The PHY domain dimer interface of bacteriophytochromes mediates cross-talk between photosensory modules and output domains

Abstract

Protein dynamics play a major role for the catalytic function of enzymes, the interaction of protein complexes or signal integration in regulatory proteins. In the context of multi-domain proteins involved in light-regulation of enzymatic effectors, the central role of conformational dynamics is well established. Light activation of sensory modules is followed by long-range signal transduction to different effectors; rather than domino-style structural rearrangements, a complex interplay of functional elements is required to maintain functionality. One family of such sensor-effector systems are red-light-regulated phytochromes that control diguanylate cyclases involved in cyclic-dimeric-GMP formation. Based on structural and functional studies of one prototypic family member, the central role of the coiled-coil sensor-effector linker was established. Interestingly, subfamilies with different linker lengths feature strongly varying biochemical characteristics. The dynamic interplay of the domains involved, however, is presently not understood. Here we show that the PHY domain dimer interface plays an essential role in signal integration, and that a functional coupling with the coiled-coil linker element is crucial. Chimaeras of two biochemically different family members highlight the phytochrome-spanning helical spine as an essential structural element involved in light-dependent upregulation of enzymatic turnover. However, isolated structural elements can frequently not be assigned to individual characteristics, which further emphasises the importance of global conformational dynamics. Our results provide insights into the intricate processes at play during light signal integration and transduction in these photosensory systems and thus provide additional guidelines for a more directed design of novel sensor-effector combinations with potential applications as optogenetic tools.

Introduction

Proteins are key players in a huge variety of biological processes of all organisms and their functions range from storage or support and scaffolding to regulatory aspects (hormones, receptors and transport) all the way to catalysis of a plethora of enzymatic activities. In order to carry out many of these functionalities, it has become more and more evident that proteins are highly dynamic complexes, rather than static actors. The dynamic nature of proteins ranges from the biological role of intrinsically disordered proteins [1] to the general and ubiquitous fluctuation between conformational substates [2] as well as the movements of individual chemical bonds and bond vectors [3]. In addition, the importance of conformational diversity has also been put forward for the evolvability of proteins in general [4]. Signalling proteins, in particular, frequently depend on minute differences between substates triggered by various stimuli [5]. In multi-domain signalling proteins, an intricate interplay of individual domains has frequently been observed. With the widespread modularity of multi-domain signalling systems, it is of special interest to understand how the dynamic properties of singular domains affect the global dynamics of the system and ultimately its biological function. Rather than rigid either-or systems that turn off and on upon a certain stimulus, biological output is often affected by complex interactions between various structural features and the signal integration mechanism at play can rarely be linearized [6] and assigned to specific interactions or single domains.

One group of signalling proteins that has recently attracted considerable attention are photoreceptors [7]. Especially red-light-sensing phytochromes promise to be valuable tools for cell biology and medical applications [8], [9], [10]. One interesting subfamily of red-light-sensitive photoreceptors are biliverdin-linked bacteriophytochromes [9], [11] with a variety of biologic output functionalities. Among the bacteriophytochromes the group of phytochrome activated diguanylate cyclases (PadCs) has recently provided interesting insights into structural aspects of red-light-associated diguanylate cyclase (DGC) activity. These phytochrome-activated DGCs respond to the light-induced isomerisation of the covalently attached biliverdin IXα chromophore from ZZZssa to ZZEssa with extensive structural rearrangements around the cofactor [12]. Ultimately, this affects output functionality, where DGC activity is characterized by the condensation of two molecules of GTP to cyclic dimeric GMP (c-di-GMP), a bacterial second messenger involved in crucial lifestyle decisions [13].

PadCs feature a PAS (Period/ARNT/Single-minded) – GAF (cGMP phosphodiesterase/adenylate cyclase/FhlA) – PHY (phytochrome-specific) architecture of the photosensory module (PSM) and a GGDEF output domain (Figure 1). The GAF domain provides the cofactor binding pocket and the PHY domain stabilizes the light state by way of a tongue element that plays an essential role in signal transduction [14], [12], [15]. Another structural feature that has recently attracted particular interest is the N-terminal segment, short NTS, which covalently binds biliverdin via a conserved cysteine residue and is involved in a distinct figure-eight knot-like structure that reduces the degrees of freedom in PAS-GAF interactions [16], [15]. Sensor and effector are joined by a characteristic coiled-coil linker, which comprises a variable number of amino acids in different family members. In fact, this “linker length” has emerged as an important feature, with its functional implications strongly dependent on the presence and composition of coiled-coil heptad repeats. The prevalent linker lengths differ by multiples of seven residues, but untypical linker lengths can be identified as well. Phylogenetic analysis of PadC homologs revealed a strong subfamily clustering according to linker length (Figure 1), further highlighting the importance of this evolutionary playground with regard to controlling PadC regulation – the majority of all PadCs identified to date feature linker length differences of -7, 0, +7 or +14. These numbers, indicative of classical heptad repeat patterns, suggest a similar mechanism of signal transduction as proposed for the prototypic member of these families, IsPadC [17], in all organisms featuring continuous coiled-coil sensor-effector linkers. However, preliminary spectral and enzymatic analyses of representative family members showed that homologs with linker lengths not equal to multiples of seven do not function as light-activated diguanylate cyclases. These observations prompted a closer investigation of the extent to which the linker length affects protein characteristics and behaviour.

Figure 1. Phytochrome-activated diguanylate cyclases.

(a) Full-length crystal structure of IsPadC in its dark-adapted dimeric form [12] in cartoon representation: N-terminal segment (NTS) = grey, Period/ARNT/Single-minded (PAS) = purple, cGMP phosphodiesterase-Adenylate cyclase-FhlA (GAF) = blue, phytochrome-associated (PHY) = green, coiled-coil (cc) linker = yellow, diguanylate cyclase (GGDEF / DGC) = red. The bound biliverdin cofactor is shown as stick model. (b) In a family tree of PAS-GAF-PHY-GGDEF systems, PadCs cluster according to linker length: The number of amino acid residues between two positions after a conserved leucine (Ile 495 in IsPadC) and the DXLT motif. Most homologs feature multiples of seven residues, e.g. -7, 0, +7 or +14 (green). A linker length of 0 corresponds to 28 amino acids. Clusters with linker lengths of +2 (purple) and -4 (pink) contain PadCs not activated by red light, as shown by preliminary analyses. For complete organism names, consult Supplementary Table 1.

