Mechanism of early light signaling by the carboxy-terminal output module of Arabidopsis phytochrome B

Plant phytochromes are thought to transduce light signals by mediating the degradation of phytochrome-interacting transcription factors (PIFs) through the N-terminal photosensory module, while the C-terminal module, including a histidine kinase-related domain (HKRD), does not participate in signaling. Here we show that the C-terminal module of Arabidopsis phytochrome B (PHYB) is sufficient to mediate the degradation of PIF3 specifically and to activate photosynthetic genes in the dark. The HKRD is a dimerization domain for PHYB homo and heterodimerization. A D1040V mutation, which disrupts the dimerization of HKRD and the interaction between C-terminal module and PIF3, abrogates PHYB nuclear accumulation, photobody biogenesis, and PIF3 degradation. By contrast, disrupting the interaction between PIF3 and PHYB’s N-terminal module has little effect on PIF3 degradation. Together, this study demonstrates that the dimeric form of the C-terminal module plays important signaling roles by targeting PHYB to subnuclear photobodies and interacting with PIF3 to trigger its degradation.

P hytochromes (PHYs) are evolutionarily conserved photoreceptors in bacteria 1 , fungi 2 , algae, and plants [3][4][5] . In plants, PHYs are red (R) and far-red (FR) photoreceptors that can be photoconverted between two relatively stable forms: the R light-absorbing inactive Pr form and the FR light-absorbing active Pfr form 6,7 . PHYs regulate almost all aspects of plant development and growth, including germination, de-etiolation, shade avoidance, plant defense, floral induction, and senescence 8,9 . The importance of PHYs in plant development and growth is best exemplified in Arabidopsis de-etiolation. When young seedlings emerge from the ground and first encounter light, photoactivation of PHYs triggers a dramatic developmental transition from etiolation, a dark-grown developmental program, to photomorphogenesis, a light-grown developmental program that restricts hypocotyl growth and promotes chloroplast biogenesis and photoautotrophic growth 10 . These diverse photo morphological responses in Arabidopsis are mediated by five PHYs, PHYA-E 11 , among which PHYB plays a prominent role 12 .
PHYs trigger photomorphogenesis by reprogramming the nuclear genome 13,14 . One of the earliest light responses at the cellular level is the translocation of photoactivated PHYs from the cytoplasm to the nucleus [15][16][17] , where PHYs interact directly with a group of nodal basic helix-loop-helix transcriptional regulators -the phytochrome-interacting factors (PIFs)-and regulate their stability and activity [18][19][20][21][22] . The PIFs belong to subfamily 15 of the bHLH protein superfamily in Arabidopsis and include eight members: PIF1, PIF3-8, and PIL1 (PIF3-Like1) 23,24 . In general, PIFs play antagonistic roles in photomorphogenesis, including promoting hypocotyl elongation and repressing chloroplast biogenesis, with different PIFs performing overlapping and distinct roles 25,26 . PIF1, PIF3, PIF4, PIF5, and PIF7 promote hypocotyl growth by activating growth-relevant genes, such as genes involved in the biosynthesis and signaling of the plant growth hormone auxin [25][26][27][28] . PIF1, PIF3, and PIF5 inhibit chloroplast biogenesis by repressing nuclear-encoded photosynthetic genes 26,[29][30][31][32] . Most PIFs accumulate to high levels in dark-grown seedlings, and their protein levels are rapidly dampened in the light by PHYs [18][19][20][21][22][23]33 . Our understanding of PIF regulation in early PHY signaling came first from extensive studies of the founding member of the PIFs-PIF3 34 . PIF3 interacts preferentially with the active Pfr forms of PHYA and PHYB 34,35 . The PHY-PIF3 interaction promotes phosphorylation and subsequent degradation of PIF3 in the light by the ubiquitin-proteasome pathway 18 . PIF3 degradation is carried out by the Cullin3-based E3 ubiquitin ligases containing the substrate recognition proteins LRB1-3 (light-response broad-complex/Tramtrack/Bric-a-brac) 36 and requires a PHY-and PIF3-interacting transcriptional coactivitor, HEMERA (HMR) 33,[37][38][39] . Because PIF3 is recruited to PHYBcontaining subnuclear photosensory domains named photobodies during the dark-to-light transition prior to its degradation 18,40 , it was proposed that PIF3 degradation occurs at photobodies. This hypothesis is supported by a tight correlation between photobody disassembly and PIF3 accumulation during the light-to-dark transition 41 and by the genetic evidence that the hmr mutantwhich is defective in photobody biogenesis-also fails to degrade PIF3 in R light 33,37,38 . However, how the PHYB-PIF3 interaction induces PIF3 degradation and transcriptional regulation of its target genes is still not fully understood.
A major task in understanding early PHY signaling is to identify and dissect the functional roles of PHY's individual conserved domains in the early PHY phototransduction events of nuclear accumulation, photobody localization, as well as PIF interaction and degradation. The prototypical plant PHY is a homodimer, each monomer contains an N-terminal photosensory module and a C-terminal output module 6,7 . The N-terminal photosensory module consists of four subdomains-an N-terminal extension that is essential to stabilize the Pfr form and can be negatively regulated by phosphorylation [42][43][44] , a PAS (period-Arnt-single-minded) domain of unknown function, a GAF (cGMP photosphodiesterase/adenylate cyclase/FhlA) domain that binds a bilin chromophore, and a PHY (phytochrome-specific) domain that stabilizes the photoactivated Pfr conformer 6,45 . Structural studies of the PAS-GAF-PHY domains of bacteriophytochromes [46][47][48] and Arabidopsis PHYB 44 show that the PHY domain contributes a hairpin loop protrusion or "tongue," which is in close contact with the bilin-binding pocket of GAF and undergoes a β-stranded to helical conformational conversion during the Pr-to-Pfr photoactivation. In addition, at the PAS-GAF interface, there is an unusual figure-eight knot called the "light-sensing knot lasso" 44,46,47,49 , which interacts directly with PIFs 50,51 . This interaction is considered as a trigger for PIF3 degradation 18,51,52 .
