Deconstructing and repurposing the light-regulated interplay between Arabidopsis phytochromes and interacting factors

Phytochrome photoreceptors mediate adaptive responses of plants to red and far-red light. These responses generally entail light-regulated association between phytochromes and other proteins, among them the phytochrome-interacting factors (PIF). The interaction with Arabidopsis thaliana phytochrome B (AtPhyB) localizes to the bipartite APB motif of the A. thaliana PIFs (AtPIF). To address a dearth of quantitative interaction data, we construct and analyze numerous AtPIF3/6 variants. Red-light-activated binding is predominantly mediated by the APB N-terminus, whereas the C-terminus modulates binding and underlies the differential affinity of AtPIF3 and AtPIF6. We identify AtPIF variants of reduced size, monomeric or homodimeric state, and with AtPhyB affinities between 10 and 700 nM. Optogenetically deployed in mammalian cells, the AtPIF variants drive light-regulated gene expression and membrane recruitment, in certain cases reducing basal activity and enhancing regulatory response. Moreover, our results provide hitherto unavailable quantitative insight into the AtPhyB:AtPIF interaction underpinning vital light-dependent responses in plants.

applicable means for the bimodal control of cellular phenomena with supreme resolution in space and time 34 . As a case in point, the expression of transgenes in yeast and mammalian cells has been subjected to red-/far-red-light control via a two-hybrid strategy 25,35,36 . To this end, a split transcription factor was engineered with one component of the AtPhy:AtPIF pair connected to a sequence-specific DNA-binding domain and the other to a transcriptional trans-activating domain. Exposure to red light prompts colocalization of the two entities and onset of expression from synthetic target promoters. In another approach 27,37,38 , the AtPhy:AtPIF pair conferred light sensitivity on plasma membrane recruitment and cellular signaling cascades in mammalian cells. Although details differ, optogenetic applications to date mostly employ the isolated PCM of AtPhyB and the N-terminal 100 amino acids of AtPIF3/6, denoted P3.100 and P6.100, that comprise the APB motif.
Despite the eminent role of the AtPhy:AtPIF interaction in nature and optogenetics, quantitative data on the interaction strength and the underlying sequence determinants are scarce. To fill this gap, we dissected and analyzed the light-dependent interaction between AtPhyB and AtPIF3/6 by several qualitative and quantitative approaches. Whereas the AtPhyB PCM bound P6.100 with about 10 nM affinity in its Pfr state and showed no detectable affinity in the Pr state, P3.100 exhibited weaker Pfrstate affinity and elevated basal affinity in Pr. By deconstructing AtPIF3/6 and engineering a wide set of shortened variants, we pinpointed APB.A as decisive for light-regulated PPIs, with a modulatory role for APB.B. Quantitative analyses informed the construction of minimal AtPIF3/6 fragments of 25 and 23 residues, respectively, that retained stringently light-regulated PPIs with AtPhyB. When deployed for the optogenetic control of gene expression and membrane recruitment, the novel AtPIF variants with a range of interaction strengths achieved stratified and enhanced light responses.

Results
Deconstructing the AtPhyB:AtPIF interaction. Starting from the AtPIF constructs P3.100 and P6.100, we generated numerous derivatives with residues deleted from the N terminus, the linker between the APB.A and APB.B segments varied, or either segment omitted or duplicated (Fig. 1e, Supplementary Table 1). All AtPIF variants were C-terminally tagged with enhanced yellow fluorescent protein (EYFP) to promote protein solubility and facilitate concentration determination. We implemented a screening assay to efficiently probe interactions of these variants with the Pfr state of the AtPhyB PCM. The screen exploits the fact that AtPIF binding stabilizes the Pfr state of AtPhyB and decelerates the thermal reversion to the Pr state in the dark 39 (Fig. 2a). For this assay, the AtPIF-EYFP variants were expressed in Escherichia coli, purified AtPhyB PCM was added to the crude cell lysate in substoichiometric amounts, and the Pfr → Pr reversion kinetics were monitored by absorption spectroscopy (Fig. 2b, c). The initial kinetics were normalized to an EYFPnegative control and provide a convenient readout for interactions (Fig. 2d). Although qualitative in nature, this first screening platform offers important advantages: (i) owing to the specificity of the AtPhyB:AtPIF interaction, the assay can be conducted in crude bacterial lysate, without the need for protein purification; and (ii) it can be easily multiplexed to test many variants in a single experiment.
