We used four tobacco species in this study: Nicotiana benthamiana Domin, Nicotiana tabacum L. cv. SR1, Nicotiana sylvestris Speg. & Comes, and Nicotiana tomentosiformis Goodsp.. Wild-type tobacco plants were sown and cultivated in soil under LD conditions in the greenhouse (16-h photoperiod, artificial light switched on if natural light fell below 700 μmol m^-2 s^-1, 22–25°C under light, 19–25°C in the dark), or under SD conditions in phytochambers (8-h photoperiod, 200 μmol m^-2 s^-1, 25–27°C under light, 20°C in the dark). For Agrobacterium-mediated transformation, wild-type N. tabacum cv. SR1 and N. sylvestris plants were germinated and grown under sterile conditions (LD, 16-h photoperiod, 23°C, 100 μmol m^-2 s^-1) on MS medium (Murashige and Skoog, 1962). For the analysis of T₁ transgenic N. tabacum lines carrying overexpression, empty vector control or promoter–reporter cassettes and transgenic N. sylvestris lines, seeds were germinated in a sterile environment under LD (N. tabacum only) or SD conditions on selective MS medium (25 mg/L hygromycin, 100 mg/L kanamycin or 3 mg/L phosphinothricin, as appropriate). Seedlings were then transferred to soil and cultivated in the greenhouse or in phytochambers as described above. In contrast, T₁ plants and subsequent generations of the CRISPR/Cas9 knockout lines were directly sown and cultivated under LD or SD conditions in soil as described above. Plant material for the isolation of gene sequences, expression analysis, and immunodetection experiments was snap-frozen in liquid nitrogen immediately after harvest and stored at –80°C. Further information on plant material, growth conditions and the harvesting time points is set out in the Supplementary Materials and Methods (incl. Supplementary Table S1 ).

Constructs for overexpression, promoter–reporter analysis, immunodetection of fusion proteins, subcellular localization and CRISPR/Cas9 knockouts are described in the Supplementary Materials and Methods (incl. Supplementary Tables S2 – S8 ).

Leaf material was ground in a mortar or MM400 bead mill (Retsch). Genomic DNA was extracted using the NucleoSpin Plant II kit (Macherey-Nagel), the protocol of Edwards et al. (1991) or, for analysis in a 96-well plate format, the Chemagic DNA Plant kit (PerkinElmer) and a PSU-2T Mini-Shaker (BioSan) for the resuspension of magnetic beads. Total RNA was isolated from leaf extracts using the innuPREP Plant RNA kit (Analytik Jena) and residual genomic DNA was digested using the TURBO DNA-free kit (Thermo Fisher Scientific). RNA quantity and quality were determined using a NanoPhotometer UV/Vis spectrophotometer (Implen) and by agarose gel electrophoresis. Complementary DNA (cDNA, final concentration 50 ng/mL) was synthesized from total RNA using Perfect Real Time PrimeScript RT Master Mix (Takara Bio Europe).

The NtCOL2a/b, NtomCOL2 and NsCOL2 gene sequences were isolated by PCR from N. tabacum, N. tomentosiformis and N. sylvestris genomic DNA, respectively. The isolated NtCOL2a/b sequence was amplified in two overlapping fragments and included about ∼2.5kb of the upstream promoter. The coding sequences of the genes were isolated by RT-PCR from N. tabacum, N. tomentosiformis or N. sylvestris RNA. Genomic and coding sequences were amplified using gene-specific primers ( Supplementary Table S2 ). The resulting amplicons were either transferred to pCRII-TOPO using the TOPO TA Cloning kit (Thermo Fisher Scientific) or to pJET1.2/blunt using the CloneJET PCR Cloning kit (Thermo Fisher Scientific) for sequencing. For P _NtCOL2a and P _NtCOL2b sequence regions, amplicons were directly inserted by ligation into pBsGFP_ER (Noll et al., 2007) and pBsGUS (Schmidt et al., 2020) for sequencing, using appropriate restriction enzymes for digestion (for details, see Supplementary Table S5 ). The full-length NtCOL2a/b genomic sequences were assembled in silico using SeqManPro and SeqbuilderPro in Lasergene v15 (DNASTAR).

