K.-H. Zhao, State Key Laboratory of Agricultural Microbiology, Huazhong Agricultural University, Wuhan 430070, China Fax: +86 27 8728 4301 Tel: +86 27 8728 4301 E-mail: email@example.com
Cyanobacteriochromes are phytochrome homologues in cyanobacteria that act as sensory photoreceptors. We compare two cyanobacteriochromes, RGS (coded by slr1393) from Synechocystis sp. PCC 6803 and AphC (coded by all2699) from Nostoc sp. PCC 7120. Both contain three GAF (cGMP phosphodiesterase, adenylyl cyclase and FhlA protein) domains (GAF1, GAF2 and GAF3). The respective full-length, truncated and cysteine point-mutated genes were expressed in Escherichia coli together with genes for chromophore biosynthesis. The resulting chromoproteins were analyzed by UV-visible absorption, fluorescence and circular dichroism spectroscopy as well as by mass spectrometry. RGS shows a red–green photochromism (λmax = 650 and 535 nm) that is assigned to the reversible 15Z/E isomerization of a single phycocyanobilin-chromophore (PCB) binding to Cys528 of GAF3. Of the three GAF domains, only GAF3 binds a chromophore and the binding is autocatalytic. RGS autophosphorylates in vitro; this reaction is photoregulated: the 535 nm state containing E-PCB was more active than the 650 nm state containing Z-PCB. AphC from Nostoc could be chromophorylated at two GAF domains, namely GAF1 and GAF3. PCB-GAF1 is photochromic, with the proposed 15E state (λmax = 685 nm) reverting slowly thermally to the thermostable 15Z state (λmax = 635 nm). PCB-GAF3 showed a novel red–orange photochromism; the unstable state (putative 15E, λmax = 595 nm) reverts very rapidly (τ∼ 20 s) back to the thermostable Z state (λmax = 645 nm). The photochemistry of doubly chromophorylated AphC is accordingly complex, as is the autophosphorylation: E-GAF1/E-GAF3 shows the highest rate of autophosphorylation activity, while E-GAF1/Z-GAF3 has intermediate activity, and Z-GAF1/Z-GAF3 is the least active state.
cGMP phosphodiesterase, adenylyl cyclase and FhlA protein domain (SMART acc. no. SM00065)
potassium phosphate buffer
Anabaena (Nostoc) sp. PCC 7120
the two photoconvertible states of CBR or Phy designated by the absorption maxima, with the stable generally 15Z state (λmax = XXX nm) preceding the light-activated generally 15E-configured state (λmax = YYY nm)
period circadian protein, Ah receptor nuclear translocator protein and single-minded protein domain (SMART acc. no. SM00091)
red–green switchable protein encoded by rgs = slr1393
Synechocystis sp. PCC 6803
Cyanobacteriochromes (CBRs) (also termed cyanochromes ) are photochromic cyanobacterial biliproteins acting as sensory photoreceptors. They are related to the red/far red responsive phytochromes (Phys)  but absorb at different wavelengths with the result that the absorptions of the different CBRs nearly cover the visible spectrum . The usually N-terminal regions that bind their photoactive linear tetrapyrrole chromophores contain one or more GAF [cGMP phosphodiesterase, adenylyl cyclase and FhlA protein domain, (SMART acc. no. SM00065)] and PAS [period circadian protein, Ah receptor nuclear translocator protein and single-minded protein domain (SMART acc. no. SM00091)] domains. The spectral variety of CBRs, compared with the more conservative Phys, probably relates to the diversity of both light quality and quantity in aquatic systems, caused not only by physical factors but also by the amount and variety of competing phototrophs that show a large variety of light-harvesting systems.
An increasing number of CBRs have been linked to regulatory processes. RcaE is required for both green and red light responsiveness in Calothrix sp. 7601. The chromophore is bound covalently to Cys198, located in a GAF domain . Another red/green photoreceptor, PixJ (All1069), has been isolated from Nostoc sp. PCC 7120 (Nostoc); the photochemistry involves a 15Z/E isomerization of the covalently bound phycocyanobilin (PCB) .
Irradiation of Nostoc with red (630 nm) and far-red light (720 nm) increased and decreased, respectively, cellular adenosine 3′,5′-cyclic monophosphate (cAMP) content. This Phy-like response is related to AphC (coded by all2699) . It is related to the photoreceptor Cph2 of Synechocystis sp. PCC 6803 (Synechocystis) that, together with TaxD1, regulates phototaxis [7,8]. UirS from the same strain has been implicated with eliciting near-UV negative phototaxis . CcaS regulates the expression of the rod-core linker polypeptide CpcG2, thereby modulating the light-harvesting capacity of the phycobilisome [10,11].
