Cyanobacteriochromes are photochromic sensory photoreceptors in cyanobacteria that are related to phytochromes but cover a much broader spectral range. Using a homology search, a group of putative blue-absorbing photoreceptors was identified in Nostoc sp. PCC 7120 that, in addition to the canonical chromophore-binding cysteine of cyanobacteriochromes, have a conserved extra cysteine in a DXCF motif. To assess their photochemical activities, putative chromophore-binding GAF domains were expressed in Escherichia coli together with the genes for phycocyanobilin biosynthesis. All except one covalently bound a chromophore and showed photoreversible photochromic responses, with absorption at approximately 420 nm for the 15Z states formed in the dark, and a variety of red-shifted absorption peaks in the 490–600 nm range for the 15E states formed after light activation. Under denaturing conditions, the covalently bound chromophores were identified as phycocyanobilin, phycoviolobilin or mixtures of both. The canonical cysteines and those of the DXCF motifs were mutated, singly or together. The canonical cysteine is responsible for stable covalent attachment of the bilin to the apo-protein at C31. The second linkage from the cysteine in the DXCF motif, probably to C10 of the chromophore, yields blue-absorbing rubin-type 15Z chromophores, but is lost in most cases upon photoconversion to the 15E isomers of the chromophores, and also when denatured with acidic urea.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
protein domain (common protein domain from cGMP phosphodiesterase, adenylyl cyclase and FhlA)
Anabaena (Nostoc) sp. PCC 7120
near ultraviolet spectral range
Cyanobacteriochromes (CBCRs) are photochromic biliproteins from cyanobacteria that act as sensory photoreceptors. These pigments have also been termed cyanochromes . In addition, in older literature, the term ‘phycochromes’ was used, although the molecular nature of such proteins was unclear . They are phylogenetically related to the red/far-red responsive phytochromes, but lack the larger PAS–GAF–PHY module and cover a much wider spectral range, extending from near-infrared to near-ultraviolet (NUV) [3-8]. Their photoactive linear tetrapyrrole chromophores are bound auto-catalytically to one or more GAF domains (SMART accession number SM00065) or PAS domains (SMART accession number SM00091), which are usually located N-terminally. Known chromophores include phycocyanobilin (PCB) [9-11] and phycoviolobilin (PVB) [1, 7, 8, 12, 13]. Some CBCRs also contain 10-thio-phycocyanorubin (TPcR) and 10-thio-phycoviolorubin (TPvR), which are derived from PCB and PVB, respectively, by addition of a second cysteine to C10 (Fig. 1, see below) [8, 14]. Only a few CBCRs have so far been linked to specific functions [3, 5, 15-18].
The photoregulatory function of CBCRs derives from photoreversible reactions of the chromophores and transmission of this signal to output domains of various types [3-5, 18, 19]. Like phytochromes, CBCRs exist in two or sometimes more [14, 20] states that are reversibly inter-convertible by light. The number of putative CBCRs, some of them containing multiple chromophore-binding GAF domains, is rapidly increasing. Thus classification is complex and context-dependent: various sub-groups may be defined by their photochemical signatures, phylogenetic relationships, domain organization, or other criteria [3-5, 21]. In the context of this work, we distinguish five types according to the spectral absorption regions of the two states. Type 1 shows the red/far-red photochromism (color of the 15Z state formed in the dark state/color in the 15E state formed by light activation) that is common in phytochromes, and these include GAF1 of AphC (AphC-GAF1) from Nostoc sp. PCC 7120 (Nostoc)  and GAF1 of Cph2 (Cph2-GAF1) from Synechocystis sp. PCC 6803 [18, 23, 24]. Due to the presence of certain phytochrome-typical domains, these two members have also been characterized as knotless or PAS-less phytochromes [5, 21]. They are included here among the CBCRs because of the capacity of the isolated GAF domains for auto-catalytic chromophore binding. Type 2 CBCRs comprise green/red inter-convertible pigments, including CcaS from Synechocystis sp. PCC 6803  and RcaE from Calothrix sp. 7601 . Like phytochromes, they show a red shift upon conversion of the 15Z to the 15E state, but with both absorption maxima blue-shifted compared to phytochromes. Type 3 CBCRs comprise red/green inter-convertible pigments, including AnPixJ (encoded by all1069) from Nostoc , and RGS (encoded by slr1393) from Synechocystis sp. PCC 6803 [10, 11]. In this case, the 15E chromophore is blue-shifted compared to the 15Z state. Such behavior is also observed for the phycobiliprotein α-phycoerythrocyanin . The kinetics of both the forward and back reactions of a member of this type, NpR6012g4 from Nostoc punctiforme, have recently been studied in detail and compared to those of phytochromes [28-30]. Type 4 CBCRs comprise the red/orange variety represented by AphC GAF3 . As for type 3, they show a blue shift on photoactivation. Type 5 CBCRs are chromoproteins in which at least the 15Z state absorbs in the blue spectral region. The absorption maxima of the dark states at 410–430 nm qualify these CBCRs as blue-light receptors, but they may also act as NUV receptors as their absorptions extend into the NUV spectral region [7, 16].