The C-terminal output module of PHYs contains two tandem PAS domains named PRD (PAS-repeat domain) and a histidine kinase-related domain (HKRD). Previous studies have shown that the C-terminal module of PHYB forms a dimer 53,54 and the HKRD in both PHYA and PHYB was postulated as a dimerization domain [54][55][56][57] . In PHYB, the C-terminal module plays important roles in nuclear accumulation and photobody localization 17,53,58 . The PRD has been demonstrated to mediate PHYB's nuclear localization, and the entire C-terminal domain is required for photobody biogenesis 53,58 . The HKRD exhibits high sequence similarity to bacterial histidine kinases 59,60 . In fact, bacterial and fungal PHYs are bona fide histidine kinase sensors, and their histidine kinase domains are the signaling-output domain that relays phototransduction through autophosphorylation of an active-site histidine and phosphotransfer to an aspartate in a cognate response regulator 1,61,62 . In contrast, PHYs in higher plants lack the conserved active-site histidine and instead have Ser/Thr kinase activity 60 . Interestingly, deleting the majority of the HKRD had only minimum effects on PHYB signaling 63 . Therefore, the prevailing model is that the C-terminal module of PHYB does not participate in signaling output, in particular the HKRD is dispensable 7,53,63 .
Although the current view of the structure-function relationship of plant PHYs has been widely accepted, a number of lines of evidence still support a signaling role for the C-terminal module. For example, many loss-of-function mutations in PHYA and PHYB fall within the C-terminal module 6 . Among the reported phyB alleles, the phyB-18 mutant is unique because it carries a D1040V mutation in the HKRD and is the only missense phyB allele identified in the HKRD 63 . More interestingly, although deleting the majority of the HKRD had a minor effect on PHYB activity, the phyB-18 mutant is severely defective in PHYBmediated light responses 63 . To further explore possible signaling roles of the C-terminal module of PHYB, we have investigated how D1040V abolishes PHYB signaling in the phyB-18 mutant. We demonstrate that the entire HKRD is a dimerization domain within the C-terminal module of PHYB and that the D1040V mutation abrogates the dimerization of HKRD and consequently attenuates the early signaling functions of PHYB in nuclear accumulation, photobody localization, interaction with PIF3, and PIF3 degradation. Our results show unexpectedly that the Cterminal module, but not the N-terminal module, of PHYB plays a direct and essential signaling-output role in PIF3 degradation.
PHYs restrict hypocotyl growth by promoting the turnover of the master growth regulators, the PIFs 18-21, 25, 26 . We therefore examined whether PBY18 affected the steady-state level of the prototypical PIF-PIF3 18,34 . In continuous R light, PIF3 is actively degraded in the wild-type Col-0 but accumulates to a high level in phyB-9 ( Fig. 1d) 18,33 . Interestingly, the defect of phyB-9 in PIF3 degradation was rescued in PBC but not in PHYB18-1 and PHYB18-2 (Fig. 1d), indicating that the D1040V mutation attenuates the function of PHYB in PIF3 degradation.
The current model suggests that PIF3 degradation requires photobody biogenesis in the nucleus 18,33,40,41,64 . Therefore, we tested whether the D1040V mutation affected the subcellular localization of PHYB. Indeed, while PBC was localized to photobodies in 100% of nuclei in R light 58 , PBY18 was localized mainly to the cytoplasm and in 49% of the cells, it was only marginally observable in the nucleus (Fig. 1e). To confirm the defect of PBY18 in nuclear accumulation, we measured the ARTICLE nuclear fractions of PHYB in light-grown PBC and PBY18-1 seedlings. Despite the similar amounts of total PHYB-CFP and PHYB18-YFP in PBC and PBY18-1, respectively, the nuclear fraction of PBY18-YFP was almost 90% less than that of PHYB-CFP (Fig. 1f). These results demonstrate that phyB-18 is defective in PHYB nuclear accumulation and indicate that the HKRD plays a critical yet unknown role in PHYB nuclear accumulation. The correlation between the defects in PHYB18's nuclear accumulation and PIF3 degradation provides new evidence supporting the notion that the PHYB-mediated PIF3 degradation occurs in the nucleus 23 .
The C-terminal module of PHYB can mediate PIF3 degradation. The C-terminal module of PHYB is both required and sufficient for mediating PHYB's nuclear accumulation and photobody localization 53,58 . Therefore, we initially wanted to test whether the D1040V mutation impairs the nuclear-and photobody-targeting of the C-terminal module of PHYB. To that end, we generated transgenic lines in phyB-9 expressing the Cterminal module of either the wild-type PHYB or PHYB18 fused to YFP 53 . We characterized four independent transgenic lines expressing the C-terminal module of PHYB fused with YFP (BCY) (Fig. 2a). Unexpectedly, these experiments revealed that  Fig. 2 The C-terminal module of PHYB is biologically active to regulate hypocotyl growth and mediate PIF3 degradation. a Schematic illustration of the domain structures of PHYB-C-terminus-YFP (BCY). b Images of 4-day-old Col-0, phyB-9, and BCY-1 to BCY-4 transgenic lines grown in 10 μmol m −2 s −1 R light. c Hypocotyl measurements of Col-0, phyB-9, and BCY transgenic lines grown in 10 μmol m −2 s −1 R light. The boxes represent from 25th to 75th percentile; the bars equal to the median values. Samples with different letters exhibit statistically significant differences in hypocotyl length (ANOVA, Tukey's HSD, P < 0.01, n > 40). d Immunoblots showing the steady-state levels of BCY, PIF1, PIF3, PIF4, and PIF5 in the indicated lines grown in 10 μmol m −2 s −1 R light. Dark-grown Col-0 and pifq samples were used as positive and negative controls for PIFs, and RPN6 was used as a loading control. The relative protein levels were normalized to the corresponding levels of RPN6 and are shown below each lanes. e Images of 4-day-old Col-0, phyB-9, and BCY-1 to BCY-4 transgenic lines grown in the dark. f Hypocotyl measurements of Col-0, phyB-9, and BCY transgenic lines grown in the dark. The boxes represent from 25th to 75th percentile; the bars equal to the median values. Samples with different letters exhibit statistically significant differences in hypocotyl length (ANOVA, Tukey's HSD, P < 0.01, n > 40). g Immunoblots showing the steady-state levels of BCY, PIF1, PIF3, PIF4, and PIF5 in the indicated lines grown in the dark. Dark-grown Col-0 and pifq samples were used as positive and negative controls, respectively, for PIFs. RPN6 was used as a loading control. The relative protein levels were normalized to the corresponding levels of RPN6 and are shown below each lanes the C-terminal module of PHYB is biologically active in mediating PHYB signaling. First, although BCY did not fully rescue the long-hypocotyl phenotype of phyB-9, three BCY lines, BCY-1 to BCY-3, were slightly but significantly shorter than phyB-9 in R light (Fig. 2b, c). The long-hypocotyl phenotype of the BCY lines correlated with the BCY expression levels. For example, BCY-4 expressed the least amount of BCY (Fig. 2d) and had the longest hypocotyl (Fig. 2b, c). More interestingly, the defect of phyB-9 in PIF3 degradation in R light was largely rescued in BCY-1 to BCY-3 (Fig. 2d). The level of PIF3 in BCY-4 was similar to that of phyB-9 (Fig. 2d), indicating that the steady-state level of PIF3 is dependent on the amount of BCY.
