Our previous characterization of the IsPadC system revealed that the phytochrome – diguanylyl cyclase (DGC) linker region populates two different coiled-coil register conformations that influence the enzymatic activity of the effector (Gourinchas et al., 2017). To test this hypothesis, amino acids at special positions of the coiled-coil were replaced to either destabilize or stabilize the presumably stimulating register conformation (Figure 1a). Previously, the corresponding sequence variants of the coiled-coil register were used in screening assays - IsPadC^Reg1 (IsPadC D504L A518L; constitutively inactive) and IsPadC^Reg2 (IsPadC S505V A526V; constitutively active) – and a detailed biochemical characterization was only provided for variants with parts of the coiled-coil deleted (Gourinchas et al., 2017). This time we focused entirely on the sequence variants with identical linker length and included an additional construct, IsPadC^Reg2.a (IsPadC S505V N512V N519V A526V), that is further stabilized in the stimulating register. To this end, we substituted the highly conserved, polar asparagine residues at the generally hydrophobic a positions of the coiled-coil in its stimulating register 2 conformation with valine residues (Figure 1 and Figure 1—figure supplement 1a–c). Employing a Congo red-based screening system for cyclic-dimeric-GMP production in E.coli (Antoniani et al., 2010), we show that IsPadC^Reg2.a, comparable to IsPadC^Reg2, is constitutively active under both dark and light conditions, whereas IsPadC^Reg1 is always inactive (Figure 1b). All three constructs were expressed in an E. coli strain coexpressing heme oxygenase to isolate holoproteins loaded with the biliverdin co-factor and to enable a more detailed biochemical characterization. It should be noted that so far none of the IsPadC constructs tested could be purified as apoprotein as they appear to irreversibly aggregate inside the cells in the absence of their cofactor. Therefore, all final protein preparations correspond to fully loaded holoproteins of each construct. A three-step purification procedure, consisting of an affinity, a reverse affinity and a size-exclusion chromatography step, allowed isolation of all full-length proteins to apparent homogeneity. Based on the elution volume of the purified proteins in the size exclusion step, all variants form stable dimers in solution. All subsequent biochemical and biophysical experiments were performed with protein samples obtained by this procedure.

Figure 1 with 4 supplements see all

Download asset Open asset

During our previous characterization of IsPadC, we observed only partial formation of the P_fr state even under constant red light illumination (Gourinchas et al., 2017), which clearly differs from absorption spectra of other well-characterized bacteriophytochromes, such as for example Deinococcus radiodurans phytochrome (Wagner et al., 2005; Wagner et al., 2007) (DrBphP, Figure 1c), or bathyphytochromes that have P_fr as their ground state (Yang et al., 2009; 2011). In an attempt to better understand the molecular basis for the incomplete P_fr spectrum under steady-state red light conditions, we analyzed the effect of actinic light intensity and the protein concentration on the light-state spectrum. As shown in Figure 1d decreasing protein concentrations correlate with a reduction of the required light intensity to increase the 750 nm contribution, however, in all cases similar steady-state light-activated spectra of IsPadC are populated. Importantly, the sigmoidal profile obtained for activation of the biliverdin cofactor correlates with the enzymatic readout of the effector domain.

Since light intensity is apparently not limiting the steady-state P_fr population under red light, we denatured IsPadC directly after illumination with saturating red light to address the isomerization state of the biliverdin cofactor. To this end, we followed the protocol of denaturation under acidic conditions described by Thümmler et al. (Thümmler et al., 1981) and observed that more than 90% of the biliverdin population is present as the 15E isomer during constant red light illumination (Figure 1e). The comparison with DrBphP clearly shows that similar levels of the isomerized biliverdin species are present upon red light activation, however, even complete isomerization of biliverdin apparently does not allow full population of the P_fr state in the case of IsPadC (cf. Figure 1c). This is either due to different biliverdin environments of the two protomers forming an IsPadC dimer or due to a steady-state equilibrium of P_fr with preceding intermediates along the activation pathway (Borucki et al., 2005; Buhrke et al., 2018; Dasgupta et al., 2009). By following the thermal reversion of 15E biliverdin to its 15Z isomer by means of quenching the protein after different dark adaptation times, mean lifetimes and amplitudes are observed that are comparable to those obtained for the native protein (inset of Figure 1e), which indicates a direct correlation of the native spectral properties with the isomerization state of biliverdin. Interestingly, at the high concentrations of 10 µM required to perform this experiment, fitting of the data required the sum of three exponentials (Table 1 in Supplementary file 1). While this fitting function and the obtained parameters differ from previously reported data for IsPadC (Gourinchas et al., 2017), we realized that these observations directly correlate with the protein concentration used for the different experiments. At low protein concentrations, fitting with the sum of two exponentials is sufficient to describe the recovery kinetics, however, increasing the protein concentration results in apparently slower recoveries until at some point the inclusion of a third exponential is indicated – eventually providing comparable values for the time constants of the first two phases (cf. Table 1 in Supplementary file 1). Importantly, also changing the salt concentration of the samples significantly impacts the obtained recovery kinetics. This, in combination with the effect of protein concentration, points towards an involvement of the protein oligomerization state and/or non-specific interactions of structural elements close to the chromophore binding site in modulating the recovery kinetics of the biliverdin cofactor.

The spectroscopic characterization of the three IsPadC register variants revealed no major differences compared to the wild type with respect to their dark- and light-state UV/Vis absorption spectra (Figure 1—figure supplement 2a–c). The absorption maxima of the P_r states are shifted slightly from 710 nm for the wild-type to 708 nm for both IsPadC^Reg1 and IsPadC^Reg2.a as well as to 707 nm for IsPadC^Reg2. Consequently, the difference spectra of light-dark (Figure 1—figure supplement 2a–c) show analogous changes in maxima of the negative P_r signals, whereas the signals associated with P_fr formation remain almost unaffected. The thermal recoveries to the ground state P_r absorption spectra, however, are pronouncedly affected by the amino acid substitutions in the coiled-coil linker (Table 1 in Supplementary file 1 and Figure 1—figure supplement 2d). We also observed that the protein concentration has a pronounced effect on the recovery kinetics of the register variants and have therefore decided to use 2 µM protein samples for the comparison of individual protein variants. Since this concentration differs from that used for the initial IsPadC characterization (Gourinchas et al., 2017) the reported values in Table 1 in Supplementary file 1 also differ slightly from previously reported data for IsPadC (Gourinchas et al., 2017). Importantly, recovery kinetics of wild-type IsPadC and the IsPadC^Reg2 variant could be fitted with the sum of two exponentials, whereas IsPadC^Reg1 and IsPadC^Reg2.a required the addition of a third exponential function to provide satisfying fits already at low protein concentrations. Interestingly, the latter variants are presumably extensively stabilized in their coiled-coil registers (inhibiting register 1 and stimulating register 2, respectively) and feature at least one phase with extremely slow recovery, whose contribution to the overall amplitude differs pronouncedly (Figure 1—figure supplement 2d). Especially IsPadC^Reg2.a appears to be severely limited in fully recovering to the P_r ground state, which shows that the conformation of the coiled-coil register and its dynamics strongly influence the biliverdin environment.

