Phytochromes from Agrobacterium fabrum^†

Introduction

This review about Agrobacterium phytochromes is dedicated to Wolfgang Gärtner on the occasion of his 65th birthday. Wolfgang worked in different fields of photobiology. He studied rhodopsins 1, phytochromes 2 and LOV domain proteins 3, among other research fields. After his move from Freiburg to the Max Planck Institute in Mülheim in 1991, where he joined the group of Silvia Braslavsky, the red light photoreceptor phytochrome became his major research topic—after he had been focusing on rhodopsin and bacteriorhodopsin the years before. His research on blue light photoreceptors started later. Initially, Wolfgang's focus was on plant phytochromes. Detailed protocols for large-scale purification of oat phytochrome from the natural source and time-resolved spectroscopy were already established by Silvia Braslavsky 4, and the respective phytochrome samples were used for biophysical studies, for example, by several fast spectroscopy techniques 5-9. The cyanobacterial phytochromes became the second “Standbein” of Wolfgang's phytochrome research 10-15. As a chemist, Wolfgang contributed by synthesis of artificial chromophores for the sake of labeling wherever it was necessary 16, 17. Another major contribution was the establishment of recombinant expression of proteins 17. From a current perspective, these are standard techniques in photoreceptor research and there is no more group left which uses purified phytochromes from natural sources. The establishment of recombinant expression, the studies on the photocycle and the use of synthetic chromophores must be regarded as important pioneering work in the field of photoreceptor biochemistry.

Compared to the middle 1990s, our knowledge about phytochrome has now drastically broadened. We know that plant phytochromes enter the nucleus upon photoconversion 18, 19. We know a lot more about the roles of plant and fungal phytochromes in signal transduction 20, 21. We know that phytochromes exist in algal species which were not commonly known 22, 23. We know in which phylogenetic groups phytochromes exist 24 and where they do not exist, for example, in red algae and animals. And we know about photoreceptors with phytochrome-like properties like a special group of chromoproteins in cyanobacteria, called cyanobacteriochromes 25-28. The characteristic feature that led to the discovery of plant phytochromes 29, photoreversibility (see Figs. 1 and 2), holds also for fungal 30, bacterial 31, 32, prasinophyte 22, or heterokont 33 phytochromes. The cyanobacteriochromes have copied from canonical phytochromes the principle of a photoreversible switch between two spectrally distinct forms but have modified their spectral features 28, 34. However, despite the evolutionary flexibility of the cyanobacteriochromes, in terms of the diversity of spectral features and domain arrangements that can be found within this group of photoreceptors, phytochromes were more successful in eukaryote evolution.

This review is focused on phytochromes Agp1 and Agp2 from Agrobacterium fabrum (former: Agrobacterium tumefaciens) 32, 35. We discuss our work in the context of phytochrome studies on bacterial, cyanobacterial, plant and fungal phytochromes. Agrobacterium fabrum is an important model organism for gene transfer from bacterium to plant 36 and many other issues; both its phytochromes can be expressed in recombinant systems in high amounts; these phytochromes are representatives of two major subgroups of phytochromes; and biophysical analyses have provided good insight into their molecular functions. We start off with a discussion on the evolution of phytochromes, which is still an open issue in phytochrome research.

Evolution of Phytochromes

Domain arrangements in bacterial, plant and fungal phytochromes are comparable insofar, as the three N-terminal domains are always PAS, GAF and PHY domains. The C-terminal domains are more divergent, and some phytochromes have pronounced N-terminal extensions (see Fig. 3 for domain arrangements of some phytochromes and related proteins). The PAS-GAF-PHY tridomain or photosensory core module (PCM) incorporates the chromophore, undergoes light-induced photoreversible absorbance changes and controls the state of the rest of the protein. The PCM is also the fraction of the entire protein of which several X-ray structures could be obtained. Phylogenetic studies on the overall evolution of phytochromes are most often based on PCM sequences.

