Two CONSTANS-LIKE genes jointly control flowering time in beet

Breeding vegetative crops (e.g. beets, cabbage, forage grasses) is challenged by two conflicting aims. For field production, flowering must be avoided while flowering and seed set is necessary for breeding and seed production. The biennial species sugar beet makes shoot elongation (‘bolting’) followed by flowering after a long period of cold temperatures. Field production in northern geographical regions starts in spring. A thickened storage root is formed only during vegetative growth. It is expected that winter beets, which are sown before winter would have a much higher yield potential. However, field production was not possible so far due to bolting after winter. We propose a strategy to breed winter beets exploiting haplotype variation at two major bolting time loci, B and B2. Both genes encode transcription factors controlling the expression of two orthologs of the Arabidopsis gene FLOWERING LOCUS T (FT). We detected an epistatic interaction between both genes because F2 plants homozygous for two B/B2 mutant alleles did not bolt even after vernalization. Fluorescence complementation studies revealed that both proteins form a heterodimer in vivo. In non-bolting plants, the bolting activator BvFT2 was completely downregulated whereas the repressor BvFT1 was upregulated which suggests that both genes acquire a CONSTANS (CO) like function in beet. Like CO, B and B2 proteins house CCT and BBX domains which, in contrast to CO are split between the two beet genes. We propose an alternative regulation of FT orthologs in beet that can be exploited to breed winter beets.

The transition from the vegetative to the generative phase is of major interest to crop breeders due to its high relevance for yield and quality. Crop plants show great variation regarding their phenological development. If vegetative parts of the plant are harvested (leaves, roots) they must not enter the reproductive phase, a major step in plant development commonly referred to as floral transition. Sugar beet (Beta vulgaris L.) is a typical vegetative crop with a biennial life cycle. After sowing in spring, it produces huge leaf and root mass until harvest in autumn. As a result of secondary thickening, a storage root is produced with sucrose contents between 17-20% 1 . As a biennial plant it enters the reproductive phase only after exposure to a long period of cold temperatures (<4 °C). Then, the shoot is elongated ('bolting') and flowers are produced. Early bolting under field conditions must be strictly avoided because it gives rise to flowering plants with small roots and low sucrose content. For seed production, plants must bolt and flower early after winter. This follows, that conventional sugar beet cannot be cultivated as a vegetative crop over winter, commonly referred to as 'winter beet' 1 .
Quantitative trait loci (QTL) and major genes controlling bolting time have been mapped to the nine beet chromosomes 2 . The bolting time QTL SEASONAL BOLTING-4 and -9 (SBT-4, SBT-9) accounts for up to 52% of the phenotypic variation 3 . The phenotypic effect of SBT-4 is likely caused by the major flowering time regulator BvFT2 because they were mapped to the same position on chromosome 4. SBT-9 was precisely mapped to the position of BR1. This QTL was recently fine mapped by a sequencing approach and a gene similar to CLEAVAGE AND POLYADENYLATION SPECIFICITY FACTOR 73-I (CPSF73-I) from Arabidopsis was suggested as a candidate gene for this QTL 4 .
Sugar beet has two sequences which share high homology to FLOWERING LOCUS T (FT) a major integrator of signals from different regulatory pathways triggering floral transition in Arabidopsis 5 . BvFT1 is a floral repressor which is transcriptionally active before winter and prevents bolting. In contrast, BvFT2 is a floral inducer which is activated during vernalization. A high BvFT2 activity is indicative for generative (bolting) beet plants 5 .