A closer inspection of the recovery kinetics by comparison of data measured at 710 nm (reformation of P_r) and 750 nm (depletion of P_fr and potential non-P_fr photoproducts) revealed that the time constants of individual recovery phases match between the 710 and 750 nm analyses. Interestingly, also the relative amplitudes contributing to individual phases at 710 and 750 nm appear to correlate for all constructs. While this observation might suggest that the recoveries do not involve photoproducts that significantly differ in their spectra from the P_r or P_fr absorbing species, it does not exclude the presence of closely related species such as for example meta-R states (Borucki et al., 2005; Buhrke et al., 2018; Wagner et al., 2008). The observation of a non-isosbestic recovery (Figure 1—figure supplement 2e) further supports the involvement of at least one colored species in addition to P_r and P_fr. In fact, the fairly constant amplitude contributions of the fastest phase of the thermal recoveries (~36%, cf. Table 1 in Supplementary file 1) suggest that this phase might correspond to the recovery of a biliverdin species with lower extinction coefficient bound to one PadC protomer, whereas the remaining phase(s) reflect the recovery of the second biliverdin species – most likely the classical P_fr state. A closer inspection of the thermal recoveries in the region of the expected isosbestic points between P_r and P_fr as well as P_r and meta-R (cf. ref [Borucki et al., 2005]) confirmed that the relative amplitudes of the individual contributions change significantly in this spectral region and that the maximal contribution of the meta-R and P_fr species lies close to the isosbestic point of the respective other contribution. Therefore, our initial spectroscopic characterization indicates that two different biliverdin species are stabilized upon red light illumination of IsPadC in the two protomers of the functional parallel phytochrome dimer.

Such an asymmetric activation of IsPadC and its variants is also supported by the observations made for a construct lacking the PHY-tongue element (IsPadC^{Δ442-477::SG}) (Gourinchas et al., 2017). The spectral characteristics of the illuminated IsPadC^{Δ442-477::SG} construct are reminiscent of a composite of P_r and meta-R states considering meta-R species characterized for D207 and H260 variants of DrBphP (Wagner et al., 2008) or low temperature trapped forms of Agp1 from Agrobacterium tumefaciens (Borucki et al., 2005). The meta-R contribution is also supported by the characteristic changes in the light-dark difference spectrum compared to the other IsPadC variants (Figure 1—figure supplement 2f and Figure 1—figure supplement 3a). Importantly, acidic denaturation of IsPadC^{Δ442-477::SG} revealed that in comparison to wild-type IsPadC only half of the biliverdin population can be isomerized to the 15E form (Figure 1—figure supplement 3b). The biliverdin cofactor bound to the respective other protomer of the dimeric species radiatively decays to the ground state rather than isomerizing, as shown by the significantly increased fluorescence of IsPadC^{Δ442-477::SG} (Figure 1—figure supplement 3c). Interestingly, the spectral properties of this partial light-activated state are retained over hours (Figure 1—figure supplement 3a) and even over-night incubation does not allow recovery to the P_r ground state (Gourinchas et al., 2017).

These data are consistent with a mechanism of IsPadC light activation involving light-induced isomerization of both biliverdin molecules of dimeric IsPadC. However, only one biliverdin moiety can undergo all subsequent structural rearrangements to obtain a classical P_fr state. The second biliverdin is most likely trapped in an early meta-R like conformation. This asymmetry is more pronounced for the register variants stabilized in register 2, where the P_fr environment is significantly stabilized relative to the wild-type and the IsPadC^Reg1 variant. Considering that the stimulating register and activation of enzymatic activity by illumination are coupled in wild-type IsPadC, we also had a closer look at the effect of stabilizing the coiled-coil in the two registers on catalytic turnover of GTP to c-di-GMP.

In order to quantify the light regulation capacities of individual IsPadC register variants we employed an HPLC-based assay that enables baseline separation of the substrate GTP, the pppGpG intermediate and the product c-di-GMP. None of the analyzed register variants showed significant levels of intermediate formation as observed for IsPadC homologs or some linker deletion variants characterized previously (Gourinchas et al., 2017). Similar to IsPadC the conversion of substrate to product tends to be inhibited by increasing substrate concentrations, however, this phenomenon is less pronounced for both variants designed to stabilize the stimulating register 2 (Figure 1—figure supplement 4a–c). Table 1 summarizes the obtained rate constants for product formation at a substrate concentration of 200 µM, which corresponds to the maximum of GTP conversion for IsPadC^Reg2. At this concentration, no light regulation can be observed for IsPadC^Reg1 and IsPadC^Reg2.a, whereas IsPadC^Reg2 still shows residual stimulation of DGC activity by red light (roughly 2-fold). Considering that IsPadC^Reg2 activity can be increased whereas IsPadC^Reg2.a cannot, indicates that a too strong stabilization of the coiled-coil in the stimulating register can have negative effects for c-di-GMP formation. Apparently, the coiled-coil conformation of register 2 has a positive effect on enzymatic activity, but also the conformational dynamics of this structural element influence overall enzymatic turnover of the IsPadC system.

Addressing the structural implications of the stimulating register 2 conformation was enabled by the crystallization of IsPadC^Reg2 under dark conditions. Solving the crystal structure by molecular replacement (MR) using data to 2.85 Å (Table 2 in Supplementary file 1) was only possible after splitting the dark-state wild-type IsPadC structure into individual subdomains, due to substantial structural rearrangements of the overall IsPadC^Reg2 structure compared to the wild-type (Gourinchas et al., 2017). Even though IsPadC^Reg2 and wild-type IsPadC form crystals with the same overall lattice, we observed interesting rearrangements in the DGC effector domains that are coupled to the transition of the coiled-coil into the expected register 2 conformation. In addition, one PHY domain and its associated tongue element differ substantially from the dark-state IsPadC structure described previously (Gourinchas et al., 2017).

In fact, already after the initial placement of all IsPadC domains by MR, the density of one protomers PHY-tongue extension did not match the β-hairpin structure characteristic for the dark-state P_r environment of biliverdin. Rather the electron density supported an α-helical element that is usually only observed for P_fr-conformations (Takala et al., 2014; Yang et al., 2009; Burgie and Vierstra, 2014) (Figure 2a–b and Figure 2—figure supplement 1). A closer inspection of the cofactor binding site revealed that also biliverdin in its classical ZZZssa conformation would not fit into the observed electron density. However, models of biliverdin in its isomerized and rotated 15E conformation nicely explain the observed electron density as supported by the polder OMIT maps (Liebschner et al., 2017) of the co-factor region (Figure 2—figure supplement 2). The structural rearrangements of the biliverdin environment are, however, only observed for one protomer, whereas the other protomer features the expected dark-state P_r environment observed for both protomers in the wild-type IsPadC structure (Gourinchas et al., 2017). Therefore, the IsPadC^Reg2 structure features an asymmetric dimer (Figure 2a) that is composed of one P_fr biliverdin environment (protomer A) and one P_r environment (protomer B). In addition to the helical tongue and the ZZEssa biliverdin configuration, also the side chains of residues Y168, H193 and F195 in the proximity of the rotated D-ring change their rotamer positions as described for other phytochrome P_fr states (Takala et al., 2014; Yang et al., 2009; Burgie and Vierstra, 2014) (Figure 2b and Figure 2—figure supplement 2).