The genomic structures of NtCOL2a, NtCOL2b, NtomCOL2 and NsCOL2 were determined by aligning the previously isolated genomic sequences with the corresponding coding sequences using SeqManPro and SeqBuilderPro in Lasergene v15. To determine sequence identities genomic, coding or protein sequences were aligned to determine sequence identities using EMBOSS Needle Pairwise Sequence Alignment (Madeira et al., 2019). The ClustalW module within MEGA-11 (Tamura et al., 2021) was used to align the CO(L)/BBX proteins of Nicotiana tabacum (Nt), Oryza sativa (Os), Arabidopsis thaliana (At), Solanum lycopersicum (Sl) and Solanum tuberosum (St) (for corresponding accession numbers, see Supplementary Tables S9 – 11 ), and the phylogenetic tree was constructed using the Neighbor-Joining (NJ) method with 1000 bootstrap-replications and generated with the iTOL online tool (https://itol.embl.de/). Domain analysis was carried out using a Clustal OMEGA multiple sequence alignment (Madeira et al., 2022) including CO homologs from Oryza sativa (Os), Arabidopsis thaliana (At), Solanum lycopersicum (Sl) and Solanum tuberosum (St) (for corresponding accession numbers, see Supplementary Tables S9 – S11 ). InterProScan (Jones et al., 2014) was used to detect conserved protein domains. The identified domains were annotated manually according to Robson et al. (2001).

Transgenic lines of N. tabacum cv. SR1 and N. sylvestris Speg. & Comes were generated via the leaf disc method (Horsch et al., 1985) using the A. tumefaciens strain LBA4404 (Hoekema et al., 1983), in which the appropriate binary vectors were introduced by electroporation. Transgenic plants were selected on MS medium supplemented with 25 mg/L hygromycin, 100 mg/L kanamycin or 3 mg/L phosphinothricin, as appropriate. Independent transgenic plant lines were regenerated from callus tissue in sterile culture media and were tested for genomic transgene integration. To increase the probability of induced mutations, callus passage of the NtCOL2 knockout lines was repeated by placing leaves of transgenic T₀ plants on appropriate sterile MS medium and shoots were regenerated from callus tissue. After rooting, transgenic plants were transferred to the greenhouse and cultivated in soil under LD conditions as described above.

For localization studies, Venus-NtCOL2a and Venus-NtCOL2b fusion proteins were expressed in the leaves of 3–4-week-old N. benthamiana plants cultivated under LD conditions. For this purpose, A. tumefaciens strain GV3101 pMP90 (Koncz and Schell, 1986) was transformed with the appropriate binary vectors. Transient expression was then achieved by the co-infiltration of leaves with strains GV3101 pMP90 and C58C1, carrying the pCH32 helper plasmid and a pBin61 derivative expressing tomato bushy stunt virus RNA silencing suppressor p19 (Hamilton et al., 1996; Garabagi et al., 2012). After infiltration, plants were cultivated for 3 days with continuous illumination before proteins were localized by confocal laser scanning microscopy.

Regenerated T₀ transgenic plants representing independent transformation events were identified by PCR using MangoTaq DNA polymerase (Bioline) with the primers listed in Supplementary Table S2 and genomic DNA as the template. T₀ and T₁ generation knockout plants were screened to determine whether a genomic cas9 gene was present, using N. tabacum GLYCERALDEHYDE-3-PHOSPHATE DEHYDROGENASE (NtGAPDH) as a template control. Genome editing of NtCOL2a and NtCOL2b was analyzed by PCR using MyTaq DNA polymerase (Bioline) and the primers listed in Supplementary Table S2 for the amplification of exon I from genomic DNA. Starting with the T₀ generation, plants were screened by direct sequencing of purified amplicons. Because the transgenic T₀ plants were chimeras, selected individuals were analyzed in more detail by sequencing the amplicons following transfer to pCRII-TOPO (TOPO TA Cloning kit, Thermo Fisher Scientific).

Gene expression in wild-type and transgenic plants was analyzed by qPCR using the CFX 96 Real-Time System in a C1000 Touch Thermal Cycler (Bio-Rad Laboratories) combined with KAPA SYBR FAST qPCR Master Mix (Merck) and gene-specific primers ( Supplementary Table S2 ). The reactions contained 500 nM of each primer and 2.5 µL template cDNA (diluted 1:10, equivalent to ∼12.5 ng). After denaturation (95°C, 3 min), the qPCR program comprised 40 cycles of denaturation (95°C, 3 s) and annealing/extension for 30 s at primer-specific temperatures ( Supplementary Table S12 ). Melt curve analysis (5 s, 58–95°C, ΔT = 0.5°C) was carried out to ensure amplicon specificity. Each sample was tested in technical triplicates for each gene, along with duplicate no-reverse-transcriptase (NRT) and no-template controls (NTC). Data were analyzed using CFX Manager v3.1 (Bio-Rad Laboratories). Quantification cycle (Cq) values of technical triplicates were averaged and used to determine the mean of each biological replicate. The target gene expression ratio was calculated as previously described (Livak and Schmittgen, 2001). The reference gene ELONGATION FACTOR-1α (EF-1α) was used for normalization (Schmidt and Delaney, 2010).