A multiple sequence alignment of the N-terminal regions of AtPIF1-8 delineates two regions of conservation that define the A and B segments of the APB motif ( Supplementary Fig. 1 24 ). The APB.A segment shows stronger conservation and comprises around 20 residues centered around the consensus core sequence ELXXXXGQ 24 ; by comparison, the APB.B region is considerably shorter and less conserved. As the very N-terminal region preceding APB.A varies substantially among the AtPIFs in length and sequence, we deemed it non-essential for AtPhyB interactions and removed it from P3.100 and P6.100. The resultant Px variants (here and in the following, x = 3, 6) retained interaction with the AtPhyB PCM, and all subsequent AtPIF variants were thus based on these N-terminally truncated forms ( Fig. 2d and Supplementary Fig. 2). Next, we interrogated the linkage between the constituent APB.A and APB.B segments, which is of heterogenous length and sequence across AtPIF1-8. We generated a set of variants, including (i) Px.L1 and Px.L2 in which the linkers of P3/P6 are shortened by 10 residues at their N and C termini, respectively; (ii) Px.LP1 in which said linker is substituted for the corresponding segment of AtPIF1, the shortest among all AtPIFs; and (iii) Px.LS in which the linker is replaced by a repetitive glycine-serine stretch of 10 residues. As gauged by their effect on dark-reversion kinetics (cf. Supplementary Fig. 2), all these variants still interacted with the Pfr state of the AtPhyB PCM. These results imply that the linker connecting the APB.A and APB.B segments is dispensable, which is confirmed in the Px. fus variants that directly link these two segments without any linker and still exhibit interaction with the AtPhyB PCM (cf. Fig. 2d and Supplementary Fig. 2). To assess whether productive AtPhyB binding mandates a specific topology of the APB segments, we generated the variants Px.BA and Px.BAfus with a The light-adapted Pfr state (brown) of AtPhyB thermally recovers to the dark-adapted Pr state (red) in a moderately paced reaction. When binding to an AtPIF variant, the recovery reaction is delayed. b AtPIF variants were C-terminally tagged with EYFP, expressed in Escherichia coli, cells were lysed, and AtPhyB PCM was added to the crude lysate. Samples were exposed to red light, and the recovery reaction was monitored over time by absorption measurements. c Normalized absorption of the AtPhyB PCM measured at 720 nm after red-light absorption in the presence of P3.100 (red) or the EYFP-negative control (gray). d The initial rates of the recovery reaction were determined and normalized to the reading obtained for the EYFP-negative control. Data indicate mean ± SEM of n = 3 independent biological replicates.
inverted sequential order of APB.A and APB.B, and the original linker sequence kept or removed, respectively. Again, these variants retained interactions with the Pfr state of the AtPhyB PCM (cf. Fig. 2d and Supplementary Fig. 2). Site-directed mutagenesis had previously ascribed a dominant role to APB.A in mediating the light-dependent interaction with AtPhyB 24 , and we hence probed the two segments of the composite APB motif separately. Both the APB.A-containing variants Px.A and the Px. As, with or without the N-terminal half of the respective linker, still showed interactions with the AtPhyB PCM as judged by the effect on dark-reversion kinetics (cf. Fig. 2d and Supplementary  Fig. 2). By contrast, neither the APB.B-based Px.B nor the Px.Bs variants, with or without the C-terminal half of the linker, respectively, exhibited interactions in this assay. Duplication of the A part in the variants Px.AA and Px.AAfus preserved interactions with the AtPhyB PCM, and vice versa, duplication of the B segment in Px.BB and Px.BBfus failed to restore them (cf. Supplementary Fig. 2). Taken together, our findings emphasize the dominant role of APB.A for mediating interactions with AtPhyB. To further characterize the APB.A segment, we successively trimmed residues flanking its ELXXXXGQ core sequence. However, even the removal of five weakly conserved C-terminal residues in the variants Px.A19 abolished interactions with AtPhyB, as judged by their inability to slow down the AtPhyB-PCM Pfr → Pr reversion kinetics (cf. Fig. 2d and Supplementary Fig. 2). Likewise, no interaction with the AtPhyB PCM was detected for more extensive truncations of the APB.A segment (cf. Supplementary Fig. 2).
Biochemical analyses of the AtPhyB:AtPIF interaction. The above screening platform affords a qualitative first-pass assessment of the AtPIF variants but does not quantify the strength of interactions with AtPhyB. Moreover, the assay is limited to interactions within the Pfr state but not the Pr state. We hence selected several of the above AtPIF candidates for in-depth analysis. Following expression and purification, we assessed the oligomeric state of these variants and of the AtPhyB PCM by sizeexclusion chromatography (SEC). In its Pr state, the isolated AtPhyB PCM elutes as a monomer with a minor homodimeric fraction, consistent with a recent SEC analysis 40 (Fig. 3a). In the Pfr state, the predominantly monomeric state is maintained but the retention from the SEC column is slightly delayed, which arguably reflects light-induced conformational changes, i.e., a compaction, of the PCM that may resemble those observed in bacterial Phys [6][7][8]10 (Fig. 3b). At a concentration of 10 µM, P3.100 and P6.100 largely eluted as homodimers with a minor monomeric population (Fig. 3c, Supplementary Fig. 3 and Table 1). Dimerization is not caused by the EYFP tag as the fluorescent protein itself eluted as a monomer (Fig. 3d, Table 1). Notably, the homodimeric state of AtPIFs is also observed in nature and critical for their physiological function as basic helix-loop-helix transcription factors 41 . Size reduction of the AtPIFs impaired homodimerization in several variants to different extent (Supplementary Fig. 3, Table 1). If the APB.A segment was truncated, as in P3.A and P6.A, or excluded altogether, as in P3.Bs and P6. Bs, homodimerization was lost completely. Taken together, these findings point toward a contribution of the APB.A segment to homodimerization of the current AtPIF variants and, by extension, of the intact AtPIF3 and AtPIF6 proteins 41 .