Because BCY is constitutively localized to the nucleus independent of light 53 , we asked whether BCY could confer its signaling activities in the dark. Indeed, the levels of PIF3 in BCY-1 to -3 were 4-to 48-fold less than that in phyB-9 in the dark (Fig. 2g). These results demonstrate that PHYB's C-terminal module is sufficient to mediate PIF3 degradation.
Although PIF3 was largely reduced BCY-1 to -3 seedlings, the BCY lines showed minimum hypocotyl phenotype in the dark-

BCY BCY18
Red Dark . c Immunoblot analysis of the levels of PHYB and PIF3 in 4-day-old Col-0, phyB-9, BCY-2, BCY18 seedlings grown in 10 μmol m −2 s −1 R light. A sample from 4-day-old dark-grown pifq seedlings was used as a negative control for the PIF3 immunoblot. RPN6 was used as a loading control. The asterisk indicates nonspecific bands in the PIF3 immunoblot. The relative levels of PHYB and BCY were normalized against the corresponding levels of RPN6 and are shown below the blots. d Immunoblot analysis of the levels of PHYB and PIF3 in 4-day-old Col-0, phyB-9, BCY-2, BCY18 seedlings grown in the dark. RPN6 was used as a loading control. The relative levels of phyB and BCY were normalized against the corresponding levels of RPN6 and are shown below the blots. e Representative confocal images showing the subcellular localization patterns of BCY and BCY18 in hypocotyl epidermal cells of 4-day-old seedlings grown either in 10 μmol m −2 s −1 R light or darkness. The percentage of nuclei showing nuclear localization of BCY or BCY18 in each transgenic line is shown. f Immunoblots showing the total and nuclear fraction of BCY and BCY18 in BCY-2 and BCY18 lines grown either in 10 μmol m −2 s −1 R light or darkness. RPN6 and Pol II were used as loading controls for the total and nuclear fraction, respectively. The relative protein levels of BCY and BCY18 were normalized to RPN6 (total) or Pol II (nuclear) and are shown below the blots. g BCY18 is defective in nuclear accumulation. The relative nuclear fractions of BCY and BCY18 were estimated based on the ratio between the relative amounts of nuclear and total proteins obtained in c. The relative nuclear fractions of BCY18 is 3.01 and 11.99 fold less than those of BCY in R light and darkness, respectively   . e Immunoblots showing the levels of PHYB and PIF3 in 4-day-old Col-0, phyB-9, PBC, PBY18-1, PBY18-2, and PBY18N seedlings grown in 10 μmol m −2 s −1 R light. The sample of dark-grown pifq was used as a negative control for the PIF3 Immunoblots. RPN6 was used as a loading control. The relative levels of PHYB, PHYB-FP, and PIF3 were normalized against the corresponding levels of RPN6 and are shown below the blots. The asterisks in the PIF3 blots indicate nonspecific bands. f Immunoblots showing the levels of PHYB and PIF3 in 4-day-old Col-0, phyB-9, BCY-2, BCY18, and BCY18N grown either in 10 μmol m −2 s −1 R light or darkness. Dark-grown pifq samples were used as negative controls for the PIF3 immunoblots. RPN6 was used as a loading control. The relative levels of PHYB and PIF3 were normalized against the corresponding levels of RPN6 and are shown below each blot only BCY-2 and BCY-3 were slightly shorter than Col-0 and phyB-9 (Fig. 2e, f), suggesting that BCY is not able to degrade all the PIFs. We therefore tested whether BCY could mediate the degradation of PIF1, PIF4, and PIF5. We used a homemade antibody against PIF1 37 and two commercially available antibodies against PIF4 and PIF5 (Supplementary Fig. 1) to detect the endogenous levels of PIF1, PIF4, and PIF5 in Col-0, phyB-9, and the BCY lines in continuous R light (Fig. 2d). Interestingly, PIF4 and PIF5, but not PIF1, accumulated to higher levels in phyB-9 ( Fig. 2d)    ARTICLE were similar to those in phyB-9, suggesting that the C-terminal module of PHYB is not sufficient to mediate the degradation of PIF4 and PIF5 in R light. To confirm these results, we then examined the levels of PIF1, PIF4, and PIF5 in the dark (Fig. 2g). PIF1 accumulated in all dark-grown BCY lines, confirming that BCY does not mediate PIF1 degradation (Fig. 2g). However, the results for PIF4 and PIF5 were surprising. First, PIF4 accumulated fivefold less in phyB-9 compared with Col-0 ( Fig. 2g), suggesting that PHYB is required for the accumulation rather than degradation of PIF4 in the dark. BCY-1, -2, and -3 lines rescued the defect in PIF4 accumulation of phyB-9 (Fig. 2g). PIF5 accumulated to similar levels in Col-0 and phyB-9, and the accumulation of PIF5 was also greatly enhanced in the BCY-1, -2, and -3 lines (Fig. 2g). These results suggest that BCY can promote the accumulation of PIF4 and PIF5 in the dark. Taken together, these results indicate that the C-terminal module of PHYB can mediate the degradation of only PIF3 but not PIF1, PIF4, and PIF5. We therefore focused on PIF3 degradation in the rest of this study.