NtCOL2a-3xc-myc and NtCOL2b-3xc-myc fusion proteins were detected by cultivating transgenic plants expressing P_Q35S:NtCOL2a-3xc-myc or P_Q35S:NtCOL2b-3xc-myc under LD conditions and grinding harvested leaf tissue in liquid nitrogen using a mortar. Proteins were extracted from 50 mg ground tissue per sample in 50 µL 5× SDS-PAGE buffer (60 mM Tris/HCl, 50% (v/v) glycerol, 10% (w/v) SDS, 500 mM DTT, 0.1% (w/v) bromphenol blue, pH ∼6.8) by vortexing (2 min) and boiling at 95°C (10 min). Mixtures were centrifuged to remove cell debris (10,000 × g, 2 min, room temperature) and the supernatant was fractionated by SDS-PAGE (Laemmli, 1970) on 10% (v/v) SDS polyacrylamide gels. The proteins were then transferred to nitrocellulose membranes (Towbin et al., 1979) and visualized by incubating for ∼1 min in 0.1% (w/v) Ponceau S, 5% (v/v) acetic acid, as a loading control (Romero-Calvo et al., 2010). After documentation, the stain was removed by soaking in phosphate-buffered saline (PBS; 140 mM NaCl, 2.7 mM KCl, 1.8 mM KH₂PO₄, 10 mM Na₂HPO₄, pH ∼7.2) containing 0.1% (v/v) Tween-20 (PBST). PBST containing 5% (w/v) skimmed milk powder was used for antibody dilution to prevent nonspecific binding. The 3xc-myc-tagged versions of NtCOL2a and NtCOL2b were detected using a mouse monoclonal anti-c-myc antibody (diluted 1:5000, Sigma-Aldrich #M4439). After further washing in PBST, the bound primary antibody was detected using a secondary goat anti-mouse IgG antibody coupled to horseradish peroxidase (Thermo Fisher Scientific #32430). After a final wash, the signal was revealed using the SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) and G:BOX Chemi XX6 gel documentation system running GeneSys v1.5.2.0 (Syngene).

Transgenic plants expressing P _NtCOL2a :uidA or P _NtCOL2b :uidA were cultivated under LD conditions. Stem and leaf petiole sections and small leaf discs were infiltrated in a vacuum with GUS staining solution and incubated for up to 24 h at 37°C as previously described (Schmidt et al., 2020). Chlorophyll was extracted by incubating the samples in methanol (37°C for up to 3 h). Samples were stored in deionized water at 4°C before imaging with a MZ 16 F stereomicroscope (Leica Microsystems).

Promoter activity in transgenic P _NtCOL2a :GFP_ER and P _NtCOL2b :GFP_ER plants was analyzed by CLSM using a Leica TCS SP5 X microscope (Leica Microsystems). Longitudinal sections of stem and leaf petioles were prepared from plants cultivated under LD conditions. Callose was stained with 0.1% (w/v) aniline blue in a 1:1 (v/v) mix of glycerol/deionized water for ~5 min to visualize phloem sieve tube plates. Sections were washed in the same solution without dye before microscopy. Fluorescence was measured at excitation and emission wavelengths of 488 and 500–600 nm (GFP_ER), or 405 and 479–533 nm (aniline blue), respectively. Small discs punched from infiltrated N. benthamiana leaves were analyzed by CLSM to determine the subcellular localization of Venus-NtCOL2a and Venus-NtCOL2b fusion proteins after transient expression. Venus fluorescence signals were measured in abaxial epidermal cells at excitation and emission wavelengths of 514 and 525–600 nm, respectively.

The accession numbers of gene and protein sequences used in this study are listed in Supplementary Tables S9 – S11 .