Increasing numbers of chromophorylated holo-Phys  and holo-CBRs , or domains thereof, are becoming accessible by heterologous co-expression, in Escherichia coli, of apoprotein genes with those coding for chromophore biosynthesis. PCB is generated from endogenous heme by expression of two genes, ho1 and pcyA, coding for heme oxygenase and PCB:ferredoxin oxidoreductase, respectively [13,14]. Similarly, phytochromobilin (PΦB) is generated by expression of ho1 and the plant PΦB:ferredoxin oxidoreductase gene HY2 . As in plant-type Phy, these chromophores are bound covalently at C-31 to the apoprotein by thioether linkages by SH-addition reactions. (Plant Phys and the similarly structured bacterial Phys like CphA are denoted in the following as plant-type Phys; they show red/far red photochromism and carry a PCB or a PΦB chromophore at the GAF domain flanked N- and C-terminally by a PAS and the PHY domains, respectively. Bacterial Phys like CphB are denoted bacteriophytochromes; they also show red/far red photochromism but carry a BV chromophore near the N-terminus that is attached at C-32. Phy-related photochromic proteins from cyanobacteria absorbing in other spectral regions are denoted CBR . Both bacterial Phy and CBR may contain additional chromophore(s).) Most of the currently known chromophore binding sites of CBRs contain PCB  but for some, including TePixJ of Thermosynechococcus elongatus, a doubly linked phycoviolobilin (PVB) chromophore has been proposed: the first bond is again to C-31 and the second to either C-5 or C-10 [1,16]. A singly bound PVB chromophore is generated in α-phycoerythrocyanin from PCB by a lyase-catalyzed isomerization concomitant with the addition to C-31 [17,18]: the chromophore generation in TePixJ may be similar  but autocatalytic . Molecular constructions using truncated proteins are also of increasing interest as molecular biomarkers [1,21–23].
The genomic sequence of Synechocystis  contains more than 10 ORFs coding for proteins that are related to Phy or CBR. A single GAF domain of RGS (coded by sll1393) has recently been shown to bind PCB autocatalytically, resulting in a red/green switchable chromoprotein . We have now studied the full-length protein and its three isolated GAF domains, and compared it with the homologous AphC from Nostoc . The lyase domain and PCB binding sites were identified by chromophorylation of RGS and AphC mutants in E. coli. The photochromism of the resulting RGS and AphC constructs was studied by absorption, CD and fluorescence spectroscopy. Although it relates in both constructs to the photoreversible Z/E isomerization of the Δ15,16 double bond of PCB, the spectroscopy and kinetics of the mono-chromophorylated RGS and the bi-chromophorylated AphC are very different. In both RGS and AphC, however, the autophosphorylation activity of the inherent kinase is photoregulated.
GAF3 is sufficient for auto-chromophorylation of RGS
A domain search (CDD, ) of RGS detected a hydrophobic motif between amino acids 1 and 20 and three GAF domains between amino acids 51 and 202 (GAF1), 248 and 398 (GAF2) and 441 and 596 (GAF3) with an embedded hydrophobic transmembrane motif (amino acids 541–562) (analyzed by tmpred, ). Also detected was a multi-domain region (amino acids 605–963, hereafter called the NtrB domain) that is similar to the nitrogen-specific signal transduction histidine kinase NtrB  as well as a phosphate regulon sensor kinase PhoR  and contains an embedded PAS domain (amino acids 608–658) (Figs 1 and S1). It has already been shown  that GAF3 alone is chromophorylated in the E. coli system. To further identify the chromophore binding site(s), lyase domain(s) and inter-domain interactions of this basic unit, genes were constructed that code for the following apoprotein mutants: RGS(21–974/Δ56–197) lacking GAF1, RGS(21–974/Δ251–393) lacking GAF2, RGS(441–596) corresponding to GAF3, RGS(21–974/Δ542–559) lacking the hydrophobic motif in GAF3, RGS(21–543) lacking the hydrophobic motif and the NtrB domain, and RGS(21–974/Δ612–656) lacking the PAS motif in NtrB (Fig. S1, Table S1). In addition, the individual GAF domains were expressed and a point mutant, RGS(21–974/C528I), served to identify the chromophore binding site (see below). The respective constructs (Table S2) were expressed in E. coli, together with HO1 and PcyA that generate PCB from endogenous heme. This heterologous chromophorylation system for biliproteins in E. coli [13,14] is well suited for poorly soluble apoproteins (Figs 2 and S2) . To facilitate over-expression in E. coli, we nonetheless omitted in all cases the N-terminal hydrophobic motif (amino acids 1–20) to produce RGS(21–974). This motif is probably responsible for membrane docking of the CBR and unrelated to its photochromism: it is absent in the homologous AphC from Nostoc (see below).
Under optimal conditions, RGS(21–974) was smoothly chromophorylated (Fig. 2A). The E. coli cells were green, with maximal absorption at 650 nm (Figs 2 and S2), and the same is true for all mutants that contained GAF3 (Fig. 2, Table S3): chromophorylation of the ‘naked’ GAF3 has already been shown before . Since, by contrast, all mutants lacking this domain were colorless, GAF3 is the single chromophore binding domain containing both the autocatalytic lyase activity and the binding site. Interestingly, the C-terminal section of GAF3 was necessary for chromophorylation, because RGS(21–543) was inactive. The lyase function seems more specifically located in the embedded helix motif which distinguishes GAF3 from the other two GAF domains, because RGS(21–974/Δ542–559) was also inactive (Fig. S3C); this will be discussed below in more detail. The binding site was identified as Cys528 by the lack of binding of RGS(21–974/C528I) (Fig. S3A; see also below). Affinity purified GAF3 also binds PCB in vitro, but the spectroscopic properties are quite different: the product absorbs at 645 nm and has only a residual photochromic response (Fig. S4). Differences between the E. coli and the in vitro reconstitution products have occasionally been noted before [5,13,29], but it is still unclear if the advantage of the former system is due to a bulk medium effect, or certain factors like chaperonines that assist the reaction, and/or the correct folding of the product.