Two upstream regulators of the two BvFT orthologs have been cloned. BOLTING TIME CONTROL 1 (BTC1) belongs to the PRR3/7 clade of PSEUDO RESPONSE REGULATOR (PRR) genes that are components of the Scientific  photoperiod pathway in Arabidopsis 6 . A dominant allele which is highly abundant in wild beet (B. vulgaris ssp. maritima) populations from the Mediterranean causes early bolting (without vernalization) resulting in an annual life cycle. Another PRR7 homolog, BvPRR7, is a cold responsive gene with a clock function in beets but not involved in bolting time regulation 7 . The second bolting time gene, BvBBX19 encodes a putative transcription factor with two B-Box zinc finger motifs but lacking a CCT domain 8 . Recently, haplotype variation of the four major bolting time genes from beet have been studied in wild and cultivated beet accessions 9 . For BTC1 and BvBBX19, 14 and 7 haplotypes were found, respectively 6,9,10 . They were classified as annual or biennial bolting time regulators. BTC1 and BvBBX19 share homology with the transcription factor CONSTANS (CO), which regulates floral transition in Arabidopsis in a long day (LD) dependent manner 11 . It has two consecutive Zn finger domains which are called B-Boxes 12 . Mutants with amino acid alterations in conserved residues of the B-Boxes are late flowering. At the C-terminus, the CO protein has a CCT (CO, CONSTANS-LIKE, and TIMING OF CAB EXPRESSION1) domain which includes a nuclear import signal. By its CCT domain, CO binds to the ubiquitin ligase COP1 and to the FT promoter by forming complexes with other transcription factors 13 . This sequence is strictly conserved in proteins which are constituents of the circadian clock 12 . CDF (CYCLING DOF FACTORS) transcription factors bind to the CO promoter and inhibit its expression during the morning. Later, they are degraded by the proteasome when GIGANTEA (GI) interacts with FLAVIN BINDING, KELCH REPEAT, F-BOX PROTEIN 1 (FKF1) and ZEITLUPE (ZTL) resulting in strong transcriptional upregulation of CO 14 . The CO protein is stabilized by light and degraded in darkness after ubiquitination and proteolysis by the 26S proteasome 15 . In Arabidopsis, apart from CO there are at least 31 genes encoding proteins with B-Box and CCT domains, 16 are CO-Like (COL) proteins with one or two B-Boxes and one CCT domain, the remaining ones are either lacking the CCT domain, or one B-Box and the CCT domain 16 . BBX19 and CO are forming dimers which jointly regulate FT in an antagonistic way 17 . BBX32 physically interacts with COL3 to form a dimer which targets the FT promoter 15 . Interestingly, beet has a large CONSTANS-LIKE gene family but is lacking a functional CO ortholog with both domains 18 . BTC1 is lacking a B-Box and BvBBX19 is lacking a CCT domain.
The purpose of this work was to understand the genetic and physical interaction between BTC1 and BvBBX19 and to lay the foundations to breed winter beets. We assumed that both proteins work together to acquire a CO-like function. To test our hypothesis, we studied an F 2 population segregating for both genes. We found an epistatic interaction between both loci which resulted in three different life cycle regimes. Combining two mutant alleles resulted in plants which completely lost their competence to bolt after vernalization. The genetic data were confirmed by yeast-two-hybrid interaction and in planta bimolecular fluorescence complementation studies. Double mutant plants are proposed as prototypes for winter beet breeding which requires complete bolting control after winter.

Results
The B2 locus is epistatic to B. We produced an F 2 population from a cross between two biennial beet genotypes, seed code 093187 (B a B a B2 f B2 f ) and 056822 (B d B d B2 h B2 h ) which differed by their B and B2 alleles. 145 plants were grown under long day conditions together with their parents and the annual and biennial controls. We determined the genotypes of the B and B2 loci for all F 2 plants using the markers CAU4234 and CAU4235 (Supplementary Tables 1 and 2). In accordance with their position on different chromosomes, the observed genotypic segregation fitted a random segregation ratio (χ 2 = 15.77; α = 0.05) (Supplementary Tables 3 and 4).
