An Unorthodox Bacteriophytochrome from Rhodobacter sphaeroides Involved in Turnover of the Second Messenger c-di-GMP*

Bacteriophytochromes are bacterial photoreceptors that sense red/far red light using the biliverdin chromophore. Most bacteriophytochromes work as photoactivated protein kinases. The Rhodobacter sphaeroides bacteriophytochrome BphG1 is unconventional in that it has GGDEF and EAL output domains, which are involved, respectively, in synthesis (diguanylate cyclase) and degradation (phosphodiesterase) of the bacterial second messenger c-di-GMP. The GGDEF-EAL proteins studied to date displayed either diguanylate cyclase or phosphodiesterase activity but not both. To elucidate the function of BphG1, the holoprotein was purified from an Escherichia coli overexpression system designed to produce biliverdin. The holoprotein contained covalently bound biliverdin and interconverted between the red (dark) and far red (light-activated) forms. BphG1 had c-di-GMP-specific phosphodiesterase activity. Unexpectedly for a photochromic protein, this activity was essentially light-independent. BphG1 expressed in E. coli was found to undergo partial cleavage into two species. The smaller species was identified as the EAL domain of BphG1. It possessed c-di-GMP phosphodiesterase activity. Surprisingly, the larger species lacking EAL possessed diguanylate cyclase activity, which was dependent on biliverdin and strongly activated by light. BphG1 therefore is the first phytochrome with a non-kinase photoactivated enzymatic activity. This shows that the photosensory modules of phytochromes can transmit light signals to various outputs. BphG1 is potentially the first “bifunctional” enzyme capable of both c-di-GMP synthesis and hydrolysis. A model for the regulation of the “opposite” activities of BphG1 is presented.

The BphG proteins contain GGDEF and EAL domains arranged in tandem (GGDEF-EAL). Theoretically, these proteins could possess either DGC or PDE activity or both. However, prior to this study, no proteins containing both activities have been identified. Therefore, it was unclear whether the "opposite" activities, i.e. c-di-GMP synthesis and c-di-GMP hydrolysis, could coexist in a single protein, and if they could, how they would be regulated. To gain insights into these questions, we undertook biochemical and photochemical analyses of the R. sphaeroides BphG proteins. We found that (i) similar to conventional Bphs, these proteins undergo reversible photoconversion between the P r and P fr forms, i.e. they are photochromic; (ii) they possess two enzymatic activities; however, only one activity is light-dependent; and (iii) they may employ an apparently unique switch between the opposite enzymatic activities.
Protein Purification-The overnight cultures of BL21(DE3) (pLysS; pT7-ho1-1) containing the BphG1 and BphG2 constructs were grown at 25°C in 50 ml of LB supplemented with 100 g/ml ampicillin, 25 g/ml kanamycin, 34 g/ml chloramphenicol, and 0.5% glucose. These cultures were transferred into larger flasks containing 1 liter of medium of the same composition and grown to an A 600 of 0.5. Isopropyl 1-thio-␤-Dgalactopyranoside was added to a final concentration of 0.5 mM to induce expression of bphG1 (or bphG2) and ho1, and growth was continued at 25°C for an additional 7 h. The cells were collected by centrifugation at 5,000 ϫ g for 15 min and resuspended in 20 ml of Ni 2ϩ binding buffer (50 mM Na 2 HPO 4 / NaH 2 PO 4 , 300 mM NaCl, 5 mM MgCl 2 , 10% glycerol, 0.2 mM n-dodecyl-␤-D-maltoside (Sigma), 20 mM imidazole, and 0.2 mM protease inhibitors (phenylmethylsulfonyl fluoride, Sigma), pH 8.3. The cells were disrupted using a French pressure cell, and the resulting crude extract was sonicated. The pH of the extract was increased to 8.5 followed by centrifugation at 25,000 ϫ g for 45 min to remove cell debris. The volume of soluble extract was increased to 20 ml with Ni 2ϩ binding buffer. Two milliliters (bed volume) of Ni 2ϩ resin (Novagen) pre-equilibrated with Ni 2ϩ binding buffer were added to the soluble cell extract and gently agitated for 1 h at 4°C. The mix was loaded onto a column, and the resin was washed with 140 ml of washing buffer (same as Ni 2ϩ binding buffer plus 60 mM imidazole and minus the phenylmethylsulfonyl fluoride). Fractions were eluted with 8 ml of elution buffer (same as Ni 2ϩ binding buffer but 200 mM NaCl instead of 300 mM plus 250 mM imidazole and minus the phenylmethylsulfonyl fluoride). The BphG1-containing fractions were pooled and dialyzed twice (4 h each) at room temperature against 2 liters of PDE assay buffer (50 mM Tris-HCl, 200 mM NaCl, 5 mM MgCl 2 , 10% glycerol, 0.2 mM n-dodecyl-␤-D-maltoside, 1 mM EDTA, 1 mM dithiothreitol, pH 8.6). The protein solution was aliquoted and stored at Ϫ70°C. The typical yield of the purified BphG1 holoprotein was ϳ0.5 mg L Ϫ1 , although the typical yield of the apoprotein was 0.1-0.15 mg L Ϫ1 . The PAS-GAF-PHY-GGDEF apo-and holoproteins were purified essentially as described for the full-length BphG1 with two deviations: (i) all buffers contained 5 mM dithiothreitol, which prevented protein precipitation; (ii) the concentration of imidazole in the wash buffer was 40 mM.