One homolog that has been characterised in extensive detail is the previously mentioned phytochrome-activated diguanylate cyclase from the organism Idiomarina species A28L, short IsPadC [9,12,15,17]. With a linker length of +7 and pronounced red-light-induced regulation of enzymatic activity, IsPadC is considered an exemplary representative of prototypical PadCs. Most importantly, in-solution spectra indicate an asymmetric P_fr state when compared to those of canonical bacteriophytochromes [17]. Based on the crystal structure of a constitutively active variant [17] as well as denaturation under acidic conditions, it was concluded that biliverdin is present in its 15E isoform in both binding pockets; however, only one of the dimer’s two PHY-tongues is rearranged from a β-hairpin to an α-helix [15]. Upon chromophore isomerisation, local reorganisation of the cofactor binding region is accompanied by conformational rearrangements of the PHY-tongue and the NTS. These changes affect the conformational dynamics of the system all the way to the PHY domain dimer interface, which extends into the sensor-effector linker and correlates with a “quasi-translational” reorientation of the coiled-coil element. As a result, the effector domains can sample additional conformational space of the DGC dimer interface and a higher frequency of productive encounters is suspected to be the cause for increased enzymatic turnover.

DGC output domains are naturally linked to a variety of sensory modules. These dimeric or even tetrameric signalling proteins bind one GTP molecule per protomer, requiring the GGDEF domains to occupy positions advantageous for intermediate (pppGpG) and eventually product (c-di-GMP) formation at the interface. Regulation of diguanylate cyclase activity depends strongly on the preceding structural element, either controlled by the degree of oligomerisation as seen in Pseudomonas aeruginosa WspR [18], Caulobacter crescentus PleD [19], Mycobacterium smegmatis DcpA [20] and Bdellovibrio bacteriovorus DgcB [21] or preventing productive encounters in inhibiting conformations of constitutive dimers as observed in zinc-sensory DGCs [22,23] – most likely by similar coiled-coil register switching mechanisms as described for IsPadC [12].

In order to better understand the mechanism of signal integration at the photosensory module of PadCs, we decided to focus on a homolog featuring prominent differences compared to IsPadC. MpPadC, from the organism Marinobacter persicus, raised our attention due to its linker length of +2 and the fact that initial spectral analysis demonstrated a prototypical bacteriophytochrome light state spectrum with no signs of heterogeneity in the chromophore environment. Interestingly, preliminary examination of enzymatic turnover revealed constitutively low activity. Consequently, these observations led us to believe that structural asymmetry of the light state might not only be beneficial with respect to enzymatic turnover, but in fact necessary for red-light-induced upregulation of enzymatic activity in PadCs. Moreover, the apparent correlation between linker length and regulation of DGC activity prompted us to investigate this particular structural feature more closely. At the same time, one should keep in mind that the complex multi-domain nature of PadCs provides multiple structural elements other than the coiled-coil linker that might also influence the functional differences between various homologs.

In an attempt to correlate individual structural elements with characteristic properties of the two contrasting family members, such as regulation of enzymatic activity or spectral properties, we designed several chimeric constructs between IsPadC and MpPadC. Biochemical characterisation of the IsMpPadC chimaeras revealed unexpected compensatory mechanisms between different PadC elements that define structural and functional properties. We further show that the dimer interface of the PHY domain plays a key role in mediating signal integration between photosensory module and enzymatic effector via a fine-tuned interplay with the coiled-coil linker. This linker extends from the PHY dimer interface to control DGC activity by conformational switching between either an inhibiting or a stimulating heptad register [12].

Results

MpPadC – a PadC family member with unexpected linker length

MpPadC shares essential residues and motifs with IsPadC; the conserved biliverdin-binding cysteine in the NTS [16,24], the DIP motif in the PAS domain [25], a PRXSF motif in the PHY-tongue [14] and the GGDEF/GGEEF motif in the output domain [26]. Figure 2 provides an overview of experimentally determined characteristics of MpPadC as well as a comparison to IsPadC. As in-solution spectra show (Figure 2A), the highly stable P_fr state (as represented by the time indicated in Figure 2B) of MpPadC features a typical symmetrical BV environment characterized by the canonical P_fr state spectrum (as opposed to IsPadC, Figure 2D). Our preliminary analysis of enzymatic turnover by means of an in vivo activity assay highlighted MpPadC as a representative of interest. As shown in Figure 3B, a high c-di-GMP concentration, represented by the intensity of red staining, is present under both dark and red-light conditions. However, as corroborated by HPLC-based activity measurements, this attribute is due to considerably higher expression levels rather than specific turnover rates; furthermore, MpPadC even shows a slightly lower DGC activity in red light than under non-actinic conditions (Figure 2C). Moreover, MpPadC features a comparatively stable P_fr state with a mean life time in the range of 30 hours, whereas dark state recovery of IsPadC is characterized by mean lifetimes of around 2 minutes (Figure 2E) [17]. In an attempt to identify potential other roles of MpPadC in its natural context, the effect of BV binding on enzymatic activity was tested. To this end, apoprotein was expressed and purified (in contrast to IsPadC, the apo form can be isolated). Interestingly, while no light regulation was observed, a distinct difference in activity between apo and holo protein enzymatic turnover could be measured (Figure 2C). As previously outlined, MpPadC displays a linker length of +2 compared to the reference organism Thioalkalivibrio species ALMg3 (TsPadC). Counting from a conserved leucine (TsPadC Leu 492, IsPadC Ile 495, MpPadC Leu 489) to the equally conserved aspartic acid of the DXLT motif of the GGDEF domain, TsPadC features 28 amino acid residues and is defined to have a linker length of 0, the number of amino acid residues in the MpPadC linker thus amounts to 30.

Figure 2. Characterisation of MpPadC and comparison to IsPadC.