The CBCRs include TePixJ from Thermosynechococcus elongatus [1, 12], the C-terminal CBCR domain of Cph2 , and UirS from Synechocystis sp. PCC 6803 [16, 17]. In the 15Z state, the chromophore has a second covalent cysteine bond to C10 that interrupts the conjugation system. Depending on the location of this cysteine relative to the chromophore plane, this class has been sub-divided [5, 7] and may require further sub-division according to the various types of chromophore present, and their response to photoactivation. In types 1–4, photoconversion involves a Z←→E isomerization at the C15 methine bridge of the covalently bound PCB or PVB chromophore that is followed by conformational rearrangements of the chromophore and the apoprotein [9, 11, 31, 32]. The reaction of type 5 CBCRs is more complex (see below).
Chromophorylated and photoactive phytochromes , holo-CBCRs  and domains thereof may be generated in Escherichia coli by expression of the apo-protein genes together with genes coding for chromophore biosynthesis from endogenous heme. For example, PCB is generated by expressing two genes, ho1 and pcyA, that encode heme oxygenase and PCB:ferredoxin oxidoreductase, respectively [10, 33, 34]. Most of the currently known chromophore-binding sites of CBCRs contain PCB , which is bound covalently to the apoprotein at C31 by thioether linkages to a conserved cysteine . In other CBCRs, a covalently bound PVB chromophore  is generated auto-catalytically from PCB during binding [4, 13].
The type 5 chromoproteins, including TePixJ of Thermosynechococcus elongatus [1, 12], contain doubly linked chromophores in their 15Z states. The first bond is again via the canonical cysteine, CC, to C31. The second, weaker, bond is formed by an ‘extra’ cysteine (CX) to C10 [7, 8, 14, 20]. An alternative binding to C-5 has been proposed , but all current evidence favours addition to C-10 [8, 14, 20]. A definitive structure is still missing.In one sub-group, this second cysteine is located in a DXCF (Asp-Xaa-Cys-Phe) or derived motif. Although it is not strictly conserved, the context and additional sequences are sufficiently well-defined that these proteins group together, and are referred to here as DXCF proteins [7, 8, 36]. The presence of a DXCF motif does not necessarily imply formation of the second bond, as indicated by NpR5113g1 . By addition of this cysteine to C10, rubin-type chromophores (10-thio-rubins) are formed. As they differ structurally from the parent chromophores PCB and PVB, by interruption of the conjugation system at C10, we refer to them as 10-thio-phycocyanorubin (TPcR) and 10-thio-phycoviolorubin (TPvR), respectively (Fig. 1). Due to this interrupted conjugation, TPcR and TPvR contain the same dipyrrolic chromophore (rings C/D), which absorbs at approximately 420 nm. The chromophore derived from rings A/B absorbs in the NUV range in TPcR. In TPvR, the conjugation is even shorter, comprising ring B only, and the absorption thus occurs even farther into the ultraviolet region [8, 14, 16, 20]. The initial photoreaction of these rubin chromophores is a Z→E isomerization of the Δ15,16 double bond of the C/D chromophore, as in phytochromes and all other CBCRs. Such a photoreaction is also known for bilirubin where, due to the symmetry of the system, it may occur at C5 and/or C15 . However, the Z→E isomerization is generally followed by reversible detachment of the second cysteine [8, 14, 20]. The ensuing extension of the conjugation system results in extraordinarily large red shifts after photoconversion to the 15E state; for example, from 418 to 602 nm . The photoactivated state may be converted back to the blue-absorbing 15Z state by irradiation into the red-shifted absorption band, due to re-binding of the extra cysteine, CX, to C10 in the 15Z state. Nucleophilic thiol addition to C10 of bilins such as PCB is a reversible process [38, 39], and this equilibrium is obviously shifted by isomerization at C15 [8, 14, 20]. A scheme summarizing the diversity of reactions of these CBCRs is shown in Fig. 1.
The relatively large genome of Nostoc sp. PCC 7120 (Nostoc)  contains more than 50 ORFs encoding putative CBCRs with GAF domain(s) . Among them are All1280-GAF2, All1688, All2239, All3691-GAF2, Alr1966-GAF2, Alr2279-GAF2, Alr3120-GAF2, Alr3356 and Alr5272-GAF1, which contain the DXCF motif or a related motif, indicating the presence of several blue-absorbing CBCRs of the DXCF sub-group in Nostoc. We expressed these candidates in E. coli, and report a group of seven photochromic GAF domains possessing 15Z states showing absorption at 410–430 nm, and light-activated 15E states with maxima in the 490–600 nm range.