D1040V abolishes activity of the C-terminal module. To examine whether D1040V affects the activities of PHYB's Cterminal, we generated transgenic lines expressing the C-terminal module of PHYB18 fused to YFP (BCY18). A BCY18 line expressing a similar level of BCY18 as the level of BCY in BCY-2 was chosen for further analysis. In contrast to BCY-2, BCY18 showed the same hypocotyl length as phyB-9 in R light and darkness (Fig. 3a, b). Consistent with the hypocotyl phenotypes, BCY18 failed to degrade PIF3 in R light and the dark (Fig. 3c, d).
We then examined whether D1040V affected the nuclear accumulation of BCY. While BCY was localized to the nucleus in 100% of cells in both R light and darkness, BCY18 was localized to the nucleus in only 63% of cells in R light and 42% of cells in the dark (Fig. 3e). Additionally, in the nuclei with detectable BCY18, nuclear BCY18 failed to localize to photobodies (Fig. 3e). When the nuclear fractions of BCY and BCY18 were compared, the nuclear BCY18 was 3-and 12-fold less than the nuclear BCY in R light and darkness, respectively (Fig. 3f, g). Together, these results demonstrate that D1040V disrupts the function of the Cterminal module of PHYB in nuclear accumulation, photobody localization, and PIF3 degradation.
Nuclear localization largely rescues PHYB18 but not BCY18.
We then tested whether the phenotype of phyB-18 is primarily due to the defect of PHYB18 in nuclear accumulation. To that end, we fused an SV40 NLS to the C-termini of PBY18 and BCY18 and named them PBY18N and BCY18N, respectively. We generated transgenic lines expressing either PBY18N or BCY18N in the phyB-9 mutant. Both PBY18N and BCY18N were localized to the nucleus but with different photobody localization patterns (Fig. 4a). PBY18N was localized to photobodies in the light (Fig. 4a). However, compared with PBC, PBY18N had fewer large photobodies and more small photobodies (Fig. 4b), suggesting that PBY18N is defective in the biogenesis of large photobodies. In contrast, BCY18N was evenly dispersed in the nucleoplasm in both R light and dark conditions, indicating that BCY18N is unable to initiate photobody biogenesis. Consistent with the photobody localization patterns, PBY18N rescued the longhypocotyl and PIF3 accumulation phenotypes of phyB-9, whereas BCY18N was unable to rescue phyB-9 in either R light or dark conditions (Fig. 4c-f). These results indicate that the main defect caused by the D1040V mutation in full-length PHYB is the defect in nuclear accumulation. However, in the context of the Cterminal module, D1040V blocks its photobody localization and PIF3 degradation in addition to nuclear localization.
The C-terminus regulates a subset of PIF-regulated genes. It was surprising that dark-grown BCY seedlings accumulated much less PIF3 but more PIF4 and PIF5 (Fig. 2g). To understand how the C-terminal module of PHYB alters the expression of PIF-dependent genes, we compared the global transcriptomic profiles of dark-grown BCY-2 and phyB-9 by RNA-seq. These experiments identified 338 genes differentially expressed and statistically significant by 1.5-fold in BCY-2 compared with phyB-9 (Supplementary Data 1). Of these 338 BCY-regulated genes, 173 were BCY-induced and 165 BCY-repressed ( Fig. 5a and Supplementary Data 1).
To determine how many BCY-regulated genes are PIF-dependent, we compared the BCY-regulated genes with a previously defined set of 3775 PIFregulated genes in 4-day-old seedlings 65 . Among the 173 BCY-induced and 165 BCY-repressed genes, 83 and 91 genes were PIF-dependent, respectively. Interestingly, 72.3% of the BCY-induced/PIF-dependent genes were PIF-repressed genes (Fig. 5a, left panel); 93.4% of BCY-repressed/PIF-dependent genes were PIF-induced (Fig. 5a, right panel). Therefore, the changes in PIF-dependent genes in BCY-2 reflect the decrease in PIF3. Together, these data provide evidence that the C-terminal module of PHYB alone can alter the expression of a subset of PIFdependent genes in the dark.
We then performed GO enrichment analysis on the BCYregulated and PIF-dependent genes. For the BCY-induced/PIFrepressed genes, 13 GO categories were significantly enriched (Supplementary Data 2). The top five GO categories with the lowest p values are related to light responses and photosynthesis (Fig. 5b, left panel), particularly genes encoding the lightharvesting apparatus in chloroplasts (Supplementary Data 1). We selected four representative BCY-induced/PIF-repressed genes to verify their expression by quantitative RT-PCR, these include chlorophyll A/B-binding protein 2 (CAB2), lightharvesting chlorophyll-protein complexII subunit B1 (LHB1B1), photosystem I light-harvesting complex gene1 (LHCA1), and photosystem II light-harvesting complex gene 2.2 (LHCB2.2). The expression of all four representative genes was indeed highly induced in dark-grown BCY-2 seedlings, similar to their expression levels in pifq and YHB-a line expressing constitutively active PHYB 66 (Fig. 5c). In addition, the induction of the photosynthetic genes in the BCY-2 line is consistent with the previously published transcriptome data of the pif3 mutant [67][68][69] . In contrast, the expression of these four genes was less activated in BCY18 and BCY18N (Fig. 5c), indicating that BCY18 and BCY18N were unable to derepress these genes in the dark. These results are consistent with the results that BCY18 and BCY18N failed to degrade PIF3 in the dark (Fig. 4f).
Among the seven enriched GO categories of the BCYrepressed/PIF-induced genes, none was related to auxin (Supplementary Data 2; Fig. 5b, right panel). To further confirm these results, we examined the expression of two auxin signaling genes, IAA19 and IAA6, and two auxin responsive genes, SAUR23 and SAUR19. As shown in Fig. 5d, the expression of these four genes was not altered in BCY-2, BCY18, and BCY18N compared with phyB-9. These data are consistent with the published result that knocking out PIF3 alone has minimum effect on hypocotyl length in the dark 25 and the fact that the BCY lines had only minor effects on hypocotyl growth in the dark (Fig. 2e, f).
