. H. investigation; T. K., M. U., and Y. H. visualization; T. K. and Y. H. formal analysis; T. K., T. F., and M. U. methodology; T. K., T. F., M. U., and Y. H. writing-original draft; T. K., T. F., M. U., T. E., and Y. H. writing-review and editing; Y. H. conceptualization; T. E. and Y. H. supervision; Y. H. funding acquisition; Y. H. project administration.