Results and Discussion
Seven blue-absorbing DXCF photoproteins in Nostoc
Blue-absorbing DXCF biliproteins have been found in Synechocystis sp. PCC 6803 , Thermosynechococcus elongatus [12, 36] and Nostoc punctiforme . As Nostoc sp. PCC7120 has a comparatively large genome, one might also expect several such photoreceptors to occur in this species. Homology and phylogenetic analyses indicated genes encoding putative CBCRs that contain the DXCF motif: all of them are clearly distinct from GAFs that bind the chromophore only by the canonical CC (Fig. S1). In known DXCF photoreceptors, such as TePixJ [1, 12], Cph2  and UirS [16, 17], the DXCF motif contains the ‘extra’ conserved cysteine, CX, forming the second bond to the chromophore. GAFs containing this motif were found in All1280 (GAF2), All1688, All2239, All3691 (GAF2), Alr1966 (GAF2), Alr2279, Alr3120 (GAF1), Alr3356 and Alr5272 (GAF1). Some of them, namely All1280, All3691, Alr1966, Alr3120 and Alr5272, contain additional GAF domains. One is even a GAF domain-only ORF with no candidates for a flanking kinase or response regulator domain (Fig. 2 and Tables S1 and S2). A similar chromophore-binding GAF domain-only protein, NpF4973, has been characterized previously , and genes encoding homologous proteins are found by BLAST in other cyanobacterial genomes. Although all eight GAFs are found in the clade of DXCF photoreceptors (Fig. S1), the DXCF motif is not strictly conserved in three cases: it is replaced by QTCF in All1688, by ELCF in Alr3120-GAF1, and by the even more divergent sequence DPCL in Alr2279. Two GAFs, All1688 and Alr2279, form blue-absorbing chromoproteins (see below), indicating that other sequence traits contribute to the annotation. This reflects the findings of DXCF photoreceptors from Nostoc punctiforme, where the motif is replaced by DRCL, PECF and KNCF , and from Synechocystis PCC6803, where it is replaced by NDCF .
The relevant GAF-coding domains (Fig. 2) were cloned individually into E. coli, and then co-expressed with genes encoding PCB-generating enzymes. Seven GAFs were chromophorylated, and the resulting chromoproteins had various reversible photochemistries (Figs 3-5), in which the originally formed 15Z states show absorption at approximately 410–430 nm, and the 15E states formed in blue light absorb at wavelengths between 490 and 600 nm. These chromoproteins are therefore putative photochromic blue-light receptors [7, 8]. Covalent binding of the chromophores was verified in all cases by SDS/PAGE combined with Zn2+-induced fluorescence  (Fig. S2). In Nostoc punctiforme, as many as 13 putative photoreceptors showing blue absorption in the 15Z state and a range of longer-wavelength absorptions in the 15E state have been characterized [7, 8]; it is therefore likely that this photoreceptor group is common to cyanobacteria, and that they function as light sensors in the 300–600 nm region, of which the 450–600 nm region in particular is relevant for cyanobacterial photosynthesis, which relies on phycobiliproteins as light-harvesting pigments [42, 43]. The absorption of the 15Z state may also be shifted, albeit to a lesser extent. For at least one member, UirS from Synechocystis sp. PCC6803, a function in NUV-induced negative phototaxis has recently been confirmed [16, 17].
Alr3120-GAF1 was chromophorylated but lacked long-lived photochromism (τ > 15 s) that could be detected using our irradiation scheme (Fig. S3). Six other GAFs (All1280-GAF1, All3691-GAF1, Alr1966-GAF1, Alr3120-GAF2, Alr5272-GAF1 and Alr5272-GAF2) were not chromophorylated (Fig. S4). With one exception, Alr3120-GAF2, these GAFs lack the canonical CC. Only two of the non-bonding GAFs carry the DXCF motif, of which Alr3120-GAF2 did not show chromophore binding despite the presence of both this motif and CC (Fig. S1).