The biennial controls bolted within 3-4 weeks after vernalization whereas the annual controls bolted early (4-6 weeks after sowing) without vernalization required (Fig. 1A). Most F 2 plants carrying at least one B d and one B2 f allele were lacking a vernalization requirement because they bolted within 114 days after sowing like plants from the annual controls (Fig. 1A ) do not bolt after vernalization giving rise to a phenotypic segregation of 9:6:1 (annual: biennial: non-bolting after vernalization). This hypothesis was rejected after a χ² test for goodness of fit to a 9:6:1 ratio (χ 2 = 150.03; α = 0.01). The second hypothesis is based on the assumption that the B2 h allele acts epistatically over the B locus. In this case, a 9:4:3 phenotypic segregation was to be expected (Supplementary Table 4). As this segregation rate was not rejected (χ 2 = 2.24; α = 0.01), we assume that the B2 h allele which was derived from an EMS mutagenesis acts epistatically to B resulting in a non-bolting (after vernalization) phenotype (Supplementary Table 4). However, this interaction does not fully explain phenotypic variation because biennial plants were found in the B d B2 h parent 056822 and among the corresponding F 2 genotypes (Fig. 1A). In conclusion, genetic analyses are clearly pointing at a joint activity of both loci to control the onset of bolting. We reasoned that the BTC1/BvBBX19 genotype impacts the expression of the two FT paralogs BvFT1 and BvFT2. It had been demonstrated that floral transition in beet is promoted through downregulation of the floral repressor BvFT1 and therewith upregulation of the floral inducer BvFT2, which both are downstream targets of BTC1 and BvBBX19 5,6,8 . We observed that the transcriptional activity of BvFT1 and BvFT2 follows the anticipated expression pattern (Fig. 2B,D). As expected, BvFT2 was highly upregulated and BvFT1 completely downregulated after vernalization in the biennial controls and in biennial F 2 plants. Interestingly, a contrasting expression pattern was observed in F 2 plants which did not bolt after vernalization. The transcriptional activity of BvFT1 was two times higher in non-bolting F 2 plants homozygous for B a /B2 h as compared to F 2 plants homozygous for B d /B2 h .
BvBBX19 and BTC1 physically interact with each other. The absence of a gene in sugar beet encoding a canonical CO protein suggested that one or several other proteins fulfill the function of CO in this plant. The most likely candidates are BvBBX19 and BTC1 since they contain two B-Box domains and the CCT domain, respectively resembling the CO domain structure 8 . One likely scenario how BvBBX19 together with BTC1 can replace CO is direct physical interaction between the two proteins resulting in a functional CO ortholog. To test this hypothesis, we performed yeast-two-hybrid studies. Constructs were made containing the full-length coding regions of BvBBX19 and BTC1 fused to either the GAL4 DNA-binding domain (BD) or the GAL4 activation domain (AD) at the N-terminus of the respective proteins. In addition, we included constructs of the previously identified BvBBX19 mutant (BvBBX19 h ), which contains a premature stop codon resulting in a BvBBX19 variant with only one B-Box 8 . Again, AD or BD-domains were fused to the N-terminus of BvBBX19 h . Wild type BvBBX19 a and BTC1 d constructs showed no autoactivation and were thus useful to study interaction between the two proteins. For both combinations of wild type BvBBX19 a with BTC1 d we observed growth of yeast cells on selective plates (-Leu, -Trp, -His) as well as induction of the α-galactosidase reporter in the quantitative assays (Fig. 3). Thus, BvBBX19 a and BTC1 d interact. In case of BvBBX19 h we observed autoactivation for the BD-BvBBX19 h construct. Thus, this construct was not useful for further interaction studies. However, AD-BvBBX19 h did not result in autoactivation. In combination with BD-BTC1 d , colonies were formed on selective medium and α-galactosidase activity induced. This result implies that the second C-terminally located B-box in BvBBX19 a , which is missing in BvBBX19 h , is not essential but supportive for the interaction with BTC1 d . Y2H data strongly suggested direct physical interaction between BvBBX19 and BTC1. For further confirmation, we applied ratiometric bimolecular fluorescence complementation assays (rBiFC). We used constructs where either the 5′-or the 3′ region of BvBBX19 a was fused with the 5′-terminal half of YFP (nYFP). Accordingly, BTC1 d was fused with the C-terminal part of YFP (cYFP) at its N-or C-terminus. These constructs were co-transfected into Nicotiana benthamiana leaves using Agrobacterium-mediated infiltration. All four combinations resulted in YFP signals (Fig. 4A) in contrast to co-expression of the non-fused nYFP controls with BTC1 d fused to cYFP at its N-terminus or C-terminus (Fig. 4B). The truncated BBX19 version nYFP-BvBBX19 h co-transfected with BTC1 d carrying cYFP at the N-terminus or C-terminus also gave a clear YFP signal in contrast to BvBBX19 h constructs carrying nYFP at the C-terminus of BvBBX19 h (Fig. 4C). The latter result is expected since BvBBX19 h contains a stop codon upstream of the second B-Box and thus does not allow expression of the C-terminal YFP half. Interestingly, in all cases complemented YFP signals were observed in nuclear bodies. Quantification of the YFP against the RFP fluorescence signals from at least 20 images (Fig. 4D) are consistent with the representative pictures presented in Fig. 4A-C.