We next investigated the interactions between the AtPIF3/6 variants and the AtPhyB PCM by SEC (Fig. 3e, Supplementary  Fig. 4 and Table 1). To this end, we first converted the AtPhyB PCM to its Pfr state by illumination with red light (640 nm), incubated it at a 5:1 molar ratio with the different AtPIF variants, and analyzed the mixture by SEC. In full agreement with the firstpass screening (cf. Fig. 2 and Supplementary Fig. 2), all variants that we had identified as binding-competent exhibited interactions with AtPhyB PCM at an apparent 1:1 stoichiometry. Vice versa, the AtPIF variants that had failed to decelerate AtPhyB reversion kinetics (cf. Fig. 2 and Supplementary Fig. 2) lacked any interactions ( Supplementary Fig. 4). We also assessed interactions between the AtPIF variants and the AtPhyB PCM in the Pr state following exposure to far-red light (720 nm) ( Fig. 3f and Supplementary Fig. 5). None of the variants showed interactions under these conditions. Insofar red-light-activated binding to the AtPhyB PCM had been retained in the truncated AtPIF variants, far-red light hence abolished it.
Having engineered a suite of AtPIF variants undergoing lightregulated PPIs with the AtPhyB PCM, we next sought to quantify the strength of these interaction in both the Pr and Pfr states.  3 Oligomeric state of the AtPIF variants and light-dependent interactions with the AtPhyB PCM. a 50 µM AtPhyB PCM were exposed to red light and analyzed by size-exclusion chromatography (SEC), where the yellow and red lines represent absorption at 513 and 650 nm, respectively. b As in a but the AtPhyB PCM was exposed to far-red light prior to chromatography. c 10 µM P3.100-EYFP were analyzed by SEC. Elution profiles were independent of illumination. d 10 µM of the negative control EYFP were analyzed by SEC. Elution profiles were independent of light. e A mixture of 10 µM P3.100-EYFP and 50 µM AtPhyB PCM was exposed to red light and analyzed by SEC. f As in e but samples were illuminated with far-red light, rather than red light. Experiments were repeated twice with similar results.
Notably, detailed quantitative data of that type are largely unavailable but would tremendously improve our understanding of the AtPhyB:AtPIF PPI and inform its optimization. To this end, we resorted to fluorescence anisotropy measurements of the EYFP moiety C-terminally appended to all the AtPIF variants. Binding of a given AtPIF-EYFP variant to the AtPhyB PCM would increase its effective hydrodynamic radius, slow down rotational diffusion, and thus increase fluorescence anisotropy (Fig. 4a). We hence incubated a constant 20 nM of the AtPIF-EYFP variants with increasing amounts of AtPhyB PCM under red or far-red light and recorded binding isotherms. The reference construct P6.100 exhibited strong binding to the AtPhyB PCM under red light but no detectable binding under far-red light even at AtPhyB-PCM concentrations of 2 µM (Fig. 4b). When calculating dissociation constants (K D ), one must consider that red light not only drives the Pr → Pfr transition of Phys but also the reverse Pfr → Pr process. Consequently, continuous illumination with red light (640 nm) leads to population of a photostationary state with a mixed Pfr/Pr population at a ratio of~0.56/0.44 42 (Fig. 4c). Correcting for the actual fraction in the Pfr state, we determined a K D for the P6.100:AtPhyB-PCM pair of 10 ± 8 nM (Table 1). This value is in good agreement with an earlier estimate for this pair of 20-100 nM within mammalian cells based on fluorescence microscopy 27 . In comparison to P6.100, P3.100 exhibited a weaker K D of 200 ± 70 nM in Pfr and an elevated basal affinity in Pr, with an estimated K D on the order of low micromolar ( Fig. 4d and Table 1). This residual interaction could in principle be due to partial population of the Pfr state of the AtPhyB PCM under the chosen illumination conditions; however, the absence of basal affinity in case of P6.100 strongly argues against this notion. The slightly weaker affinity and much less pronounced light effect in P3.100 compared to P6.100 may account for the previously reported inability to detect light-regulated interactions of AtPIF3 with the AtPhyB PCM in mammalian cells 27 . We then recorded binding isotherms under red and far-red light for all the AtPIF variants we had purified and analyzed by SEC (Supplementary Figs. 4 and 5, Table 1). Consistent with our first-pass assessment (cf. Fig. 2d and Supplementary Fig. 2), the removal of the nonconserved N-terminal residues preceding the APB.A segment had no influence on the Pfr interaction. Unexpectedly, omission of these residues in the AtPIF3 context substantially attenuated the basal Prstate affinity. For the AtPIF3 variants, removal of the linker and the APB.B part had no or at most modest effects on affinity to the Pfr state ( Supplementary Fig. 6, Table 1). By contrast, in AtPIF6, the removal of the linker and the APB.B part more severely attenuated the affinity to the Pfr state to values between 200 and 700 nM. In addition, the affinity