Initially, we searched for potential CO/BBX-related proteins in tobacco by using the 17 Arabidopsis CO/BBX proteins from subclades I–III ( Supplementary Table S9 ) as BLAST queries against tobacco protein sequences in the National Center for Biotechnology Information (NCBI) non-redundant protein sequences (nr) database. This revealed numerous, mainly predicted tobacco sequences (for accession numbers see Supplementary Table S9 ) clustering with the different AtCO/AtBBX proteins ( Figure 1A ). We also mapped these tobacco proteins against the recently identified NtBBX/NtCOL sequences (Song et al., 2022; Zhao et al., 2022) and found that we could expand the list of NtBBX proteins (NtBBX44-53; Supplementary Table S10 ). Based on the phylogenetic tree containing all AtCO/AtBBX-family members, NtBBX44-53 were assigned to subgroup II ( Figure 1A ; Supplementary Figure S1 ). In terms of flowering control, AtCO is the first BBX protein characterized and NtBBX1 and NtBBX2 are the most closely related to AtCO ( Figure 1A ). Therefore, we focused on the characterization of these two proteins and refer to them as NtCOL2a (NP_001311813) and NtCOL2b (XP_016462705). Using NtCOL2a and NtCOL2b as BLAST queries, we specifically searched in Nicotiana tomentosiformis and Nicotiana sylvestris protein sequences in the NCBI nr database for ancestral orthologs and identified two closely related proteins, NtomCOL2 (XP_009630583) and NsCOL2 (XP_009765376). Pair-wise alignments of the protein sequences revealed that NtCOL2a evolved from the predicted NtomCOL2 sequence and NtCOL2b evolved from the predicted NsCOL2 sequence. NtCOL2a and NtCOL2b share 99.8% and 100% identity with the corresponding ancestral proteins, respectively, and are 94.6% identical to each other. Based on this, we identified and verified the genomic and coding sequences of NtCOL2a, NtCOL2b, NsCOL2 and NtomCOL2 and isolated the corresponding genomic DNA and cDNA sequences from cultivar SR1 and the progenitor species, respectively. Overall, genomic NtCOL2a and NtCOL2b shared 99.5% (NtCOL2a to NtomCOL2) and 99.8% (NtCOL2b to NsCOL2) identity with their ancestral genes, confirming the progenitor genomes are highly conserved in N. tabacum.

An alignment of genomic sequences with the corresponding isolated coding sequences revealed that each gene features three exons (I–III, Supplementary Figure S3 ). Further insight in the regulation of tobacco COL proteins was achieved by comparison of NtCOL2a/b amino acid sequences with CO(L)/BBX proteins from Arabidopsis, rice, potato and tomato (Yano et al., 2000; Robson et al., 2001; Ben-Naim et al., 2006; González-Schain et al., 2012; Abelenda et al., 2016; Zhao et al., 2022). A phylogenetic tree ( Figure 1A ) emphasized the evolutionary origin of NtCOL2a and NtCOL2b and the close relationship between the tobacco proteins and homologs from potato and tomato (solanaceous species) and Arabidopsis. NtCOL2a and NtCOL2b featured the domain structure typical for CO proteins, which includes two N-terminal B-box zinc finger domains and a C-terminal CCT domain ( Figures 1B–D ; Supplementary Figure S2 ), suggesting that the proteins are functional floral regulators (Putterill et al., 1995; Yano et al., 2000; Robson et al., 2001; Crocco and Botto, 2013). The screening of several protein databases verified the presence and location of the two B-box-type zinc finger domains (InterPro accession number IPR000315) and the CCT motif (IPR010402) in each of the four tobacco proteins ( Figure 1B ). The two B-boxes (designated B-box1 and B-box2) are directly adjacent to each other at the N-terminus, and the CCT motif is located in the characteristic C-terminal position (Robson et al., 2001; Griffiths et al., 2003; Crocco and Botto, 2013). The predicted B-boxes and CCT domain were near identical in all four tobacco proteins, with only one amino acid differing between NtCOL2a and NtCOL2b, and between NtomCOL2 and NsCOL2, at the seventh position in B-box 1 ( Figures 1C, D ). Moreover, the tobacco domains were highly similar to those in the reference proteins, apart from distantly-related OsHd1 ( Figures 1C, D ). Both B-boxes featured a CO-typical consensus structure consisting of five cysteine and two histidine residues separated by a defined number of amino acids (Robson et al., 2001; Griffiths et al., 2003; Crocco and Botto, 2013). The CCT motif included a nuclear localization signal (NLS) also found in other, diverse CO homologs (Crocco and Botto, 2013). Indeed, N-terminal fusions of NtCOL2a and NtCOL2b with Venus (Nagai et al., 2002) expressed in N. benthamiana leaf epidermal cells were detected in the nucleus ( Figures 1E, F ; Supplementary Figure S4 ). The unfused Venus protein (control) was localized in the cytoplasm and nucleus ( Figure 1G ; Supplementary Figure S4 ).