Discussion
We have performed a genetic study with all combinations of BTC1 (B) and BvBBX19 (B2) alleles in an F 2 population. B d is a typical annual allele only found in wild beet populations from the Mediterranean. B a carries six non-synonymous SNPs and a large insertion in the promoter compared to 'annual' alleles 6 . B2 f is found in annual beets 9 while B2 h is a nonsense EMS mutant allele 8 .
Beet is a typical long day plant. Bolting even in the presence of the early bolting alleles or after vernalization is strongly delayed in short days 19 . Therefore, all experiments were performed under long day conditions. An annual life cycle requires an annual B allele and a functional B2 allele (B d /B2 f ). The competence for early flowering is   6 . Double mutants homozygous for the biennial B and the non-functional B2 allele completely lost their competence to bolt which confirmed our initial hypothesis that B and B2 jointly regulate the onset of bolting in sugar beet. By combining two mutant alleles, we could select plants that did not bolt even after cold treatment. We found an incomplete epistatic interaction between bolting resistant plants among all B2 h homozygous genotypes, but the presence of the B d alleles modified the B2 h effect because biennial plants were present in the B d B2 h parent and F 2 plants. We assume that apart from these two major bolting time regulators, additional genes can modify bolting time. Moreover, the rare occurrence of spontaneously bolting plants in production fields points at environmental factors modifying the activities of B and B2. These factors together may explain the presence of biennial plants in the 056822 parent.
The genus Beta comprises iteroparous perennials with an annually repeated requirement for vernalization 20 . Future studies with perennial wild beets will resolve the question whether the BTC1/BvBBX19 module and the BR1 QTL 4 control the perennial life cycle. However, strictly non-bolting genotypes are likely to be a dead end of  evolution because they cannot reproduce sexually in contrast to iteroparous plants which flower and set seeds in subsequent years after winter. Thus, it is no surprise that despite of extensive screenings BTC1/BvBBX19 double mutants have not been found in nature so far. The B2 h genotypes exhibited a strong requirement for vernalization even in the presence of the early bolting allele B d . This indicates that there are upstream regulators of the BTC1/BvBBX19 module which respond to cold temperatures and to alterations of the B2 h protein. This makes B2 a primary target of a putative vernalization regulatory pathway. However, no further mutants have been detected so far. Searching for orthologous genes from Arabidopsis has not been successful and beet lacks a functional ortholog of FLOWERING LOCUS C which is a major integrator of signals from the vernalization pathway in Arabidopsis. Seemingly, divergent vernalization pathways have evolved in both species. Because vernalization has an epigenetic basis, genes responding to methylation might be interesting candidates. Consequently, two genes, SHORT VEGETATIVE PHASE (BvSVP) and BvVIN3 come into focus as upstream regulators of the BTC1/BvBBX19 module because they are hypomethylated and/or differentially expressed after cold exposure 21,22 .