Although such studies provide many clear results, they also leave several open issues, such as the evolutionary origin of eukaryotic phytochromes. The amino acid sequences of bacterial phytochromes are very diverse; the most distantly related sequences have amino acid identities of 15%. This broad diversity is combined with diverse domain arrangements. The major domain pattern of bacterial phytochromes is that with a PCM coupled to a histidine kinase, but combinations with GGDEF, EAL, PAS and STAS domains are also possible. Bathy phytochromes, which in contrast to other phytochromes have a Pfr ground state or dark adapted state (Fig. 1), are found in a separate group; these proteins are also characterized by a C-terminal response regulator as an intrinsic module 37. This broad diversity of bacterial phytochromes reflects the overall diversity of bacterial proteins which arose during an evolution of 3.5 billion years. Phytochromes of eukaryotic groups, which are fungi, heterokonts (including diatoms) and archaeplastida (comprising the plants), are less diverse 24, 38, 39. All plant phytochromes are of the PCM-PAS-PAS-histidine kinase type (Fig. 3), except the neochromes that exhibit some domain rearrangement. Fungal phytochromes show a PCM-histidine kinase-response regulator domain arrangement (Fig. 3). The archaeplastida and the fungi form clear monophyletic groups 24, 38, 39, 22; within the group of plant phytochromes, the PhyA/C and PhyB/C/D branches can be clearly distinguished. An evolutionary hallmark is the chromophore-binding Cys residue. This Cys is found in the N-terminus of the PAS domain in those phytochromes which have a biliverdin (BV) chromophore (see Fig. 1 for chromophore structure and Fig. 3 for domain arrangement), that is, most bacterial, fungal and heterokont phytochromes. Agrobacterium Agp1 was in fact the first phytochrome for which this binding site was identified by mutagenesis 32 and mass spectrometry 40. If phytochromobilin (PϕB) or phycocyanobilin (PCB) is used as chromophore (see Fig. 1 for structures), the chromophore-binding Cys lies in the GAF domain. In both cases is the respective Cys highly conserved, and it is also mutually exclusive 40. The choice of a bilin with a ring A ethylidene side chain as chromophore and the chromophore-binding Cys in the GAF domain correlate in archaeplastida and typical cyanobacterial phytochromes. This speaks for a cyanobacterial origin of archaeplastida phytochromes and thus for an engulfment of the phytochrome gene via the plastid endosymbiont. However, a change of a single amino acid could as well have happened independently—in the cyanobacterial and in the archaeplastida lines. It has indeed been reported several times that the introduction of a cysteine at the right position of the GAF domain of a BV-binding phytochrome results in covalent attachment of a PCB chromophore 32, 11. In most published phylogenetic trees, plant phytochromes do not appear as subgroup of cyanobacterial phytochromes 39, 22. The construction of a phylogenetic tree has many degrees of freedom: choice of sequences, alignment algorithms and parameters, or choice of phylogeny program and the respective parameters. All these vary from one study to the other. We have constructed many different trees using different algorithms in order to improve the quality of the predictions. In many cases, we obtained trees in which cyanobacterial and archaeplastida phytochromes appeared as sister groups, as long as the trees were constructed with the PCM sequences. A tree in which such a relationship is given is shown in Fig. 4. We therefore favor a close phylogenetic relationship between archaeplastidal and plant phytochromes. This is in line with a pathway via primary endosymbiosis. In studies where the archaeplastidal phytochromes branch at other positions of the phylogenetic tree, the bacterial groups closest to the archaeplastida consistently differ between individual studies. In other words, there is no consensus alternative to the cyanobacteria that would be supported by two independent studies. Such studies are probably based on misalignments or other “errors”.

The histidine kinase modules of phytochromes can also be used for phylogenetic studies, because most phytochromes have such modules or share some homologies with these. Our results show a completely different evolutionary pattern for the histidine kinase (see Fig. 4 as example). In those analyses we never observed a close relationship between plants and cyanobacteria. Our conclusion was that the PCMs and histidine kinase modules have undergone several rearrangements in the evolution 24. Such a rearrangement can explain the insertion of two PAS domains between PCM and histidine kinase as found in plant phytochromes. There are many protein sequences in the databases with multiple PAS domains and a histidine kinase (see bottom of Fig. 3 as example). A fusion of the phytochrome PCM with a fragment of such a protein could have generated the first typical plant phytochrome.