Figure 2 with 3 supplements see all

Download asset Open asset

The P_fr/P_r heterodimer of the final IsPadC^Reg2 model is further supported by UV/Vis absorption spectra acquired for IsPadC^Reg2 crystals in comparison to wild-type IsPadC crystals. Averaging multiple spectra collected at different orientations of crystals relative to the incident measuring light, revealed a clear shoulder around 750 nm for the IsPadC^Reg2 crystal spectra (Figure 2—figure supplement 3). While a quantitative comparison to wild-type IsPadC crystal spectra is complicated by the preferred orientation of molecules in the crystal lattice and the resulting effect on extinction coefficients, the almost identical crystal lattices and morphologies of the two crystals enable a qualitative comparison of their spectra after baseline subtraction and scaling to their Soret band absorption. The comparison of both crystal spectra clearly demonstrates that the IsPadC^Reg2 crystals contain a mixture of P_r and P_fr populations that reflects the refined P_fr/P_r heterodimer.

The asymmetry of the IsPadC^Reg2 structure extends from the biliverdin environment and the associated PHY-tongue to the PHY domain itself. Refolding of the tongue hairpin provides the structural flexibility for the PHY domain of chain A to reposition relative to chain B resulting in pronounced rearrangements at the PHY dimer interface. This repositioning of the PHY dimer correlates with the structural rearrangements of the coiled-coil to populate the register 2 conformation. In fact, the terminal PHY-helix directly extends into the coiled-coil sensor-effector linker and the whole structural element undergoes a hinge-like motion that is related to the structural transitions observed for various parallel phytochrome dimers (Gourinchas et al., 2017) (Figure 2c). Considering that more than 40 residues are part of this extended helical region, the global structural rearrangements are a complex mixture of individual sub-regions. Especially for the central part of the coiled-coil, the hinge-like separation of the corresponding residues is not very pronounced (Figure 3 and Figure 3—figure supplement 1) and, due to the absence of any rotational contribution to the transition between registers 1 and 2, is approximated as a quasi-translational movement throughout the manuscript. Interestingly, the hinge-like rearrangement at the central helical spine of the phytochrome module breaks the Arg327-Asp494 salt bridge at the PHY interface. This places the sidechains of Arg327 of both protomers in close proximity potentially contributing to the dynamic nature of the PHY interface upon illumination (Figure 3—figure supplement 1c).

Figure 3 with 2 supplements see all

Download asset Open asset

The quasi-translational reorientation of the coiled-coil register extends to the very end of the sensor-effector linker, which is directly linked to the strictly conserved DXLT motif of DGCs (Römling et al., 2017). The structural consequences of the coiled-coil rearrangement are a different spacing of the two structural elements corresponding to the functionally important DXLT sequence in the dimeric assembly. Since these structural elements are directly linked to the open active site at the interface of the two GGDEF domains, the coiled-coil register switching apparently has direct implications for DGC activity. In this context, it should be emphasized that the IsPadC^Reg2 structure has been solved in almost the same crystal lattice as the wild-type IsPadC dark state structure. This implies that the conformational rearrangements of the DGC dimer are associated with the coiled-coil register transition and the asymmetric activation of the sensory module rather than being affected by lattice specific crystal contacts. Interestingly, superimposing the DGC dimer onto itself by aligning one protomer with the corresponding other protomer (residues 529–683 chain A over chain B) reveals a pronounced asymmetry of the DGC dimer interface in IsPadC^Reg2 (Figure 2c), which significantly differs from the DGC dimer interface of the IsPadC dark state (RMSD 0.370 over 141 α-carbon and RMSD 0.267 over 138 α-carbon of chain A superimposed over chain B for IsPadCReg2 and IsPadC, respectively). Overall we observe a rotational movement of the DGC protomer linked to the repositioned coiled-coil linker helix. This leads to an opening of the DGC dimer at one side and a closing at the other side (Figure 2c and Figure 3—figure supplement 2a–b).

In the absence of substrate or product bound to the IsPadC^Reg2 structure it is, however, not possible to associate a specific functional state to this DGC architecture. Nevertheless, the fact that IsPadC^Reg2 features constitutively high GTP turnover suggests that the conformational opening of the DGC dimer might either increase the affinity for GTP and/or bring two GTP moieties after binding into an energetically favorable position to catalyze the first phosphodiester bond formation. Interestingly, the DGC dimer architecture does not resemble any deposited GGDEF dimer in an inhibited conformation (Chan et al., 2004; Wassmann et al., 2007; De et al., 2008; De et al., 2009; Navarro et al., 2009), suggesting that the observed dimer might correspond to a functionally important conformation along the catalytic cycle of DGCs. In fact, the relative position of the two catalytic GGEEF sites at the center of the DGC dimer is affected by the asymmetric rearrangements in IsPadC^Reg2 and the net effect is a shortened distance between the GGEEF sites potentially facilitating the contact between the C3 ribose hydroxyl group of one GTP and the alpha phosphate group of the respective other GTP molecule (Chan et al., 2004; Schirmer and Jenal, 2009) (Figure 3—figure supplement 2c). This could provide the geometry required for the initial phosphodiester bond formation step, generating the pppGpG intermediate. Subsequent structural rearrangements would then be required to complete the second phosphodiester bond formation step and ultimately the release of the c-di-GMP product. That the observed architecture corresponds to one functionally important state along this pathway is further supported by the observed asymmetry of specific conserved residues close to the binding site of the GTP base (e.g. Lys569 and Tyr670) that change their rotamer positions and relative orientation to the other protomer (Figure 3—figure supplement 2d). In fact, these structural rearrangements are similar to changes induced by substrate binding as observed in the GTP soaked wild-type structure (Gourinchas et al., 2017).

Considering the importance of dynamic rearrangements during the catalytic cycle of converting two molecules of GTP to c-di-GMP via the pppGpG intermediate, it is unlikely that a single conformation observed in a crystal structure will correspond to ‘the active’ conformation of DGCs. To further address the importance of conformational dynamics on DGC activity and on the coupling of the coiled-coil region with both the enzymatic effector and the red-light sensor, we performed hydrogen-deuterium exchange experiments analyzed by mass spectrometry (HDX-MS).