The spectroscopic properties of all mutated chromoproteins were very similar to that of the wild-type protein lacking only the short N-terminal α-helix (Table S3). This includes, in particular, the reversible photochemistry and the fluorescence yield, which are very sensitive to the protein environment of the chromophore. Clearly, therefore, GAF3 not only contains the lyase function and the binding site, but all significant chromophore–protein interactions also reside in this domain and are not influenced by the interactions with other domains. There is one possible exception to this notion, however, that may be relevant for signal transduction to the C-terminal NtrB section. In RGS(21–974/Δ612–656) lacking the embedded PAS domain (Fig. S1), the extinction coefficient is ∼ 25% lower, and the fluorescence yield is ∼ 25% larger than in the other chromoproteins (Table S3). None of the constructs of RGS showed noticeable dark reversion of the state containing E-PCB.
The isolated PCB-GAF3 chromoprotein had a maximal absorption at 662 nm after denaturation with acidic urea (8 m) in the dark  (Figs 3, 4A); this is identical to the absorption of 15Z-PCB . SDS/PAGE combined with Zn2+ induced fluorescence  showed that PCB was covalently bound to the apoprotein. This was verified by digestion with trypsin: the chromophores were retained, and the absorption (λmax = 660 nm, Fig. 4A) is characteristic for 15Z-PCB chromopeptides [17,31,32]. After HPLC of the digest, the chromopeptide was identified via mass spectrometry as NHESLAVGDVETAGFTDC(PCB)HLDNLR (Fig. 4B), in agreement with a covalently attached PCB at Cys528.
The chromophore specificity of GAF3 was tested with phycoerythrobilin and PΦB, generated by replacing pcyA in the E. coli system with pebAB and HY2 [13,29], respectively. PΦB, which differs from PCB only by the C-18 substituent (vinyl instead of ethyl), is smoothly attached to GAF3 at Cys528 (Fig. S2F; see also ). Cells producing PEB remained colorless: this chromophore also contains an 18-vinyl group but lacks the Δ15,16 double bond, indicating that the autocatalytic lyase activity or the binding site is incompatible with the latter.
AphC has two chromophore binding sites
The domain structure of AphC is very similar to that of RGS except for the absence of the N-terminal hydrophobic motif. It has three cysteines at positions 138, 478 and 554 that are part of typical chromophore binding sequences. To identify which of them actually binds PCB, we constructed the three single-site mutants AphC(C138L), AphC(C478A) and AphC(C554L), as well as the double-site mutant AphC(C138L/C554L) (Fig. S1). After introduction of the respective mutant plasmid together with the PCB-synthesizing plasmid in E. coli, AphC, AphC(C138L), AphC(C478A) and AphC(C554L) were all chromophorylated (Fig. S5A–D), but AphC(C138L/C554L) was inactive (Fig. S5F). This indicates that AphC has two binding sites, one in GAF3 at C554 (equivalent to C528 in Sll1393) and an additional one in GAF1 at C138. To check whether each of the three GAFs of AphC could be individually chromophorylated, we subcloned separately gaf1, gaf2 and gaf3. After introduction of the respective gaf plasmid together with the PCB-synthesizing plasmid in E. coli, GAF1 and GAF3 were found to be chromophorylated (Fig. 3A,B), but GAF2 was colorless. When these chromoproteins were denatured in acidic urea (8 m), the absorption peak at 662 nm (dotted lines in Fig. 3A) showed that PCB was bound unmodified to both sites; further, covalent attachment was verified by the Zn2+ induced fluorescence on SDS/PAGE (insets in Fig. 3). PCB binding to the isolated GAF domains supports the notion that each binding domain also possesses its own autocatalytic lyase activity. Unlike RGS, the in vitro reconstitution of both GAF1 and GAF3 from AphC generated photochemically competent chromoproteins (Fig. S4).
Photochromism of RGS and AphC is assigned to Z/E isomerizations of PCB
RGS(21–974) and all chromophorylated derivatives of this N-terminally truncated wild-type protein show pronounced photochromism, no matter whether they are in solution or in E. coli cells (Figs 2 and S2). After irradiation with white or green light (500 nm), the isolated chromoproteins are green (λmax = 648 nm, P650). They turned pink (λmax = 536 nm, P535) under red light (650 nm) and could be switched back many times to the P650 state with green light. Distinct absorption differences were retained after denaturation with acidic urea (8 m) in the dark where interactions between the chromophore and apoprotein are minimized and the absorption spectra reflect the free chromophore. The absorption maxima of the denatured P650 at 662 nm and P535 at 596 nm (dotted lines in Fig. 3A, see also ) match those of protein-bound PCB in the 15Z and 15E configuration, respectively [10,32]. The CD spectra of the native chromoproteins support this assignment (Fig. 5A, D) . Furthermore, irradiation of the denatured P535 state resulted in conversion to the denatured P650 state, which is typical for the irreversible, unidirectional E→Z isomerization of free PCB (Fig. S6).
PCB (and related chromophores) can principally isomerize at three double bonds (Δ4,5, Δ9,10 and Δ15,16). Only the Δ15,16 isomers have been identified positively in Phy [2,34–36]; this double bond also isomerizes in the PVB-containing α-phycoerythrocyanin . Our current assignment based on optical spectroscopy is also consistent with an isomerization at Δ15,16. In view of recent proposals of alternative [19,38] or additional photoreactions , a re-evaluation beyond optical spectroscopy is desirable, in particular with the CBR showing unusual photochemistries. Reconstitution attempts with PEB that lacks the Δ15,16 double bond were, unfortunately, unsuccessful (see above).