Taken together, these results show that the C-terminal module of PHYB is sufficient to regulate a subset of PIF-dependent genes. However, we do acknowledge that all differentially expressed genes may not be directly regulated by the decrease level of PIF3. The increased amount of PIF4 and PIF5 may also contribute to these changes. As the nuclear-encoded photosynthetic genes but not the growth-relevant auxin-related genes were regulated by the C-terminal module of PHYB, this may explain why the function of the C-terminal module of PHYB had received relatively little attention. This is because inhibition of hypocotyl growth has been used as a standard readout for PHYB activity and the C-terminal module of PHYB alone does not have a major impact on hypocotyl growth (Fig. 2b, e) 53,54,63 .
The entire HKRD is a dimerization domain. The HKRD exhibits high sequence similarity to bacterial histidine kinases 59,60 , which consist of a dimerization and histidine phosphotransfer (DHp) domain and a catalytic and ATP-binding (CA) domain 70 . Because PHYB lacks the conserved histidine in the DHp domain 59, 60 , we refer the DHp domain in PHYB as the "D domain" (Fig. 6a). The CA domain of PHYB contain all the conserved N, G1, F, and G2 subdomains for ATP binding 59,60 . Asp 1040 is conserved among plant, algal, and bacterial PHYs ( Supplementary Fig. 2), residing in a linker region between the D and CA domains (Fig. 6a). Therefore, D1040V could disrupt the function of the CA and/or dimerization. Because mutations in conserved residues in the G1 and G2 ATP-binding motifs have little effect on the function of PHYA, it was suggested that ATPbinding of the CA domain is not required for PHY function 71 . Therefore, D1040V might affect a property other than CA activity, one possibility would be dimerization of the HKRD. The HKRD in both PHYA and PHYB was postulated as a dimerization domain [54][55][56][57] , but the contribution of D and CA to dimerization has not been determined. To examine whether Asp 1040 is in a region required for dimerization, we examined which region of HKRD is required for dimerization by yeast two-hybrid assays. As predicted, the HKRD interacted with itself in yeast (Fig. 6b, c), confirming that the HKRD contains a dimerization domain. Interestingly, deleting the majority of either CA (HKRD-D) or D (HKRD-CA1) lost the dimerization activity (Fig. 6b, c). The loss of dimerization activity for HKRD-CA1 was surprising, because the same fragment, which contains the C-terminal 210 amino acids, was previously shown to dimerize in yeast 57 . We reasoned that this discrepancy could be due to the high stringency of the yeast two-hybrid assay using the antibiotic aureobasidin A (AbA) as a selection marker. Therefore, we re-examined the interactions by using the HIS3 reporter-a less stringent selection method. The alternative selection method indeed showed that HKRD-CA1 interacted with itself despite its basal self-activation activity (Fig. 6d). In addition, HKRD-CA2, which lacks the entire D domain, also showed weak dimerization activity (Fig. 6b, d). In contrast, HKRD-CA3, which contains the CA domain alone, had no dimerization activity (Fig. 6b, d)  In vitro co-immunoprecipitation results showing that the D1040V mutation disrupts the interaction between the C-terminal domain of PHYB and attenuates the interaction between full-length PHYB. HA-tagged PHYB, PHYB18, PHYB's C-terminal domain (BC), or PHYB18's C-terminal domain (BC18) was co-expressed with the corresponding Myc-tagged version using in vitro transcription/translation and subsequently immunoprecipitated using anti-HA affinity matrix. Bait and prey proteins were detected by immunoblots using anti-HA and anti-Myc antibodies, respectively. The relative amounts of immunoprecipitated Myc-tagged PHYB fragments are shown below the immunoblots. c In vitro co-immunoprecipitation results showing that the D1040V mutation in the C-terminal module of PHYB reduces its interaction with the C-terminal modules of PHYD and PHYE. HA-tagged C-terminal module of PHYB (BC) or PHYB18 (BC18) was in vitro co-translated with Myc-tagged C-terminal module of PHYA (AC), PHYC (CC), PHYD (DC) or PHYE (EC) and immunoprecipitated using anti-HA affinity matrix. Bait and prey proteins were detected by immunoblots using anti-HA and anti-Myc antibodies, respectively. The relative amounts of immunoprecipitated Myc-tagged PHY C-terminus are shown below the immunoblots D1040V disrupts homo and heterodimerization of PHYB. We then tested whether D1040V could disrupt the dimerization of HKRD. When a D1040V mutation was introduced into either the bait or the prey HKRD, the dimerization activity was abolished (Fig. 7a), supporting the notion that D1040 is involved in the dimerization of HKRD. We then asked whether D1040V could affect dimerization in the context of the C-terminal module and the full-length PHYB. Because the C-terminal module and fulllength PHYB have self-activation activity in yeast, we performed pull-down assays instead using in vitro translated PHYB and PHYB's C-terminal module (BC) fused to either HA or Myc tag. As expected, HA-tagged PHYB C-terminal module (HA-BC) was able to pull down Myc-tagged PHYB C-terminal module (Myc-BC) (Fig. 7b)  ARTICLE interacts directly with itself. In contrast, when the C-terminal module of PHYB18 (BC18) was tested in the pulldown assay, HA-BC18 failed to pull down Myc-BC18 (Fig. 7b), indicating that D1040V disrupts the dimerization of the C-terminal module of PHYB. These data also provide evidence that the HKRD is the only dimerization domain in the C-terminal module of PHYB. In the context of full-length PHYB, although the interaction between HA-PHYB18 and Myc-PHYB18 was reduced by fivefold compared with that between the wild-type PHYB, HA-PHYB18 could still pull down Myc-PHY18 (Fig. 7b), suggesting that although HKRD plays a predominant role in the dimerization of full-length PHYB, there is an additional dimerization domain present in the N-terminal photosensory module. One candidate is the GAF domain, which has been shown to interact directly in the crystallized dimer of PHYB's N-terminal photosenory module 44 . PHYB also heterodimerizes with PHYC, PHYD, and PHYE 57,72 , and the heterodimerization is mediated by the C-terminal module, likely the HKRD 57 . We tested whether D1040V attenuates PHYB heterodimerization by pulldown assays using in vitro co-translated HA-tagged C-terminal module of PHYB or PHYB18 and Myc-tagged C-terminal module of PHYA, PHYC, PHYD, and PHYE. These results show that the Cterminal module of PHYB could pull down PHYD and PHYE but not PHYA and PHYC (Fig. 7c). The D1040V mutation in PHYB dramatically reduced its affinity with PHYD and PHYE (Fig. 7c). Together, these results, combined with the results from previous studies 57 , indicate that the HKRD also mediate PHYB heterodimerization and D1040V disrupts PHYB heterodimerization with PHYD and PHYE.