Next, we characterized the spatial expression of NtCOL2a/b by qPCR and promoter–reporter analysis. NtCOL2a and NtCOL2b expression was monitored by qPCR at two developmental stages under LD and SD conditions, focusing on the apical, medial, and basal leaves, as well as the stem. The tissues were harvested at dawn from vegetative and reproductive plants, the latter with visible floral buds. NtCOL2a and NtCOL2b were expressed at similar levels under LD ( Figure 2A ) and SD ( Figure 2B ) conditions, and during vegetative and reproductive growth, but levels were highest in the mature medial and basal leaves and lowest in the stem. Previous studies have shown that CO regulates the expression of FT in the phloem companion cells of leaves (Takada and Goto, 2003; An et al., 2004; Wigge et al., 2005; Chen et al., 2018). We therefore used promoter–reporter analysis to define the cell-specific spatial expression profile by fusing the ~2.5 kb P _NtCOL2a or P _NtCOL2b promoter sequences to either uidA encoding GUS or to GFP_ER . Stable transformation of N. tabacum cv. SR1 plants resulted in several independent T₀ transformants per construct, which were cultivated and analyzed under LD conditions. Promoter activity was investigated in the medial leaves due to their high expression levels. GUS activity was analyzed in the petiole ( Figures 2C, D, F, G ) and lamina ( Figures 2E, H ), revealing P _NtCOL2a and P _NtCOL2b acted predominantly in the vascular bundles, specifically in the phloem, as reported for the AtCO promoter (An et al., 2004). This expression pattern was observed in at least three transgenic T₀ plants per construct, but P _NtCOL2a :uidA L8 ( Figures 2C–E ) and P _NtCOL2b :uidA L14 ( Figures 2F–H ) are shown as representative examples. Additional but less intense staining of parenchymal tissue regions was detected in petiole cross-sections (e.g., Figures 2C, F ), indicating that NtCOL2a and NtCOL2b expression may not be restricted to the vascular bundles. Promoter activity in the P _NtCOL2a :GFP_ER and P _NtCOL2b :GFP_ER lines was determined by CLSM using longitudinal petiole sections, confirming gene expression in the phloem. In at least five T₀ plants per construct, GFP_ER was detected in phloem companion cells adjacent to the sieve elements, which were visualized by callose-specific aniline blue staining of the sieve plates, but representative sections are shown for P _NtCOL2a :GFP_ER L16 ( Figure 2I ) and P _NtCOL2b :GFP_ER L14 ( Figure 2J ). These experiments confirmed NtCOL2a and NtCOL2b expression at the site of FT transcription, as also observed for NtFT3 in N. tabacum (Harig et al., 2012).

CO expression is precisely controlled by the internal circadian clock and photoperiod. CO transcript abundance in diverse plant species thus follows a diurnal rhythm, and the CO protein is stabilized at the post-translational level only when high expression levels coincide with the light period (Song et al., 2014). To examine the temporal expression profile of NtCOL2a and NtCOL2b, qPCR was carried out at 4-h intervals during one day under LD and SD conditions. Again, we selected medial leaves of vegetative and reproductive individuals due to the high general expression in this tissue. NtCOL2a and NtCOL2b expression followed similar oscillating patterns ( Figures 3A–D ), with a different daily course under LD ( Figures 3A , B ) and SD ( Figures 3C , D ) conditions. However, NtCOL2a ( Figures 3A , C ) and NtCOL2b ( Figures 3B , D ) expression were comparable in vegetative and reproductive plants grown under the same conditions. The general profile comprised a peak of expression around dawn followed by a steep drop during the light period and a renewed increase in the dark. Under LD conditions ( Figures 3A, B ), NtCOL2a and NtCOL2b expression peaked sharply at the beginning of the light period. Despite a slight decrease, relatively high expression levels were still detected 4 h after dawn, indicating the coincidence of high transcript abundance with the morning light. After reaching the expression minimum 8 h after dawn, a transient and less intense second peak was detected at the end of the light period (12–16 h after dawn), which was more obvious during reproductive growth. Nevertheless, the expression of both genes remained relatively low until the middle of the dark period (20 h after dawn), suggesting that transcript levels increased only late in the night. In contrast, the highest levels of NtCOL2a and NtCOL2b mRNA under SD conditions were detected between 20 h and dawn, indicating that the mRNA accumulated mostly in the dark ( Figures 3C, D ). This peak shift towards the late night went along with an earlier and more intense depletion of transcript levels in the morning light, which was already detected 4 h after dawn. The lowest expression levels were observed 8 h after dawn, corresponding with the end of the light period. The transcript abundance then increased continuously during the dark, beginning to peak late at night (20 h after dawn).