RGS(21–974), RGS(21–974/Δ56–197), RGS(21–974/Δ251–393) and RGS(21–974/Δ612–656) were insoluble in 20 mm potassium phosphate buffer at pH 7 (KPB7) but could be solubilized at pH ≥ 8 in the presence of 0.05% SDS (Fig. 2). The predicted hydrophobic transmembrane helix (amino acids 542–559) in the GAF3 domain  is present in all constructs; however, RGS(441–596), which also contains this hydrophobic motif, was soluble in KPB7, thereby facilitating spectral measurements. The P535 state of RGS(441–596) had an intense CD signal, which is typical for twisted chromophores like the 15E-PVB in α-PEC ; the P650 state, however, possessed only weak CD in the visible and strong positive CD in the near ultraviolet (Fig. 5A,D), which is similar to that of 15Z-PCB bound to Cys155 in the β-subunit of phycoerythrocyanin . This type of CD seems to be associated with the conformation of PCB proteins, or states, with intense fluorescence .
The photochromism of AphC is also indicative of Z/E isomerization(s) of PCB, but it is more complex because the two binding domains show a very different behavior (Fig. 3). PCB-GAF1 can be reversibly transformed by light between two states, i.e. P635 absorbing at 635 nm and P685 absorbing at 685 nm (solid lines in Fig. 3A). The same photoreaction is also observed with the C554L mutant of the full-length protein, indicating no cross-talk among the GAF domains (Fig. S5D). P685 was always in a mixture with P635 due to spectral overlap and the relatively rapid dark reversion (see below). In this case, however, the long-wavelength state has the 15E-PCB chromophore and the short-wavelength state the 15Z-PCB chromophore, as shown by denaturation with acidic urea (8 m) in the dark: the absorption maximum of the denatured red state was at 660 nm and that of the far-red state at 606 nm (dotted lines in Fig. 3A). This red-shift of the E state is reminiscent of Phy [6,32]. Their CD spectra also showed the pattern found with Z/E isomers (Fig. 5B, E) : except for the red-shift of the E state, they show the typical weak signal for 15Z-PCB and the strong signal for the more twisted 15E-PCB in native chromoproteins [10,32,39].
The photochromism of PCB-GAF3 of AphC was much less pronounced when studied by the standard irradiation protocol. The state absorbing at 590 nm, when irradiated with > 610 nm light, developed only a shoulder at shorter wavelengths (Fig. 3B) which disappeared when irradiated with 700 nm light. Again, the same photoreaction is observed with the full-length protein lacking the second binding cysteine at position 138, indicating no cross-talk among the GAF domains (Fig. S5B). The absorptions of the two states are similar to those of PCB-GAF3 from Synechocystis, but the short-wavelength state of GAF3 from AphC is always in a mixture with the long-wavelength state due to its fast dark reversion (see below). Their CD spectra also showed only slight differences (Fig. 5C, F) . No significant absorption differences were retained after denaturation with acidic urea (8 m) in the dark (dotted lines in Fig. 3B). The maximal absorption at 660 nm of both denatured states indicated that the chromophore was Z-PCB in both cases (dotted lines in Fig. 3B). This is most likely due, again, to the very rapid E→Z dark reversion that is complete in the time needed for denaturation by conventional mixing (see below). Judged from the identification of 15Z-PCB in both denatured states, we speculate that the phototransformation also involves a 15Z/E isomerization of the chromophore. This is supported by the blue-shift of PCB-GAF3 upon irradiation of the Z state (see below), but this remains to be proven.
The spectral features described above indicate that both PCB-GAF1 and PCB-GAF3 of AphC were photochromic, but the states resulting from phototransformation of the thermally stable states (λmax = 635 and 645 nm, respectively) show a moderately fast and very rapid dark reversion of PCB-GAF1 and PCB-GAF3, respectively. The 15E state of PCB-GAF1 (λmax ∼ 685 nm), obtained by irradiation with 570 nm light, reverts back in the dark to the 15Z state (λmax = 635 nm) within 30 min (Figs 3C and S7A). (P685 is also accessible by irradiation with a 610 nm low-pass filter; obviously, the Z→E photoconversion is faster than the E→Z conversion.) The dark reversion of PCB-GAF3 was much faster. Curve resolution of the spectrum, obtained after saturating irradiation of the dark-stable red state (λmax = 645 nm), revealed that the second state absorbs in the orange spectral region (λmax = 595 nm): these wavelengths indicate a novel type of photochromic system. Irradiation of the red state with a low-pass filter (≥610 nm), which allows for a faster transformation than interference filters, produces the 595 nm state of PCB-GAF3 that reverts back to 645 nm with a rate constant τ of ∼ 20 s (Figs 3D and S7B). This was too fast for our equipment to record a full spectrum; a narrow-range scan is shown in Fig. 3B, D. Both the moderately fast and the very rapid dark reversion of GAF1 and GAF3, respectively, were retained in AphC (see below). Also, site-directed mutants with only a single chromophore, namely AphC(C138L) and AphC(C554L), showed comparable kinetics with the respective isolated PCB-bearing domain (Figs S5 and S7–S9); consequently, there are no interactions with other domains that stabilize the 15E states in GAF1 and (putative) GAF3.