(a) Dark (black), red-light (red) and difference (blue, Light – Dark) in-solution spectra of MpPadC. The light state spectrum corresponds to classic phytochromes. (b) Thermal recovery of MpPadC over a timespan of hours; eventually, the chromophore is reverted to its P_r state by far-red light (780nm). (c) Turnover of GTP per dimer in MpPadC, a slight downregulation is measured upon illumination (red). Apoprotein (grey) shows barely any activity. (d) Comparison of IsPadC (dashed) and MpPadC (continuous) UV/vis spectra normalized to maximal P_r Soret band absorption, with the former featuring an asymmetric and the latter a symmetric light spectrum. (e) Relative thermal recovery of IsPadC and MpPadC. (f) Turnover of GTP per dimer in IsPadC (dashed) and MpPadC (continuous). IsPadC, while featuring a lower basal activity than MpPadC, is strongly upregulated upon illumination.

Figure 3. Overview of the various chimeric IsMpPadC constructs.

(a) Schematic representation of domain arrangements (IsPadC = green; MpPadC = purple). In Is^NPGY_(c)Mp^cD C4, the linker of MpPadC was expanded by the first five residues of the IsPadC linker, resulting in an overall linker length of +7 (35 residues). The tongues of the IsPadC-PHY domains of constructs Mp^NPGIs^YMp^tIs^cD C8 and Is^NMp^PGIs^YMp^tIs^cD C9 were exchanged for those of MpPadC. For the exact residue numbers of the fusion points consult Supplementary Table 2. (b) A congo red-based assay for qualitative screening of in vivo DGC activity. High c-di-GMP concentrations are represented by red staining of the colonies. Bacteria were grown in dark or under constant red-light illumination, the complete screening plates are shown in Supplementary Figure 1. AppA from Rhodobacter sphaeroides was used as negative control.

Due to the pronounced differences in spectral but also functional aspects between IsPadC and MpPadC, these two homologs were selected as parent proteins for a series of chimeric constructs to unravel the origin of the characteristic functional and structural behaviour observed in both proteins.

The design of PadC chimaeras

For unambiguous description of the individual chimeric constructs, a uniform nomenclature is used in this paper. The respective domains are represented by single letters: N, P, G, Y, c and D for NTS, PAS, GAF, PHY domain, the coiled-coil linker and DGC / GGDEF, respectively. In two constructs, the PHY-tongue alone was altered, represented by the letter t. The abbreviations Is (for IsPadC) and Mp (MpPadC) represent the origin of the corresponding amino acid sequences. Is^NPGYMp^cD C1, for example, features IsPadC NTS, PAS, GAF and PHY domains; whereas the sequences of the coiled-coil linker and the output domain belong to those of MpPadC. For specific residue range numbers consult Supplementary Table 2. To provide a qualitative overview of enzymatic turnover, all constructs were screened in vivo using a cell based DGC assay [12], the results of which, as well as the domain structures of the various chimaeras, are shown in Figure 3. In an attempt to assign specific spectral and functional characteristics to individual domains, we created a total of nine chimaeras intended to highlight the influence of particular structural features.

Origin of the GGDEF domain affects substrate inhibition but not capability for signal integration

The first element we considered to potentially influence regulation of enzymatic activity was the diguanylate cyclase effector itself. One feature that turned out to be solely affected by the origin of the GGDEF domain was that of substrate inhibition. In IsPadC, turnover rates decrease with increasing substrate concentrations [12], whereas a normal saturation curve was observed for MpPadC. Among the chimaeras with an effector domain matching that of MpPadC (Is^NPGYMp^cD C1, Is^NPGMp^YcD C2, Mp^NIs^PGMp^YcD C3 and Is^NPGY(c)Mp^cD C4) no substrate inhibition was measured (Supplementary Figure 2). In contrast, the chimaeras featuring an IsPadC-based GGDEF domain (Is^NMp^PG Is^YcD C5, Mp^NPGIs^YcD C6, Mp^NPGYIs^cD C7, Mp^NPGIs^YMp^tIs^cD C8 and Is^NMp^PGIs^YMp^tIs^cD C9) showed either initial turnover rates independent of substrate concentration or the substrate inhibition observed in many phytochrome-linked DGCs [13].

However, no additional effects of GGDEF domain origin could be identified. Construct Is^NPGY(c)Mp^cD C4 is strongly upregulated by red-light illumination (Figure 5A), demonstrating that the composition of the MpPadC effector does not per se impede signal integration and light-induced regulation of enzymatic turnover.

Figure 5. Biochemical characterization of chimeric constructs.

(a) Relative thermal recoveries of IsPadC (green), MpPadC (purple) and all chimeric constructs (using absorbances at the maximum and minimum wavelengths of the respective difference spectra). Dashed lines represent PadCs with an asymmetric, continuous lines PadCs with a symmetric light state. (b) The specific DGC activity per PadC dimer was determined for every construct at 200 μM GTP. Turnover rates were calculated from four timepoints and measured under non-actinic (black) and red-light (red) conditions for all constructs (except “a”, the apo-form of MpPadC). The standard error of the estimate from the linear regression is shown as an error bar. Approximate fold-factors between dark and red light (black) as well as holo- and apoprotein (grey) measurements are indicated by brackets above the bars.

Characteristic differences in signal transduction are not defined by the chromophore environment

In light of these observations, we considered the protein environment in direct contact with the chromophore to potentially play an essential role in signal integration in PadCs. With respect to the photosensory module, three structural features interact with the biliverdin cofactor: the GAF domain provides the biliverdin binding pocket, the PHY-tongue extends back towards the cofactor environment and a conserved cysteine of the N-terminal segment covalently tethers the chromophore to the protein. The latter element also tightly interacts with the tongue and residues of the cofactor binding site in the P_r state. All three functional elements were shown to be involved in light sensing and signal transduction, undergoing structural rearrangements following the isomerisation of biliverdin [14,15,27–29].