The day-neutral flowering behavior of N. tabacum emerged as a result of the tetraploidization of N. tomentosiformis (facultative SD flowering) and N. sylvestris, which strictly flowers under LD conditions (Aoki and Ito, 2000; Murad et al., 2002). To determine whether the diurnal expression patterns of NtCOL2a and NtCOL2b differ from those of their progenitor genes, we measured the abundance of NtomCOL2 ( Figures 3E, G ) and NsCOL2 ( Figures 3F, H ) mRNA by qPCR. Given the flowering behavior of N. tomentosiformis and N. sylvestris, NtomCOL2 transcript levels under LD conditions ( Figure 3E ) and NsCOL2 transcript levels under SD conditions ( Figure 3H ) were only tested in vegetative plants. Under both conditions, NtomCOL2 and NsCOL2 expression profiles generally resembled those of the N. tabacum genes, showing nearly the same peaks and troughs during the course of the day. However, one minor variation we occasionally observed was a rapid depletion of NsCOL2 mRNA 20 h after dawn in the reproductive N. sylvestris plants under LD conditions. In summary, the diurnal profile of the Nicotiana COL genes, with expression levels peaking around dawn, is similar to that reported for the potato CONSTANS-LIKE 1 (StCOL1) gene (Abelenda et al., 2016).

To determine the effects of NtCOL2a/b overexpression, we generated transgenic P_35S:NtCOL2a and P_35S:NtCOL2b N. tabacum lines, cultivated T₁ individuals representing six independent lines each under LD and SD conditions, and compared them with empty vector control plants (VC_{pBin19 Hyg} L1) carrying pBin19 Hyg T-DNA (Bevan, 1984, modified by Dr. Lena Grundmann, Münster, Germany), hereafter abbreviated to VC ( Figure 4 ). All overexpression lines accumulated higher levels of the corresponding transcript than the VC ( Figure 4A ). NtCOL2a mRNA levels were at least ~1252-fold higher (L14) and up to ~4752-fold (L7) higher in the P_35S:NtCOL2a lines, and NtCOL2b mRNA levels were generally at least ~543-fold higher (L14) and up to ~1122-fold higher (L17) in the P_35S:NtCOL2b lines, except L22 (38-fold). The accumulation of NtCOL2b mRNA in the P_35S:NtCOL2a lines and vice versa was comparable to the VC plants, indicating no cross-regulation between NtCOL2a/b. The overexpression of neither NtCOL2a nor NtCOL2b obviously affected the flowering behavior of the T₁ plants (representative examples under LD conditions shown in Figures 4B–D ). Under LD conditions, all NtCOL2a and NtCOL2b overexpression lines flowered at the same average time as the VC plants ( Figure 4E ), although most of the lines tended to produce ~1–2 fewer leaves on the main shoot before flowering ( Figure 4F ), which is statistically significant (suggesting the higher NtCOL2a or NtCOL2b transcript levels have a slight negative impact on plant development) but not biologically relevant. The P_35S:NtCOL2a and P_35S:NtCOL2b plants cultivated under SD conditions tended to flower up to 5 days earlier than the VC ( Figure 4G ) with the exception of P_35S:NtCOL2b L10 and L22. This may indicate a slight acceleration of floral transition when high NtCOL2a/b transcript levels are abundant under SD conditions. However, the slight variations between the lines did not correlate with the transcript levels of the overexpressed gene. Moreover, the slightly shorter vegetative growth phase was not reflected in the number of leaves on the main shoot, which was on average comparable to the VC ( Figure 4H ). In conclusion, overexpression of NtCOL2a and NtCOL2b had no obvious effect on flowering behavior. In line with these observations, AtCO overexpression in tobacco also has no impact on flowering time (Ben-Naim et al., 2006).