Spectral Analysis-Absorbance spectra were recorded with a UV-1601 PC UV-visible spectrophotometer (Shimadzu) at room temperature. Protein solution (100 l) in a 10-mm light path quartz cuvette was irradiated directly in the spectropho-tometer from the top of the cuvette. The light originated from a halogen lamp (EKE 21 V 150 W, General Electric) with a flexible light guide to which optical filters were attached. The 12.5-mm diameter interference band filters (Andover Corporation) with center wavelengths of 694.3 and 767.6 nm and a 50% bandwidth of 9.8 Ϯ 0.5 nm were used to generate red and far red light, respectively (approximate fluency 3.5 mol m Ϫ2 s Ϫ1 ).
Enzymatic Assays-All enzymatic assays were performed at 30°C in a water bath. A standard reaction mixture (1 ml) contained 0.5-4.0 M enzyme in the PDE or DGC assay buffer (23). The protein kept in the dark was irradiated through a flexible light guide for 5 min prior to reaction with red or far red light, and the light was kept "on" for the duration of the assay. The reaction was started by the addition of 1:100 (v/v) of c-di-GMP for PDE assays or GTP for DGC assays. Aliquots (100 l) were withdrawn at different time points, mixed with 10 l of 0.2 M CaCl 2 (PDE reactions only), and boiled for 5 min. The precipitated protein was removed by centrifugation at 15,000 ϫ g for 5 min. The supernatant was filtered through a 0.22-m pore size filter (Millipore) and analyzed by reversed-phase high-pressure liquid chromatography (HPLC) as described earlier (23). Protein assays were performed using the BCA method (Pierce) with bovine serum albumin as the protein standard. Proteins were analyzed using SDS-PAGE or the Protein 200 Plus LabChip kit on an Agilent 2100 Bioanalyzer (Agilent Technologies).
Mass Spectroscopy-The 31-kDa band corresponding to the protein copurified with BphG1 was purified by size exclusion chromatography and run on SDS-PAGE. The protein band was excised from the gel and sent to Proteomic Research Services, Inc. (Ann Arbor, MI) for peptide fingerprinting and protein identification. There, the protein was subjected to in-gel trypsin digestion and MALDI-TOF mass spectroscopy analysis on a Voyager DE-STR instrument (Applied Biosystems). The m/z values of the trypsinolysis products were analyzed by ProFound software.
When purified, the BphG1 apoprotein was colorless, whereas the holoprotein had a blue-green color indicative of bound BV. By size exclusion chromatography, we determined that BphG1 was present in several forms, i.e. oligomers of the apparent molecular masses Ͼ443 kDa, and a presumed tetramer of the apparent mass of 390 kDa. A very small amount of BphG1 was present as an apparent monomer (92 kDa) (Fig. 2B).
The UV-visible spectrum of the BphG1 holoprotein in the dark was dominated by the absorption maximum of 709 nm, characteristic of the P r form (Fig. 3A). This suggests that P r is the ground state of BphG1. Upon irradiation with red light (694 nm) or with white light, we observed decreased absorption at 709 nm and a new absorption maximum at 755 nm (Fig. 3A). The peak at 755 nm corresponds to the P fr form. When the red light-irradiated BphG1 was exposed to the far red light (768 nm), the P fr form quickly reverted to the P r form. The difference spectrum (P r Ϫ P fr ) was similar to the difference spectra of the Bphs that function as protein histidine kinases (3,6,15,19) (Fig.  3A). It is known that the photoexcited forms of many phytochromes can convert to the ground state in the dark, and so could BphG1. It converted from the P fr to the P r form in the dark slowly, taking ϳ40 min for a 95% conversion (Fig. 3B).