How can the bolting-resistant phenotype be explained by protein-interaction and expression studies? BTC1 requires a functional BvBBX19 protein consistent with our data that beets carrying the annual B d allele do not bolt in the presence of the mutant B2 h allele. Loss of competence to bolt is due to downregulation of the floral inducer BvFT2 and upregulation of the floral repressor BvFT1 6 . The non-bolting phenotype of the B2 mutant is not caused by a lack of protein-interaction because binding between BvBBX19 h and BTC1 d was demonstrated (Figs 3 and 4). Y2H data imply that BBX19 and BTC1 do not require another beet protein for their interaction. However, it cannot be excluded that BTC1 and BvBBX19 interact with other proteins as CO does with PHYTOCHROME INTERACTING FACTOR 4 23 . We propose a model where BTC1 and BvBBX19 (mutated and wild type) dimerize to bind to the BvFT2 promoter. Likewise, the interaction between Arabidopsis COL proteins and CO has been demonstrated. B-BOX 32 binds to CONSTANS-LIKE3 15 and BBX19 binds to CO to suppress its function as an activator of FT 17 . Consequently, the heterodimer cannot bind to the Box 1 Motif of the FT promoter which is essential for binding of the CO protein. We reason that BTC1 or BvBBX19 alone are not able to bind to the BvFT2 promoter and that the heterodimer of mutant BvBBX19 h with the BTC1 d protein cannot bind to the BvFT2 promoter. The importance of intact domains for their binding activities was recently reported for Arabidopsis where a truncated COβ variant lacking the CCT domain lost its DNA-binding affinity 24 . The variant protein results from alternative splicing of the CO mRNA. Moreover, the truncated protein inhibits the function of the full-size COα protein by reducing its protein abundance and preventing its DNA-binding. A similar mechanism of CO-BBX functional interaction has been reported for rice where OsBBX14 activates the CO ortholog Hd1 which is a repressor of the rice FT ortholog Heading date 3a (Hd3a) under LD conditions 25 . In Arabidopsis, CO was shown to bind to a tandemly repeated sequence element of the FT promoter [consensus TGTG(N2-3)ATG motif] 26 . A promoter analysis of the beet FT genes revealed that this element is lacking from the 5′ regions of BvFT1 and BvFT2 (Supplementary  Table 5). Moreover, overexpression of BvBBX19 and BTC1 in Arabidopsis CO mutants did not accelerate flowering (data not shown). Evidently, plants have evolved different mechanisms to control FT expression. In Arabidopsis and rice, CO-like genes and CO orthologs gained different functions as both activators and suppressors of their downstream target. We propose an alternative mechanism for beet, where two FT paralogs are differentially regulated by two CO-like genes whose function depends on vernalization. The upstream regulators responding to external cues are still unknown. Moreover, the involvement of other homologs of CO binding proteins such as TARGET OF EAT1 (TOE1) or small B-BOX protein (MiP1a and MiP1b) 11 from Arabidopsis remains to be demonstrated.
Simon et al. 27 suggested, that CO has been derived from COL genes and that the function of the CO protein, which is specific to Brassicaceae species gave Arabidopsis an adaptive advantage during its expansion to northern geographical regions. Also B. maritima spread to northern regions after the last ice age. It is tempting to speculate that the flexible B/B2 module was an important factor for its adoption to winter climates and LD conditions. We reason that BTC1 and BvBBX19 must also perceive signals from the photoperiod and the vernalization signaling pathways because bolting initiation depends on long days and exposure to cold temperatures. A recent study with Arabidopsis demonstrated that the PRR proteins play an important role in stabilizing the CO protein. They suppress the proteasomal degradation of CO and contribute to light-mediated accumulation of CO during the day 28 . In beet, one member of the PRR clade has been further studied. Interestingly, BvPRR7 is a cold responsive gene with a clock function and caused late flowering after overexpression in Arabidopsis 7 . Future studies are needed to show if this gene plays a role in beet as an upstream regulator of the BTC1/BvBBX19 module.