to the Pr state, non-detectable for the variants P6.100 and P6, increased as well. As a corollary, AtPIF3 and AtPIF6 variants lacking the APB.B segment exhibited closely similar K D values for a given construct topology. As a case in point, the P3.As and the P6.As variants, comprising 25 and 23 residues, respectively, both interacted with the AtPhyB PCM with an affinity of~700 nM in the Pfr and weaker than 10 µM in the Pr state. These data for P6As are consistent with a recent report that demonstrated lightdependent PPI for an AtPIF6 construct of closely similar length and sequence 43 . Duplication of the APB.A segments in the AtPIF3/6 backgrounds resulted in variants with affinities in the range of 200-400 nM for Pfr and weaker than 2 µM for Pr. We also analyzed several AtPIF3/6 variants entirely lacking the APB.A segment or possessing shortened versions of it, neither of which showed any interaction with AtPhyB PCM when probed by SEC nor by their effect on Pr reversion kinetics. In almost all these variants, fluorescence anisotropy failed to detect interactions either (Supplementary Fig. 6 and Table 1); merely, the P3.A19 and P6.A19 variants with C-terminally trimmed APB.A segments exhibited weak affinity for the Pfr state in the low micromolar range (Supplementary Fig. 6 and Table 1). In summary, these results n.d. not detectable a As determined by size-exclusion chromatography b A "+" sign indicates that an interaction could be detected by size-exclusion chromatography, a "−" sign denotes that no interaction was observed c As determined by fluorescence anisotropy confirm the APB.A segment as the main interaction epitope in both AtPIF3 and AtPIF6. Intriguingly, AtPIF6 differs from AtPIF3 by higher affinity for Pfr and much reduced affinity for Pr. As the removal of the APB.B segment largely cancels these differences, we conclude that APB.B in AtPIF6, but not in AtPIF3, enhances the affinity for Pfr and diminishes that for Pr. In AtPIF3, the Nterminal amino acids contribute to elevated basal affinity for Pr.
Repurposing the AtPhyB:AtPIF interaction for optogenetics. Through sequence variations and quantitative analyses, we generated modules for light-regulated PPIs spanning an affinity for the Pfr state from around 10 to 700 nM. We next investigated whether this set of novel AtPIF variants can be leveraged for optogenetics in mammalian cells. In a first line of experiments, we embedded the variants into a previously reported system for red-/ far-red-light-regulated gene expression that provides an in-cell readout of relative PPI affinities 36,44 . To this end, the AtPhyB PCM was covalently attached to a VP16 trans-activating domain, and the different AtPIF variants were linked to the E-protein DNA-binding domain, which binds to a cognate operator sequence upstream of a minimal promoter driving expression of secreted alkaline phosphatase (SEAP) (Fig. 5a). Through light-induced AtPhyB:AtPIF interactions, the trans-activating domain localizes to the DNA-binding domain and the promoter and thereby induces SEAP expression. SEAP activity levels are quantified and normalized to the levels of constitutively expressed Gaussia luciferase to correct for variations of cell density, transfection efficiency, and overall expression. We found that the P3.100 and P6.100 reference constructs upregulated normalized SEAP expression by tenfold and fourfold, respectively, under red light compared to darkness when expressed in Chinese hamster ovary cells (CHO-K1). The comparatively small regulatory effect for P6.100 results from substantial basal SEAP expression. We then subjected all the AtPIF3/6 variants we had previously characterized to the same analysis (Fig. 5b, c Fig. 8). When the cells were first exposed to red light for 24 h, followed by far-red illumination for another 24 h, they exhibited basal SEAP expression levels comparable to cells incubated in darkness throughout. Given that gene expression for the different sequence variations followed similar trends in both the AtPIF3 and the AtPIF6 backgrounds, we wondered whether the emerging underlying principles extend to other AtPIF orthologs. We hence generated the corresponding sequence variations in the AtPIF1 background and assessed their impact on light-regulated gene expression ( Fig. 5d and Supplementary  Fig. 7). Several of the resultant AtPIF1 variants supported lightactivated SEAP expression, although generally with slightly attenuated maximal levels and regulatory effects. Nonetheless, the AtPIF1 variants conformed to the general activity pattern observed for the AtPIF3/6 variants; specifically, only the AtPIF1 variants preserving an intact APB.A segment were capable of upregulating SEAP expression under red light. Taken together, these experiments demonstrate the utility of the cellular set-up for the efficient appraisal of light-regulated PPIs in mammalian cells. By capitalizing on this set-up, we obtained derivative systems with enhanced dynamic range and reduced leakiness that outperformed the original reference systems.