Cofactor interactions in the GAF domain are centred around the characteristic DIP motif of phytochromes, most residues in the direct vicinity of which are strongly conserved as well. In the design of chimeric constructs aimed at identifying the influence of individual structural elements, PAS and GAF were considered only as a complex since they are believed to have co-evolved in the context of bacteriophytochromes and feature a characteristic knot structure with the linker between NTS and PAS domain threaded through an extension of a GAF domain loop. Early assumptions that the difference in symmetry of the P_fr state between IsPadC and MpPadC might correlate with the origin of the PAS-GAF bidomain are contradicted by construct Is^NMp^PGIs^YcD C5, which demonstrates asymmetric red-light-induced activation despite its MpPadC PAS-GAF complex. However, while the matter of symmetry cannot simply be assigned to the origin of the PAS and GAF domains, another characteristic appears to depend entirely on it: IsPadC and MpPadC feature distinct Q-bands of 709 nm and 687 nm, respectively. Examination of the light and dark spectra measured for all chimaeras confirmed that the Q-bands of constructs Is^NPGYMp^cD C1, Is^NPGMp^YcD C2, Mp^NIs^PGMp^YcD C3 and Is^NPGY(c)Mp^cD C4 range from 706 nm to 710 nm, whereas wavelengths of 686 nm and 687 nm were measured for Is^NMp^PGIs^YcD C5, Mp^NPGIs^YcD C6, Mp^NPGYIs^cD C7, Mp^NPGIs^YMp^tIs^cD C8 and Is^NMp^PGIs^YMp^tIs^cD C9 (Figure 4). Not entirely unexpectedly, the PAS-GAF core is the key player with respect to influencing absorption properties of individual PadCs in their P_r states. In contrast to the Q-bands of the P_r state, maxima of the difference spectra in the P_fr region are very similar for IsPadC and MpPadC (751 nm and 749 nm, respectively).

Figure 4. Representation of IsMpPadC chimaeras and their corresponding UV/Vis absorption spectra.

Domains derived from IsPadC and MpPadC are coloured green and purple, respectively. Dark spectra are represented in black, light state spectra in red and difference spectra (light – dark) in blue. Maxima of positive and negative differences are indicated.

The second structural element to closely interact with biliverdin, the NTS, did not show any major influence on spectral characteristics in IsMpPadC chimaeras. However, previous conclusions that it strongly affects P_fr stabilities [15] are corroborated by measurements of thermal recoveries from P_fr to P_r. In two of three instances, chimeric constructs with an N-terminal segment corresponding to that of IsPadC show shorter mean lifetimes than the equivalent chimaeras with an MpPadC NTS (Is^NMp^PGIs^YcD C5 vs. Mp^NPGIs^YcD C6, Mp^NPGIs^YMp^tIs^cD C8 vs. Is^NMp^PGIs^YMp^tIs^cD C9), as represented in Figure 5A. In IsPadC, the N-terminal segment closely interacts with the PHY-tongue [15] and their interplay strongly affects P_fr state stabilities and thereby thermal recoveries. In the absence of any structural data for MpPadC it is difficult to prove that the interaction of NTS and PHY-tongue close to the BV-environment differs greatly between IsPadC and MpPadC. However, variations in such an interaction might be one explanation for the drastically different recovery kinetics and the sequence alignments also show some characteristic differences in NTS composition and the length of the tongue element in the interacting region (Supplementary Figure 3).

The PHY-tongue itself, as described previously, plays a critical role in signal transduction. To determine the influence of the tongue as a specific structural feature rather than the PHY domain in its entirety, the PHY-tongues of constructs Is^NMp^PGIs^YcD C5 and Mp^NPGIs^YcD C6 were substituted with the corresponding MpPadC sequences, yielding the chimaeras Is^NMp^PGIs^YMp^tIs^cD C9 and Mp^NPGIs^YMp^tIs^cD C8, respectively. In comparison to the original constructs, both Is^NMp^PGIs^YMp^tIs^cD C9 and Mp^NPGIs^YMp^tIs^cD C8 showed a slower recovery, corroborating the assumption that an IsPadC PHY-tongue provides a more efficient return to the ground state. The chimaera with the most stable excited state, Mp^NPGYIs^cD C7, also features an MpPadC PHY-tongue, further linking the observed mean lifetimes to this specific structural feature. Signal transduction, on the other hand, is not influenced by the origin of the PHY-tongue. Both MpN^PGIs^YMp^tIs^cD C8 and Is^NMp^PGIs^YMp^tIs^cD C9 show clear upregulation after red-light illumination (Figure 5B). This demonstrates that the PHY-tongue is not a determining factor for successful signal integration when comparing IsPadC and MpPadC.

Heptad repeats in the coiled-coil are a prerequisite for signal propagation

The feature we considered most likely to strongly affect red-light-induced regulation of enzymatic activity is the coiled-coil linker. To address the effect of linker length on signal integration, construct Is^NPGY_(c)Mp^cD C4 was designed. In comparison to Is^NPGYMp^cD C1, the first five residues of the IsPadC linker (IVADS) were inserted before the sequence of the MpPadC linker, adding up to a total linker length of +7. With a continuous heptad repeat pattern being the only difference between the two aforementioned chimaeras, comparison of structural characteristics and enzymatic behaviour allowed conclusions on the exact influence of the linker length on these features. In vivo screening of DGC activity as well as a HPLC-based activity assay both verified red-light-induced upregulation of enzymatic turnover in Is^NPGY_(c)Mp^cD C4 (Figure 3B and Figure 5B, respectively), while Is^NPGYMp^cD C1 demonstrated no modulation of c-di-GMP levels. The fact that Is^NPGY_(c)Mp^cD C4 is capable of integrating the light-triggered signal shows that the residues necessary to form a fully functioning coiled-coil are present in the sequence of MpPadC despite its lack of light activation. Any IsMpPadC construct featuring a canonical MpPadC linker, however, is not upregulated upon red-light illumination (Is^NPGYMp^cD C1, Is^NPGMp^YcD C2 and Mp^NIs^PGMp^YcD C3), clearly demonstrating that linker lengths not obeying full heptad repeats suffice to inhibit upregulation of DGC activity upon cofactor isomerisation. The capability to form a continuous coiled-coil linker is apparently required for translational switching between the two different heptad registers, as shown for IsPadC [12].

While no chimaera with a linker length of +2 shows upregulation after red-light illumination, the presence of full heptad repeats in a +7 linker background alone is not enough to guarantee signal integration, either. In fact, construct Mp^NPGYIs^cD C7 shows no upregulation of enzymatic turnover induced by red light (Figure 5B) – despite its IsPadC linker of appropriate length, it is unaffected by illumination. The only explanation for this unexpected behaviour is a characteristic difference in the PHY domain interface of MpPadC compared to IsPadC, most likely incompatible with the requirements for a seamless structural continuation into the coiled-coil linker.

