B-box transcription factor 28 regulates flowering by interacting with constans

B-box transcription factors (BBXs) are important regulators of flowering, photomorphogenesis, shade-avoidance, abiotic and biotic stresses and plant hormonal pathways. In Arabidopsis, 32 BBX proteins have been identified and classified into five groups based on their structural domains. Little is known about the fifth group members (BBX26–BBX32) and the detailed molecular mechanisms relevant to their functions. Here we identified B-box transcription factor 28 (BBX28) that interacts with Constans (CO), a transcriptional activator of Flowering Locus T (FT). Overexpressing BBX28 leads to late flowering with dramatically decreased FT transcription, and bbx28 deficient mutant displays a weak early flowering phenotype under long days (LD), indicating that BBX28 plays a negative and redundant role in flowering under LD. Additionally, the interaction between BBX28 and CO decreases the recruitment of CO to FT locus without affecting the transcriptional activation activity of CO. Moreover, the N-terminal cysteines, especially those within the B-box domain, are indispensable for the heterodimerization between BBX28 and CO and activation of CO on FT transcription. Genetic evidences show that the later flowering caused by BBX28 overexpression is compromised by CO ectopic expression. Collectively, these results supported that BBX28 functions with CO and FT to negatively regulate Arabidopsis flowering, in which the N-terminal conserved cysteines of BBX28 might play a central role.


Results
BBX28 negatively regulates flowering. To uncover the biological functions of BBX28 in addition to its role in the regulation of photomorphogenesis 42 , we obtained a bbx28 mutant with a T-DNA insertion in the first exon without a detectable BBX28 transcript level (Fig. S1a,b), and two independent transgenic lines overexpressing BBX28 fused to yellow fluorescent protein (YFP) under the control of a CaMV 35S promoter in wild type (Col-0) (BBX28OE, 35S-BBX28-YFP/Col-0; lines 1# and 4#) (Fig. S1b), while Col-0 and plants overexpressing YFP (YFPOE, 35S-YFP/Col-0) were used as controls. When these plants were grown under LD, bbx28 mutant flowered two days earlier with one rosette leaf less than Col-0, while BBX28OE lines showed severely late flowering phenotype, with about twenty days later and ten more rosette leaves than Col-0 ( Fig. 1a-c). As the two independent BBX28-overexpressing lines display similar phenotypes (Fig. 1a-c), we used BBX28OE-1# for further analysis.
We crossed BBX28OE with bbx28 mutant to obtain BBX28OE/bbx28 plants. We found that BBX28OE/bbx28 plants flowered similarly to BBX28OE plants ( Fig. S1c-e). The weak phenotype of bbx28 mutant might be due to the functional redundancies of Arabidopsis group V BBX proteins in flowering regulation (Fig. S1f).
BBX28 decreases FT transcription. To reveal genes involved in the negative role of BBX28 in regulating flowering, we tested the transcript levels of key genes involved in photoperiodic flowering, floral identity and FT transcription, including CO, FT, SOC1, LEAFY (LFY), FRUITFUL (FUL), TEMPRANILLO 1 (TEM1), FLC and SHOT VEGETATIVE PHASE (SVP) 16 , in 7-day-old LD-grown Col-0, bbx28 and BBX28OE plants. Quantitative real-time PCR (qRT-PCR) of seedlings collected at zeitgeber time (ZT) 4 showed that FT transcript level reduced most significantly in BBX28OE, followed by SOC1, LFY, FUL and SVP, while CO, TEM1 and FLC transcript levels remained unchanged (Fig. 2a).
As the expressions of CO and FT show rhythmic patterns during 24 h of a day in flowering pathway 4 , we tested whether the diurnal expression profiles of FT and CO are affected by BBX28. To this end, 7-day-old Col-0, bbx28 and BBX28OE seedlings grown under LD were harvested with an interval of three hours from dawn up to a 24 h-period. We found that FT transcript level was greatly decreased by BBX28 with more dramatically in late afternoon and dark (Fig. 2b). Interestingly, CO transcription was not affected by BBX28 during the day but decreased about half in BBX28OE plants after dusk (Fig. 2c).