Extending our previous HDX-MS characterization of wild-type IsPadC (Gourinchas et al., 2017), we performed a complementary analysis with the IsPadC^Reg2 system, which is almost constitutively active but still shows residual light regulation of enzymatic activity. Comparing just the differences in deuterium incorporation between light- and dark-state conditions for the two systems reveals no major differences. For both the wild-type and the IsPadC^Reg2 variant we observed an increase in conformational dynamics in the PHY-tongue region, the central helical spine and the coiled-coil sensor–effector linker upon illumination (Figure 4). This increase in conformational dynamics, especially for the coiled-coil linker, has recently been implicated in increasing DGC activity by providing the conformational flexibility to switch between the two coiled-coil registers (Gourinchas et al., 2017). A closer inspection of the deuterium exchange kinetics, however, revealed that the conformational dynamics of IsPadC^Reg2, compared to the wild-type protein in either the dark-adapted state or during constant red light illumination, are reduced in several functionally important regions. This is especially true for the PHY-tongue region, the coiled-coil and a GGDEF element close to the GTP binding site (Figure 4 and Figure 4—figure supplement 1). The stabilization observed at the DGC dimer interface directly reflects the molecular structure of IsPadC^Reg2 with respect to the close interactions between the α0 helices and the DXLT motifs as well as the interaction of the α1/α2 and α4 helices on the closed side of the DGC interface. Obviously, the general stabilization of the GGDEF dimer also affects the conformational flexibility of the GGEEF motif.

Figure 4 with 2 supplements see all

Download asset Open asset

Of special interest are the reduced conformational dynamics of the coiled-coil, since we previously correlated an increase in conformational dynamics with an increase in activity. Because IsPadC^Reg2 is more active than the wild-type in the dark, the reduced dynamics observed for this variant suggest that it already populates a stable register 2 conformation in solution. The increase in deuterium incorporation upon illumination therefore only increases the dynamics of the system rather than switching to a specific register. Since IsPadC^Reg2 features a classical P_r state in solution, the structural rearrangements in the coiled-coil can apparently occur uncoupled from structural rearrangements in the biliverdin environment, for example linked to P_fr formation as observed in the crystal structure. These findings support the proposed switching mechanism between the two coiled-coil registers and show that both a structural component and a dynamic contribution are responsible for regulating enzymatic activity of the output module (Figure 5).

Figure 5

Download asset Open asset

As far as the proposed asymmetry of IsPadC upon illumination is concerned, the HDX-data cannot give a decisive answer. However, it is interesting to note that several peptides at the central helical spine, the PHY-tongue and the coiled-coil feature broad isotope distributions during deuterium incorporation (Figure 4—figure supplement 2). These broad distributions are indications for bimodal deuterium uptake that originates from two molecular species with different deuterium exchange kinetics, which would be in line with two biliverdin environments coupled to different PHY-tongue conformations under steady-state light conditions.

In the IsPadC (Gourinchas et al., 2017) system (Figure 1a), the bacteriophytochrome module is linked to a diguanylyl cyclase (DGC) effector (Ryjenkov et al., 2005) and controls formation of the bacterial second messenger cyclic-dimeric-GMP (Römling et al., 2013) by red light (Figure 1a–b). A closer inspection of illumination properties of IsPadC revealed red light activation already at low photon flux densities, but also showed only partial P_fr formation at saturating light intensities (Figure 1c–d). Consequently, light-activated absorption spectra differ significantly from those of systems featuring highly stable P_fr states (Yang et al., 2009; Burgie et al., 2016). Nevertheless, acidic denaturation at saturating light conditions revealed more than 90% of the biliverdin to be isomerized to the 15E state (Figure 1e). In comparison to the D. radiodurans phytochrome (DrBphp) dimer that features two isomerized chromophores under saturating light (Figure 1e) and allows both phytochrome-tongue (PHY-tongue) elements to be refolded (Takala et al., 2014; Burgie et al., 2016), we hypothesized that differences in the local environment of the bilin cofactors or at the dimeric interface of IsPadC destabilize the establishment of P_fr environments on both protomers. Therefore, a fraction of chromophores, albeit being isomerized, might be trapped in a global P_r-like phytochrome environment that does not allow refolding of the tongue to populate the P_fr state. Based on the spectral characteristics of IsPadC in comparison to several variants of DrBphP (Wagner et al., 2008) and the importance of P_fr/P_r heterodimers in plant phytochrome-B signaling (Klose et al., 2015) the local structural environment of the biliverdin cofactor in the non-P_fr protomer most likely reflects a meta-R like state (Borucki et al., 2005) where side chains close to the biliverdin adapt their rotamer positions, however no tongue refolding can take place.

In general, excitation of the biliverdin chromophore (Mroginski et al., 2007; Toh et al., 2010) is followed by fast isomerization and slower local structural changes around the chromophore binding pocket that are propagated to the effector domains in full-length proteins (Gourinchas et al., 2017; Björling et al., 2016). Refolding of the PHY-tongue region from a β-hairpin to an α-helical structure (Takala et al., 2014; Yang et al., 2009; Anders et al., 2013; Burgie et al., 2016) also results in rearrangements at the phytochrome dimer interface and highlights the cross-talk between biliverdin activation and conformational dynamics of the dimeric assembly. Along the line of the relevance of the dimer interface, disruption of the PAS-GAF bidomain dimerization has been shown to impact thermal reversion of P_fr in the case of DrBphP (Takala et al., 2015). Similarly, allosteric effects of the dimer interface were shown to influence the mode of biliverdin interaction in engineered fluorescent proteins (Stepanenko et al., 2016). The influence of the dimeric interface on properties of the two biliverdin cofactors is further supported by the stabilization of the coiled-coil sensor-effector linker in different constructs of IsPadC. Following thermal relaxation of the variants revealed a pronounced effect of stabilization of register 2 on the recovery of the P_fr contribution. At the same time, the fast reverting meta-R component is only slightly influenced by the conformational dynamics at the dimer interface. Most likely, the molecular mechanism of meta-R recovery follows a shunt-pathway similar to that described for Agrobacterium fabrum, Agp1 and Agp2 (Buhrke et al., 2018). The different effects of the dimer interface on the recovery of the different biliverdin species are an additional indication for the structural asymmetry of light-activated IsPadC. In fact, the correlation of increased P_fr lifetimes with increased stabilization of the register 2 conformation of the linker at the same time as sustained prevention of P_fr formation on the second protomer, as indicated by constant amplitudes for the recovery of the meta-R component (Table 1 in Supplementary file 1), shows that the biliverdin environments are functionally coupled via cooperative effects at the dimeric assembly (Klose et al., 2015). This special case of negative cooperativity indicates that P_fr associated conformational changes in one protomer break the global symmetry of the dimer and result in an asymmetric assembly that prevents reorganization of the second biliverdin binding site and/or associated functional elements such as the PHY-tongue region.

Further support for the stabilization of heterodimers via the dimer interface originates from the characterization of an IsPadC^{Δ442-477::SG} variant (Gourinchas et al., 2017) in which the PHY-tongues have been replaced by a short loop. The increased fluorescence and the fact that only half of the biliverdin population can be isomerized to the 15E state (Figure 1—figure supplement 3b–c) support the importance of the PHY-tongue elements for stabilization of the biliverdin environments. At the same time, the observed asymmetry in the IsPadC^{Δ442-477::SG} construct shows that structural changes at the dimer interface can stabilize different chromophore environments upon light activation of IsPadC; based on the reduced amplitude and the blue-shifted maximum of the light-dark difference spectrum (Figure 1—figure supplement 3a), we assume a composite of P_r and meta-R states to be present upon light activation of IsPadC^{Δ442-477::SG}.