This study has importance for breeding vegetative crops which are sown in spring under winter climate conditions. After early sowing, cold temperatures can pose a risk because they cause early bolting which drastically reduces yield (e.g. cabbage, carrots, salad, beet root) 29 . Therefore, breeders have been selecting for bolting resistant mutants, many of these carry mutations in functional orthologs of FLC (only Brassica species), CO or FT. Breeding winter beets requires full bolting control after winter. In contrast to traditional 'spring beets' , they must not bolt after winter. This can be achieved by selecting for non-bolting (after vernalization) alleles from the BR1 QTL on chromosome 9 4,30 . Alternatively, we propose a haplotype-based breeding strategy using well defined BTC1 and BvBBX19 alleles. But how can we harvest seeds from the parents if they are already bolting resistant? This problem could be overcome by introducing the early bolting B allele (e.g. B d ) into non-bolting parents turning them into biennials which can flower and set seeds after winter. We propose a haplotype swapping strategy where different B and B2 alleles 9 are combined with each other. A second approach relies on conditional bolting of B h parents. Non-bolting plants can enter the reproductive phase under extreme environments. We have obtained seeds from B h genotypes after cultivation in a climate chamber under 24 hours light and largely extended vernalization period. As an alternative, the bolting resistance of B h seed parents could be overcome by field cultivation in southern regions under high temperatures. It was recently shown, that CO expression increases under high temperatures 23 . Although this was observed under SD conditions, it is tempting to speculate that B/B2 allele combinations display different temperature sensitivity before and after vernalization.

Materials and Methods
Plant material and growth conditions. We performed a cross between two single plants of the biennial beet lines seed code 056822 (plant #15) and 093187 (plant #8). The female parent 056822 carries the BTC1 d allele only found in annual beets which confers early bolting without vernalization 6 and a mutated BvBBX19 allele 8 which we termed BvBBX19 h following the haplotype nomenclature described by Höft et al. 9  For phenotyping, 145 plants of the F 2 population 142063 were grown in a climate chamber under long day conditions (16 h light/8h dark, 320 µmol m −2 s −1 ) for 325 days. The two parent lines 056822 and 133703 (selfing progeny of 093187), three biennial (seed codes 092492, 930184, 930176) and two annual genotypes (001684, 991971) were grown as controls (five plants per line). Plants were first grown in 9 cm pots for 135 days at 20 °C and then cold treated at 4 °C for 12 weeks, followed by an acclimatization phase at 12 °C for three days. For the rest of the experiment, they grew again in 11 cm pots at 20 °C for another 102 days. Every second day, plants were randomized and the onset of bolting was recorded (BBCH scale code: 51) according to Meier et al. 31  DNA techniques. For DNA isolation, leaves were harvested from six-weeks-old F 2 plants and freeze dried.
Genomic DNA was isolated applying the CTAB method 32 . A 10-fold dilution of the extracted DNA was later used for PCR using Taq DNA Polymerase (Invitrogen). We used the InDel marker CAU4234 and the CAPS marker CAU4235 for genotyping the BTC1 and BvBBX19 locus, respectively (Supplementary Table 2). PCR products were separated on 1% agarose gels.  B a B2 a B2 a ). Total RNA was isolated from young leaves that were harvested 23 days after cold treatment in a 4 hours interval over 24 hours (first measurement at ZT 0, the time of lights on). Total RNA was extracted with the peqGOLD Plant RNA Kit (PeqLab) and subsequently treated with DNase. 500 ng of total RNA was reverse transcribed using a First Strand cDNA Synthesis kit (Fermentas). Resulting cDNA was diluted 10-fold and 2 µl of the dilution were used as a template for qRT-PCR. Three independent biological and three technical replicates were analyzed. qRT-PCR was performed with a Platinum SYBR Green Mastermix (Invitrogen) on a CFX96 Real-Time PCR detection system (Bio-Rad) with a final reaction volume of 20 µl and a final primer concentration of 20 pM. The housekeeping gene BvGAPDH was used as a reference. Data were analyzed with the CFX Manager Software v2.1 (Bio-Rad). Expression levels were first calculated with the comparative CT (Δ CT ) method and then normalized to the geometric mean of BvGAPDH to calculate the relative expression levels.