In a second set of experiments, we deployed several of the newly generated AtPIF6 variants for light-regulated recruitment of target proteins to the plasma membrane of NIH-3T3 cells. To this end, we equipped the AtPhyB PCM with a C-terminal CAAX prenylation motif for membrane targeting and the AtPIF6 Fig. 4 Quantitative analyses of the light-dependent protein:protein interaction between AtPIF variants and the AtPhyB PCM. a In its Pr state, the AtPhyB PCM exhibits weak or no affinity to AtPIF, but upon red-light exposure, the affinity is enhanced. Binding to the AtPhyB PCM increases the effective hydrodynamic radius of the AtPIF variants and slows down rotational diffusion. In turn, the fluorescence anisotropy of an EYFP tag C-terminally appended to the AtPIF increases. b Titration of 20 nM P6.100-EYFP with increasing concentrations of dark-adapted (gray) or red-lightexposed AtPhyB PCM (red), as monitored by anisotropy of the EYFP fluorescence. Data points show mean of n = 3 biological replicates. The red line denotes a fit to a single-site-binding isotherm. c Absorption spectra of the AtPhyB PCM in its dark-adapted Pr state (red line) and as a Pfr/Pr mixture following red-light exposure (blue). The dashed line denotes the absorption spectrum of the pure Pfr state, calculated according to ref. 42 . d As in b but for P3.100-EYFP rather than P6.100-EYFP. Experiments were repeated twice with similar results.
variants with an N-terminal EYFP tag 27,37,38 (Fig. 6a). Cell lines stably expressing both the AtPhyB PCM and one of the AtPIF6 variants, linked by an internal ribosome entry site (IRES), were created through lentiviral transduction. Cells were exposed to red (650 nm) and far-red light (750 nm), respectively, and the subcellular distribution of the EYFP-AtPIF6 variants was monitored by fluorescence microscopy (Fig. 6b-e). Under farred light, the reference variant P6.100 mostly localized to the cytoplasm, but under red light it partially translocated to the plasma membrane (Fig. 6c-f). Whereas the variants P6.A, P6.As, and P6.AA exhibited overall similar subcellular distribution under red and far-red light as P6.100, the variant P6.fus failed to show any light response of subcellular localization. Although subtle performance differences between the individual AtPIF6 variants cannot be ruled out, these are exceeded by the cell-to-cell variability of light-dependent translocation (Fig. 6f). Nonetheless, the experiments show that the new AtPIF6 variants with a much smaller footprint support light-regulated plasma membrane recruitment at similar efficiencies as the reference P6.100. This notion is further supported by the overall comparable expression level of the AtPIF6 variants and its effect on the magnitude of light-regulated membrane recruitment (Fig. 6g).

Discussion
In this study, we have dissected the light-regulated PPIs between the AtPhyB PCM and the AtPIFs 3 and 6, which underpin diverse adaptive responses in planta and multiple applications in optogenetics. To this end, we implemented a set of complementary experimental approaches ranging from SEC and fluorescence anisotropy to reporter assays in mammalian cells that deliver both qualitative and quantitative information on the PPIs. At a qualitative level, these assays consistently showed the APB.A segment to be necessary and sufficient for AtPhyB-PCM interactions, in line with previous reports 24 . By contrast, the APB.B segment alone did not promote detectable interactions. Our quantitative analyses put concrete numbers on the affinity of the AtPhyB:AtPIF3/6 pairs, information that hitherto was largely lacking. Strikingly, P6.100 exhibited a K D of only~10 nM for AtPhyB PCM in its Pfr state but entirely lacked interaction with the Pr state, from which we estimate an at least 1000-fold affinity difference. By contrast, the light dependence of the P3.100: AtPhyB-PCM interaction was less pronounced, with dissociation constants of~200 nM in the Pfr state and low micromolar in the Pr state. We tied the more stringent red-light response in AtPIF6 to its APB.B segment, which enhances affinity for the Pfr state of the AtPhyB PCM while simultaneously attenuating basal affinity for the Pr state. We speculate that these inherent differences between AtPIF3 and AtPIF6 might reflect their natural roles in planta. Whereas AtPIF3 predominantly serves as a negative regulator of photomorphogenesis by modulating the abundance of AtPhyB 45-47 , AtPIF6 acts as a positive regulator by inhibiting hypocotyl elongation under red light, at least when overexpressed 48 . To prevent untimely inhibition of hypocotyl growth, a more stringent light response with very low basal affinity in Pr may be required for this particular PIF. Recently, it has been reported that PIFs, and in particular AtPIF3, are constantly turned over both in darkness and under red light as a mechanism of achieving optimal levels for tight regulation of the skotomorphogenic and photomorphogenic responses 14 . A more permissive binding of AtPIF3 to the Pr state of AtPhyB as observed here  49 . The differential affinities of the individual PIFs might therefore contribute to the fine-tuning of physiological responses 14,49 . In fact, our study now provides a means of gradually adjusting the interaction strength of a given PIF, which could benefit the analysis of signal transduction mechanisms in planta. In a similar vein, the quantitative data on the AtPhyB:AtPIF PPI may help rationalize the phenotypes of pertinent pif mutant alleles. Finally, the comparatively smaller regulatory effect in AtPhyB:AtPIF3, compared to AtPhyB:AtPIF6, may explain why this PPI pair proved inferior for generic optogenetic applications 27 .