In IsPadC, structural rearrangements at the dimer interface modulate the conformation of the coiled-coil sensor-effector linker and thereby regulate DGC activity (Gourinchas et al., 2017). More precisely, the coiled-coil sequence of the linker region adopts two conformational registers leading to either DGC activation or inhibition (Figure 1b, and Figure 1—figure supplement 1). In fact, the stabilization of the coiled-coil in its inhibited register (IsPadC^Reg1) led to low, light-independent basal activity comparable to the IsPadC dark state. By contrast, stabilizing the stimulating register (IsPadC^Reg2) led to a highly active variant with a residual light-induced increase in enzymatic activity. The detailed characterization of variants of IsPadC featuring different degrees of stabilization of the inhibiting or stimulating linker registers revealed significant differences in functional regulation of the enzymatic effector (Figure 1b, and Table 1). In combination with the effects of the coiled-coil variants on properties of the photocycle described above, their direct influence on the enzymatic functionality of the output module highlights the central role of the coiled coil linker in both integrating the light signal and regulating DGC activity. Interestingly, the IsPadC^Reg2 variant features a high dark state activity and can be further stimulated by red light illumination, whereas IsPadC^Reg2.a, for which two conserved Asn residues in the coiled-coil element were exchanged by hydrophobic valines, is constitutively active and can no longer be affected by illumination (Figure 1b, and Table 1). The observed effect of substituting the Asn residues supports their functional relevance at central positions of the heptad units (Gourinchas et al., 2017; Fletcher et al., 2017) and indicates that a conformational switching between coiled-coil registers is required for optimal activation of DGC activity (Figure 5).

The crystal structure of the IsPadC^Reg2 variant stabilized in the stimulating register shows that the structural rearrangements of the coiled-coil transition between the two registers, as required for signal integration (Gourinchas et al., 2017), are compatible with P_fr state formation in a single protomer. In combination with the spectral signature of a P_fr/meta-R heterodimer observed upon illumination of all IsPadC constructs, this further supports the possibility of an IsPadC heterodimer as a central functional species (Figure 1d). Upon isomerization and D-ring rotation of one biliverdin the conformational changes induced at the dimer interface might prevent structural reorganization of the PHY-tongue region close to the second biliverdin binding site. Depending on the degree of coupling between various functional elements, different equilibria of P_r, lumi-R, meta-R and P_fr states can influence the ultimate structural rearrangements in the phytochrome systems. Based on the assignments of meta-R transitions that involve structural rearrangement of the tongue region in other phytochrome systems (Björling et al., 2016; Stojković et al., 2014; Anders et al., 2014), we suspect that an early meta-R like state is enriched during steady state light illumination in IsPadC. The different stabilities of P_fr states in various dimeric phytochromes might therefore have evolved to optimize specific action spectra as well as to enable the formation of functional heterodimers. Eventually, the possibility to tune the relative amounts of various biliverdin environments by functionally important structural elements might also be involved in the adaptability of the sensory module to different output domains.

The influence of the dimeric interface on the spectral photoisomerization properties has previously also been observed in inhibited photoconversion of the dimeric chromophore-binding domain of D. radiodurans phytochrome that can partially be reestablished upon monomerization (Auldridge et al., 2012). In fact, including the PHY domain in D. radiodurans phytochrome PSM constructs or addressing the full-length system results in full conversion to a P_fr/P_fr homodimer (Takala et al., 2015; Auldridge et al., 2012). Apparently, nature employs both P_fr/(meta-R or P_r) heterodimers and P_fr/P_fr homodimers as functional units for regulating downstream effectors. This appears to be true for bacterial phytochromes as well as evolutionary distant plant phytochromes, where asymmetric properties on the level of different protomers as well as different functional states of individual phytochromes are being employed to regulate complex cellular processes (Clack et al., 2009; Liu and Sharrock, 2013; Klose et al., 2015; Hennig and Schäfer, 2001; Sharrock and Clack, 2004). However, when considering other organisms, the description of functional asymmetry in photoreceptor oligomers is rather scarce (Heintz and Schlichting, 2016; Winkler et al., 2013), although it apparently represents an interesting evolutionary approach to modulate photochromic behavior and signaling diversities. In general, asymmetry is emerging as a critical aspect of signal integration (Neiditch et al., 2006; Narayanan et al., 2014; Chervitz and Falke, 1996) as well as regulation of output functionalities (Kim et al., 2017; Hu et al., 2013). Based on spectral data provided for several phytochrome systems, functional activation through P_fr state formation in a single protomer might be a prevalent mechanism conserved in the family of phytochrome-like sensors that has so far been underestimated (Tarutina et al., 2006; Yang et al., 2007; 2015; Loughlin et al., 2016; Essen et al., 2008).

Interestingly, both the wild-type and the IsPadC^Reg2 crystal structures have been obtained from crystals grown under constant dark conditions. Even though both constructs display very similar P_r state spectra in solution (Figure 1c and ref [Gourinchas et al., 2017].), IsPadC^Reg2 always features a small shoulder around 750 nm indicating a subpopulation of species present in the P_fr conformation. Apparently the crystallization conditions favor the assembly of asymmetric P_fr/P_r heterodimers existing in equilibrium with the predominant P_r/P_r homodimer in solution. These heterodimers feature the coiled-coil in its stimulating register, which stabilizes one protomer in the P_r and the other in the P_fr conformation (Figure 2 and Figure 3). The enrichment of the P_fr species only occurs during crystal growth and not in solution, confirming that the amino acid substitutions in IsPadC^Reg2 alone are not responsible for the stabilization of P_fr under dark conditions, as observed in bathyphytochromes (Zienicke et al., 2013). Nevertheless, the P_fr/P_r heterodimer should not be considered a crystallization artifact, but rather reflects one conformational assembly of the full-length system that is energetically compatible with the stabilization of the coiled-coil in its stimulating register.