The origin of the other eukaryotic phytochrome PCMs can be discussed in a similar manner. Heterokonts (encompassing diatoms and brown algae) are derived from secondary endosymbiosis 41. These phytochromes could either derive from the primary (cyanobacterial) and the secondary endosymbiont (probably red alga) or from the host of this endosymbiont. The latter alternative is the more likely one because heterokont phytochromes have a BV-binding site 33, and in phylogenetic analyses they never appear related to cyanobacterial or archaeplastidal phytochromes. The domain arrangement with a C-terminal response regulator domain points also to a rearrangement of PCM and C-terminus. Bacterial phytochromes with an intrinsic response regulator domain have also a HWE histidine kinase domain, which is not too closely related to the corresponding domains in heterokont phytochromes.

As fungi do not have plastids and are also not derived from ancestors that contained a plastid, the pathway via plastid endosymbiosis was not possible. Accordingly, these phytochromes also have a BV-binding site and a C-terminal response regulator domain, a chromophore and a domain arrangement as found predominantly in non-cyanobacterial bacterial species. Fungal and heterokont phytochromes could have the same prokaryotic origin, and in a recent study, they do indeed appear as sister groups 33. The last eukaryotic group which has phytochromes, the amoebozoa, encompasses the slime molds 42, 43. This group is the sister group of ophistokonts in which fungi, animals and few other subgroups are united. The evolutionary position of slime mold phytochromes would therefore be highly interesting. Presently two slime mold phytochrome sequences are known, both from Physarum polycephalum 11, 42. In our own preliminary studies, however, it was not possible to perform reliable alignments with these sequences.

Phytochromes of Agrobacterium fabrum

The positions of Agp1 and Agp2 in the phylogenetic trees are indicated by blue boxes in Fig. 4. The two phytochromes of A. fabrum are phylogenetically not closely related to each other. Agp1 is a normal bacterial phytochrome with a prototypical domain arrangement. Depending on the topic, it can either be taken as representative of the majority of bacterial phytochromes which have a Pr ground state and a light regulated His kinase, or as a model for all phytochromes, if the mechanism underlying intramolecular signal transduction within the PCM is considered. The unusual domain arrangement of Agp2 with a HWE histidine kinase and a response regulator is found in several other phytochromes of rhizobiales 37, a group of related α-proteobacteria that live in the soil. As noted above, Agp2 belongs to the bathy phytochromes which have a Pfr dark form (Fig. 1). In our phylogenetic studies, the PCM and the histidine kinase module of Apg2 are placed quite at the base of the trees (Fig. 4); Agp2 could be an ancient form of bacterial phytochromes.

Biological Functions

Phytochromes have been discovered in plants, where they are the dominating photoreceptors that participate in the control of almost any light dependent developmental step 44; seed germination, induction of photosynthetic pigments, elongation growth, stomata development to mention just a few examples. Fungal phytochromes that have been discovered later are known to control transition of sexual to asexual development and the induction of stress responses 21, 45; phytochrome dependent responses of diatoms and slime molds 43 are also known 33. In comparison to the broad variety of phytochromes in bacteria, knowledge of the respective biological responses is rare. The role of cyanobacterial phytochromes is unclear, although many responses controlled by cyanobacteriochromes are known. In prokaryotes, the first phytochrome effect was found in photosynthetic proteobacteria. The synthesis of bacteriochlorophyll and photosystem complexes is light regulated via phytochrome in Bradyrhizobium and Rhodopseudomonas palustris species 46. Bradyrhizobium and Agrobacterium share a closer relationship within the Rhizobiales. Other related phytochrome effects have been found in these bacteria 47 and the swarming of Pseudomonas syringae is also mediated by phytochrome 48, 49.