BBX28 interacts with CO through its N-terminus.
Since FT is the direct target of CO, the reduction of FT transcript level by BBX28 encouraged us to test whether BBX28 interacts with CO. Co-localization analyses revealed that BBX28 and BBX28NT but not BBX28CT co-localized with CO in photobodies (Fig. 3a). We performed NoTS assay 43 to test the interaction between BBX28 and CO in vivo. We found that CO was successfully recruited to the periphery of nucleolus by Nuc2-BBX28 but not Nuc2-mCherry (Fig. 3b) www.nature.com/scientificreports/ interacts with CO in nuclei in vivo. LUC complementation imaging and yeast two-hybrid interaction assays also supported the interaction between BBX28 and CO through BBX28NT (Fig. 3c,d). The in vivo interaction between BBX28 and CO was further confirmed by Co-IP as CO was detected in BBX28-YFP-immunoprecipitated samples but not in the control (Figs. 3e; S2a). To study whether the B-box domain within N-terminus of BBX28 mediates its interaction with CO, BBX28 B-box (1-46aa) and BBX28NT with a deletion of the B-box domain (BBX28NTΔB-box, 47-112aa) were constructed (Fig. S2b). Yeast two-hybrids showed that the B-box domain but not BBX28NTΔB-box interacts with CO ( Fig. S2c), suggesting a central role of B-box domain in mediating the interaction between BBX28 and CO.
BBX28 inhibits CO targeting to FT locus without affecting the transcriptional activation activity of CO. To investigate the functional links among BBX28, CO and FT, we first compared the expression patterns among BBX28, CO and FT. Transgenic plants expressing GUS driven by the BBX28 promoter in wild type (pBBX28-GUS/Col-0) were generated and GUS staining showed that the expression pattern of BBX28 is similar to those of CO and FT 44,45 , with higher levels in vascular tissues (Fig. 4a). Dual-LUC assays (Fig. S3a) showed that the FT expression level was increased by CO but decreased by BBX28 (Fig. 4b), indicating a negative effect of BBX28 on FT transcription. Since CO activates FT transcription and BBX28 interacts with CO (Figs. 3 and 4b), we co-expressed BBX28 and CO and performed Dual-LUC assays to test the effect of coexpressing BBX28 and CO on FT expression. The results indicated that the FT transcript level was reduced upon co-expressing BBX28 and CO (Fig. 4b,c). Moreover, BBX28NT but not BBX28CT displayed a similar effect on FT transcription to that of full-length BBX28 upon co-expressing with CO (Fig. 4c). It was known that CO binds to FT promoter to activate its expression 5,17 . We asked whether BBX28 affects the DNA binding ability of CO to FT locus. Plants overexpressing CO (COOE; 35S-CO-YFP/Col-0) were crossed with BBX28OE line (BBX28OE; 35S-BBX28-3 × FLAG-mCherry/Col-0) to obtain CO and BBX28 co-expressing lines (BBX28OE × COOE). The line 1#, 3#, and 5# of BBX28OE × COOE were generated by crossing BBX28OE-2# with line 1#, 2#, and 6# of COOE respectively (Fig. S3b,c). COOE-2# and BBX28OE × COOE-3#, having a similar CO protein level which excluded the potential effects of different CO levels on our analysis (Fig. S3d), were selected for chromatin immunoprecipitation (ChIP)-qPCR assay. The level of CO recruited to the transcription start site of FT (FT2 fragment) was much more than that to other fragments of FT in COOE plants, exhibiting enrichment over four folds compared to the non-immune control (NIC) (Figs. 4d, S3e). When CO was co-expressed with BBX28, the level of CO targeted to FT2 fragment decreased to less than two folds compared to NIC (Fig. 4d), indicating that BBX28 decreases the recruitment of CO to FT locus. www.nature.com/scientificreports/ Next we tested whether BBX28 affects the transcriptional activation activity of CO (Fig. S3f). The expression of LUC (LUC/REN ratio) increased in Gal4BD-CO-YFP and Gal4BD-VP16-YFP (Fig. S3g), consistent with the previous report 5 . When Gal4BD-CO-YFP was co-expressed with BBX28 or TOE1, a positive control which was shown to repress the transcriptional activation of CO 18 , we found that the transcriptional activation activity of CO was not affected by BBX28 while inhibited obviously by TOE1 (Fig. 4e). Without Gal4BD-CO-YFP, BBX28 and TOE1 had no effects on the reporter LUC (Fig. S3h). Together, these results indicated that BBX28 does not affect the transcriptional activation activity of CO.