The pronounced asymmetry in the IsPadC^Reg2 structure is characterized by one protomer in the P_r-state while the second protomer features a P_fr-conformation with its biliverdin chromophore in the 15E configuration and the associated PHY-tongue element rearranged into the characteristic P_fr α-helical conformation. Obviously in the refolded state, the PHY domains exhibit greater flexibility, leading to a less defined electron density for this region compared to the rest of the protein. The directly linked PHY-tongue elements that are stabilized by interactions with the GAF domains, however, feature a well-defined electron density leaving no ambiguity regarding their differences in secondary structure (Figure 2—figure supplement 1). The P_r-state protomer, with its characteristic β-hairpin conformation in the tongue element, superimposes very well with the corresponding protomer of the IsPadC dark state structure (RMSD of 0.26 Å over 347 Cα-atoms for the PSM alignment for the respective chains B, Figure 3—figure supplement 1). The superposition is only disturbed by some flexibility towards the end of the coiled-coil linker and the associated DGC. The P_fr protomer deviates from the P_r state of wild-type IsPadC as well as the P_r protomer of IsPadC^Reg2 by refolding of the PHY-tongue region into an α-helical structure that leads to a bending at the helical spine region triggering a quasi-translational movement of the associated PHY – coiled-coil – DGC region (RMSD of 1.56 Å over 402 Cα-atoms for the PSM alignment of IsPadC^Reg2 on IsPadC of the respective chains A). The directly linked GGDEF protomer repositions relative to the second DGC protomer increasing the asymmetry within the GGDEF dimer. By calculating the free energy of assembly dissociation (ΔG^diss) by PISA analysis (Krissinel and Henrick, 2007), we observed that the energetic stability of the PAS-GAF-PHY interface (residues 8–500) in the IsPadC^Reg2 structure is only slightly affected compared to the one calculated for IsPadC (Gourinchas et al., 2017) (ΔG^diss = 6.5 kcal mol⁻¹ and 8.8 kcal mol⁻¹, respectively). Moreover, the stability of the PAS-GAF-PHY-coiled-coil assembly in the IsPadC^Reg2 structure is in the range of the same assembly calculated for the IsPadC structure (Gourinchas et al., 2017) (ΔG^diss = 20.0 kcal mol⁻¹, and ΔG^diss = 22.6 kcal mol⁻¹, respectively), which suggests that the overall stability of the phytochrome sensor is only slightly affected by the structural rearrangement of the PHY-tongue and the coiled-coil linker. A closer look at the dimeric PHY interface shows that upon linker helix translation the Arg327 - Asp494 intermolecular salt bridges observed in the dark state structure of IsPadC are absent in the IsPadC^Reg2 structure thereby enhancing the conformational freedom at the PHY interface (Figure 3—figure supplement 1c). Interestingly, the asymmetry of the IsPadC^Reg2 structure is also reflected in several other amino acids at the dimeric interface, for instance Phe132, Gln137, Arg140, Met320 for the PSM interface, and His577, K569, Y670 at the DGC interface. Upon isomerization of the biliverdin cofactor, Tyr168 and Phe195 change their rotamer conformation to accommodate the D-ring rotation as already described for other phytochrome species (Yang et al., 2009; Anders et al., 2013; Burgie and Vierstra, 2014). Also His193 changes its rotamer to interact with the carbonyl oxygen of the rotated D-ring (Figure 3b–c). In addition, the tryptophan switch scenario (Anders et al., 2013; Burgie and Vierstra, 2014) is also observed for refolding of the IsPadC PHY-tongue.

As far as downstream signaling of the activated PSM heterodimer is concerned, the crystal structure supports the switching mechanism between two coiled-coil registers proposed recently (Gourinchas et al., 2017). In contrast to the anticipated rotation mechanism, we observe a quasi-translational rearrangement (Figure 3a,d, and Figure 3—figure supplement 1) that is linked to a hinge movement of the whole PHY-coiled-coil linker region (Figure 2d). This type of functional rearrangement of the phytochrome dimer interface has also been observed by comparison of various phytochrome structures already at the PSM level (Gourinchas et al., 2017; Yang et al., 2015). More precisely, the refolding of the PHY-tongue is linked to a tilting of the central helical spine – PHY – coiled-coil region. This eventually causes the C-terminal parts of the coiled-coil linker to interact more tightly and repositions the directly following DXLT motifs at the dimer interface of the GGDEF domains. This highly conserved motif is directly linked to structural elements that are part of the open active site at the dimer interface of the GGDEF domain (Schirmer, 2016). In the absence of substrate, the coiled-coil linker rearrangement induces a different DGC dimer architecture due to a displacement of the DGC protomer linked to the photoactivated phytochrome protomer (Figure 2c). This results in a more open interface at one side of the dimer that potentially facilitates substrate binding or product release. In addition, we observed that the GGEEF motifs of IsPadC^Reg2 are positioned closer to each other, indicating that the two GTP substrates could be more closely interacting and potentially initiating the chemical reaction between the two molecules in an asymmetric fashion, as opposed to the symmetric binding of GTP in the dark inhibited conformation (Gourinchas et al., 2017). It should be noted, however, that the complex reaction mechanism of DGCs involves coordination of two GTP molecules, initial bond formation to form the asymmetric pppGpG intermediate, structural rearrangements to form the second linkage of c-di-GMP and subsequent product dissociation. Therefore, a single crystallographic snapshot cannot correspond to ‘the active’ conformation of DGCs; the dynamic nature of GGDEF dimerization (Gourinchas et al., 2017; Chan et al., 2004; Schirmer, 2016; Zähringer et al., 2013) rather supports a model in which the coiled-coil register switching influences the conformational sampling of different dimeric interfaces. In the case of the stimulating register 2 conformation of the coiled-coil, the sampling of dimers compatible with the geometry required for the DGC reaction to proceed is facilitated. Since an additional increase in activity can be achieved by stimulating the conformational dynamics of the coiled-coil by illumination, not a static dimer interface, but rather a dynamic assembly of GGDEF protomers, enables optimal DGC activity.

Interestingly, the superposition of the IsPadC^Reg2 structure with the GTP soaked IsPadC structure (Gourinchas et al., 2017) suggests that GTP binding within the DGC dimer induces a slight translation of the coiled-coil linker towards the stimulating register (Figure 3—figure supplement 1a,b). Therefore, the full-length system involves a complex cross-talk between sensor and effector domains that influence each other based on finely tuned dimer interfaces. The coiled-coil as central communication hub integrates the functional state of input and output modules, indicating that the rational design of artificial sensor-effector couples needs to consider this complex interplay of dynamic dimeric interfaces of sensor and effector domains as well as properties of the coiled-coil linker.

The importance of conformational dynamics was also confirmed by hydrogen-deuterium exchange coupled to mass spectrometry (HDX-MS) experiments (Figure 4, and Figure 4—figure supplement 1). Interestingly, phytochrome activation by red light results in similar effects for both IsPadC and IsPadC^Reg2. Both constructs showed an overall increase in deuterium uptake upon red light illumination indicating a light-induced increase of conformational dynamics. However, the relative amount of deuterium exchanged under dark and light conditions is lower for peptides in functionally important regions of IsPadC^Reg2 compared to IsPadC. This general decrease of conformational dynamics of IsPadC^Reg2 highlights the molecular cross-talk between different functionally important elements (Gourinchas et al., 2017). The decrease of deuterium uptake of coiled-coil peptides correlates with the stabilization of the linker in the stimulating register conformation as anticipated for the amino acid substitutions in IsPadC^Reg2. The fact that this variant has a higher turnover rate constant in the dark compared to IsPadC indicates that the coiled-coil conformation in the stimulating register has a positive effect on enzymatic activity of the DGC (Figure 5a). However, HDX-MS shows that illumination increases the dynamics of the coiled-coil and thereby destabilizes the coiled-coil arrangement, which results in an additional increase in enzymatic activity; this implies that, in addition to the register conformation, the overall dynamics of the system play an important role in modulating DGC activity (Figure 5a).