Since PCB-GAF3 of AphC fluoresced slightly, we trapped the transient state by fluorometry at liquid nitrogen temperature. PCB-GAF3 of AphC(C138L) was irradiated with 700 nm light (interference filter) and ≥610 nm light (low-pass filter), and the samples were then quickly immersed in liquid nitrogen. The emission and excitation spectra are shown in Fig. S8. As expected from the kinetics, no pure spectra were obtained, but the data indicate that both states of PCB-GAF3 are fluorescent. The Z state emitted at 665 nm and the putative E state at 643 nm, with a lower yield if judged from the relative intensities in Fig. S8. In their excitation spectra, the Z state of PCB-GAF3 had a maximum at 646 nm and the E state at 587 nm, which coincided with their respective absorption maxima (Fig. 3B,D).
Histidine kinase activity is light regulated
The C-terminal section of RGS in Synechocystis is annotated as an autophosphorylating histidine kinase . The light effects on the autophosphorylation activity of PCB-RGS(21–974) were studied by saturating irradiation of the isolated chromoprotein with different light qualities and then incubating it with [γ-32P]ATP followed by autoradiography (Fig. 6). Labeled phosphate was covalently incorporated into the E state (P535) more efficiently (∼ 3.3- and 3.6-fold after 20 and 30 min incubation, respectively) than into the Z state (P650) (Fig. 6A). These results support the function of RGS as a (histidine) kinase that is upregulated by red light and downregulated by green light.
Due to the presence of two PCB chromophores and their differential photochromism in AphC, a more complex autophosphorylation activity may be anticipated. AphC can principally assume four different states, but due to the spectral properties of the GAF1 (P635 ↔ P685) and GAF3 domains (P645 ↔ P595), and the differential kinetics, only three states can be accumulated in significant amounts: state I, obtained by irradiation with 570 nm light, contains mainly 15E-GAF1 and 15Z-GAF3; state II, obtained with light > 610 nm, contains mainly 15E-GAF1 and 15E-GAF3; and state III, obtained with 700 nm light, contains mainly 15Z-GAF1 and 15Z-GAF3 (Fig. S9). For these three states, the autophosphorylation activity was measured by incorporation with [γ-32P]ATP. Unlike with RGS, however, this was done already during the irradiation, because of the rapid thermal reversion of the 15E states. Under these conditions, state II is 3.5-fold more active than state I, and state III has comparable activity to state I (Fig. 6B). AphC(C138L), which has only GAF3 chromophorylated, is only poorly soluble. The data still indicate that the putative 15E state, enriched by irradiation with light > 610 nm, is 2.4-fold more active (Fig. 6C) than the 15Z state. In the reverse experiment with AphC(C554L) where only GAF1 is chromophorylated, the 15E state obtained with light > 610 nm is 1.6-fold more active (Fig. 6D) than the Z state. Obviously, both chromophore-bearing domains regulate autophosphorylation in AphC. Since the 15E state of the PCB on GAF3 is never highly enriched under our irradiation conditions, the larger contribution of this small fraction, compared with GAF1, indicates that the effect of the fully isomerized GAF3 domain may be much larger than that of GAF1. Due to the rapid dark reversion of 15E-GAF3, its contribution would increase with increasing light intensity and strongly depend on the relative contributions of the green and orange bands.
Autocatalytic chromophore binding
The chemical and spectroscopic properties identify RGS as a CBR. Among this growing number of photochromic kinases, AnPixJ  and RGS share similar photochromic properties. AnPixJ has four consecutive GAF domains, of which isolated GAF2 and GAF4 could be chromophorylated well but GAF3 only to a small extent. Only chromophorylated GAF2 of AnPixJ was photochromic, and its properties are similar to PCB-GAF3 of RGS. Both isolated GAF1 and GAF3 from AphC are readily chromophorylated, and so are the two respective full-length proteins carrying only a single active cysteine at position 138 or 554. As judged from these and other CBRs , individual GAF domains are capable of autocatalytic chromophore addition and the development of (nearly) native spectroscopic properties. These domains (∼ 15 kDa) are much smaller than both chromophore binding domains of Phy , which also bind the chromophores autocatalytically, and they are even smaller than most phycobiliprotein subunits that require lyases for correct binding of the chromophore . Possibly, the requirements for rendering the chromophores highly fluorescent are more demanding than for rendering them photochromic. However, ∼ 3 kDa of the phycobiliproteins take the N-terminal helices that carry most of the interaction of the individual subunits in the αβ-monomers, and further parts of the remaining ∼ 14 kDa are involved in interactions with other monomers and with linker proteins that eventually result in generating the highly organized phycobilisome .
Autocatalytic binding has been defined, and distinguished from spontaneous binding, as formation of a functionally competent chromoprotein that is indistinguishable from the native chromoprotein . Strictly speaking, our chromophorylation experiments meet only the first part of this definition, because the native chromoproteins isolated from the cyanobacteria are still unknown. The general autocatalytic binding in Phy and CBR, as well as the spectroscopic data and, in particular, the similarities with AnPixJ, suggest an autocatalytic mechanism. This is also supported for both GAF domains of AphC by generation of spectroscopically indistinguishable products by in vitro binding of affinity purified proteins. Interestingly, this situation is different with RGS, which supports again the advantages of the E. coli system for reconstitution. Ikeuchi and Ishizuka  have pointed out, however, that the chromophorylation results in E. coli can also differ from those in the parent cyanobacterium; it is not clear if this is due to the different environment or to specific protein–protein interactions that are absent in E. coli. A final proof of an autocatalytic mechanism has to await comparison with the native chromoproteins, however, that are at present unavailable. While phycobiliproteins require lyases for correct chromophore binding , this function resides on the apoprotein in Phy and CBR. Wu and Lagarias  proposed a glutamate within a 200 amino acid domain as essential for lyase activity. In RGS-GAF3, deletion of the putative hydrophobic helix embedded in GAF3 prevents autocatalytic chromophore binding. It contains a glutamate (E550) in a FVAEQ sequence that bears similarity to the region containing the consensus glutamate. By comparison with other CBRs, this helix appears not to be a generally necessary element for chromophore binding, however. For RGS, only one such element is predicted, i.e. in the chromophorylated GAF3. For AnPixJ (All1069), one hydrophobic helix is predicted in GAF1 and a second one in GAF2, but GAF1 is not chromophorylated. Conversely, for AphC only one hydrophobic helix is predicted in GAF3, but GAF1 lacking this element is also chromophorylated, and the product is photochromic.