Next, we checked the abundance of NtCOL2 proteins during the day. In Arabidopsis, light mediates the post-translational stabilization of CO, causing diurnal variations in protein abundance (Valverde et al., 2004). To determine whether NtCOL2a and/or NtCOL2b are stabilized in N. tabacum, we measured the protein levels by expressing NtCOL2a-3xc-myc and NtCOL2b-3xc-myc under the control of the quadruple constitutive CaMV 35S promoter (P_Q35S). Given that NtCO-specific antibodies are not available for immunodetection, we added a C-terminal (3xc-myc) epitope tag (Evan et al., 1985). We recovered several independent N. tabacum T₀ transformants per construct expressing the transgene cassette. Two of the corresponding T₁ lines (P_Q35S:NtCOL2a-3xc-myc L5 and P_Q35S:NtCOL2b-3xc-myc L11) were chosen for the analysis of protein abundance under LD conditions compared to the pBin19 (Bevan, 1984) VC line (L1). Total protein extracts from young medial leaves were analyzed by SDS-PAGE and western blotting at six different time points during the day ( Figures 5A, B ), each sample representing a plant pool (P1–P6) that consisted of individuals from the same line. NtCOL2a-3xc-myc ( Figure 5A ) and NtCOL2b-3xc-myc ( Figure 5B ) levels varied during the day and the band size (50–75 kDa) appeared slightly larger than calculated in silico (~49.5 kDa). Nevertheless, the absence of any corresponding signal in the VC samples confirmed the specificity of these bands. NtCOL2a-3xc-myc was present at all time points, except 20 h after dawn in the middle of the night. Remarkably, the protein strongly accumulated in the morning light (1 h after dawn) and remained at relatively low levels for the rest of the day, despite the constitutive expression of the transgene. In contrast, NtCOL2b-3xc-myc was only present at low levels, and was detected in the early morning (1 h after dawn) and the evening (15 h after dawn), suggesting differences in the post-translational stability of these proteins. The presence of both proteins in the light indicated that the post-translational stabilization of NtCOL2a and NtCOL2b in N. tabacum might be dependent on light and/or the circadian clock. This was supported by the accumulation of NtCOL2a-3xc-myc 1 h after dawn, the time when NtCOL2 transcript levels peak under LD conditions ( Figures 3A, B ). At this time, both fusion proteins were also detected when the plants formed visible floral buds ( Figure 5C ), indicating that protein stabilization also occurs during reproductive growth. However, more proteins tended to accumulate in the vegetative plants ( Figure 5C ).

To determine whether the varying protein levels reflected post-translational modifications or post-transcriptional regulation of the P_Q35S:NtCOL2-3xc-myc transcripts, we analyzed NtCOL2a and NtCOL2b expression by qPCR in the leaves of the vegetative plant pools 4 h after dawn. Compared to the VC, all pools of the overexpression lines accumulated higher levels of the overexpressed transcript ( Figure 5D ). NtCOL2a levels varied between ~15.0-fold (P6) and ~80.2-fold (P1) higher in the P_Q35S:NtCOL2a-3xc-myc (L5) plants, whereas NtCOL2b levels were elevated by ~1.7-fold (P4) to ~6.3-fold (P3) in the P_Q35S:NtCOL2b-3xc-myc (L11) plants. These slight variations in expression level did not correlate with the observed diurnal oscillations of the NtCOL2a-3xc-myc or NtCOL2b-3xc-myc proteins. For example, the highest level of NtCOL2 mRNA was detected in P1 of P_Q35S:NtCOL2a-3xc-myc (L5), where the protein abundance was low. Nevertheless, the low NtCOL2b mRNA levels ( Figure 5D ) in all T₁ plants of P_Q35S:NtCOL2b-3xc-myc (L11) might explain the general low abundance of the NtCOL2b-3x-myc protein ( Figures 5B, C ).

Given that the lack of an obvious phenotype caused by the constitutive overexpression of NtCOL2a or NtCOL2b, we generated single and double knockout mutants in N. tabacum using the CRISPR/Cas9 system (Fauser et al., 2014). Gene specific protospacers were designed in silico using CCTop (Stemmer et al., 2015) and used as part of the sgRNA to generate the corresponding binary vector constructs for stable plant transformation. Proximal frameshift mutations were induced by designing protospacers (NtCOL2a _PS1 and NtCOL2b _PS1) within the antisense strand of exon I ( Figure 6A ). For the single knockouts, we used one protospacer per gene, and off-target effects were avoided by ensuring that the corresponding sites in the other NtCOL2 gene featured at least three mismatches in the protospacer region or lacked a protospacer adjacent motif (PAM) of the “NGG” or “NRG” type recognized by SpCas9 endonuclease (Hsu et al., 2013). For the double knockout (NtCOL2a/b _PS1), we targeted a region that was identical in both genes. Other potential off-target sites were identified by computational screening of the N. tabacum genome (cv. Basma Xanthi) as a reference (Sierro et al., 2014). This revealed that all three protospacers showed at least four mismatches when compared to any other putative exonic off-target region.