BBX28 N-terminal cysteines mediate heterodimerization between BBX28 and CO and affect the activation of CO on FT transcription. As BBX28 interacts with CO through its N-terminus (Fig. 3), we then asked whether the conserved cysteines in N-terminal domain (Fig. S4a) play a role in BBX28-CO interaction. We found that mutations of BBX28 N-terminal cysteines greatly blocked BBX28-CO interaction with BBX28 C5,8A , BBX28 C16A and BBX28 C24,27A displaying more obvious blocking effects, followed by BBX28 C47,50A and BBX28 C70,73A (Fig. 5a). Consistently, the co-localization signals in photobodies between CO and BBX28 C5,8A or BBX28 C16A decreased significantly, followed by BBX28 C24,27A , BBX28 C47,50A and BBX28 C70,73A mutants. BBX28 C5-C73A with all nine conserved cysteines mutated diffused in the nucleus without photobodies formed when co-expressing with CO (Fig. S4b,c). These results indicated that the N-terminal cysteines are essential for BBX28-CO interaction and C5, C8, C16, C24 and C27 sites located in the B-box domain (1-46aa) are more important. As BBX28 and CO (BBX1) both belong to BBX proteins, we concluded that BBX28 N-terminal cysteines mediated its heterodimerzation with CO.
To test whether these cysteines affect FT transcription, Dual-LUC assays showed that BBX28 cysteine mutations did not change the effects of BBX28 on FT expression, except BBX28 C5,8A mutation with less than one-fold increase (Fig. 5b). When those mutations were co-expressed with CO, BBX28 C5,8A , BBX28 C16A , BBX28 C24,27A and BBX28 C5-73A mutations, which affect the interaction between BBX28 and CO ( Fig. 5a; Fig. S4b,c), largely restored FT expression (Fig. 5c), supporting that these cysteines play an inhibitory role in the activation of CO on FT transcription, possibly by affecting the heterodimerization between BBX28 and CO. www.nature.com/scientificreports/

BBX28 genetically regulates flowering through CO and FT. To test the genetic relationships among
BBX28, CO and FT, we first generated bbx28co-9 double mutant by crossing bbx28 with co-9 mutant and analyzed their flowering phenotypes under LD. We found bbx28co-9 double mutant delayed flowering similarly to co-9 single mutant ( Fig. S5a-c). As the early flowering phenotype of bbx28 is weak ( Fig. 1) while co mutant (co-9) displays severe late flowering ( Fig. S5a-c), we then tested the genetic relationship between BBX28 and CO through analyzing the flowering phenotypes of plants overexpressing BBX28 (BBX28OE), CO (COOE) or BBX28 and CO (BBX28OE × COOE). The line 1#, 3#, and 5# of BBX28OE × COOE were generated by crossing BBX28OE-2# with line 1#, 2#, and 6# of COOE respectively, and the transcript levels of CO and BBX28 were confirmed by qRT-PCRs (Fig. S3b,c). We found that BBX28OE × COOE-5# flowered early with about twelve days earlier and five rosette leaves less than Col-0, similar to COOE-6# plants ( Fig. 6a-c); BBX28OE × COOE-1# flowered early with about eleven days earlier and four rosette leaves less than Col-0, similar to COOE-1# plants ( Fig. S5d-f); BBX28OE × COOE-3# flowered early with about ten days earlier and four rosette leaves less than Col-0, similar to COOE-2# plants (Fig. S5g-i), indicating the later flowering caused by BBX28 overexpression was compromised by CO overexpression. BBX28OE × COOE plants flowered no later than COOE plants, which might be due to that the high levels of CO transcripts in both genotypes, resulting in a saturated level of FT transcript for promoting flowering. Moreover, FT transcripts were much less in BBX28OE × COOE plants than that in COOE plants (Fig. 6d), although the CO transcript and CO protein levels were comparable in BBX28OE × COOE plants and corresponding COOE plants (Fig. S3b,  www.nature.com/scientificreports/ by BBX28OE × COOE (1# and 5#) (Fig. S3b), accordantly, the FT transcript level displayed similar patterns to CO, with highest in BBX28OE × COOE-3#, followed by BBX28OE × COOE (1# and 5#) (Fig. 6d). Together, these results suggested that BBX28 delays flowering in a CO and FT-dependent manner.