Importantly, HDX-MS measurements derive from an ensemble measurement of all peptides in the protein dimer. Therefore, it is not possible to specifically address asymmetric stabilization or destabilization in the context of the full-length protein dimer in solution. However, the broad deuterium uptake distributions of PHY-tongue peptides upon illumination (Figure 4e) are indicative of a deuterium exchange regime that considers two conformational states with pronounced differences in deuterium uptake kinetics. Even if these observations are no direct proof of a heterodimer in solution, they suggest that different conformational states of the PHY-tongue elements co-exist upon light activation of the PadC system. This conformational coupling is also reflected in the crystal structure of IsPadC^Reg2 and its pronounced differences to the inactive wild-type structure. Even though the P_fr/P_r heterodimer conformation is not the energetically most favorable in solution, it does represent one functional form of the dimeric assembly that highlights the cross-talk between the site of light activation and the central regulatory element for enzymatic activity – the coiled-coil linker. Additional experiments to address the directionality of signal transduction by producing apoproteins of IsPadC^Reg2 and wild-type IsPadC could provide interesting additional insights, however, the isolation of apoprotein is not feasible in the IsPadC system.

Figure 5 summarizes the observed influence of the coiled-coil register conformation and the dynamics of the system for regulating output activity of the DGC in a simplified model. To that end, the structure of IsPadC^Reg2 and the structural changes associated with activation will help to readdress the activation mechanism of other phytochromes that are commonly used for designing optogenetic tools (Tischer and Weiner, 2014; Shcherbakova et al., 2015). In the context of phytochrome function, our study confirms that functional P_fr/meta-R or P_fr/P_r phytochromes exist and that, while chromophore isomerization coupled to reorganization of the PHY-tongue region appears to be conserved in phytochromes, the extent to which this affects the output domain is tuned by the dimeric phytochrome interface via modulation of the effective population of P_fr-states as characterized by the structural rearrangements of the PHY-tongue and the PHY dimer interface. Ultimately, the flexibility to stabilize different functional states via the dimer interface enables phytochromes to be adapted to various effectors and also enables tuning of the action spectra by reducing the cross-section with inactivating far-red light.

Linker register variants were produced by site-directed mutagenesis following the protocol described by Liu et al. (Liu and Naismith, 2008) using pETM-11 IsPadC based plasmids as templates (Gourinchas et al., 2017) (used primers are listed in Table 3 in Supplementary file 1).

IsPadC variants were expressed and purified as described in detail previously (Gourinchas et al., 2017). Briefly, (His)₆-tagged holoproteins were expressed in BL21 (DE3) containing the previously generated pT7-ho1 helper plasmid (Gourinchas et al., 2017) to co-express heme oxygenase (HO-1). Cells grown at 37°C in LB medium completed with 0.36% (w/v) glucose and 10 mM MgCl₂ were cooled to 16°C and the medium was supplemented with 10 mg liter⁻¹ δ-aminolevulinic acid prior to induction with 0.25 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for ~15 hr. After cell harvesting, the soluble lysates obtained by combined treatment with lysozyme (100 µg ml⁻¹) and sonication (2 × 5 min, 100 W, continuous mode; Labsonic L, Sartorius, Göttingen, Germany) were separated by ultracentrifugation (206,000 g), and affinity-purified on a Ni²⁺-Sepharose matrix (Ni Sepharose 6 Fast Flow, GE Healthcare, Little Chalfont, UK). After a washing step (lysis buffer containing 30 mM imidazole), proteins were eluted (lysis buffer containing 250 mM imidazole) and dialyzed overnight at 4°C in presence of histidine-tagged tobacco etch virus (TEV) protease (~1:16 TEV/substrate). Afterwards TEV and cleaved histidine tag were removed from the protein solution by reloading the dialyzed protein onto the Ni²⁺ column and collecting the flow through. An additional step of size-exclusion chromatography on a 16/60 Superdex 200 prep grade column (GE Healthcare) was performed before proteins were concentrated in the buffer listed in Table 3 in in Supplementary file 1 (Amicon Ultra-15; Merck Millipore, Burlington, MA), aliquoted, flash-frozen in liquid nitrogen and conserved at −80°C until needed.

For characterization of steady-state kinetics of the IsPadC variants, the proteins were subject to an additional size-exclusion chromatography run with a 10/300 Superdex 200 increase analytical grade column (GE Healthcare) in order to exchange the protein into HPLC buffer (Table 3 in in Supplementary file 1) and to remove potential protein aggregates caused by sample thawing.

To screen for diguanylyl cyclase activity, we followed our previously described protocol (Gourinchas et al., 2017) adapted from the method described by Antoniani et al. (Antoniani et al., 2010). E. coli BL21 (DE3) co-expressing the pT7-ho1 plasmid and transformed with pETM-11 IsPadC based plasmids were cultivated at 30°C to OD⁶⁰⁰ 0.5 in YESCA medium (casamino acids 10 mg ml⁻¹ and yeast extract 1.5 mg ml⁻¹) supplemented with MgSO₄ (0.05 mg ml⁻¹), FeSO₄ (0.5 µg ml⁻¹), kanamycin (30 µg ml⁻¹), and chloramphenicol (34 µg ml⁻¹). After 4 hr induction at 16°C with 0.25 mM IPTG in the presence of 10 mg liter⁻¹ δ-aminolevulinic acid, 3 µl of the induced culture were spotted on YESCA agar plates containing 30 µg mL⁻¹ kanamycin, 34 µg mL⁻¹ chloramphenicol and 0.01 mg mL⁻¹ congo-red, and incubated at 20°C for 16 hr in the dark or under constant red light illumination (630 nm, 75 µW cm⁻²). As negative control we used a pETM-11 AppA construct (Winkler et al., 2013) that does not show any DGC activity.

UV-Vis absorption spectra were acquired with a CCD-based Specord S300 Vis spectrophotometer (Analytic Jena, Jena, Germany) using protein samples equilibrated at 20°C and diluted in their corresponding storage buffers (Table 3 in Supplementary file 1). Dark-adapted P_r-state absorption spectra were measured under non-actinic conditions by minimizing the contact time with measuring light using a neutral density filter (ND = 1.0; Thorlabs, Newton, NJ) between the light source and the sample cuvette. P_fr-enriched spectra were recorded under constant red light irradiation (660 nm, 45 mW cm⁻², Thorlabs) in the presence of the same neutral density filter for the measuring light.

To record changes in P_fr/P_r ratio under various photon flux densities, the red LED (660 nm, Thorlabs) was adjusted to different light intensities at a distance of 60 cm from the sample. After establishing a steady-state at each light intensity, spectra were measured and absorbance values at 750 nm were extracted. Each light intensity measurement was acquired from the corresponding lower to the maximal intensity setting before allowing the sample to recover to its ground state and repeating the measurement in a technical triplicate. The sample standard deviation of individual measurements is used as error indicator in Figure 1d.