In Agrobacterium fabrum we have studied light regulation of wild type and mutants of differential gene expression, motility, infectivity (T-DNA transfer from bacterium to plant) and conjugation (plasmid DNA transfer among bacteria). Only eight out of 4500 genes showed differential expression between light and dark. This difference was lost in an agp1⁻ agp2⁻ double mutant 50. A light effect on swimming was attributed to a photolyase which also functions as a photoreceptor here, not to phytochromes 51. Tumor induction of Arabidospis root segments was found to be down regulated by red and far-red light, and in phytochrome knockouts, there was very weak infection only. This points to a phytochrome effect on light regulation of infection 50. A clear phytochrome effect was found for the conjugation of Agrobacterium fabrum. Driven by results of a computational study 52, a coaction of phytochrome with the TraA protein was suggested 53. Conjugation is a process during which DNA in a single-stranded form is transferred to another cell. TraA is a relaxase which is thought to initiate the formation of single-stranded DNA. Red and far-red light diminished conjugation of Agrobacterium fabrum wild type donor cells (see also Fig. 5 as an example 53). Phytochrome knockout donor strains also showed a reduced DNA transfer efficiency. The double knockout donor did not undergo conjugation at all. The evolutionary advantage of such a light inhibition of conjugation could be a protection against UV damage. Single-stranded DNA is more sensitive against DNA damage than double-stranded DNA, because the most efficient repair requires the complementary strand—conjugation is one example in which DNA is in a single-stranded mode. The T-DNA transfer to plant cells occurs also in a single-stranded mode. Although the effect of Agp1 and Agp2 on differential gene expression in Agrobacterium fabrum seems much less prominent than in plants or fungi, both phytochrome effects are processes in which DNA, but solely plasmid DNA, is processed.

Biochemical Studies

Both Agp1 and Agp2 have been subject of a number of biochemical and biophysical studies. Using locked chromophores, the stereochemistry of the chromophores in the Pr and Pfr states of Agp1 and Agp2 was proposed to be ZZZssa for Pr and ZZEssa for Pfr 54-57. (A cartoon on the universal phytochrome photocycle and the stereochemistry of the chromophore in the Pr and Pfr forms is depicted in Fig. 2.) Adducts with locked chromophores that cannot undergo stereochemical changes at the methine bridge between rings A and B exhibit light-induced spectral changes, but the spectra of the photoproducts differ largely from Pfr 56, 58. Therefore, structural changes involving ring A are probably also required during photoconversion. Rearrangements of the local environment around ring A are also postulated by fluorescein labeling at the chromophore-binding Cys20 of Agp1, combined with time-resolved spectroscopy: spectral changes in fluorescein were observed upon chromophore incorporation and by photoconversion 59. These measurements can be correlated with proton release and uptake that has also been measured using fluorescein as indicator 60. During Pr-Pfr photoconversion, proton release is related to the formation of meta-Rc and proton reuptake corresponds with the meta-Rc to Pfr transition. Polarity change of the ring A environment is also observed during the meta-Rc to Pfr transition.

The early time steps that lead to the formation of Pfr have been investigated for both Agp1 and Agp2 by ultrafast spectroscopy 61-63. During measurements on the Agp1 Pfr to Pr photoconversion an interesting and yet unresolved observation was made. The overall photoconversion from Pfr to Pr was found to have a very low quantum yield of 0.4% 32, whereas the quantum yield for photoisomerization, the first step in Pfr-Pr conversion, was found to be in the same range as observed for other phytochromes 63. It therefore seems that after E-Z photoisomerization, a major fraction of chromophores isomerizes back to the E configuration. The chromphore structures in the Pfr states of various bathy and prototypical biliverdin-binding phytochromes were studied in more detail using a combined spectroscopic-theoretical approach 64. A special feature of the bathy phytochrome Agp2 is its spectral property in the Pr form. Unlike other phytochromes, the Pr spectra are pH sensitive, indicating chromophore deprotonation at alkaline pH 65. A reaction scheme starting from the excited state of the Pr chromophore for Agp2 under acidic/alkaline conditions was deduced from the ultrafast measurements 61. Within 1.5/6 ps after excitation a twist around C5 of the chromophore is induced. During the next 22/58 ps another twist around C15 occurs; during this step the deprotonated chromophore becomes protonated at alkaline pH. From here, the Z-E isomerization to lumi-R or the back reaction to Pr occurs. This bifurcation of the reaction pathway and the distribution between both pathways are the determinants of the photoconversion quantum yield.