D1040V attenuates the interaction between PHYB and PIF3. The interaction between PHYB and PIF3 is a key early signaling event that triggers PIF3 degradation 18,34,35,51,73 . PIF3 interacts with both the N-and C-terminal modules of PHYB 73 . However, because the interaction between PIF3 and the N-terminal module is stronger and photoconvertible, this interaction has been considered to be a trigger for PIF3 degradation 18,51,52 , whereas the significance of the interaction between PIF3 and the C-terminal module is less understood. Because the D1040V mutation resides in the C-terminal module, we asked whether D1040V has any impact on its interaction with PIF3. We examined the interaction between HA-PIF3 and the C-terminal module of PHYB and PHYB18 using in vitro immunoprecipitation assays. As shown in Fig. 8a, D1040V reduced the interaction between PIF3 and the C-terminal module by approximately two-thirds. The reduced interaction between PHYB18 and PIF3 may explain the defect in PIF3 degradation in BCY18N (Fig. 4f).
We then examined the effect of D1040V on PIF3 binding in the context of the full-length PHYB. These experiments showed that D1040V dramatically reduced the binding of PIF3 to the PHYB apoprotein, Pfr, and Pr forms (Fig. 8b). The interaction between PIF3 and PHYB18-HA was only about one-third of that between PIF3 and PHYB-HA (Fig. 8b). These results suggest that the defect in PIF3 degradation by PHYB18 could be due to its reduced affinity with PIF3. Because D1040V affects the dimerization of the C-terminal module, these results also suggest that dimerization of the HKRD facilitates the PHYB-PIF3 interaction. Together, these results, combined with the results that the C-terminal module is sufficient to mediate PIF3 degradation (Fig. 2) support a new model that PIF3 degradation depends on its interaction with PHYB's C-terminal module.
PIF3 degradation does not need PHYB N-terminal interaction.
The new hypothesis that PIF3 degradation is dependent on PHYB's C-terminal module is contradictory to the current model that PIF3 degradation is mediated by the interaction with the Nterminal photosensory module of PHYB 51,52 . The N-terminal module interacts with PIF3 in a photoreversible manner; PIF3 preferentially interacts with the biologically active Pfr conformer 34,52 . The residues in the N-terminal module of PHYB involved in the Pfr-specific binding to PIF3 have been identified in the knot lasso 50,51 . In particular, three mutations, R110Q, G111D, and R352K, were shown to abolish the interaction between PHYB and PIF3 without altering PHYB photoconversion 50,51 . Moreover, the G111D and R352K mutations also disrupt the binding with other PIFs, including PIF1, PIF4, PIF5, and PIF7 51 . Consistent with their defects in PIF3 binding, these three PHYB mutants were unable to rescue the long-hypocotyl phenotype of phyB-5 50,51 . We reasoned that if the C-terminal module of PHYB was required and sufficient for PIF3 degradation, disrupting the interaction between PIF3 and the N-terminal knot lasso of PHYB should not affect PIF3 degradation. To test this hypothesis, we examined whether the three transgenic lines expressing the full-length PHYB with individual R110Q, G111D, and R352K mutations-PBG-R110Q, PBG-G111D, and PBG-R352K 51 -could still mediate PIF3 degradation. As previously reported, all three lines showed long-hypocotyl phenotypes in R light (Fig. 8c, d) and no hypocotyl phenotypes in the dark ( Supplementary Fig. 3a, b). However, in contrast to the long-hypocotyl phenotype, the levels of PIF3 in these lines in R light were greatly reduced compared with phyB-5 (Fig. 8e), indicating that these phyB mutants were still able to mediate PIF3 degradation. As expected, these lines did not affect the PIF3 level in the dark ( Supplementary  Fig. 3c). These results indicate that PIF3 degradation does not require the interaction between PIF3 and the N-terminal module of PHYB.

Discussion
A central mechanism in PHY signaling involves direct PHY-PIF interaction and subsequent degradation of PIFs 23 . PIF3 interacts with both the N-and C-terminal modules of PHYB 73 . The widely accepted model indicates that PIF3 degradation is triggered by the light-induced interaction with the N-terminal photosensory module of PHYB through specific residues in the knot lasso 23,[50][51][52] , whereas the C-terminal output module is considered to participate in only subcellular localization and dimerization but not directly in signaling, and the HKRD is thought to be dispensable for PHYB function 53,63 . In this study, we demonstrate a contrasting light signaling mechanism in which the C-terminal domain of PHYB is the signaling-output module essential and sufficient for mediating PIF3 degradation (Fig. 8f). The light-induced degradation of PIF3 in continuous light does not depend on the photoconvertible interaction between PIF3 and PHYB's N-terminal photosensory module but rather relies on its interaction with the dimeric form of the C-terminal output module. Our results have demonstrated that the HKRD of PHYB acts as a dimerization domain in the C-terminal output module and participates in early light signaling events, including PHYB nuclear accumulation, photobody biogenesis, and PIF3 interaction and degradation. Therefore, similar to bacterial PHYs, the plant phytochrome can also transduce signals through the HKRD not by its histidine kinase activity but rather via dimerization and protein-protein interaction.