Three distinct sub-types of DXCF photoreceptors
The chromophorylated DXCF GAF domains showed a variety of photochromic effects (Figs 3-5). All show absorption at approximately 420 nm that is reversible bleached to various degrees by irradiation with blue light (420 nm); the concomitant increases occur at approximately 600 nm (Alr3356), 530–560 nm (All1280, Alr1966 and Alr2279) and even below 500 nm (All3691). The bound parent chromophore(s) were identified after removing non-covalent interactions with the protein by denaturation with acidic urea (8 m, pH 2.0) , and then converting any 15E isomers to the respective 15Z isomers by irradiation with 500–570 nm light. All spectra are consistent with the presence of only PCB and/or PVB under these denaturing conditions: the reduction of the band at approximately 420 nm upon denaturation shows that the bond to the second cysteine is lost during the treatment (Fig. S5). The relative contributions of PCB and PVB were determined using the known extinction coefficients [35, 45], and the results are summarized in Table 1. All3691-GAF2 contained only PVB (λmax = 592 nm, Fig. 4), as verified and quantified by the typical reversible photochemistry of the bound PVB under acidic denaturing conditions . All1688, All2239, Alr3120-GAF1 and Alr3356 contained only PCB (Fig. 3), while the other chromoproteins contained mixtures of both chromophores (Fig. 5). Light minus dark difference spectra obtained after denaturation also allowed determination of the Z/E ratio of the chromophores in these chromoproteins after various irradiation conditions (Table 1 and Fig. S5) .
Table 1. Photochemical properties of the DXCF photoreceptors in potassium phosphate buffer (20 mm, pH 7.0). The 15Z state was obtained by irradiation of All3691-GAF2 and Alr1966-GAF2 at 510 nm or All1688, All2239, Alr3356, All1280-GAF2 and Alr2279 at 570 nm. The 15E state was obtained by irradiation of 420 nm (for more details, see Experimental conditions). Under denaturing conditions, absorptions at approximately 662–666 nm are derived from Z-PCB, absorptions at approximately 548–556 nm are derived from E-PVB, and absorptions at approximately 598 nm are derived from E-PCB and/or Z-PVB; the latter were differentiated by the difference spectra (Fig. S5)
Maximal absorption (nm)
Parent bilin ratio (PVB/PCB)
The spectral data suggest that there are three distinct sub-types of type 5 CBCR photoreceptors in Nostoc. Among sub-type 5a chromoproteins carrying PCB alone or the respective rubin (TPcR), only Alr3356 showed a significant photochromism (Fig. 3). The absorption by the light-activated state at 598 nm is in the range of other known 15E-PCB-containing chromoproteins (Fig. S5). All1688 showed little but partly reversible photochemistry, but the difference spectrum indicates chromophore heterogeneity (Fig. 3D). The 545 nm band is at a considerably shorter wavelength than expected for 15E-PCB. Possibly, the blue-shifted absorption is due to twisting of the chromophore in the native protein . It is also conceivable that PCB is regenerated from a PVB-like chromophore; a case of reversible inter-conversion between these chromophores has recently been reported by Rockwell et al. . The much reduced photochemistry of All1688 and All2239, and its absence in Alr3120-GAF1 (Figs S3 and S5), indicate increasingly tight binding pockets compared with Alr3356. At the relatively slow (5–10 s) time limit of our spectrophotometer, we found no indication that the reduced photochemistry is due to a fast thermal reversion. The photochemistry of All1688 requires special attention: the changes in the native chromoprotein are similar to those found in NpF1000 , SyCikA  and NpAF142g3, although the latter contains a PCB/PVB mixture [8, 14]. All samples show a minor red shift (shoulder) and broadening of the 430 nm band that is more obvious in the difference spectrum (Fig. 3D). Similar changes are seen during Z→E isomerization of bilirubins . The spectral changes of chromoproteins such as All1688 (Fig. 3A) are thus compatible with 15Z→15E photo-isomerization of TPcR, in which only a small fraction reacts to produce 15E-PCB in which the second bond to CX is lost. However, under denaturing conditions, both the originally formed and light-activated states showed maximal absorption at 662 nm (Table 1 and Fig. S5). This is typical for 15Z-PCB, indicating that the 15E chromophore of the light-activated state is isomerized during work-up and denaturation : the E→Z isomerization of bilirubins in the dark is known to be quite rapid [47, 48].
The single example of sub-type 5b, All3691-GAF2, shows a narrow, intense absorption band at 490 nm in the light-activated state (Fig. 4). Under denaturing conditions, the dark and light-activated states absorbed maximally at 594 and 548 nm, respectively (Fig. S5), as does PVB in the 15Z and 15E configurations, respectively . The 430 nm band for the native dark state thus contains a 15Z-TPvR chromophore generated by thiol addition to 15Z-PVB at C10, and the native light-activated state possesses a 15E-PVB chromophore lacking the second bond. The pigment is similar to the blue/teal receptors, NpR5113g3 and NpR1597g1, from N. punctiforme , and Tlr1999 from T. elongatus . The absorption band for the 15E state of All3691-GAF2 is blue-shifted (λmax = 490 nm, Fig. 4) compared with that of the 15E state in α-phycoerythrocyanin (λmax = 506 nm) . The structure of the light-activated state of α-phycoerythrocyanin shows a twist of the 15E-PVB chromophore between rings C and D that largely uncouples ring D from conjugation, which is relevant for the short-wavelength shift to 506 nm ; such a twist has also been suggested for NpR5113g3 and NpR1597g1 from N. punctiforme  and Tlr1999 from T. elongatus . The even greater blue shift in the light-activated state of All3691 may thus be due to an even stronger distortion of the chromophore, by which a phycourobilin conjugation system is effectively generated. An alternative explanation involves binding of the second cysteine to C5 and twisting ring D out of conjugation. Binding at C5 has previously been suggested to occur in another CBCR . However, such a structure seems unlikely in All3691-GAF2, given ready loss of the bond to the DXCF cysteine during denaturation . This binding site has been ruled out for several other blue/teal receptors by binding studies using modified chromophores .