Proton exchange reactions during photoconversion were also analyzed by Resonance Raman spectroscopy in solution 65-67 and on crystals 68. In a recent study on the bathy phytochromes Agp2 and Pseudomonas aeruginosa phytochrome PaBphP, a proton exchange reaction between a conserved histidine and the chromophore was found which enables a keto enol tautomerism of ring D that drives the thermal back reaction (dark reversion) from Pr to Pfr 69.

An interesting temperature effect on kinase activity and spectral properties was found for Agp1: with increasing temperature from 20°C to 35°C the autophosphorylation activity dropped to almost zero. The maximum activity is found at 20–25°C 70, 71. A similar effect was observed in cyanobacterial phytochrome Cph1 but with a temperature optimum at 5°C; that is, the colder the temperature, the more active the kinase. The tested temperatures are in the physiological range; thus, the observed effect can be expected to be relevant for the in vivo situation as well. Such a temperature effect has also been described for other histidine kinases such as the VirA protein that regulates infectivity and motility in Agrobacterium fabrum and a thermosensor two-component system of an Antarctic Archaebacterium 72. At 35°C, the photoconversion properties of Agp1 did not follow the general scheme, because besides Pr and Pfr, a third form exists with a bleached Pr-like spectrum that is formed during prolonged irradiation from Pfr. This temperature effect is dependent on the histidine kinase and not found in truncated Agp1-M15. The results imply that phytochrome could not only act as light but also as temperature sensor. In this context, a simple experiment on Arabidopsis wild-type and phytochrome mutants was performed to see whether the phytochrome in this plant also responds to temperature. An Arabidopsis mutant defective in phytochrome B had a shorter hypocotyl than the wild type when grown at 32°C, but not at 29°C or 23°C. This experiment was performed in darkness 71. It was suggested that plant phytochrome could serve as thermosensor also in darkness 71. Phytochrome-mediated temperature effects in Arabidopsis seedlings were indeed found subsequently 73, 74.

Crystal Structure

There is a long list of crystal structures of bacterial phytochromes (see, for example, PDB entries 1ZTU, 2O9C, 2O9B, 3C2W, 3IBR, 3G6O, 3NHQ, 3S7P, 3S7O, 3S7N, 3S7Q, 3NOP, 3NOU, 3NOT, 4IJG, 4O0P, 4O01, 4CQH, 4GW9, 4E04, 4QOJ, 4QOI, 4QOH, 4O8G, 4Y5F, 4Y3I, 4XTQ, 4ZRR, 5Z1W, 5C5K, 5AJG, 5AKP, 4RQ9, 4RPW, 5HSQ, 5I5L, 5LBR, 5L8M, 5K5B, 5MBP, 5MBO), plant phytochrome (PDB entry 4OUR), or cyanobacterial phytochromes (PDB entries 2VEA, 3ZQ5, 3VV4, 4GLQ, 4BWI, 5DFY, 5DFX). These structural data are in some cases supported by electron microscopy 75, 76, small angle X-ray scattering (SAXS) 77-79 and NMR measurements on the chromophore 16, 80-85 or on the entire protein (PDB entries 2K2N, 2KOI, 2KLI, 2LB9, 2LB5, 2M7U and 2M7V). Almost all crystal structures are from protein fragments and not from full-length proteins. These fragments encompass the PAS-GAF-PHY tridomain or PCM (Photosensory Core Module), PAS-GAF or GAF-PHY domains or, as in the cyanobacteriochromes, GAF domains. One structure of a longer fragment 86 (PDB entry 4E04) and another structure of a full-length phytochrome 87 (PDB entry 5AKP), both with atypical C-terminal domains, are known as well, but there is no full-length structure of a phytochrome that contains a His kinase as part of the C-terminal region. Because there are PCM structures of bathy phytochromes in the Pfr form, normal phytochromes in the Pr form and—in the case of DrBphP—also in both Pr and Pfr forms, structural changes that occur during photoconversion can be well described for the PCM. His kinase structures have on the other hand been obtained from several other model proteins 88. It is nevertheless still unclear how the signal is transmitted from the PCM to other parts of the protein and how, for example, the enzyme activity is modulated.