Photochemistry and dark reversion of bi-chromophoric AphC
AphC has two chromophore binding GAF domains, and one that lacks the respective cysteine and does not bind a chromophore. This is similar to Cph2 , but the organization of the two proteins is different. In Cph2, one of the binding domains is located at the N-terminus and the other at the C-terminus, while both are located in the N-terminal half of AphC. The photochemistry of the AphC is also unusual. Both the GAF1 and GAF3 domains contain PCB chromophores. Judging from the optical spectra, the photoreaction of the one on GAF1 involves a 15Z/E isomerization. The spectral properties of the two states resemble those of PCB binding cyanobacterial Phy like Cph1 , but the spectra of both states are blue-shifted by ∼ 15 nm. Shifts of this size are frequently observed among phycobiliprotein chromophores . The more unusual red–orange reversible photochemistry of the second chromophore on GAF3 is reminiscent of PixJ from Nostoc , but here the putative E state is red-shifted from 543 to 590 nm. Another type of green–red photoreversible photochromism was observed with CcaS from Synechocystis or Nostoc punctiforme [10,11], and (probably also) RcaE from Calothrix sp. 7601 , but here the green-absorbing state is thermally stable. Obviously, both states of PCB-GAF3 are blue shifted compared with Phy but, as in the latter, the putative E state is red shifted with respect to the Z state: the thermostable state P644 of GAF3 absorbing at 644 nm carries a Z-PCB chromophore. The rapid thermal reversion, which is complete during the denaturation procedure with acidic urea, has so far prevented the positive identification of the photoactivated chromophore; judging from other photoactive biliproteins we speculate, however, but cannot currently prove, that it is also generated by a 15Z/E isomerization. In solution or under denaturing conditions, 15E bilins are more or less blue shifted compared with their 15Z isomers; in the case of PCB proteins in acidic urea, the blue-shift amounts to ∼ 70 nm [10,44–47]. By contrast, both large red- and blue-shifts are realized in the various native biliproteins: in plant-type Phy, the 15E chromophore is considerably red shifted, while it is blue shifted in RpBphP3 from Rhodopseudomonas palustris , in several CBRs  and also in α-phycoerythrocyanin, which carries a PVB chromophore . Its 15E state shows a highly twisted C–D methine bridge thereby almost removing ring D from the conjugation system [37,49]. In RpBphP3, an ‘E-pocket’ around amino acids 207–212 has been identified that largely determines the spectral properties of the 15E chromophore, with a particularly striking contribution of a single residue, L207 . The molecular basis for the ‘phytochrome shift’, i.e. the large red-shift of the native 15E compared with the 15Z chromophore is, nonetheless, still controversial [48,50,51]. Irrespective of the red- or blue-shift of the 15E relative to the 15Z state of the chromophore, the fluorescence is quenched, indicating an increased mobility.
Thermal reversion has also been observed with both red and blue shifted states containing 15E chromophores; the kinetics are furthermore influenced by the environment and by domain interactions within the holoprotein [52–56]. The reversion of the 585 nm state of GAF3 from AphC is unusually fast, which may relate to a function for light intensity sensing (see below). Fast back reactions («1 s) are frequently observed among retinal proteins, where the chromophores also undergo reversible Z/E isomerizations, resulting in negligible net photochemistry when studied by conventional spectrophotometers . A comparably fast back reaction could also prevent substantial accumulation of the 15E isomer in seemingly photochemically inactive biliproteins; an example may be CikA from Synechocystis .
AphC acts as a photoregulator in Nostoc for cellular cAMP levels which are decreased and increased by red light (630 nm) and far-red light (720 nm), respectively ; this spectral dependence correlates with the photochemistry of the GAF1 domain. The additional 645–595 nm photochromism of GAF3 might extend the photoresponsive region; its signaling activity is indicated by the autophosphorylation of the histidine kinase in isolated AphC (see below). If the rapid thermal reversion persists in Nostoc, it may be difficult to detect in vivo. Since substantial amounts of P595 are only formed under continuous red light, it may relate to a high irradiance response.