Sub-type 5c includes All1280-GAF2, Alr1966-GAF2 and Alr2279 (Fig. 5), which bind both PCB and PVB (Fig. S5). PCB prevails in All1280-GAF2, while PVB prevails in Alr1966-GAF2 and Alr2279 (Table 1). One way to generate such chromoprotein mixtures is by incomplete conversion of the initially generated PCB to PVB. Rockwell et al.  and Ishizuka et al.  reported that similarly incomplete isomerization may be improved by photocycling between the stable state carrying the 15Z-rubin chromophore and the light-activated state carrying the 15E-PCB/PVB chromophores. Rubin intermediates have also been implicated in lyase-catalyzed concomitant binding and isomerization of the chromophore in phycobiliproteins ; cycling through this state assists isomerization of PCB to PVB in the GAFs of sub-type 5c. The presence of a mixed population of PCB/PVB-containing chromoproteins renders their photochemistry rather complex because of the presence of two parallel, inter-connected photocycles [14, 20]. Alr1966-GAF2 shows a similar sequential photoconversion to that shown for GAFs 2, 3 and 4 of NpF1883 and Tlr0924 : after irradiation of the 15Z-TPcR/TPvR state with 420 nm light, its absorption peaks at 514 nm with a long-wavelength shoulder at approximately 600 nm (Fig. 5), which are assigned to 15E-PVB and 15E-PCB, respectively (Fig. S5G). Irradiation at the long-wavelength edge of this photoproduct using red light resulted in depletion of the shoulder at approximately 600 nm with an increase of absorption at 428 nm. Subsequent irradiation with 510 nm light resulted in depletion of the residual absorption at 514 nm, with a further increase of absorption at 428 nm (Fig. 5).
The chromoproteins of sub-type 5c show various ratios of PCB to PVB (Table 1). The PCB/PVB ratio varied further with photochromism (Table 1), implying that the second cysteine linkage at C10 of PCB/PVB is (partly) transient and reversible, and/or cooperated with the isomerization of PCB/PVB [8, 49]. After denaturation, the light-activated 15E states of the last three entries in Table 1 have more PCB than the dark states, indicating PVB to PCB conversion on photoconversion of the 15Z to the 15E states. Loss of the second cysteine linkage resulted in increased formation of PCB upon chromophorylation, as shown by mutations of DXCF motifs (see below). In the few cases where chromophorylation of full-length proteins and isolated DXCF-type GAF domains has been compared, there was no evidence for an influence of other domains in the holoprotein on chromophorylation . However, it is unclear whether the blue-absorbing CBCRs of the DXCF type also carry biliprotein mixtures in their cyanobacterial environment. Rockwell et al.  suggest a role in photoproduct tuning because of incomplete TPvR formation even at extended incubation times, and the possibility of partly inter-converting PCB and PVB during the photocycle. Holoproteins such as NpF1883 with several mixed-population GAF domains thus present a bewildering number of states. Alternatively, it is possible that the chromoproteins of sub-type 5c are in fact of sub-type 5b, with incomplete conversion of PCB/TPcR to PVB/TPvR in E. coli and in vitro. In other multi-chromophoric biliproteins, site selectivity for certain chromophores is observed in the native situation, even if they lack such selectivity in vitro or in E. coli [50-52].
Taken together, there is an emerging pattern of recurring biochemical and photophysical traits among the various CXCF-type GAFs. Most of the chromoproteins from Nostoc sp. PCC7120 studied here show one or more matches with those from Nostoc punctiforme and other cyanobacteria [14, 20]. However there is no obvious correlation among the protein sequences and certain biochemical and photochemical traits, as shown in Fig. 6. Among pairs with the highest sequence homologies, there are matches with a high degree of biochemical similarity, for example Alr3356/NpF4973 or All1688/NpF1000, but also more disparate pairs such as NpF6001/All1280-GAF2. Likewise, many pairs with lower homology show very similar biochemistry.