Three-dimensional structures of homologous proteins or protein domains are always highly similar. This is also true for the protein folds of the different PCMs. The common most relevant features are (also see summary in Table 1): (1) The chromophore pocket is mainly formed by the GAF domain, in which the chromophore-binding amino acids are more conserved than those further apart from the chromophore. The chromophore-binding site of cyanobacterial and plant phytochromes (which bind a phycocyanobilin or phytochromobilin chromophore, respectively) is also part of the GAF domain; (2) The N-terminal PAS domain interacts with the GAF domain in an unusual manner: The N-terminus of the PAS and the N-terminus of the GAF form a so-called figure of eight knot. In cyanobacteriochromes that lack the PAS domain (e.g. 89), such a knot cannot be formed. However, these chromoproteins are also photoactive, indicating that the knot is not required for photoreversibility; (3) In superposed structures, the chromophore-binding Cys of typical bacterial phytochromes (at the N-terminus of the PAS domain) comes very close to the chromophore-binding Cys of plant and cyanobacterial phytochromes. The orientation of the chromophores within the proteins is thus very similar in all PCM structures, no matter which chromophore is incorporated; (4) Crystal structures together with NMR and vibrational spectroscopy confirmed the stereochemistry of the chromophores in Pr and Pfr forms (Fig. 2B) to be the same as suggested by the studies with locked chromophores; (5) The PHY domain may be regarded as a mediating domain in intramolecular signal transduction; it is linked with the histidine kinase or other output domains on its C-terminal end and linked with the chromophore-binding GAF domain at its N-terminus. The so-called tongue of the PHY domain is a region that folds back onto the chromophore pocket, also resulting in an interaction of the PHY with the PAS domain; and (6) Other interesting features of phytochrome structures are the long helices, one connecting GAF and PHY domains and the other connecting the PHY domain with the His kinase module. These helices could have evolved to establish structurally stable domain connections that restrict relative movements of the domains within narrow ranges and put the domains at their correct positions (although the GAF-PHY connecting helix is not rigid). Moreover, structural changes as triggered by light via the chromophore could be stabilized by these long helices, which according to this view should play central roles in intramolecular signal transduction from the sensory module to the output module.

When PCM structures of normal phytochromes in the Pr form and bathy phytochromes in the Pfr form became available, comparative analyses revealed interesting differences in the overall folds and conformations between these two types of proteins (see Table 2 for a summary) 90, 91. Based on these observations, it was possible to speculate about conformational changes that occur during Pr-Pfr photoconversion. Crystal structures of Deinococcus PCM in the Pr and Pfr forms showed that the observed differences between normal and bathy phytochromes do indeed reflect the protein conformational changes that occur during photoconversion 78, 92, 93. In the present view, two segments of the tongue that form an antiparallel β-sheet structure in the Pr form swap their positions relative to the GAF domain during photoconversion as described by a “tryptophan switch” mechanism 89. The same segments refold into an α-helix and a loop during this transition. In addition, the long helix between the GAF and the PHY domain is bent in the Pr and more stretched in the Pfr form.