The PAS domain of BphG1 contains a cysteine residue, Cys-18, which is positionally conserved among Bphs and is predicted to serve as the site of BV attachment (14,15). To test whether or not BV in the BphG1 holoprotein was bound covalently, we separated the holo-and apoproteins by SDS-PAGE, stained the gel with a Zn 2ϩ solution, and exposed it to UV light. We observed the zinc-dependent fluorescence of the holoprotein (but not the apoprotein), which indicated that BV was bound covalently (Fig. 3C). Therefore, in all respects, BphG1 behaved as a typical photochromic Bph.
BphG1 Has an Essentially Light-independent c-di-GMP-specific PDE Activity-To test for enzymatic activity of BphG1, we incubated the holoprotein with GTP and c-di-GMP, which are the substrates of the DGC and PDE activities, respectively. Incubation with GTP did not result in the formation of c-di-GMP, whether the reaction was performed in the dark or light (Fig. 2C). Longer term incubation with GTP yielded some GDP and P i (not shown), which is characteristic of the enzymatically inactive GGDEF domains (23). The incubation of BphG1 with c-di-GMP resulted in formation of linear dimeric GMP, the  NOVEMBER 17, 2006 • VOLUME 281 • NUMBER 46 expected product of the PDE reaction (Fig. 2D). No hydrolysis of cyclic mononucleotides (cAMP or cGMP) was detected (not shown). Therefore, BphG1 functions as a c-di-GMP-specific PDE.

Unorthodox Bacteriophytochrome Involved in c-di-GMP Turnover
We tested the effect of irradiation on the PDE activity of the holoprotein but found no such effect. We were puzzled as to why the activity of the photochromic protein would be lightindependent. Therefore, we undertook a significant effort to identify conditions where the PDE activity would show light dependence. We modified the composition and pH of the reaction buffer, varied the protein-to-substrate ratios, tested the effects of additional nucleotides on enzymatic activity, examined the light dependence of individual oligomeric forms of BphG1, purified and assayed the activity of the BphG2 holoprotein using plasmid pBphH15, replaced the Synechocystis sp. heme oxygenase with the R. sphaeroides heme oxygenase BphO1, and tested the light dependence of the BphG1-BphO1 complex. 3 The maximum difference in the rate of PDE reaction between the P fr and P r forms that we have observed was with the tetramer fraction of BphG1. However, even there the difference was marginal, i.e. ϳ25%, which is unlikely to be physiologically significant (Fig. 2E). It appears that photoreception is decoupled from the output activity of BphG1. Therefore, 3 M. Tarutina and M. Gomelsky, unpublished data. BphG1 functions as an essentially light-independent c-di-GMP-specific PDE.
The lack of light dependence prompted us to investigate the possibility that the PDE activity responds to the presence or absence of BV, which would make BphG1 a sensor of BV, not light (12). To test this possibility, we compared the PDE activities of the apo-and holoforms of BphG1. These activities turned out to be very similar (Fig. 2E). The lack of BV dependence confirmed our conclusion that the PDE activity is decoupled from photoreception. This remained at odds with photochromicity of the BphG1 holoprotein until a serendipitous discovery described below revealed the light-dependent properties of BphG1, which were completely unexpected.
The Protein Copurified with BphG1 Is Its EAL Domain-We noticed that an ϳ31-kDa protein was copurified with BphG1 through the Ni 2ϩ affinity chromatography, whether it was expressed as an apo-or holoprotein ( Fig. 2A, lanes 4 and 8). We decided to explore the nature of the copurified protein. We separated the 31-kDa protein from BphG1 by size exclusion chromatography (Fig. 4A) and subjected it to trypsin digestion followed by peptide fingerprinting by MALDI-TOF mass spectroscopy. To our surprise, the copurified protein was found to perfectly match (16 peptides matched; expectation 5 ϫ 10 Ϫ10 ) the sequence of BphG1. The utmost N-terminal peptide was identified as 664 GELFRPSLYEETTQLVELDNDMR. It corresponds to the linker between the GGDEF and EAL domains as determined by the Pfam protein domain data base. If we assume that Gly 664 is the utmost N-terminal residue in the copurified protein, then the expected size of the copurified protein would be 273 amino acids. This correlates with 31 kDa, the observed molecular mass of the copurified protein. The copurified protein is therefore composed of the linker between the GGDEF and EAL domains and the entire EAL domain. Apparently, the 97-kDa BphG1 protein undergoes partial cleavage in E. coli. The two cleavage species, 31 and 66 kDa, are visible on the overloaded SDS-PAGE gels (not shown).