Discussion
The CO-FT module is the major determinant to regulate flowering in response to day length. In this study, we identified a new CO-interacting protein BBX28 which belongs to the Group V B-box transcription factors (BBX26-BBX32). We further dissected a specific role of N-terminal domain of BBX28 in the negative regulation of flowering.
In the light signaling pathways, it was shown that some components function through their different protein domains intelligently. The C-terminal domains of CRY1 and CRY2 (CCT1 and CCT2) interact with COP1 and mediate their signaling responses to light activation. The N-terminal domain of CRY1 (CNT1) mediates its constitutive dimerization, which is required for the light activation of CCT1 activity 46,47 . COP1 is a central switch for light signaling transduction. The nuclear localization signals (NLSs) of COP1 reside in its central core domain while the N-terminal portion acts as the major determinant for its cytoplasmic distribution. The coiled-coil Data are means ± SEM. Statistical significance was analyzed by student's t-test; **p < 0.01, *p < 0.05; ns, not significant. www.nature.com/scientificreports/ domain of COP1 mediates its homodimerization and the C-terminal WD40 domain interact with multiple transcription factors or photoreceptors for protein degradation [48][49][50] . Our results showed that the N-terminus of BBX28 mediates its interaction with CO and affects the recruitment of CO to FT locus, resulting in reduction of FT transcription with delayed flowering under LD (Figs. 3, 4 and 6). The late flowering phenotype is relieved when BBX28 N-terminus is deleted (Figs. 1d-f, 4c), supporting specific roles of the regulatory protein domains in the light signaling transduction. The protein interactions, such as BBX19-CO, BBX32-BBX4 and BBX21-HY5, indicated heterodimer formation among BBXs or between BBX and non-BBX proteins 16,27,29,51 . Arabidopsis BBX32 (AtBBX32) interacted with soybean BBX62 (GmBBX62) through the N-terminal B-box region of BBX32 52 . Those data suggest the B-box domain plays an essential and conservative role in heterodimerization, but the detailed molecular mechanism for the functioning of BBX domain is still not fully revealed. In this study, the interaction between BBX28 and CO (BBX1) provided another evidence to support heterodimerization of BBX proteins (Fig. 3). In addition, we further showed that several conserved cysteines in the N-terminus of BBX28, particularly those within the B-box domain, are essential for BBX28-CO interaction and the activation of CO on FT transcription (Fig. 5). The structural basis for the role of heterodimerization between BBX28 and CO under light/dark condition in flowering regulation will be of great interest to be investigated in the future.
The evolutionary and structural relationships of BBX proteins have been investigated in many plant species, from algae to monocots and dicots, by the phylogenetic approach 53  BBX28 was known to negatively regulate photomorphogenesis by interacting with ELONGATED HYPOCO-TYL 5 (HY5) and interfering with the activity of HY5 in its downstream target gene expression. Besides, COP1 interacts with BBX28 and mediates its protein degradation in darkness via the 26S proteasome, demonstrating key roles of BBX28 in COP1-HY5 axis to maintain proper HY5 activity and normal photomorphogenic development 42 . We showed that BBX28 interacts with CO and decreases the recruitment of CO to FT promoter and weakens the activation of CO on FT transcription. The BBX28-HY5 axis in photomorphogenesis and BBX28-CO-FT axis in flowering regulation indicated that BBX28 plays roles in two different light signaling pathways through diverse interacting proteins to regulate downstream gene expression. It was shown that COP1 interacts with BBX28 and CO and mediates their degradations in darkness 9,42 . The co-localizations of these three proteins in photobodies suggested that BBX28, COP1 and CO may act in the same complex (Fig. S6a). Genetic analyses indicated that the negative role of BBX28 in flowering is CO-dependent under LD (Figs. 6, S5). CO was shown to act genetically downstream of COP1 to regulate flowering 9 . It is therefore of interest to study if and how BBX28 affects COP1-mediated CO degradation in flowering when these three proteins interact with each other. Additionally, we found CO transcription was not affected by BBX28 during the day but decreased about half in BBX28OE plants after dusk (Fig. 2c). It is also of interest to explore the relationship between BBX28 and CO at the transcriptional level.