The PHY dimer interface acts as a bridge between sensor and effector

In fact, no IsMpPadC chimaera with an MpPadC PHY domain interface shows enzymatic upregulation upon illumination, both Is^NPGMp^YcD C2 and Mp^NIs^PGMp^YcD C3 are even downregulated (Figure 5B). As presented in a previous chapter and demonstrated by constructs Mp^NPGIs^YMp^tIs^cD C8 and Is^NMp^PGIs^YMp^tIs^cD C9, the PHY-tongue is clearly not the determining feature. While it certainly plays a crucial role for signal integration, it has been shown that an exchange of the tongue alone does not prevent successful upregulation of GTP turnover. Apart from its tongue, the PHY domain comprises of two helices, extending into the GAF domain and the coiled-coil linker respectively, which are part of the phytochrome-spanning helical spine and constitute an essential part of the characteristic 4-helix bundle dimer interface observed in many GAF architectures [30,31]. Evidently, this interface strongly affects signal transduction from sensor to effector. If both PHY interface and linker correspond to an IsPadC sequence, upregulation of enzymatic activity is guaranteed. However, if the PHY interface originates from MpPadC, no upregulation of DGC turnover is observed. Therefore, the PHY domain fulfils a central role in mediating signal integration in phytochrome-regulated DGCs by fine-tuning the communication between sensor and effector. The direct connection to the coiled-coil linker thereby enables signal integration and modulation of the dynamic range.

An asymmetric light state is not an absolute requirement for regulation

Another characteristic feature, the structural asymmetry of IsPadC, that was suspected to be important for signal transduction, however, does not correlate with the origin of the PHY domain. In fact, we found structural asymmetry of the light-activated state not to be an absolute prerequisite for successful enzymatic activation in PadCs. The asymmetric IsPadC light state population indicated by in-solution spectra together with the structural insights of a mixed P_r/P_fr dimer structure featuring the stimulating coiled-coil register was suspected to facilitate productive GGDEF encounter with respect to catalysing the first asymmetric step in pppGpG intermediate formation during GTP turnover [17]. The symmetric P_fr state observed for MpPadC coupled with its lack of upregulation and generally rather low specific activity supported the hypothesis of asymmetric activation for effective control of GTP turnover. Out of all nine chimeric constructs examined, as outlined previously, four feature symmetric P_fr states (Mp^NPGIs^YcD C6, Mp^NPGYIs^cD C7, Mp^NPGIs^YMp^tIs^cD C8 and Is^NMp^PGIs^YMp^tIs^cD C9; Figure 4). Interestingly, of these four chimaeras, only Mp^NPGYIs^cD C7 shows no upregulation of enzymatic turnover upon red-light illumination. Thus, asymmetry, while potentially advantageous with respect to switching between inhibiting and stimulating coiled-coil registers, does not appear to be an absolute prerequisite for red-light-induced upregulation of DGC activity in artificial PadCs. Moreover, asymmetry cannot be assigned to any specific structural element; however, exchange of a single domain or feature can be enough to deactivate or restore asymmetry: While Is^NMp^PGIs^YcD C5 features an asymmetric light state, neither Mp^NPGIs^YcD C6 nor Is^NMp^PGIs^YMp^tIs^cD C9 show any spectral indication of asymmetry. Exchanging either the NTS or the PHY-tongue to that of MpPadC is enough to establish symmetry, highlighting the complexity of domain-interplay in phytochrome-regulated diguanylyl cyclases.

In light of these results, the asymmetric nature of the IsPadC light state spectrum might result from the stability of the P_r environment as determined by a tighter coordination of biliverdin in the PAS-GAF environment influenced by the interaction with the NTS and the PHY tongue elements. As shown in Figure 2D, the extinction coefficient of the IsPadC Q-band is significantly higher than that of MpPadC, considering that the Soret band absorption should be less influenced by changes in the cofactor environment (as shown by the marginal effect of biliverdin binding to phytochromes [32] as well as the influence of the level of protonation in the cofactor environment of DrBphP [33] on Soret band properties). Apparently, the MpPadC PAS-GAF bidomain on its own does not stabilize the biliverdin configuration in ZZZssa as effectively as its IsPadC counterpart, but upon incorporation of either the IsPadC NTS or the IsPadC PHY tongue the biliverdin geometry is stabilized in such a way that its spectral properties approach those of wildtype IsPadC. Since it is apparently no specific interaction that influences this characteristic phytochrome property, we suspect the conformational dynamics of the BV environment to be a key factor in influencing spectral properties and eventually also long-range signal transduction.

Discussion

PadCs are highly complex systems with their biochemical function and enzymatic behaviour being determined by an intricate interplay of various domains and structural features. This insight is in line with previously published observations that photoactivation in bacteriophytochromes is governed by a dynamic equilibrium of all essential modules and structural elements ranging from the chromophore surroundings to the output module [11,12,29,34]. According to this model of signal integration, the net activity of bacteriophytochromes is fine-tuned by the interplay of changes in all involved modules and structural features [11], corresponding to the violin model of allosteric activity regulation [6,12,35]. It has further been suggested that multiple signalling pathways – in this particular case through the PHY-tongue as well as the characteristic PAS-linked knot structure as observed for Deinococcus radiodurans – could work either in concert or constitute redundant processes [36], further highlighting the complexity and the interplay of multiple structural elements of signal integration in bacteriophytochromes.