Role of conserved cysteines in chromophore binding and photochemistry
A conserved cysteine, CC, is common to all CBCRs: it binds to C31 of the chromophores. Blue-absorbing biliprotein photoreceptors have an ‘extra’ cysteine, CX, that is responsible for a second bond to the chromophore, by thioether bond formation to C10 of PCB or PVB [7, 8, 13]. Formation of this bond interrupts the conjugation at the central methine bridge, leading to blue-light absorbing rubins, but both TPcR and TPvR retain the photo-isomerizable Δ15,16 double bond. Reversibility of thiol addition at C10 has been demonstrated for C-phycocyanin and model pigments of the dihydrobilin type [38, 53], and more recently for several blue-absorbing biliproteins of the DXCF-type [14, 20]. We tested the function(s) of the two cysteines in the CBCRs from Nostoc using single site variants, or variants in which both CC and CX were replaced. The spectroscopic data for these variants are given in Figs S6–S8.
Proteins in which the extra cysteine, CX, was replaced in the DXCF or related motif (see above) in the blue-absorbing CBCRs from Nostoc generally retained the chromophore, but such replacement resulted in loss of the absorption at approximately 420 nm (Fig. S6) and a concomitantly increased absorption in the 560–650 nm region. These changes are consistent with binding of an unmodified PCB chromophore (All1280-GAF2, Alr1966-GAF2, Alr2279 and Alr3356) or a mixture of PCB and PVB (All3691-GAF2) (Fig. S8). In all these cases, rubin formation is inhibited, confirming that CX in these systems is the residue forming the bond to C10. Interestingly, the chromophorylation patterns differed from those of the native proteins (analyzed after loss of the second cysteine bond by denaturation), supporting a reaction scheme in which the cysteine in the DXCF motif participates in the PCB/PVB isomerization [8, 49]. This effect goes both ways: in Alr3356, the wild-type protein and the CX variant bind PCB, while wild-type All3691-GAF2 binds only PVB and the variant binds a mixture of PCB and PVB. CX has been implicated in the PCB/PVB isomerization , but All3691-GAF2 allows at least partial isomerization despite the missing CX. In All2239, loss of the second cysteine results in a much reduced absorption, possibly indicating that chromophore binding is strongly inhibited. However, the small visible absorption and somewhat larger NUV absorption are also consistent with the presence of a free, non-protonated bilin chromophore [54, 55]. This would mean that the chromophore is bound, but without the conformational change to an extended chromophore that is typical for biliproteins. Non-covalent chromophore binding of this type is seen in bilin-binding insect lipocalins [56, 57]. In most DXCF variants, photochemistry is also affected (Fig. S8). Only All3691-GAF2 shows a pronounced and reversible photochromism resembling that of α-phycoerythrocyanin, as shown by alternating irradiation with 570 and 510 nm light : this relates to the fraction bearing the PVB chromophore, while the fraction carrying a PCB chromophore appears to be inert. Alr3356 shows only a minor blue shift upon irradiation. Nonetheless, the changes after denaturation indicate a sizeable inter-conversion: in acidic urea, the dark state has a spectrum indicating 15E-PCB, while the light-activated state has a spectrum indicating a mixture of 15Z-PCB and 15E-PCB. The chromophores of these denatured samples were uni-directionally photoconverted to 15Z-PCB (Fig. S8), confirming the light-induced 15E→15Z isomerization of PCB. A minor shift upon 15Z→E photo-isomerization was previously observed in the type II photoreaction of α-phycoerythrocyanin . Reversal of the chromophore configurations in the dark- and light-activated states is also not unknown among the phytochromes . Finally, a DXCF-type GAF, NpR5113-GAF1, described in N. punctiforme, binds a chromophore only at CC, without formation of a rubin .
Classical phytochromes with PCB or phytochromobilin chromophores  and all CBCRs, including the DXCF sub-group [7, 8], have a conserved cysteine, CC, that auto-catalytically forms a thioether bond to C31 of the chromophore. This is usually contained in a CH motif, but exceptions are known: for example, CY is found in Tlr0924 and Tlr1999 , as well as in All2239 and Alr2279 studied here, and CL is found in RcaE and CcaS [25, 26], as well as in All1688 studied here; these cysteines are nonetheless regularly found by homology searches. Replacement of this cysteine resulted in loss of the chromophore in All2239, Alr2279 and All3691-GAF2, and in strongly reduced absorption in All1280-GAF2 and Alr1966-GAF2 (Figs S6 and S7). This confirms that this cysteine is the conservative bilin-binding CC. There is one exception: Alr3356(C99I) still binds bilin (Figs S6F and S7F,G), although lack of the stable covalent thioether bond was confirmed by the very weak Zn2+-induced fluorescence on the SDS/PAGE gel (Fig. S2). Such weak fluorescence may arise from a minor residual fraction in which the C10 bond is retained under denaturing conditions: the sample contains much more chromophore than the others. Small amounts of rubinoid pigments (shoulders at approximately 430 nm) are also often found in phycobiliprotein reconstitutions , and have even be discussed as intermediates in chromophore binding . While All1280-GAF2 and Alr1966-GAF2 are photo-inactive, Alr3356(C99I) retained modest blue/orange photochromism (Fig. S7F). The unidirectional photoreaction to 690 nm in the denatured state (Fig. S7G) verifies that at least part of this reactivity corresponds to Z→E isomerization of PCB. Photochemical activity of a non-covalently bound chromophore is not uncommon [61, 62].