We proposed a working model for BBX28 in flowering regulation (Fig. S6b). CO binds to FT promoter to activate FT transcription. The interaction between N-terminus of BBX28 and CO decreases the recruitment of CO to FT locus. The N-terminal cysteines play an indispensable role in BBX28-CO heterodimerization and activation of CO on FT transcription. In wild type (WT/Col-0), the balance between BBX28 and CO maintains precise FT expression, which leads to normal flowering. When BBX28 is over-expressed (BBX28OE), the recruitment of CO on FT locus is decreased by the overdosed BBX28 protein, resulting in a significant reduction of FT transcript which delays flowering under LD. Considering the similar protein structures of B-box transcription factors, the working model of BBX28 presented here might be applied to other BBX proteins in some degree. In addition, as BBX proteins have roles not only in flowering regulation, but also in photomorphogenesis, shadeavoidance response, abiotic and biotic stresses and plant hormonal pathways 27 , it is therefore of interest to study if and how the working model of BBX28 applies to those signaling pathways.

Methods
Plant materials and growth conditions. Arabidopsis thaliana ecotype Columbia-0 (Col-0) was used as the wild type for all experiments. The T-DNA insertion mutants of bbx28 (SAIL_828_G11) and co (co-9, SAIL_24_H04) were obtained from the Arabidopsis Biological Resource Center (ABRC), and the homozygous lines were identified by PCR using the primers listed in Table S1. The seeds were surface sterilized and plated on Murashige and Skoog (MS) plates. After stratification for 4 days in dark at 4 °C, the plates were transferred to the growth chambers at 22 °C with different light conditions based on experimental requirements. The long day (LD) condition was 16 h light/8 h dark.
Yeast two-hybrid assays. Yeast two-hybrid interaction assays were performed according to the Yeastmaker Yeast Transformation System 2 User Manual (Clontech). The full lengths of CDS, truncated fragments or mutated sequences of genes of interest were subcloned into pGADT7 or pGBKT7, respectively. The constructs Scientific Reports | (2020) 10:17789 | https://doi.org/10.1038/s41598-020-74445-7 www.nature.com/scientificreports/ were then co-transformed into yeast (AH109) according to the user manual. The yeast cells containing the bait and the prey constructs were grown on selective plates (SD-Leu-Trp, SD-Leu-Trp-His, SD-Leu-Trp-His-Ala and SD-Leu-Trp-His + 3-AT) for analysis. The concentration of 3-AT was 3 mM or 10 mM.
Microscopy. Tobacco (N. tabacum) leaves were used for transient expression assays. The constructs were introduced into Agrobacterium tumefaciens strain GV3101 by electroporation and infiltrated into tobacco leaves with an injection syringe. 48 h later, the infiltrated tobacco leaf disks were subjected to microcopy analyses as described previously. Image stacks of nuclei were subjected to deconvolution by using softWoRx software (Applied Precision) [54][55][56] . Co-localization or Nucleolus-tethering System (NoTS) assays were performed as previously described 43 .
Firefly luciferase (LUC) complementation imaging assays. The firefly luciferase (LUC) complementation imaging assay was performed as previously described 57 . BBX28, BBX28NT (1-112 amino acids, aa), BBX28CT (113-223 aa) and CO were fused with the N-or C-terminal fragment of LUC (NLuc and CLuc) respectively. The fused plasmids were introduced into GV3101 by electroporation and then co-infiltrated into tobacco (N. benthamiana) leaves with an injection syringe. 48 h later, the infiltrated leaves were injected with 100 mM luciferin (Sango, dissolved in water) and the luciferase signals were detected by the PMCapture software (Version 1.00) of a Chemiluminescence Imaging System (Tanon 5500, Shanghai, China).