Our results indicate that the dimeric form of the C-terminal module of PHYB interacts directly with PIF3 to mediate its degradation (Fig. 8f). This conclusion is supported by the results that the C-terminal module alone was sufficient to trigger PIF3 degradation and activate a distinct set of PIF-repressed photosynthetic genes (Figs. 2 and 5). We do not believe that this occurred via heterodimerization of the C-terminal domain with another PHY because degradation occurred in the dark when other PHYs would be expected to be inactive and localized outside the nucleus. The signaling function of the C-terminal module depends on its dimer form because the D1040V mutation, which disruptd dimerization of the C-terminal module (Fig. 7a, b), reduced the interaction with PIF3 (Fig. 8a) and abrogated PIF3 degradation in the nucleus (Fig. 4f). The conclusion that PIF3 degradation is mediated by the C-terminal module but not the N-terminal module of PHYB is consistent with the previous results that the N-terminal photosensory module alone is not sufficient for PIF3 degradation 22,41 . Here, we provide additional evidence that the phyB mutants defective in PIF3 binding with the knot lasso of PHYB-R110Q, G111D, and R352K 51 -had little effect on PIF3 degradation (Fig. 8e), demonstrating that PIF3 degradation does not require the photoconvertible interaction with the N-terminal photosensory module of PHYB. Instead, the interaction with the N-terminal module was shown to regulate the transcriptional activity of PIF3 22 . Together, the results presented in this study, combined with previous studies 22,41 , indicate that PIF3 degradation is mediated by the interaction with PHYB's C-terminal output module.
PIF3 binds to the C-terminal module of PHYB relatively weakly compared to the Pfr form of the N-terminal photosensory module 52,73 . We can not exclude the possibility that in context of full-length PHYB the interaction between PIF3 and the C-terminal module can be enhanced by the light-induced interaction with the N-terminal photosensory module 73 . PIF3 interacts with full-length active PHYB much more strongly than the N-and C-terminal modules separately, suggesting that PIF3 might bind to the two modules cooperatively 73 . However, the PIF3-PHYB interaction does not seem to be regulated by lightinduced conformational changes of the C-terminal module per se, because the PHYB-R110Q, PHYB-G111D, and PHYB-R352K mutant proteins, which presumably bind PIF3 through only their C-terminal modules, had similar affinities to PIF3 between their respective Pr and Pfr conformers 51 , suggesting that the PIF3 and C-terminal-module interaction is enhanced in the Pfr and therefore not lightdependent. Our results indicate that this weak, nonphotoconvertible interaction is capable of mediating PIF3 degradation (Fig. 8f). This conclusion implies that the lightdependence of PIF3 degradation is not specifically due to its lightconvertible binding to PHYB but rather is contributed mainly by the light-triggered accessibility of PIF3 in the nucleus through regulation of PHYB's nuclear accumulation. Here, we show that PHYB18, which was defective in nuclear accumulation, was unable to mediate PIF3 degradation (Fig. 1). In addition, fusing an NLS to PHYB18 largely rescued the defects of PHYB18 in mediating PIF3 degradation (Fig. 4). Our results provide additional evidence supporting the notion that PIF3 degradation occurs in the nucleus and that nuclear accumulation of PHYB is a major switch that triggers PIF3 degradation 74 .
Our data show that the C-terminal module of PHYB mediates the degradation of PIF3 specifically. BCY-1 to -3 lines failed to degrade PIF1 in the dark (Fig. 2g), indicating that BCY does not mediate PIF1 degradation. These results are consistent with the previous findings that PIF1 and PIF3 are degraded through distinct mechanisms. For example, PIF3 is ubiquitylated by the CUL3 LRB E3 ubiquitin ligase 36 , whereas PIF1 is ubiquitylated by the CUL4 COP1-SPA E3 ubiquitin ligase 75 , and COP1 and SPA1 play opposing roles for PIF1 and PIF3 degradation 40,76 . PHYB also promotes the degradation of PIF4 and PIF5 in continuous R light (Fig.  2d).
Our results indicate that the C-terminal module of PHYB is not sufficient to mediate PIF4 and PIF5 degradation in the light (Fig. 2d). Surprisingly, our results also show that PHYB is required for PIF4 accumulation in the dark and the C-terminal module could promote PIF4 and PIF5 accumulation in the dark (Fig. 2g). The mechanism by which PHYB imposes opposing effects on PIF4 accumulation in the light and dark is still unknown. Taken together, these results provide evidence that the stability of PIFs are differentially regulated via distinct mechanisms. We further characterized the role of HKRD in dimerization and defined its function in signaling. The HKRD consists of D domain and CA domain (Fig. 6a) and was shown to be a dimerization domain for both homo and heterodimerization [54][55][56][57] . Based on its sequence similarity to bacterial histidine kinases 59, 60 , the D domain but not the CA domain is predicted to be involved in dimerization. However, our data show that the dimerization of the C-terminal module require both the D and CA domains as well as the linker region between the them (Fig. 6b-d). D1040V disrupted HYB homodimerization and its heterdimerization with PHYD and PHYE (Fig. 7), suggesting that Asp 1040 could be directly involved in dimerization. Heterodimerization of BCY with active PHYD and PHYE could promote their functions in the light and therefore may explain the enhanced activity of the BCY line in the light vs darkness (Fig. 2). Supporting this notion, when the D1040V mutation was introduced into the C-terminal module, the BCY18 line lost the light-dependent discrepancy in the hypocotyl response ( Fig. 3a-d). PHYB also heterodimerizes with PHYC 57, 72 , but this interaction was not detected by our pulldown assay (Fig. 7c), therefore it is not clear whether Asp 1040 is also involved in the PHYB-PHYC dimerization. In the full-length PHYB, the defect in HKRD dimerization in PHYB18 could be partially compensated by the N-terminal photosensory module (Fig. 7b). Also, PBY18N, but not BCY18N, is active in the nucleus (Fig. 4). These results provide evidence that the N-terminal module facilitates dimerization, likely through the GAF domain (Fig. 8f) 44 .
Both the nuclear accumulation and photobody localization of PHYB depend on its C-terminal module 53,58 . Within the C-terminal module, the PRD is required and sufficient to mediate the translocation of a bulky protein reporter, GUS-YFP, to the nucleus 58 . Although it is still not clear whether the PRD mediates the nuclear import of PHYB by interacting directly with an importin, it is certain that the PRD provides the molecular basis for nuclear accumulation. The previous results suggest that the PRD mediates the nuclear accumulation of PHYB independent of the HKRD. Here, we show that the D1040V mutation in the HKRD abrogates the nuclear accumulation of both full-length and the C-terminal module of PHYB (Figs. 1 and 3), suggesting that the HKRD can communicate with the PRD and modulate its activity in nuclear accumulation, likely via dimerization (Fig. 8f). The HKRD was indicated to regulate the nuclear accumulation of PHYA 77 . The phyA-402 mutant carries a L946F mutation in the HKRD of PHYA and is defective in PHYA nuclear accumulation 77 . Coincidentally, the L946F mutation lies in PHYA's D domain, it would be interesting to test whether this mutation affects PHYA dimerization.