By deconstructing and quantitatively analyzing AtPIF3/6, we devised a suite of interaction modules with several beneficial traits (Table 1): First, the AtPIF variants span an affinity range from 10 to 700 nM, thus enabling the precise tuning of light-regulated PPIs as demanded by a specific application. Second, the AtPIFs can be reduced to around 23-25 residues while largely retaining lightregulated PPIs with the AtPhyB PCM. As we demonstrate, the smaller size facilitates the construction of tandem repeats of the APB.A motif, which, depending upon context, may enhance lightdependent responses. Third, the reduction in size also affected the oligomeric state of the AtPIFs, which are homodimeric at full length 41 but predominantly monomeric in several of the truncated variants studied presently. As we showcase for the scenarios of light-regulated gene expression and membrane recruitment, the set of novel AtPIF variants can indeed improve absolute activity and degree of light regulation in optogenetics. As a case in point, despite stringently light-regulated PPIs with the AtPhyB PCM, the original P6.100 variant promoted substantial basal gene expression in darkness, thus degrading the regulatory effect of light. We tentatively ascribe the relatively poor performance of P6.100 to its high Pfr-state affinity; even limited population of the AtPhyB Pfr state, e.g., due to light pollution or temperature changes 50 , may hence activate the PPI to considerable extent and over prolonged periods 39 . In support of this notion, the attenuation of the Pfr-state affinity in the shortened AtPIF6 variants led to reduced basal activity and enhanced regulatory efficiency. Duplication of the APB.A segment improved the performance for light-regulated expression, although the Pfr-state affinity of the Px.AA variants is almost unchanged relative to the corresponding Px.A variants. We hence ascribe this improvement to avidity and cooperativity effects. Our analyses readily extended to the AtPIF1 context, where shortened variants exhibited similar patterns of activity and light regulation as the AtPIF3/6 variants (cf. Fig. 5b-d). We speculate that the underlying principles can be generalized to APBcontaining PIF proteins from A. thaliana and other plants 51,52 . The performance of individual AtPIF variants in a given experiment can considerably vary and may be difficult to gauge upfront, not least because it likely depends on application context. We thus consider it an advantage to have now a set of AtPIF variants with known interaction strengths and varying properties. With this suite of AtPIF variants in hand, additional processes may be unlocked for optogenetic control by red and far-red light. As recently summarized 53 , numerous cellular parameters and pathways depend on PPIs and can thus be controlled by certain photoreceptors that associate or dissociate under blue light. The underlying regulatory strategy should readily extend to the present AtPhyB:AtPIF pairs and thereby to red and far-red light. Other potential use cases for the new AtPIF variants include immunoreceptor signaling 30 and light-regulated biomaterials 54 . As one shortcoming, optogenetic applications of plant Phys currently require the exogenous addition of PCB or PΦB chromophores, which do not widely occur outside cyanobacteria and plants. This contrasts with bacterial Phys, which utilize biliverdin (BV) that is available in mammals as a heme degradation product [55][56][57][58] . In particular, a recently described bacterial Phy undergoes PPIs depending on red and far-red light and has been harnessed for light-regulated gene expression [59][60][61] . The reliance on BV in this system obviates exogenous chromophore addition, which may prove advantageous for applications in vivo.
In summary, we have constructed and characterized a toolkit of novel AtPIF variants with varying interaction strength, size, and oligomeric state. Beyond application in optogenetics, the availability of these variants also stands to benefit the biophysical analyses of the Phy:PIF interaction. Although previous studies had localized this interaction to the N-terminal extension of Phys, atomically resolved information on the Phy:PIF complex is lacking 40,[62][63][64] . Minimized AtPIFs may well facilitate X-ray crystallographic analysis and thus pave the way toward elucidation of the complex structure. Moreover, the qualitative and quantitative interaction assays presently established can be deployed to chart Phys and interacting factors from A. thaliana and other plants.

Methods
Molecular biology and protein purification. Genes encoding A. thaliana PhyB PCM (residues 1-651), PIF3 (1-100), and PIF6 (1-100) were synthesized with codon usage adapted for expression in E. coli (GeneArt, Invitrogen, Regensburg, Germany). Via Gibson assembly 65 , the AtPhyB PCM was furnished with a Cterminal hexahistidine tag and subcloned onto the pCDFDuet1 vector (Novagen, Merck, Darmstadt, Germany) under control of a T7-lacO promoter; the plasmid, designated pDG282, additionally harbors a bicistronic cassette of Synechocystis sp. heme oxygenase 1 and pcyA 66 , also under the control of T7-lacO. For the expression of AtPIF3/6, the corresponding genes were subcloned onto a pET-19b vector (Novagen) under the control of a T7-lacO promoter by Gibson assembly or AQUA cloning 67 and thereby equipped with an N-terminal His 6 -SUMO tag 68 and a C-terminal EYFP tag, attached via a short linker (DSAGSAGSAG). For interaction studies in bacterial lysate, the AtPIF3/6 genes were subcloned onto a pET-28c vector (Novagen) under the control of a T7-lacO promoter, again with C-terminal linkers and EYFP. Variants of the AtPIF proteins were generated in both plasmid contexts, and the identity of all constructs was confirmed by Sanger DNA sequencing (GATC, Konstanz, Germany or Microsynth Seqlab, Göttingen, Germany).