Transgenic plants. The constructs were introduced into GV3101 by electroporation and transformed into
Col-0 by the floral-dip method 58 . To obtain transgenic plants in different mutant backgrounds, the plants in Col-0 background were crossed with mutants and the backgrounds were identified by PCRs. Primers used were shown in Table S1.
Site-direct mutagenesis. Site-direct mutagenesis was generated by the two-step overlap PCR. We designed a pair of mutant primers by artificially changing the base sequences of interest. When introducing cysteine to alanine (Cys-to-Ala) substitutions, we changed the bases to GCC. Then PCRs were performed by using the wild type forward primer and mutated reverse primer or the mutated forward primer and wild type reverse primer. The same amounts of the purified PCR products were mixed and used as the template for final PCR, which was performed by using the wild type forward and reverse primers. The site-direct mutagenesis was confirmed by sequencing. Primers used were shown in Table S1. Quantitative real-time PCR. Expression analyses were performed by quantitative real-time PCR (qRT-PCR) as described previously 43  www.nature.com/scientificreports/ powder). Lysed proteins were placed on ice for 20 min and mixed several times. Then proteins were centrifuged at 14,000×g for 10 min at 4 °C. The supernatants were transformed to new tubes and 5 × SDS-PAGE loading buffer (250mMTris-HCl, pH 6.8, 10% (w/v) SDS, 0.5% (w/v) BPB, 50% (v/v) Glycerol, 5% (v/v) β-mercaptoethanol) was added. Samples were boiled at 100 °C for 5 min and subjected to SDS-PAGE for Western blotting assays. Blotting signals were detected by the PMCapture software (Version 1.00) of a Chemiluminescence Imaging System (Tanon 5500, Shanghai, China).
Dual-luciferase (dual-LUC) assay. The Dual-LUC assay was performed as described previously 59 . The FT promoter (1800 bp) was subcloned into pGreenII 0800-LUC vector (a kind gift from Prof. Hongtao Liu) as the reporter and full-length of BBX28 CDS, BBX28 truncates or point mutations and full-length of CO CDS were subcloned into pCambia131-35S-N1-YFP 54 as the effectors. pCambia131-35S-YFP (YFP) was used as the control effector. The reporter construct was introduced into the GV3101 harboring pSoup-P19. The effecter constructs were introduced into GV3101. The cells were harvested when grown to OD600 = 0.8-1.0 and washed once and then resuspended by sterilized water to OD600 = 0.8. The reporter and the effecter were mixed together at 1:2 ratio for co-infiltration of the tobacco leaves. For assays containing two effectors, the reporter and effectors were mixed together at 1:2:2 ratio. The control and the experimental samples were infiltrated into the same tobacco leaf and more than three independent leaves were served as technical replicates. 48 h later, the tobacco leaf disks in the infiltrated areas were collected by a puncher with 1.1 cm in diameter for protein extraction. Extracted proteins were subject to Dual-LUC assays by the Dual-Luciferase Reporter Assay System (Promega E1910) with GloMax 20/20 Luminometer software (version 1.10) (Promega). The samples were ground in liquid nitrogen in 1.5 ml EP tubes and extracted by 100 µl 1 × Passive Lysis Buffer (5 ×), incubated on ice for 15 min and mixed several times for efficient lysis. The extracted proteins were centrifuged at 14,000×g for 10 min at 4 °C and supernatants were transformed into new tubes. 8 µl supernatant was transformed to a new tube and 40 µl luciferase assay buffer was added and mixed for recording the LUC value. Then 40 µl Stop&Glo buffer was added and mixed for recording the REN value. The LUC/REN ratios were calculated. Data from the luminometer were exported by the GloMax Spreadsheet Interfacer (Promega).
Chromatin immunoprecipitation (ChIP) assay. ChIP assay was performed as previously described 60 .