Knowing that EAL domains are sufficient for PDE activity (27), we tested the activity of the 31-kDa fragment. We found that it indeed retained a c-di-GMP-specific PDE activity. It is noteworthy that, on the size exclusion column, the EAL domain fragment ran with an apparent mass of 58 kDa, corresponding to that of a dimer (Fig. 4B). The apparent dimerization is unlikely to be required for PDE activity, because individual EAL domains that we analyzed earlier existed predominantly in the monomeric forms (27). Therefore, the tendency to homodimerize must be somewhat specific to the EAL from BphG1 and/or to the linker between GGDEF and EAL. The relative specific PDE activity of the EAL domain fragment was ϳ30% of that of the full-length BphG1 (Fig.  4C). Puzzled by the observation that the enzymatic activity of BphG1 was concentrated in its EAL domain, we decided to investigate the function of the remainder of the protein PAS-GAF-PHY-GGDEF.   Fig. 1A, lane 4). B, size exclusion chromatography showing an apparent EAL domain dimer. C, kinetics of the PDE activity. NOVEMBER 17, 2006 • VOLUME 281 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 34755
The PAS-GAF-PHY-GGDEF Fragment of BphG1 Is a Lightdependent DGC-The PAS-GAF-PHY-GGDEF fragment was not copurified along with BphG1 as was the EAL domain fragment, because the His tag used for affinity purification is located at the C terminus of BphG1. We therefore engineered a C-terminal His tag to the PAS-GAF-PHY-GGDEF protein fragment (Fig. 1B, plasmid pPhyGGDEF) and overexpressed and purified this fragment in the holo-and apoforms ( Fig. 5A and not shown). We found that, similar to the full-length BphG1, the PAS-GAF-PHY-GGDEF holoprotein contained covalently bound BV and was photochromic. Its spectral parameters were identical to those of the full-length BphG1 (not shown). We tested the enzymatic activity of this fragment. Intriguingly, it displayed DGC activity (Fig. 5B). Furthermore, the DGC activity was strongly activated by light (Fig.  5C). Therefore, the PAS-GAF-PHY-GGDEF fragment functions as a bona fide photoactivated DGC. The PAS-GAF-PHY-GGDEF apoprotein showed no DGC activity suggesting that BV is essential for this activity (not shown).
Based on the observations presented above, we hypothesize that BphG1 has a potential to work as a "bifunctional" enzyme. Depending on conditions, it may act either as PDE, which degrades c-di-GMP essentially independently of light, or as DGC, which synthesizes c-di-GMP in the light-dependent manner. This makes BphG1 the first photochromic Bph that has a non-kinase light-dependent enzymatic output(s) and the first potentially bifunctional enzyme involved in c-di-GMP synthesis and hydrolysis. The switch in "enzymatic identity" from PDE to DGC employed by BphG1 may involve a cleavage of the C-terminal EAL domain.

DISCUSSION
In this report, we have presented spectroscopic and biochemical characterization of BphG1, the unconventional Bph from R. sphaeroides. The second, highly similar (97% identity) Bph protein present in this bacterium, BphG2, proved to be indistinguishable from BphG1 in vitro. We found that BphG1, which is different from the majority of Bphs in that it contains the GGDEF-EAL output module, responds to light in a manner similar to that of the Bphs containing kinase outputs. We also found that one of the two enzymatic activities of BphG1, DGC, is light-dependent in vitro, whereas the other one, PDE, is essentially light-independent.
Until recently, it has remained unclear to what extent the complex photosensory modules present in phytochromes can operate with the non-kinase outputs (8). A study by Chen et al. (34) on a plant phytochrome PhyB has shown that light-induced conformational changes transduced to an output PAS domain result in the exposure of the cryptic nuclear localization signal, which promotes translocation of PhyB to the nucleus. Our data on the PAS-GAF-PHY-GGDEF fragment of BphG1 suggest that light-induced conformational changes are efficiently transduced to the GGDEF domain to result in activation of the DGC activity. Therefore, photosensory modules of phytochromes can regulate different outputs. Hence, phytochromes are no exception to modular organization, which is common in sensory transduction proteins (35). It is reasonable to anticipate that many Bph-like proteins identified through microbial and environmental sequencing and containing nonkinase outputs, e.g. methyl-accepting domain, phosphatase 2C domain, GGDEF, and EAL, will prove to function as bona fide photoreceptors (36). These advances and the recent progress with structural characterization of the photosensory core of the Deinococcus radiodurans Bph (16) provide an exciting possibility to rationally design phytochrome-based photoswitches with desired outputs. It is noteworthy that the PAS-GAF-PHY-GGDEF fragment from BphG1 can already be employed for photoactivated c-di-GMP synthesis. It may be useful for studying the molecular mechanisms of action of this novel second messenger in bacteria or for investigating pharmacological effects of c-di-GMP in mammalian cell cultures and tissues (37)(38)(39).