Yeast-2-Hybrid assays. Yeast-2-Hybrid experiments were performed using the Matchmaker Gold
Yeast-Two-Hybrid System (Clontech). The proteins of interest were fused at their N-terminus to either the DNA binding domain (BD) or the activation domain (AD) of the GAL4 transcription factor by insertion of the full-length coding sequences of BvBBX19 a , BvBBX19 h (BvBBX19 mutant) and BTC1 d into the NcoI and XhoI sites of the vector pACT2 or the NcoI and SalI sites of the vector pAS2-1. Full-length coding sequences of BvBBX19 a , BvBBX19 h and BTC1 d were obtained by PCR with primers listed in Supplementary Table 2. The correctness of the amplified sequences was verified by sequencing. Yeast cells (strain Y2H Gold, Clontech) were transformed according to the supplier's manual. Screening for histidine auxotrophy was done with nine clones of each transformant which were spread on non-selective (-Leu, -Trp) or selective (-Leu, -Trp, -His) plates and incubated for two days at 28 °C.
Quantification of interaction was determined by the α-galactosidase-assay 33 with minor modifications. For this purpose, yeast transformants were cultured overnight in 3 mL selective medium. After measuring OD 600 of the overnight culture and pelleting the cells, 200 µL of the overnight medium were mixed with 600 µL assay buffer (0.33 M sodium acetate, pH 4.5, 10 mg mL −1 p-nitrophenyl-alpha-D-galactopyranoside) and incubated at 29 °C. After 21 h of incubation 200 µL stopping buffer (2 M sodium carbonate) were added and OD 410 was measured. α-galactosidase activity was calculated as: α-galactosidase units = 1,000 × OD 410 /(t × V × OD 600 ), where t = time of incubation in min, V = volume of culture, OD 410 = absorbance by p-nitrophenol, OD 600 = cell density at the beginning.

Ratiometric Bimolecular Fluorescence Complementation (rBiFC) and immunoblots. For rBiFC,
BvBBX19 a , BvBBX19 h and BTC1 d were C-or N-terminally fused to either the N-terminal or the C-terminal half of YFP (n/cYFP). Thus, eight different construct combinations were obtained and the unfused N-terminal half of YFP was used as negative control in combination with cYFP-BTC1 d or BTC1 d -cYFP.
For generation of constructs, the full-length coding sequences of BvBBX19 a , BvBBX19 h and BTC1 d were PCR amplified with att sites allowing recombination into the entry vectors pDONR221-P1P4 and pDONR221-P3P2 followed by recombination into the Gateway vector pBiFCt-2in1 that also provides RFP as an internal standard 34 . The obtained constructs were transformed into Agrobacterium tumefaciens strain GV3101(pMP90) 35 . 4-weeks-old Nicotiana benthamiana plants were transiently co-transformed by Agrobacterium infiltration 36 with the construct combinations mentioned above and with p19 to suppress gene silencing 37 . Three days after infiltration YFP complementation was analyzed using a Leica TCS SP5 Confocal Laser Scanning Microscope (Leica). YFP was excited with a 488 nm laser and RFP with a 561 nm laser. YFP fluorescence was detected between 535 nm and 560 nm, RFP fluorescence between 600 nm and 625 nm. Quantification of the mean fluorescence of split-YFP was done by normalization against RFP. Data were calculated as means of at least 20 images selected at random. Relative fluorescence was determined using ImageJ estimating the mean grey value of the different pictures within an area of around 5 pixels. The maximum grey value per pixel of YFP fluorescence was set as 225. Expression of BvBBX19 a or BvBBX19 h fused to nYFP or unfused nYFP (as negative control for rBiFC) was detected via the HA-tag positioned at the C-terminus of nYFP (Fig. S1). Proteins were extracted from infiltrated leaf tissues by TCA precipitation 38 . 40 µg of proteins per lane were separated by SDS-PAGE. After blotting, the nitrocellulose membrane was blocked with 7% milk powder in TBS and probed with rat anti-HA antibody (Roche, 11867423001, 1:1,500 in TBS-T) and secondary αRat-HRP antibody (Millipore, NMM1767593, 1:10,000 in TBS-T) using the ECL detection assay (Bio-Rad) according to supplier's manual.