In line with such a complex interplay of multiple functional elements, we observed that the asymmetric property of the light state in many PadCs cannot be assigned to any specific structural element. The PAS-GAF bidomain, even though strongly suspected to be a key player in determining symmetric activation of phytochrome dimers [15], proved to only affect spectral properties of the P_r state rather than specific PHY-tongue interactions or rearrangements that are required for P_fr stabilization. For example, construct Is^NMp^PGIs^YcD C5 features an asymmetric light spectrum despite its MpPadC PAS-GAF core; however, symmetry can be established by exchanging either the NTS (Mp^NPGIs^YcD C6) or the PHY-tongue (Is^NMp^PGIs^YMp^tIs^cD C9) to that of MpPadC. This observation is an example of the intricate interplay between the different key structural elements of the cofactor environment that ultimately define the systems’ function. We have shown that an asymmetric light state is not an absolute prerequisite for enabling productive encounters between the GGDEF effectors and thus successful light-dependent upregulation of DGC activity. In fact, three upregulated chimeric constructs (Mp^NPGIs^YcD C6, Mp^NPGIs^YMp^tIs^cD C8 and Is^NMp^PGIs^YMp^tIs^cD C9) feature symmetric light spectra. While these findings prove that asymmetry is not a strict requirement for upregulation of GTP turnover, we suspect it to nonetheless be advantageous to regulated PadCs; especially considering the reports of varying examples and instances of asymmetry in GGDEF-linked dimeric proteins [21] as well as phytochromes [28,37,38]. In PadCs, a partial structural rearrangement, as facilitated by asymmetry, could suffice to enable faster recoveries to the dark-adapted conformation, thus allowing for more effective regulation. In bacteria, important lifestyle decisions such as the transition between motility and biofilm formation are governed by c-di-GMP on a transcriptional, posttranscriptional and posttranslational level [13]. In light of the influence of this second messenger, tight control of its availability is essential and extensively regulated [39]. Moreover, it should be considered that the chimaeras created are artificial systems rather than naturally occurring proteins. Presently, of the 17 native PadCs characterized, no symmetric homolog has been shown to feature upregulation of enzymatic activity (unpublished data) – nonetheless, some homologs with a linker length of +2 are known to downregulate rather than upregulate GTP turnover upon red-light illumination (MpPadC). In IsPadC and IsNMp^PGIs^YcD C5, the tightly controlled interplay of NTS and PHY-tongue is suspected to push the system towards asymmetry. Exactly how this interplay affects properties of the activated bacteriophytochrome species will be interesting to follow up on in future research efforts.

Between IsPadC and MpPadC, the sequences of the N-terminal segments differ substantially, with the NTS of the former being four residues longer than that of MpPadC (Supplementary Figure 3). Moreover, the NTS regions in general feature barely any conservation apart from the cysteine that covalently binds the biliverdin cofactor [12,24], even among members of the same linker length clusters in PadCs. In IsPadC, the residues preceding Cys17 form an α-helix and semi-conserved hydrophobic residues are involved in binding to the PHY-tongue and the GAF domain. Since no crystal structure is yet available for MpPadC, the exact arrangement of its NTS is unknown, but the apparently conserved involvement of structural rearrangements of the NTS during the P_r to P_fr transitions even in distantly related bacteriophytochromes [40,41] indicates the importance of the structural composition of this functional element. Similarly, the sequence preceding the characteristic PRXSF motif in the PHY-tongue is barely conserved and features unexpectedly strong variations in amino acid properties. Since this part of the tongue region contacts the NTS in several crystal structures [12,14,41,42], these interactions might serve as an evolutionary playground for fine-tuning light responses and functional integration in phytochromes in general. As confirmed by IsMpPadC chimaeras (Mp^NPGIs^YMp^tIs^cD C8 and Is^NMp^PGIs^YMp^tIs^cD C9), the PHY-tongue is not a determining factor for the loss of upregulation in MpPadC, highlighting that it is not absolutely essential for signal integration in bilin-binding photoreceptors – as also exemplified by naturally occurring cyanobacteriochromes, which constitute stand-alone GAF domains [43–45] or tongue-less constructs that still retain light regulation [12]. The concept of evolutionary tunability in these regions is further supported by the fact that the length of PHY-tongues varies extensively among the phytochromes of different species [11].

Considering the central role of the PHY domain interface observed in our analysis of IsMpPadC chimaeras, we also looked at the sequence conservation in this area. The sequences of the PHY dimer interfaces differ greatly among various PadCs (Supplementary Figure 3). Some negative charges (D494, D498) that interact with arginines (R327) in IsPadC are missing in MpPadC (R488, A492) as well as other homologs featuring linker lengths of +2 and symmetric P_r spectra (Marinobacter species T13-3, Marinospirillum celere). These differences are hypothesized to be involved in influencing PHY dimer interface stability and potentially providing more conformational flexibility, which would in turn allow the tongues in both protomers to rearrange and fold into the α-helical element required for P_fr stabilization, which would eventually enable symmetric formation of a prototypical bacteriophytochrome P_fr state. Specifically, MpPadC features three arginines (R488, R490, R491) in close spatial proximity at the PHY dimer interface and near the start of the coiled-coil linker region. These positive charges might destabilize the helix-bundle at the dimer interface. Potentially, the PHY domains are thereby also incapable of establishing the appropriate architecture to facilitate successful register switching in the coiled-coil linker as required for upregulation of DGC activity [17]. For the D. radiodurans bacteriophytochrome, in spite of extensive protomer contacts along the protein [46], a comparatively loose conformation of the PHY domain dimer interface has been reported [27,47,48], suggesting that the MpPadC PHY domains might similarly feature a weaker interaction than those in IsPadC. It has also been shown that the PHY domain interface architecture of D. radiodurans remains in the same loose conformation even after attachment of a cyclase coiled-coil, rather than adopting a more compact structure as observed in IsPadC [49], confirming that addition of an IsPadC-linker does not necessarily negate or overrule the effects of an MpPadC-PHY dimer interface. At the same time, the D. radiodurans bacteriophytochrome has potentially developed other mechanisms to regulate histidine kinase or phosphatase activity [11], and different output functionalities of bacteriophytochromes may be regulated by PSM cores in various ways.

Different origins of the photosensory module and the linker were reported to produce strongly decreased photoactivation [15], potentially impeding correct dimer interface establishment, which is corroborated by IsMpPadC chimaeras. Relay of information such as the presence of light between functionally different domains like a sensory module and an enzymatic effector requires evolutionary fine-tuning to ensure efficient signal integration. The PHY domain, the interface of which appears capable of bridging that language barrier, and the coiled-coil linker need to work in concert to enable successful signal transduction, highlighting the powerful tool evolution has at its disposal to effectively tune light-induced activity regulation of diverse output functionalities.

The importance of the linker element is a recurring feature in signalling proteins – in nitric oxide activated soluble guanylyl cyclases, instability in the coiled-coil element linking in- and output domains appears to be conserved across all identified sGCs to enable conformational dynamics throughout signal transduction [50] and the helical spine has been termed a “signalling hub” in phosphorylation-responsive photosensitive histidine kinases [37]. In the D. radiodurans bacteriophytochrome, the long linker helix is essential for integration of PHY-tongue rearrangement and features a characteristic hinge [27]. The artificial fusions required in the context of rational design of optogenetic tools based on the photosensory core module of bacteriophytochromes have proven to require extensive tweaking to establish linker-lengths and -sequences that are compatible with both the PSM as well as the output domain of choice [49,51,52].