Both RGS and AphC have C-terminally a putative histidine kinase domain. As in CBR and Phy , RGS and AphC show autophosphorylation activity which is higher in the 15E than in the 15Z states. The two biliproteins are therefore likely to be involved in light-dependent phospho-relay signaling, although the response regulators are yet to be identified. This may influence, or interact, with the cellular N- or P-status, because the C-terminal domain of RGS (amino acids 605–963) shows sequence homology to the N-specific signal transduction histidine kinase NtrB and the phosphate regulon sensor kinase PhoR. Both NtrB and NtrC are organized as a two-component regulatory system in bacteria, where phosphorylation of NtrC by NtrB results in transcriptional activation of nitrogen-regulated genes . PhoR is part of the two-component regulatory pair, PhoR/PhoB, that is responsible for upregulation of the Pho regulon expression in response to Pi limitation . RGS interacts with Slr1037, a two-component response regulator of the CheY subfamily, and also with a Slr2111 of unknown function , but a participation in phospho-relay signaling has not been studied. The situation with AphC is again more complicated due to the presence of two chromophores with different spectral and kinetic properties, and the poor solubility of isolated AphC carrying only the chromophore at GAF3. Under continuous light, which stabilizes the 15E-GAF1/15E-GAF3 state, the rate of phosphorylation corresponds roughly to the sum of the independent autophosphorylation rates of each of the domains. As in RGS, there are no obvious response regulators near the aphC gene.
The ready autocatalytic chromophore binding, intense fluorescence and/or photoswitching capacity of biliproteins renders these pigments candidates for applications in imaging, including optogenetics [22,23,40,60,61]. The photoswitchable GAF domains of CBR and bacterial Phy are particularly interesting due to their relatively small size (∼ 15 kDa), their spectral coverage of the entire visible spectrum, which extends even into the near infrared (reviewed by ) and their varying kinetics (see above). GAF1 of RGS as well as GAF1 and GAF3 of AphC add to these spectral variations; recently, heterologous generation of chromophorylated RGS by a single transformation has been demonstrated . For some applications in optogenetics  or STORM microscopy , the dark recovery kinetics is an important consideration; in this respect, the fast reversion of the 15E state of AphC-GAF3 is particularly interesting.
Cloning and expression
All genetic manipulations were carried out according to standard protocols . The DNA fragment of rgs(21–974), lacking the 5′-terminal region coding for the N-terminal α-helix, was PCR amplified from the genomic DNA of Synechocystis by using primers P1 and P2 (Table S1) and Taq DNA polymerase (MBI Fermentas, Beijing, China). To identify the lyase domain and PCB binding site of RGS, the mutation primers (Table S1) were designed to generate GAF3 (rgs(441–596)), GAF1-deleted mutant (rgs(21–974/Δ56–197)), GAF2-deleted mutant (rgs(21–974/Δ251–393)), part of GAF3 plus NtrB-deleted mutant (rgs(21–543)), hydrophobic motif in GAF3-deleted mutant (rgs(21–974/Δ542–559)) and PAS-deleted mutant (rgs(21–974/Δ612–656)), as well as the site-directed mutant (rgs(21–974/C528I)) lacking C528 (Fig. S1). The mutation was carried out with the mutation kit (TaKaRa, Dalian, China) according to the manufacturer’s instructions.
According to blast analyses in NCBI, AphC from Nostoc is similar to RGS. A short N-terminal helix is followed by GAF1 (amino acids 31–195), GAF2 (amino acids 301–427), GAF3 (amino acids 465–602) and a multi-domain region (amino acids 645–920) that is similar to NtrB and PhoR. The DNA fragment of AphC was PCR amplified from the genomic DNA of Nostoc by using primers P17 and P18 (Table S1). Each of the three GAF domains contains a cysteine (C138, C478 and C554) in a region that is highly conserved and typical for chromophore binding in GAFs of CBR. To identify the PCB binding site of AphC, the site-directed mutants of aphC(C138L), aphC(C478A) and aphC(C554L) were constructed with the mutation primers P19–P24 (Table S1). aphC contains a EcoRV site that was used to cleave aphC(C138L) and aphC(C554L) with EcoRV. By re-ligating the resulting two fragments containing C138L and C554L, the double mutant aphC(C138L/C554L) was generated (Fig. S1). To check whether domains GAF1, GAF2 or GAF3 can each be chromophorylated, the isolated gaf1, gaf2 and gaf3 sections from aphC were PCR amplified via primers P25–P30.
All PCR products were ligated into the cloning vector pBluescript (Stratagene, Beijing, China). After sequence verification, the gene fragments were subcloned into pET30 (Novagen, Munich, Germany). For over-expression, these pET30-derived expression vectors were transformed into E. coli TunerTM (DE3) (Novagen) containing the PCB-generating plasmid pACYC-ho1-pcyA  (Table S2). The doubly transformed cells were cultured at 18 °C in lysogeny broth medium supplemented with kanamycin (20 μg·mL−1) and chloromycetin (17 μg·mL−1). After induction with isopropyl thio-β-d-galactoside (1 mm) for 16 h, the cells were centrifuged at 12 000 g for 3 min at 4 °C, washed twice with water and stored at −20 °C until use.
While the chromophorylated mutants RGS(21–974), RGS(21–974/Δ56–197), RGS(21–974/Δ251–393) and RGS(21–974/Δ612–656) are insoluble, RGS(441–596) is water soluble. For the soluble chromoprotein, the cell pellet was resuspended in ice-cold potassium phosphate buffer (KPB, 20 mm, pH 7.0) (KBP7) containing 0.5 m NaCl and disrupted by sonication for 5 min at 50 W (JY92-II; Scientz Biotechnology, Linbo, China). The suspension was centrifuged at 12 000 g for 15 min at 4 °C, and the supernatant was purified by Ni2+-affinity chromatography on chelating Sepharose (Amersham Biosciences, Shanghai, China), developed with KBP7. Bound proteins were eluted with KBP7 containing, in addition, imidazole (0.5 m). After collection, the sample was dialyzed twice against KBP7. If necessary, the affinity-enriched proteins were further purified by FPLC (Amersham Biosciences) with a Superdex 75 column developed with KBP7 . For the insoluble chromoproteins, cell pellet was resuspended in ice-cold KBP7 and disrupted by sonication. The suspension was centrifuged at 12 000 g for 15 min at 4 °C. The highly photochromic pellet containing the chromoprotein was washed twice with KBP7. To solubilize the chromoprotein, the pellet was suspended in potassium phosphate buffer (20 mm, pH 8.0) (KBP8) containing SDS (0.05%) and gently stirred for 1 h. After centrifugation, the supernatant containing the solubilized chromoprotein was purified by Ni2+-affinity chromatography on chelating Sepharose as described above, but using KPB8 throughout. The 0.5 m imidazole eluate was dialyzed twice against KPB8 containing glycerol (10%).