The photochromic CBCRs of the DXCF type considerably expand the spectral range of photoactive biliproteins [4, 8]. Their dark states absorb in the same region as the flavoprotein photoreceptors [63-65], but their absorptions extend farther into the UV range , and their light-activated states cover the visible spectrum up to 600 nm. The first members with an identified function, CikA, Cph2 and PixJ1 [6, 18, 59, 66], suggest overlapping physiological roles . Provided this ‘rising tide’ of novel CBCR biliproteins lives up to their proposed photosensory functions, the extended phytochrome family to which they belong will provide extraordinarily rich information regarding precise acclimation of cyanobacteria to their light environment. Many carry additional putative input domains that link this information to other signals [4, 5, 68]. Only one of the CBCRs studied in this work, namely Alr3356, lacks an output domain, but a function is indicated by the presence of a highly homologous GAF domain-only protein (NpF4973) in Nostoc punctiforme that auto-catalytically binds a photochromically functional chromophore , and by genes coding for similar sequences in other cyanobacteria [4, 7].
The present work expands a recent report by Rockwell et al. , showing a similar diversity of proteins in another cyanobacterium, Nostoc sp. PCC7120, that are classified as DXCF-type CBCRs. By site-directed mutagenesis, the roles of both the canonical bilin-binding cysteine, CC, as well as that of the extra cysteine, CX, in the DXCF motif were verified. Consistent with previous work [14, 20], the results support reversible binding of the latter cysteine to C10. The divergence from this motif in three of the newly found members, including DPCL in Alr2279, as well as in chromoproteins from Nostoc punctiforme , suggest that more genes may be present in the sequenced cyanobacteria that have so far gone unnoticed. Members such as All3691-GAF2 extend the absorption of the light-activated state further into the blue range compared with other known members of this class.
Remarkably, this variety of spectral and photochemical properties is achieved through biosynthesis of a single bilin, PCB. The conversion to PVB, and to the rubins of both chromophores, TPcR and TPvR, is catalyzed by the chromophore-binding GAF domains. Cyanobacteria synthesize two bilins, PCB and phycoerythrobilin, by site-specific reduction of biliverdin , but only PCB carries the Δ15,16 double bond that is photo-isomerized in these as well as in all other known CBCRs [4, 8, 70]. This double bond is retained both after isomerization to PVB, and after the addition reaction of the second cysteine. This and previous work  indicates that light activation may also play a role in the case of PVB/TPvR-containing CBCRs, but this requires verification using the holoproteins.
The reversible nucleophilic addition of a cysteine is the key to the blue-shifted absorption of the dark states. As shown by the mutagenesis results, in the absence of this cysteine, the chromoproteins display features of photoactive biliproteins carrying the unmodified PCB or PVB chromophore. For example, the PVB-related photochromism of All3691-GAF2 (Fig. S8) is very similar to that of α-phycoerythrocyanin . The variation of the absorption maxima, including that of the light-activated state of All3691-GAF2 at 490 nm, is well within the range encountered in biliproteins. For example, depending on the apoprotein, PCB shows variations in absorption maxima of approximately 100 nm . However, some of the chromoproteins (All1688 and Alr3356) show unusual photochemistries that require further study.
In most cases, mutation of CX also inhibits conversion of PCB to PVB (see Fig. S8). This supports the proposal by Rockwell et al.  that this cysteine is involved in the conversion. All3691-GAF2 is an exception to this rule: the C59A mutant still generates considerable amounts of photoactive PVB (Fig. S8). Its photochemistry is restricted to this chromophore, as in Tlr1999 . A comparison of the GAFs from N. punctiforme  and those from Nostoc sp. PCC 7120 studied here also shows that the emerging classification of biochemical and photochemical diversity among DXCF-type receptors does not match their (overall) homologies (Fig. 6). The growing database may in future allow a definition of motifs that account for certain biochemical and photochemical traits, which is desirable for understanding the function(s) of this type of CBCRs, as well as their applications.
Cloning and expression
Amino acid sequence alignments of GAFs were generated using Clustal W (http://www.ebi.ac.uk/Tools/msa/clustalw2/) . An unrooted phylogenetic tree was constructed using MEGA version 5.0  using the neighbor-joining algorithm. The same phylogenetic tree was obtained when maximum-parsimony, maximum-likelihood and minimum-evolution methods were used. A bootstrap value of 100 was used; in each clade of similar GAFs, the bootstrap value exceeded 20.