On an evolutionary level, these findings also raise the potential of the coiled-coil linker acting as a driving force to return to P_r in IsPadC, which is apparently not necessary for the non-light-regulated homolog MpPadC. This difference is corroborated by the thermal recoveries of IsPadC and MpPadC deviating with mean lifetimes in the range of 2 minutes versus 30 hours, respectively. IsPadC, which is capable of strongly increasing c-di-GMP production, is under strong pressure to return to its ground state as soon as the trigger for enzymatic upregulation – red light – is absent. This allows Idiomarina species to base important lifestyle decisions on the presence of light of a certain wavelength. MpPadC, on the other hand, with only weak pressure to return to P_r, features a much more stable light state; its thermal recovery serves no biochemical function. Considering such diverse P_fr state stabilities, the measured stark difference in enzymatic turnover between apo- and holoprotein as well as a potential evolutionary loss of a continuous coiled-coil in the linker due to lack of functional necessity led us to suggest that proteins with linker lengths of +2 might serve as bilin rather than red-light sensors. This hypothesis is encouraged by one homolog with the same linker length missing the biliverdin-binding cysteine residue in the NTS (Marinospirillum alkaliphilum) – as covalent binding of the detected compound would render these sensors rather inefficient. Other than red light, the availability of iron is another environmental factor to be considered by bacteria in the transition from motility to sessility – Frangipani et al., for example, reported a link between iron uptake and c-di-GMP in P. aeruginosa [53]. In this sense, the levels of iron released during heme degradation, and consequently biliverdin formation, could provide a positive feedback by way of c-di-GMP production.

The enzymatic effectors of IsPadC and MpPadC show similar DGC activity dependent on the sensory system as well as the linker rather than the origin and exact sequence of the GGDEF domains themselves. This apparent similarity in enzymatic effector, especially in the context of the wildtype proteins’ behaviour, corresponds to the observation that the core fold of PadCs is always the same, regardless of whether the effector is attached to the photosensory module (see [12] for a published crystal structure of the IsPadC PSM construct), highlighting the importance of the PSM structure. The flexibility of the effector domains alone is further corroborated by SAXS measurements of IsPadC, which show a highly stable photosensory module but pronounced conformational flexibility for the GGDEF domains [12]. In IsPadC, under non-actinic conditions, the productive encounter of the two protomers’ effectors that enables efficient enzymatic turnover is minimized. These dynamics appear to relate to the stability of the inhibiting coiled-coil register of the linker, which is presumably missing – or only partially populated – in PadCs with a linker length of +2. Without the rigid framework of a functional coiled-coil linker as well as correct arrangement of the PHY dimer interface, the GGDEF domains are less restricted in their possible conformations, allowing for some basal activity and providing a possible explanation for the turnover rates observed for MpPadC. However, the quasi-translational rearrangement of the coiled-coil helices as observed for IsPadC and required for stimulation of DGC activity is not possible in the truncated linker of MpPadC. In addition to the linker sequence and length itself, the coupling with the preceding PHY domain interface is equally important. This is exemplified by the only chimeric construct that features significant levels of the pppGpG intermediate [54] during c-di-GMP formation, Mp^NPGYIs^cD C7, a chimaera that shows exceptionally low production of c-di-GMP under both dark and red-light conditions and features an MpPadC PHY domain linked to an IsPadC linker. In this case, pppGpG concentrations are actually higher than c-di-GMP levels for all GTP concentrations measured (Supplementary Figure 4), highlighting the importance of an appropriate PHY dimer interface that can integrate the light signal and translate this information to an input for the coiled-coil linker. The necessity of an operational linker element as well as tight control of upstream sensory modules with regard to activity regulation corroborates observations of the functional importance of the structural component preceding the GGDEF domain in regulation of DGC-linked sensory proteins [22,23].

As demonstrated in this work, application of chimeric proteins in determining structure-function relationships between functionally diverse homologs is challenging. Functional properties of chimeras usually cannot be classified by an all or nothing system. Therefore, defining the interplay of multiple functional elements on a molecular level and drawing specific conclusions for other unrelated systems is not straight forward in many cases. However, chimaera-based experimental designs do allow identification of structural features that have no or only negligible influences on functional properties as well as highlight the interplay between two or more specific domains. Functional redundancies (as shown here for the establishment of asymmetry) or specific interactions of functional elements at play might not be easily accessible by other more directed approaches such as individual amino acid substitutions. The strength of the IsMpPadC chimeric constructs is based on pinpointing regions that influence the regulation of the dimeric effector, which can be interpreted as hotspots for rational or evolutionary campaigns with the ultimate goal of modulating effector regulation, photoexcited state stability, and/or photochromic properties. Bacteriophytochromes are known to transition between several conformational states rather than remaining rigidly within one or two structures [12,40,55,56]. Understanding how a particular structural element is modulated by upstream and/or downstream elements also provides important insight into signal integration mechanisms. Thus, the true strength of chimaeras is to highlight elements that, when targeted for modifications, influence phytochrome properties. In the end, any novel optogenetic tool is a chimeric system generated from a photosensory module and an effector of interest. Being able to build upon insights from chimeras within closely related PadCs will therefore also be beneficial for more ambitious new optogenetic tool developments. Appreciating the fact that more than just linker length and/or linker composition is needed for the creation of a functional sensor-effector system is one of the key-findings of this study. Since apparently different PHY dimer architectures have been evolved by nature for enabling regulation of diverse functionalities, libraries of different sensory modules and screening of multiple linker lengths and compositions to functionally couple phytochrome sensors to any effector of interest should be one promising approach for the generation of new optogenetic tools.

In summary, we show that structural asymmetry of the light activated state, while potentially advantageous, is not an absolute necessity for regulation of enzymatic activity in PadCs. We further demonstrate the complexity of these sensor-effector systems and the interplay of multiple domains and structural features, showcasing that only few functional characteristics can be attributed to any singular element. The communication between photosensory module and GGDEF effector was shown to play an essential role in signal transduction, with the PHY domain dimer interface bridging the two components. Our conclusions highlight the challenge presented in understanding phytochrome signalling in PadCs in sufficient detail to enable design of artificial tools for practical application in optogenetics.