Protein concentrations were determined by the Bradford assay , calibrated with bovine serum albumin, and SDS/PAGE was performed with the buffer system of Laemmli . Proteins were stained with Coomassie brilliant blue and those containing chromophores were identified by Zn2+ induced fluorescence .
For autophosphorylation, the reconstituted and Ni2+-affinity purified RGS or AphC derivatives, whose purity was checked with SDS/PAGE in Fig. S10, were first dialyzed against Tris/HCl buffer (50 mm, pH 8.0) and then photoconverted by irradiation at the appropriate wavelengths (see below). The resulting Z or E states (2 μg) were incubated at 25 °C in the dark for 20 and 30 min, respectively, in Tris/HCl buffer (50 mm, pH 8.0) containing NaCl (50 mm), MgCl2 (10 mm) and ATP (2.5 μm, with 150 kBq of [γ-32P]ATP) . The kinase reaction was stopped by addition of an equal volume of Tris/HCl buffer (120 mm, pH 8.0) containing SDS (4%) and dithiothreitol (120 mm). Samples were then subjected to SDS/PAGE. The developed gel was washed with aqueous methanol (45%)/acetic acid (10%), dried, and then exposed onto an image film (Fuji, Shanghai, China). The intensity of each band was measured with a fluorescence and radioisotope science imaging system (Fuji FLA-5100).
Photoconversions were carried out with a fiber optical cold-light source (Volpi, 150 W, Munich, Germany) equipped with appropriate filters. When using 500, 570, 650 or 700 nm interference filters (15 nm fwhm, light intensity 15 μmol·m−2·s−1), samples were irradiated for 5 min. When using a > 610 nm low-pass filter (light intensity 45 μmol·m−2·s−1), samples were irradiated for 1 min. For RGS and derivatives reversible photochemistry is induced by a 500–650–500 nm irradiation cycle, and for AphC and derivatives by a 570–610–700 nm irradiation cycle. Spectra were recorded before and after saturating irradiation.
All chromoproteins were investigated by UV-visible absorption spectroscopy (Perkin-Elmer Lambda 25 or Beckman-Coulter DU 800, Shanghai, China). Covalently bound PCB in RGS and AphC derivatives was quantified after denaturation with acidic urea (8 m, pH 1.5) by its absorption at 662 nm, using an extinction coefficient of 35 500 m−1·cm−1 . Fluorescence spectra were recorded at room temperature with a model LS 45 spectrofluorimeter (Perkin-Elmer, Shanghai, China). Fluorescence quantum yields, ΦF, were determined in KPB7, using the known ΦF = 0.27 of phycocyanin from Nostoc  as standard. For 77 K fluorescence emission and excitation spectra, the samples were quickly frozen in quartz tubes by immersion in liquid nitrogen after the respective light irradiation and then measured with a model F-7000 spectrofluorimeter (Hitachi, Beijing, China). CD spectra were recorded at 20 °C with a model J-810 spectropolarimeter (JASCO, Shanghai, China).
Analyses of chromopeptides
Reconstituted PCB-GAFs were purified by Ni2+ chromatography and dialyzed against KBP8. Tosyl-phenylalanine chloromethyl-ketone-treated trypsin (Sigma-Aldrich, Beijing, China) was added to the sample (10 units·μg−1) together with urea (end concentration 2 m). The mixture was incubated for 2 h at 37 °C and for another 2 h after dilution with the same volume of KBP8 followed by fractionation on Bio-Gel P-60 (Bio-Rad, Munich, Germany) equilibrated with dilute HCl (pH 2.5). Colorless peptides and salts were eluted with dilute HCl (pH 2.5), and the adsorbed chromopeptides with acetic acid (30%, v/v) in dilute HCl (pH 2.5) [31,39]. The collected samples were analyzed by liquid chromatography mass spectrometry (HPLC-MS; Applied Biosystems, Foster City, CA, USA). For HPLC, a Dionex U3000 (Dionex, Beijing, China) system with a Capcell Pak C18 column was developed with a mixture of aqueous formic acid (0.1%) (A) and acetonitrile containing formic acid (0.1%) (B) (linear gradient A : B from 80 : 20 to 60 : 40). Mass spectra were determined in positive ion mode using a TSQ quantum access mass spectrometer with an ESI source (Thermo Fisher, Beijing, China).
We thank Qiong Ma, Ya-Fang Sun, Zhi-Bin Wang, Xian-Jun Wu and Nan Zhou (Wuhan, China) for valuable experimental assistance. We are grateful to R. J. Porra (CSIRO-Plant Industry, Canberra, Australia) for help in preparing the manuscript. KHZ (grant 30870541, 21072068) and MZ (grant 30870519) are grateful for support by the National Natural Science Foundation of China.