Although crystals of “Agp1-PCM” diffracting to 3.4 Å resolution were available rather early 94, only a preliminary structural model could be obtained initially as crystallographic refinement was unsatisfactory, even with improved crystals yielding diffraction data sets that were suitable to 2.7 Å resolution 95. A method of protein surface engineering termed “surface entropy reduction” (SER) 96 changed the situation 95. In this approach, clusters of amino acids with long and flexible side chains, Glu and Lys, are replaced by Ala residues to enable the formation of crystal contacts that cannot form with the wild-type protein. Several “SER” mutants yielded similar crystallization results as the wild type, but with the mutant Agp1-SER13-PCM, a combination of two cluster replacements, crystals were obtained under different conditions. With these “Agp1-SER13-PCM” crystals that belonged to a different space group and had different unit cell parameters than the crystals of the wild-type protein, diffraction data could be collected to a resolution of 1.85 Å, which is the best resolution for a PCM crystal so far (Fig. 6). Moreover, using the 3D model of Agp1-SER13-PCM as search model for molecular replacement allowed improving the wild-type Agp1-PCM model significantly such that it could be refined satisfactorily to 2.7 Å resolution. Both crystal structures of Agp1-PCM and Agp1-SER13-PCM differ in the arrangement of two subunits within a dimer. In Apg1-PCM, both subunits are arranged in a parallel manner, whereas an antiparallel arrangement was obtained for Agp1-SER13-PCM (Fig. 7A,B). Apparently, the replacement of amino acids in Agp1-SER13-PCM and the use of different precipitants caused destabilization of the parallel dimer interface while at the same time enabling formation of a new crystal contact, which led to this switch from parallel to antiparallel 95.

The overall structures of the monomers in the crystals of the wild-type and the mutant proteins are very similar, but there are differences at a more detailed level (Fig. 7C). The long helix that connects GAF and PHY domains is more stretched in the Agp1-PCM and more bent in the Agp1-SER13-PCM structure. The different bending must be compensated in another part of the protein because the tip of the tongue interacts with the chromophore pocket of the GAF domain at the same position. This compensation is found in two hinges, one on each side of the tongue itself, next to Trp445 and Trp468, respectively, the two Trp residues that according to the “tryptophan switch” mechanism should swap their roles in packing against the GAF domain after photoconversion. The bending of the long helix is also clearly different between Pr and Pfr crystal structures; in all former cases, the long helix is more bent (Fig. 8). This holds true also for Deinococcus phytochrome which has been crystallized in the Pr and Pfr forms 78, 92, 93. Structural comparisons led to the conclusion that Pr-Pfr photoconversion results in a stretching of the long helix (Fig. 7), combined with a refolding and shortening of the tongue. This view is consistent with the differences observed between the structures of different phytochromes in their dark states, that is, normal phytochromes in the Pr form and bathy phytochromes in the Pfr form 90, 91. All Pr and Pfr structures point to a rearrangement of the secondary structure of the tongue. In the Pr form, as in the two Agp1-PCM structures, there are two antiparallel β-strands at the base of the tongue and extended loops (in DrBphP-PCM) or a β-hairpin (e.g. in Agp1-SER13-PCM) in the tip of the tongue. In the Pfr form, the tongue folds into an α-helix and an extended loop (Table 2).

The high-resolution structure of Agp1-SER13-PCM shows also a close interaction with Pro461 and the ring A of the chromophore (Fig. 9A). As in the same structure the chromophore adopts a sterically strained conformation at ring A, it was suggested that this steric strain is induced by the interaction with Pro461 which belongs to the conserved PRxSF motif in the tip of the tongue 95. A similar steric strain of the chromophore has not been seen in the structures of other phytochromes, but this may be due to the lower resolutions of the respective crystal structure analyses. Consistent with our observation, however, the Pro of the PRxSF motif is in van der Waals contact with ring A in all Pr structures. In Pfr structures of other phytochrome PCMs, the homologous Pro residue interacts with ring D of the chromophore (Fig. 9B). There is sufficient evidence to suggest that Pro461 (or its homologs) flips between rings A and D during Pr-Pfr photoconversion. While the Arg and Ser residues of the PRxSF motif serve to stabilize the protein conformation in either the Pr or Pfr state (Fig. 9A,B), respectively, we propose that the steric strain induced by the Pro of the same motif in the Pr state drives conversion into the Pfr state after photoexcitation, which is consistent with formation of a bleached photoproduct after red light illumination of the P461A mutant of Agp1 95. This Pro residue may be regarded as a sensor of initial structural changes in the local environment of pyrrole ring A of the chromophore after photoexcitation of the Pr form. Equally, this residue may sense structural changes caused by photoexcitation of the Pfr state in the ring D environment and drive conversion into Pr.