Materials and Methods

Protein preparation, expression, and purification

Chimeric constructs were generated from two pETM-11-based bacteriophytochrome-linked diguanylate cyclase plasmids [12], IsPadC and MpPadC. A synthetic DNA fragment of MpPadC (sequence accession number WP_091706258) was synthesized with codon optimization for expression in Escherichia coli (GeneArt, Thermo Fisher Scientific). Using a PCR-based Gibson cloning approach, specific regions of the corresponding coding sequences were amplified using the primers listed in Supplementary Table 3. Assembly of the constructs was performed using the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs). The various chimaeras were expressed and purified as described previously [12]. Briefly, (His)₆-tagged holoproteins were expressed in E. coli BL21 (DE3), which contained a previously generated pT7-ho1 helper plasmid coding for heme oxygenase (HO-1) enabling bilin chromophore production. Transformed cells were cultivated in LB medium at 37°C to an OD₆₀₀ of 0.45, then cooled to 16°C. 10 mg L^-1 δ-aminolevulinic acid were supplied, and expression induced with 0.25 mM isopropyl-β-D-thiogalactopyranoside (IPTG) at an OD₆₀₀ of ~ 0.7. Harvested cells were treated with 100 μg mL^-1 lysozyme and protease inhibitor cocktail before sonication (5x5 min, 70%, pulsed mode; Labsonic L, Sartorius). The soluble fraction was separated by ultracentrifugation (206,000 rcf, 45 min).

Soluble proteins were purified by affinity chromatography using a Ni²⁺-Sepharose matrix (Ni Sepharose 6 Fast Flow, GE Healthcare). After a washing step (lysis buffer (10 mM HEPES pH 7, 0.5 M NaCl, 2 mM MgCl₂) with 30 mM imidazole), proteins were eluted (lysis buffer with 250 mM imidazole). The tag was cleaved during overnight dialysis at 4°C in dialysis buffer (50 mM HEPES pH 7.0, 500 mM NaCl, 2 mM MgCl₂, 20 mM imidazole) in the presence of histidine-tagged tobacco etch virus (TEV) protease (~1:16 TEV/substrate). After another step of affinity chromatography to remove cleaved tag and TEV protease, the proteins of the flowthrough were concentrated and loaded onto a size exclusion chromatography column (HiLoad 16/60 Superdex 200 prep grade; GE Healthcare) equilibrated in storage buffer (10 mM Hepes pH 7.0, 500 mM NaCl, 2 mM MgCl₂) and concentrated using ultra centrifugal filters (Amicon Ultra-15; Merck Millipore). Aliquoted proteins were flash-frozen in liquid nitrogen and stored at - 80°C.

Cell-based diguanylate cyclase assay

Screening of diguanylate cyclase activity was performed as described previously [12]. pETM-11-based plasmids were transformed in E. coli BL21 (DE3) cells co-expressing a pT7-ho1 helper plasmid. Cells were grown to an OD600 of 0.5 at 30°C in YESCA medium (10 mg mL^-1 casamino acids, 1.5 mg mL^-1 yeast extract) supplemented with 0.05 mg mL^-1 MgSO4 and 0.5 μg mL^-1 FeSO4. After induction with 0.25 mM IPTG and addition of 10 mg L^-1 δ-aminolevulinic acid, cultures were shaken at 16°C for 4 hours. 3 μL of cultures concentrated to an OD600 of 10 were spotted on YESCA agar plates containing 0.01 mg mL^-1 congo-red and incubated at 20°C in the dark or under constant red-light illumination (630 nm, 75 μW cm^-2). The pETM-11 RsAppa (light-sensing anti-repressor AppA from Rhodobacter sphaeroides, [57]) construct has no DGC activity and served as negative control.

Spectroscopic characterization

UV-vis absorption spectra were measured using a Specord 200 Plus spectrophotometer (Analytic Jena) at 20°C, with samples diluted in storage buffer to ~2 μM. Dark-adapted P_r state spectra were acquired under non-actinic light conditions, minimizing contact with light. P_fr-enriched spectra were recorded after 1 minute of red-light illumination (660nm, 20.9 μW/mm², Thorlabs).

P_r state recovery kinetics were measured at maximum and minimum wavelengths of difference spectra (P_fr – P_r) after 1 minute of red-light illumination, using a Specord 200 Plus spectrophotometer (Analytic Jena) with 10 ms integration time sampled every 10 sec, 1 min, 5 min, 10 min or 15 min, depending on photoactivated state stability. Since the thermal recovery is affected by the measuring light, theoretical endpoints were determined from P_r state spectra to enable qualitative comparison of different constructs. Recovery times estimated using these theoretical endpoints as fixed values should not be over-interpreted quantitatively, but serve to highlight different recovery kinetics and characteristics of various chimaeras.

Kinetic analysis

Turnover of GTP to c-di-GMP was measured with high performance liquid chromatography (HPLC), adapting a protocol previously described [12] and based upon a method described in [58]: Different concentrations of GTP (50, 200, 500 μM) were added to purified protein diluted in reaction buffer (10 mM Hepes pH 7.0, 500 mM NaCl, 50 mM MgCl₂) and previously incubated at 20°C for 1 min under non-actinic light or constant red-light illumination (660 nm, 20.9 μW/mm², Thorlabs). After varying times of incubation under non-actinic light for the dark state measurement, or under constant red-light illumination for the light state measurements, samples were thermally inactivated by 1 min incubation at ~90°C. After centrifugation to remove heat-denatured protein, nucleotides were separated by reversed phase HPLC (ProntoSil C18 ace-EPS column, Bischoff) using an aqueous mobile phase (10 mM K₂HPO₄ pH 8.0, 5% MeOH, 5 min) under isocratic conditions at 35°C. All activities were normalized to the dimer concentration of the respective constructs.

1. Abbreviations

BV

biliverdin

c-di-GMP

cyclic dimeric guanosine monophosphate

DGC

diguanylyl cyclase

GAF

cGMP phosphodiesterase/adenylate cyclase/FhlA

HPLC

high performance liquid chromatography

LB

lysogeny broth

NTS

N-terminal segment

PadC

phytochrome activated diguanylyl cyclase

PAS

Period/ARNT/Single-minded

PHY

phytochrome-specific

PSM

photosensory module

SAXS

small-angle X-ray scattering

TEV

tobacco etch virus