A more specific mechanism has been proposed for nuclear import of PHYB 78 . This model suggests that PHYB does not contain a bona fide NLS but rather is piggybacked into the nucleus by PIFs 78 . However, the current evidence does not favor this mechanism in planta. As mentioned previously, despite the strong interaction between the N-terminal module of PHYB and PIF3, the N-terminal module alone remains cytoplasmically localized 53,58 . In addition, the mutations in the knot lasso, R110Q, G111D, and R352K, which disrupt the strong interaction between PIFs and the N-terminal module, had little effect on PHYB nuclear localization 50,51 . Our results show that the PHYB18 mutant, which could still interact with PIF3 and presumably other PIFs through the N-terminal module (Fig. 8b), was defective in nuclear accumulation (Fig. 1). Together, these results indicate that binding to PIFs is not sufficient to mediate PHYB nuclear accumulation in Arabidopsis, and therefore argue against the PIF-mediated PHYB nuclear important model. Photobody biogenesis is mediated by the entire C-terminal module of PHYB; neither the PRD nor the HKRD alone is sufficient to mediate photobody biogenesis 17,53 . Here we show that photobody localization of PHYB requires dimerization of the HKRD (Fig. 4a). In addition, the nuclear-targeted C-terminal module of PHYB18 (BCY18N) was unable to localize to photobodies and to degrade PIF3 in the dark (Fig. 4f), supporting the hypothesis that photobody biogenesis is required for PIF3 degradation 33,41,64 . These results are also consistent with the conclusion drawn by mathematical modeling that the PfrPfr dimer of PHYB is required for both photobody localization and the signaling events to inhibit hypocotyl growth 79 . Our results suggest that the constitutively formed photobodies by the Cterminal module of PHYB retain some functions of the regular photobodies in signaling, particularly in the regulation of PIF3 stability.
Plasmid construction and generation of transgenic lines. The primers used for making constructs are listed in Supplementary Table 1. The PBY18 and PBY18N constructs were generated by subcloning full-length or the N-terminal PHYB with the D1040V mutation into the KpnI site of pCHF3-YFP and pCHF3-YFP-NLS vectors, respectively. The BCY18 and BCY18N constructs were generated by subcloning the C-terminal sequence of the PHYB cDNA (for 594-1172 a.a.) with the D1040V mutation into the KpnI site of pCHF3-YFP and pCHF3-YFP-NLS vectors, respectively. The expression of all the transgenes was driven by the constitutive cauliflower mosaic virus 35S promoter. Transgenic lines were generated by transforming phyB-9 mutant plants with Agrobacterium tumefaciens strain GV3101 harboring the above constructs. For each construct, more than 10 independent T1 lines were selected on half-strength MS medium containing 30 µg/ml kanamycin. Lines that segregated approximately 3:1 for kanamycin-resistance in the T2 generation were selected on the basis of the level of overexpression. T3 self progeny of homozygous T2 plants were used for the experiments.
For the fractionation experiments, plant nuclei were isolated from 4-day-old dark-or R-light-grown seedlings as described previously 38 with the following modifications. Tissue was ground to fine powder in liquid N 2 and dissolved in a 2× volume of nuclear isolation buffer (20 mM PIPES-KOH, pH 6.5; 2 M hexylene glycol; 10 mM MgCl 2 ; 1× EDTA-free protease inhibitor cocktail (Roche); 0.25% Triton X-100; 5 mM β-mercaptoethanol; 1 mM PMSF). The lysate was filtered through two layers of Miracloth, and the cleared lysate was collected as the "Total" sample. The rest of the filtered lysate was loaded on top of 30% Percoll (Sigma) and centrifuged at 700×g for 10 min at 4°C. The enriched nuclear pellet was dissolved in nuclear isolation buffer and collected as the "nuclear fraction." Both "total" and "nuclear fraction" samples were boiled for 5 min in 1× Laemmli buffer and used in the subsequent immunoblot assays.
Confocal imaging and quantification of photobody morphology. Four-day-old seedlings grown in continuous R light (10 µmol m −2 s −1 ) were fixed in 2% paraformaldehyde in 1× PBS under vacuum on ice for 15 min and then washed with 50 mM NH 4 Cl in 1× PBS for 2 × 5 min, 1× PBS with 0.2% Triton X-100 for 2 × 5 min, and 1× PBS for 3 × 5 min. Fixed seedlings were mounted with ProLong Gold antifade reagent (Thermo Fisher Scientific), sealed with nail polish, and stored at 4°C until imaging. Nuclei of hypocotyl epidermal cells were imaged using a Zeiss LSM 510 inverted confocal microscope (Zeiss) with a 100×/1.4 Plan-Apochromat oil-immersion objective. YFP was monitored using 514 nm excitation from an argon laser and a 505-550 nm bandpass detector. CFP was monitored using 458 nm excitation from an argon laser and a 470-500 nm bandpass detector. Images were collected using LSM 510 software version 4.2. and processed using Adobe Photoshop CC software (Adobe Systems). The proportion of cells with or without nuclear signals was manually scored. The volume and number of photobodies were analyzed using Huygens Essential software (Scientific Volume Imaging). The object analyzer tool was used to threshold the image and to calculate the volume of photobodies.
Yeast two-hybrid assay. Bait (pGBKT7) and prey (pGADT7AD) vectors described above were transformed into Y2HGold and Y187 yeast strains (Clontech), respectively. Diploid yeast cells were generated by mating single colonies from bait and prey strains and then selected on SD/-Trp/-Leu plates. For Fig. 6, overnight cultures from single yeast colonies were diluted to an OD 600 of 0.2 and spotted on SD/-Trp/-Leu, SD/-Trp/-Leu/-His, and SD/-Trp/-Leu supplemented with 125 ng/ml Aureobasidin A (AbA). For Fig. 7a, overnight cultures from single