Integrating Recent Studies on Conformational Changes

As always, new answers impose new questions. Although the structural differences between Pr and Pfr may be considered as settled, it is as yet not clear how these changes are triggered by the light-induced isomerization of the chromophore. A more detailed understanding of the photocycle intermediate structures would be crucial in this regard. The mechanism of signal transmission from the N-terminal PCM to the rest of the protein, for example, the histidine kinase module, and in particular of modulation of histidine kinase activity, is also still unclear. Because typically, phytochromes are homodimeric proteins, changes in the arrangement of both subunits could be important for intramolecular signal transduction. In a recent interdisciplinary study encompassing X-ray crystallography and time-resolved small angle X-ray scattering (SAXS), it has been proposed that within the PCM dimers, the PHY domains separate as a result of Pr to Pfr photoconversion 78, 97. It was suggested that the same change of quaternary structure could trigger conformational changes in the His kinase module of the full-length protein. Indeed, the X-ray structures of the DrBphP-PCM in Pr and Pfr 78, 92, 93 show such a disappearance of the dimer contacts between the PHY domains in the Pfr form. Although both published crystal structures of DrBphP-PCM in the Pfr state or Pfr-enriched state showed a similar Y-shaped subunit arrangement, which was also supported by the SAXS experiments, the Pfr crystal structures of bathy phytochromes exhibit a different subunit arrangement 91. Therefore, without consideration of additional experimental evidence, the subunit arrangements in the PCM crystal structures should not be taken as representative for the structure of the full-length protein in solution. In protein crystal structures, each protein monomer is always in contact with several other monomers; it is often difficult if not impossible to distinguish between biologically relevant protein–protein interfaces that are part of a quaternary structure and those interfaces which solely exist as crystal contacts. Moreover, there is another discrepancy between the data of Takala et al. 78 and former biochemical experiments on the quaternary structures of the cyanobacterial phytochrome Cph1 and of Agp1. The PCM of both bacterial phytochromes was found to be monomeric in the Pr and dimeric in the Pfr form 98-100; Takala et al.78 assumed the DrBphP-PCM to be a dimer in Pr and Pfr. A recent time-resolved SAXS study on full-length DrBphP phytochrome showed indeed a different result for light-induced protein conformational changes 79. According to this new model, the intramolecular signal is relayed by structural changes in the tongue and a twisting of the long helices connecting the PHY domain with the histidine kinase. The large number of degrees of freedom in the modeling of SAXS data demands the use of other structural techniques suitable to characterize light-induced conformational changes in full-length phytochromes. In a recent study on spin-labeled full-length Agp1 (103), distances between two subunits in the dimer were estimated by PELDOR. The position of the spin label was either in the GAF, PHY, or histidine kinase region. These measurements showed that the arrangement of subunits is different from that in the PCM crystal structure, whereas the distances in the histidine kinase region were in agreement with a homology model. No distance changes upon Pr-Pfr photoconversion were found. These results show that the arrangement of subunits does not undergo large changes upon photoconversion. Clearly, for maximal understanding of phytochrome action, more studies on full-length phytochromes are required. Differences between PCMs and full-length proteins should be considered more in the future.

Conclusions

The phytochrome system of Agrobacterium fabrum allows integrated research on light regulation in a bacterium. We know a lot about the biochemical properties of both phytochromes, we know the crystal structure of Agp1-PCM, and we know two biological functions of these photoreceptors. We propose that the different steps in signal transduction can be unraveled in the near future.