Purification of the AtPIF3/6-EYFP variants employed a similar protocol with the following differences. No δ-aminolevulinic acid was added, and incubation after induction continued at 16°C for 40 h. Following IMAC, the N-terminal His 6 -SUMO was cleaved overnight at 4°C during dialysis into 50 mM Tris/HCl pH 8.0 and 20 mM NaCl using SENP2-protease. The His 6 -SUMO tag was removed by IMAC, and the flow-through containing the AtPIF3/6 construct was collected and analyzed by PAGE. Depending upon purity, the proteins were optionally further purified by anion-exchange chromatography as described above. Pure AtPIF3/6-EYFP variants were dialyzed into storage buffer and stored at −80°C. An analysis by denaturing PAGE of the purified AtPIF3/6-EYFP constructs and the AtPhyB PCM is shown as Supplementary Fig. 9.
Spectroscopic analysis. The concentration of purified AtPhyB PCM and the AtPIF3/6-EYFP variants were determined at 22°C by absorption measurements on an Agilent 8453 UV-visible spectrophotometer (Agilent Technologies, Waldbronn, Germany). In case of the AtPIF3/6-EYFP variants, a molar extinction coefficient at 513 nm of 84,300 M −1 cm −1 was used 71 . Photoreversible Pr ↔ Pfr conversion of AtPhyB PCM was ascertained by illumination with lightemitting diodes (LED) with emission wavelengths of 650 ± 15 nm (5.6 µW cm −2 ) and 720 ± 15 nm (0.7 µW cm −2 ), respectively. Spectra recorded after illumination revealed isosbestic points at 374 and 672 nm. Absorption spectra were also recorded after denaturation in 6.5 M guanidinium hydrochloride. By referencing to the previously reported extinction coefficient for PCB under these conditions 72 , we calculated an extinction coefficient at the isosbestic point 672 nm for AtPhyB PCM in its native state of 47,600 M −1 cm −1 . The fraction of AtPhyB PCM in the Pfr state upon saturating red-light illumination (640 nm) was determined as described in ref. 42 .
Interaction assay in bacterial lysate. pET-28c plasmids harboring AtPIF3-EYFP or AtPIF6-EYFP variants were transformed into chemically competent BL21(DE3) cells. Three replicate clones were used to inoculate 3× 5 mL TB medium supplemented with 50 µg mL −1 kanamycin. Cultures were incubated at 37°C up to an OD 600 of 0.6-0.8, at which point temperature was lowered to 16°C and expression was induced by addition of 1 mM IPTG. Incubation continued overnight, and cells were harvested by centrifugation at 3000 × g for 10 min. Pelleted cells were resuspended in 300 µL lysis buffer [1× FastBreak Cell Lysis Reagent (Promega GmbH, Mannheim, Germany), 10 µg mL −1 DNaseI (PanReac AppliChem, Darmstadt, Germany), 200 µg mL −1 lysozyme (Sigma-Aldrich, Darmstadt, Germany)] and rotated at 22°C for 10 min. Cell debris was removed by centrifugation at 186,000 × g for 45 min using an Optima MAX-XP Ultracentrifuge (Beckman-Coulter, Krefeld, Germany). The concentration of a given AtPIF3/6-EYFP variant in the lysate was determined by absorption measurements at 513 nm using a CLARIOstar microtiter plate reader (MTP) (BMG Labtech, Ortenberg, Germany). AtPhyB PCM at 2.5 µM concentration was mixed with a threefold molar excess of the AtPIF3/6-EYFP variants in 384-well clear MTPs (Thermo Fisher Scientific, Waltham, USA). After illumination with red light (650 ± 15 nm, 5.6 µW cm −2 ) for 4 min, the MTPs were covered with a clear lid, and absorption at 720 and 850 nm was measured every 5 min at 28°C in an Infinite M200 PRO plate reader (Tecan, Männedorf, Switzerland) for 12 h. After background correction, data at 720 nm were normalized to the signal of the L-EYFP (Supplementary Table 1) negative control, and the relative initial velocity was determined over the data acquired during the first 4 h.
Interaction assays with purified components. Size-exclusion chromatography: The light-dependent interaction between AtPhyB PCM and the AtPIF3/6-EYFP variants was assessed by gel filtration chromatography using a Superdex 200 Increase 10/300 GL (GE Healthcare) column on an ÄKTApure system, equipped with multi-wavelength detection (GE Healthcare). To this end, a mixture of 50 µM AtPhyB-PCM and 10 µM PIF-EYFP in 67 mM sodium phosphate buffer pH 8.0 and 200 mM NaCl was prepared and illuminated with 650-or 720-nm light for 2 min before sample application. Twenty-five microliters of this mixture was applied to the column and separated at a constant 0.75 mL min −1 flow rate. Absorption of EYFP and the AtPhyB PCM was measured at 513 and 650 nm, respectively. All proteins were also tested individually, where the AtPIF3/6-EYFP and EYFP samples were not illuminated prior to application.