In addition to extending our knowledge of phytochromes, characterization of the BphG1 protein provided new insights into the function and regulation of proteins containing tan- demly arranged GGDEF-EAL domains. These proteins represent a large fraction of all GGDEF/EAL domain proteins encoded in bacterial genomes, e.g. 38% in E. coli. To date, only a handful of them have been assayed in vitro (28,40), and all showed one enzymatic activity, either DGC or PDE. On the genetic level, many more GGDEF-EAL proteins were analyzed (41,42), yet there was no evidence that any of them were bifunctional. The lack of bifunctionality in a subset of the GGDEF-EAL proteins can be explained by our hypothesis that one of the two domains in the GGDEF-EAL tandem is enzymatically inactive due to mutations (27). However, BphG1 shows that potentially bifunctional enzymes involved in c-di-GMP turnover may also exist.
How are opposite enzymatic activities (c-di-GMP synthesis and hydrolysis) regulated in a bifunctional protein? BphG1 holds an unexpected answer to this question. It possesses an essentially light-and BV-independent PDE activity. However, it can be converted into a DGC upon removal of the EAL domain. Why is DGC activity not expressed in the full-length BphG1? How does removal of EAL reveal the cryptic DGC activity? Why is the PDE activity light-independent, while the DGC activity of a truncated protein light-dependent? Although precise answers to these questions will require additional experimentation, below we propose the "EAL lock" model that is consistent with all in vitro observations (Fig. 6).
Our model is based on three elements. (i) BphG1 functions as homotetramer. It is expected that the output domains, not the photosensory module, contribute the most to protein-protein interactions (43). (ii) The observation that the EAL domain plus the upstream linker forms a dimer identifies one of the proteinprotein interfaces in the BphG1 homotetramer. Note that dimerization is not required for PDE activity (27). (iii) The DGCs are known to work as homodimers. According to the x-ray structure of the DGC from Caulobacter crescentus, PleD, each GGDEF domain monomer is likely to bind one GMP moiety of c-di-GMP (26). Our earlier observations revealed that the individual GGDEF domains readily form homodimers. However, these homodimers are nonproductive, i.e. they have no or low enzymatic activity (23). A conversion of the nonproductive GGDEF homodimer into the productive one requires conformational changes in the input (sensory) domains (23,25). Based on this information, we propose that protein-protein interactions between the EAL domains of BphG1 restrict mobility of the upstream GGDEF domains, prevent the nonproductive-toproductive conversion of the GGDEF domain homodimer, and therefore preclude expression of the DGC activity. In effect, EAL domain interactions lock BphG1 in the PDE mode. We speculate that the EAL lock also explains the lack of light responsiveness of the full-length BphG1. When the EAL domain is removed, the GGDEF domains gain the mobility necessary to undergo the nonproductive-to-productive homodimer conversion. This conversion is subject to regulation by light.
Whether or not the EAL domain of BphG1 is cleaved off in its native host, R. sphaeroides has yet to be tested. Theoretically, additional ways to unlock the EAL lock may exist. For example, one may envision an additional protein that has high affinity to EAL and that may break the EAL-EAL interactions, which would allow greater mobility of the GGDEF domains. Such protein, if it existed, would be expected to inhibit the PDE activity of EAL upon binding. Otherwise, the PDE activity of EAL would cancel the newly released DGC activity. Currently, we know of only one protein that interacts with BphG1, i.e. its cognate heme oxygenase BphO1, which supplies BV to BphG1. We tested light responsiveness and PDE-to-DGC conversion of the BphO1-BphG1 complex and found no significant differences compared with the behavior of the BphG1 protein alone. 3 Therefore, the currently available data favor cleavage as a mechanism of unlocking the EAL lock. Does BphG1 have an autoproteolytic activity as was observed for the cyanobacterial Cph1 (44)? Is a protease involved, or does the EAL domain cleavage occur through a nonenzymatic mechanism? How