P_r-state recovery kinetics were followed at 710 and 750 nm after 1 min red light illumination (660 nm, 45 mW cm⁻², Thorlabs) using a Specord 200 Plus spectrophotometer (Analytic Jena) with 10 ms integration time sampled every 5 s and the recorded time constants differ by less than 10% between 700 and 750 nm measurements.

To record the UV-Vis absorption spectra of the denatured biliverdin bound proteins, we followed the previously described method from Thümmler et al. (Thümmler et al., 1981). The concentrated protein sample was kept in the dark, or irradiated 1 min with saturating red light (660 nm, 45 mW cm⁻², Thorlabs), before diluting it 1:10 in quenching buffer (0.1% trichloroacetic acid in methanol) to a final concentration of 1–2 μM. The absorption spectra of the denatured P_r- or P_fr-states were then directly recorded. Re-illumination of P_fr-peptides by red light triggers the isomerization of biliverdin back to the P_r-state and therefore P_fr-peptides were denatured directly after switching off the red light source and kept in the dark until their measurement.

Fluorescence measurements were performed using an RF-5301 PC spectrofluorophotometer (Shimadzu, Kyoto, Japan) equipped with a concave blazed holographic grating excitation and emission monochromator, and containing a 150 W Xenon lamp (Ushio, Cypress, CA) as light source. Samples diluted to 2 μM were equilibrated at room temperature and complete darkness for the dark state measurement or 1 min under saturating red light (660 nm, 45 mW cm⁻², Thorlabs) for the light state measurement prior to data acquisition. A sampling interval of 0.2 nm was used with a slit opening of 1.5 nm for the excitation slit and 5 nm for the emission slit.

The IsPadC^Reg2 variant was crystallized at 293 K under dark conditions using a sitting-drop vapor diffusion setup. To obtain the best diffracting crystals, 2 µl drops containing equal volumes of protein solution at 5 mg ml⁻¹ and reservoir solution (0.1 M magnesium formate, 12% (w/v) PEG 3,350), were seeded using the method of streak seeding and equilibrated against 35 µl of reservoir solution. Crystals appeared after overnight incubation and reached final dimensions within 10 days.

Crystals were harvested under low intensity non-actinic green light conditions (520 ± 20 nm LED) by transferring the crystals to a cryprotectant solution (reservoir solution containing 30% glycerol) and subsequent flash-freezing in liquid nitrogen. Diffraction data for IsPadC^Reg2 were collected at beamline P11 of the Deutsches Elektronen-Synchrotron (DESY). Data were processed using the XDS program package (Kabsch, 2010) (Table 2 in Supplementary file 1).

The crystal structure of IsPadC^Reg2 variant was solved by molecular replacement using PHENIX Phaser (McCoy et al., 2007) with the PAS-GAF fragment and the PHY-domain lacking the tongue extensions of the full-length structure of IsPadC (pdb 5LLW), as well as the DGC fragment of the IsPadC GTP soaked structure (pdb 5LLX). Two molecules for the PAS-GAF, PHY, and DGC search model were successfully placed in the asymmetric unit. The initial model of a PAS-GAF-PHY-GGDEF dimer was then manually extended to build the PHY-tongue and coiled-coil linker regions in several rounds of maximum-likelihood refinement of models modified with Coot using σ_A-weighted 2mF_o – DF_c and F_o – F_c electron density maps. In addition, torsion-NCS restraints and secondary structure restraints were included together with TLS groups for the individual domains of the protein. During the final rounds of refinement reference model restraints were applied on PAS-GAF-PHY excluding the PHY β-hairpin extension using a previously solved high resolution IsPadC PSM structure (Gourinchas et al., 2017), and on GGDEF domains using the IsPadC dark state structure (Gourinchas et al., 2017) and optimization of X-ray and ADP weights was performed.

Spectral properties of the crystal were recorded at 100 K under a cryostream using a microspectrophotometer at the cryobench ID29S of the ESRF facilities (von Stetten et al., 2015).

To record the conversion of GTP to c-di-GMP for all the variants, we adapted a protocol (Gourinchas et al., 2017) for high performance liquid chromatography (HPLC) from a previously described method (Enomoto et al., 2014). Briefly, purified protein samples were mixed with GTP at various concentrations in reaction buffer (50 mM Hepes pH7.0, 500 mM NaCl, and 50 mM MgCl₂) at 20°C under non-actinic light for the dark state measurement, or under constant red light illumination (630 nm, 0.7 mW cm⁻², Luminea LED) following a 1 min pre-illumination of the sample for the light state measurements. Samples were then thermally inactivated by 1 min incubation at 95°C. After separating the substrate and products from denatured protein by centrifugation, the nucleotides were separated by a linear 7 min gradient from 2% to 20% MeOH using a reversed phase HPLC column (SunFire C18 4.6 × 100; Waters, Milford, MA) equilibrated in 10 mM triethylammonium formate (pH 6.0). For protein variants isolated in a partially activated state and featuring a slow thermal recovery a prolonged incubation of the sample under far-red light (730 ±20 nm, Thorlabs LED) coupled to a far-red light bandpass filter (750 nm, Thorlabs) at room temperature and complete darkness was performed to fully populate the P_r-state of the sample prior to the dark state measurements. All kinetic data were normalized to the concentration of the dimeric protein, and all samples were corrected for the amount of c-di-GMP formed during the inactivation step.

Deuterium labeling experiments were performed to address the effect of illumination on the conformational dynamics of IsPadC^Reg2 in order to compare it with the previously performed analysis of the IsPadC system (Gourinchas et al., 2017). Briefly, 200 µM protein sample were equilibrated at 20°C in the dark or under red light illumination (630 nm, 0.7 mW cm⁻², Luminea LED) for 1 min prior to starting the deuterium-labelling reaction by diluting the sample 20-fold in D₂O containing 10 mM Hepes (pD 7.0), 150 mM NaCl, and 10 mM MgCl₂. Reactions were quenched with ice-cold 200 mM ammonium formic acid (pH 2.5) after 10 s, 45 s, 3 min, 15 min, and 60 min and injected into a cooled HPLC setup as described previously (Winkler et al., 2015). To sum up, deuterated samples were digested on an immobilized pepsin column (Poroszyme; Applied Biosystems, Foster City, CA) operated at 10°C. Resulting peptides were desalted on a 2 cm C18 guard column (Discovery Bio C18, Sigma), and separated during a 7 min acetonitrile gradient (15–50%) in the presence of 0.6% (v/v) formic acid on a reversed phase column (XR ODS 75 x 3 mm, 2.2 µM, Shimadzu). Peptides were then infused into a maXis electrospray ionization-ultra high resolution-time-of-flight mass spectrometer (Bruker, Billerica, MA). Deuterium incorporation was analyzed and quantified using the Hexicon 2 software package (http://hx2.mpimf-heidelberg.mpg.de) (Lindner et al., 2014).

• Andreas Winkler