All genetic manipulations were performed according to standard protocols . The DNA fragments of respective gaf domains were PCR-amplified from the genomic DNA of Nostoc using the primers listed in Table S1 and Taq DNA polymerase (MBI Fermentas, Burlington, Canada). To identify the PCB/PVB-binding site of the GAFs, mutation primers (Table S1) were designed to generate site-directed variants in which the cysteine that is highly conserved and typical for chromophore binding in GAFs, or the second cysteine in the DCXF motif, was mutated to other amino acids. This was achieved using a mutation kit (TaKaRa, Otsu, Japan) according to the manufacturer's instructions.
PCR products were ligated into the cloning vector pBluescript (Stratagene, Loveland, CO, USA). After sequence verification, the gene fragments were sub-cloned into pET30 (Novagen, Darmstadt, Germany). For over-expression of GAFs and derived variants (Table S2), the pET30-derived expression vectors were transformed into E. coli Tuner™ (DE3) (Novagen) containing the PCB-generating plasmid pACYC-ho1-pcyA . The doubly transformed cells were cultured at 18 °C in Luria–Bertani (LB) medium supplemented with kanamycin (20 μg·mL−1) and chloromycetin (17 μg·mL−1). After induction with isopropyl β-d-thiogalactoside (1 mm) for 16 h, the cells were centrifuged at 12 000 g for 3 min at 4 °C, washed for 1 min twice with water, and stored at −20 °C until use.
To isolate the chromophorylated GAFs, the cell pellet was resuspended in ice-cold potassium phosphate buffer (20 mm, pH 7.0) containing 0.5 m NaCl, and disrupted by sonication for 5 min at 50 W using a JY92-II cell sonicator (Scientz Biotechnology, Ningbo, China). The suspension was centrifuged at 12 000 g for 15 min at 4 °C, and the supernatant was purified using Ni2+ affinity chromatography on chelating Sepharose (Amersham Biosciences, Uppsala, Sweden), developed using potassium phosphate buffer. Bound proteins were eluted using potassium phosphate buffer containing 0.5 m imidazole. After collection, the sample was dialyzed against potassium phosphate buffer.
Protein concentrations were determined by the Bradford assay , calibrated using bovine serum albumin, and SDS/PAGE was performed using the Laemmli buffer system . Proteins were stained using Coomassie brilliant blue, and those containing chromophores were identified by Zn2+-induced fluorescence .
Photoconversions were performed using a fibre optical cold-light source (150 W; VOLPI, Schlieren, Switzerland) equipped with appropriate filters. When using 420, 510 or 570 nm interference filters (15 nm full width at half height, light intensity 15 μmol·m−2·s−1), samples were irradiated for 5 min. For the DXCF GAFs, reversible photochemistry was induced using a 420/510/420 nm irradiation cycle (PVB/TPvR chromophores), or a 420/570/420 nm irradiation cycle (PCB/TPcR) for 5–10 min to photo-saturate the samples. In case of poorly photoconverting samples, saturation was verified by irradiation for up to 1 h. Spectra were recorded before irradiation, and immediately after each saturating irradiation.
All chromoproteins were investigated by UV-Vis absorption spectroscopy using a Lambda 25 spectrometer (Perkin-Elmer, Waltham, MA, USA) or a DU 800 spectrometer (Beckman-Coulter, Brea, CA, USA). Covalently bound chromophores in DXCF GAFs were quantified after denaturation with acidic urea (8 m, pH 2.0) on the basis of their absorption at 662 nm (Z-PCB, ε = 35 500 m−1·cm−1 ), 595 nm (E-PCB, ε = 32 000 m−1·cm−1 ), 592 nm (Z-PVB, ε = 38 600 m−1·cm−1 ) or 530 nm (E-PVB; 36 000 m−1·cm−1 ). To evaluate the ratio of bound PCB and PVB, the denatured samples were first irradiated with 510 nm light, to drive all bound PCB and PVB into the Z configuration, and then the absorbance was measured at 592 and 662 nm. As bound PVB does not absorb at 662 nm , we first determined the concentration of bound PCB from the absorption at 662 nm, and then that of PVB from the absorbance at 592 nm subtracted from that of PCB, which was calculated assuming an absorbance ratio A662/A592 = 2.1 for PCB .
We are grateful to R.J. Porra (CSIRO Plant Industry, Canberra, Australia) for help in preparing the manuscript. We are grateful for support from the National Natural Science Foundation of China (grant number 31110103912 to H.S. and K.H.Z., and grant number 21072068 to K.H.Z.).