Fluorescence anisotropy: AtPhyB PCM was illuminated with 640-or 750-nm light for 2 min immediately prior to the experiment (640 ± 15 nm; 65 µW cm −2 and 750 ± 15 nm; 420 µW cm −2 ). Samples containing 20 nM AtPIF3/6-EYFP and increasing AtPhyB-PCM concentrations between 0 and 10 µM were prepared in 20 mM HEPES/HCl pH 7.3, 10 mM NaCl, and 100 µg mL −1 bovine serum albumin, transferred into black 384-well MTPs (Brand, Wertheim, Germany), and illuminated with 640-or 750-nm light. Fluorescence anisotropy of EYFP fluorophore was measured on a CLARIOstar MTP reader (BMG Labtech) with an excitation wavelength of 482 ± 16 nm, a 504-nm long-pass dichroic filter, and a detection wavelength of 530 ± 40 nm. The fluorescence gains for the horizontal and vertical detection channels were adjusted to a fluorescence anisotropy value of 0.315, as determined for EYFP with an Olis DSM 172 spectrophotometer (On-Line Instrument Systems, Bogart, USA). Anisotropy data were evaluated with the Fit-omat software 73 using a single-site binding isotherm: where r represents the anisotropy of the PIF-EYFP fluorescence, [PhyB] is the concentration of the AtPhyB PCM in either the Pr or Pfr state, and K D is the dissociation constant. For the case of strong binding exhibited by the variants P6.100 and P6, we used a modified single-site binding isotherm that takes into account that the relevant [PhyB] concentrations are on the same order of magnitude as the constant concentration c total of the PIF-EYFP protein: Light-regulated gene expression in mammalian cells. The split transcription factor system for light-controlled gene expression in eukaryotic cells was based on a previously reported set-up 36,44 . To allow ratiometric analysis, this earlier set-up was expanded by cloning the Gaussia luciferase under the control of a constitutive promoter onto the same plasmid as the SEAP reporter gene.  36 . After addition of 20 µL 120 mM para-nitrophenyl phosphate, the absorption at 405 nm was measured for 1 h using a BMG Labtech CLARIOstar or a TriStar2 S LB 942 multimode plate reader (Berthold Technologies, Bad Wildbad, Germany) 36 . Outliers were statistically determined and excluded 75 .
Light-mediated membrane recruitment in mammalian cells. For each AtPIF6 variant tested, a lentiviral vector (pHR) was constructed containing a membranebound AtPhyB PCM (PHY-CAAX, residues 1-650) and a YFP-conjugated AtPIF6 variant. An IRES was introduced between the two coding sequences to ensure regulation of dual expression. Lentivirus was created by transfecting HEK-293T cells with pHR constructs and harvesting filtered media 48 h post-transfection. Mouse fibroblasts (NIH-3T3) were cultured in Dulbecco's Modified Eagle Medium (DMEM) containing 10% (v/v) fetal bovine serum. Fibroblasts were treated with lentivirus containing the constructs of interest. For all fibroblast experiments, cells were cultured in a 96-well glass-bottomed plate. Wells were pretreated with fibronectin for 30 min, following which fibronectin was aspirated and cells were plated and spun down for 5 min at 800 rpm. Cells were plated in 96-well glassbottom plates and allowed to adhere for at least 12 h. Imaging was performed using a ×60 oil immersion objective (NA 1.4) on a Nikon TI Eclipse microscope with a CSU-X1 confocal spinning disk, an EM-CCD camera, and appropriate laser lines, dichroics, and filters. DMEM was supplemented with phycocyanobilin 30 min prior to the start of the experiment. Cells were exposed to infrared light followed by red light to cause membrane recruitment and the resulting change in cytoplasmic fluorescence was measured using ImageJ by selecting a cytoplasmic region and computing the average pixel intensity before and after photostimulation. The change in cytoplasmic YFP-PIF level was normalized to the total YFP-PIF fluorescence in the nucleus under infrared conditions, to normalize to total expression level differences caused by lentivirus. In these experiments, light was delivered through the microscope using a Mightex Polygon digital micromirror device (DMD), X-Cite XLED1 LED light sources at 635 ± 20 and 730 ± 20 nm, and a ×40 objective lens. The duration of LED illumination was 1 min. To estimate the light dose delivered to the cell, we measured the light intensity using a ThorLabs power meter (PM100D) when the DMD was set to 100% transmission and obtained 100 µW for 635-nm light and 20 µW for 730-nm light, over a field of view of about 100 µm squared. For all experiments, we set the DMDs to 5% dithering (so each region was only illuminated for 5% of the time), translating into a final calculated intensity of 5 µW 635-nm light and 1 µW of 730-nm light. The light was delivered over an approximately 100 µm × 100 µm field of view, leading to an overall LED power density of 50 mW cm −2 at 635 nm and 10 mW cm −2 at 730 nm. Notably, these values are slightly higher but of comparable magnitude to those used by Pathak et al. for the AtPhyB:AtPIF3/6 system in the context of lightregulated gene expression 76 .
Statistics and reproducibility. Data are reported as mean ± SD or as mean ± SEM of n ≥ 3 biologically independent replicates. Details are specified in the legends to the figures and tables. All experiments could be reproduced with similar results.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.