Several independent T₀ transformants were generated and screened for induced mutations. T₁ lines from three selected T₀ transformants (NtCOL2a _PS1 L1, NtCOL2b _PS1 L2 and NtCOL2a/b _PS1 L3) carrying mutated allelic variants (hereafter Ntcol2a^– and Ntcol2b^– ) were cultivated under LD conditions, and three cas9-free plants per line were randomly chosen for further analysis. We screened for induced mutations by amplicon sequencing of NtCOL2a and NtCOL2b genomic DNA. All T₁ individuals solely carried mutated alleles of the target gene(s), which were identical in all plants of the same line ( Figure 6B ) and had already been found in the T₀ generation. In the single knockout Ntcol2a^– and Ntcol2b^– plants, the corresponding off-target region in the other NtCOL2 gene was not mutated, confirming the specificity of the protospacers. The Ntcol2a^– T₁ plants were homozygous NtCOL2a knockouts with deletions of the dinucleotide AC ( Figure 6B ). Similarly, the Ntcol2b^– T₁ plants were homozygous NtCOL2b knockouts with deletions of the dinucleotide TG ( Figure 6B ). In the double knockout Ntcol2a^–/b^– , the NtCOL2a and NtCOL2b genes carried the same deletion of a single cytidine residue ( Figure 6B ). All mutations resulted in highly truncated NtCOL2a and/or NtCOL2b proteins ( Figure 6C ). The –2 bp deletions in the nullizygous Ntcol2a^– and Ntcol2b^– single knockout T₁ plants generated proteins of 37 amino acids, containing only the N-terminus with the first 23 amino acids of B-box 1. Due to the position of the protospacer, the nullizygous Ntcol2a^–/b^– double knockout T₁ plants generated proteins of 120 amino acids, which were nevertheless highly likely nonfunctional because they lack the CCT domain including the NLS and only contain the N-terminal B-box domains.

To determine the phenotype of the NtCOL2 knockouts, we examined the flowering behavior of the mutants in the T₂ generation. The offspring of one self-fertilized T₁ plant of Ntcol2a^– L1, Ntcol2b^– L2 and Ntcol2a^–/b^– L3 were cultivated under LD and SD conditions and compared to wild-type controls. Representative genotyping of at least six individuals per line confirmed the nullizygous genotypes of the T₂ offspring. Under SD conditions, the loss of NtCOL2 appeared to have no significant impact on flowering behavior. The single and double knockout individuals flowered on average at the same time as wild-type controls ( Figures 6D, G ) and there was no difference in the number of leaves ( Figures 6E, H ). Under LD conditions, the single Ntcol2a^– knockout plants flowered slightly later than controls (~2 days) but this was not observed for the nullizygous Ntcol2b^– plants ( Figure 6D , representative plants in Figure 6F ). However, both Ntcol2a^– and Ntcol2b^– plants tended to produce ~2 fewer leaves on the main shoot compared to controls ( Figure 6E ), indicating a mild effect on development that was statistically significant but not biologically relevant. Interestingly, this trend was not confirmed in the double knockouts, which produced a wild-type phenotype (representative plants in Figure 6I ) with a similar flowering time ( Figure 6G ) and a comparable number of leaves ( Figure 6H ). We also measured by qPCR the expression level of four tobacco FT genes in medial leaves (harvested 4 h after dawn) in nullizygous Ntcol2a^– , Ntcol2b^– and Ntcol2a^–/b^– plants grown under LD and SD conditions. However, we found no differences in expression levels compared to wild-type controls ( Supplementary Figure S5 ). These experiments suggested that the loss of NtCOL2 activity has only a marginal effect on flowering behavior, and only under LD conditions.

COL proteins have little or no influence on floral transition in some species, including homologs in day-neutral flowering tomato and potato varieties (Ben-Naim et al., 2006; González-Schain et al., 2012). As shown for NtCOL2a and NtCOL2b, the overexpression of the tomato COL genes SlCOL1 and SlCOL3 had no obvious impact on the flowering of transgenic tomato plants, and they are unlikely to be key floral regulators (Ben-Naim et al., 2006). In potato andigenum genotypes, the CO homologs StCO and StCOL1 also have only a weak influence on flowering (González-Schain et al., 2012; Abelenda et al., 2016). The close phylogenetic relationship between NtCOL2a, NtCOL2b and these tomato and potato proteins ( Figure 1A ) strengthens the hypothesis that the two tobacco homologs have little or no activity as floral regulators. Furthermore, phylogenetic analysis of CO in Arabidopsis and related Brassicaceae species revealed that CO and its homologs evolved by gene duplication from one common ancestral gene. However, the function of CO as a key regulator of photoperiodic flowering seems to have emerged after this duplication event. The regulation of photoperiod-dependent processes by CO homologs in other plant families may reflect the convergent evolution of gene function (Simon et al., 2015), which appears not to be the case for the NtCOL2 genes.

The published allotetraploid N. tabacum genome contains five FT-like genes (NtFT1–NtFT5), which can be associated with their ancestral genes in the two diploid progenitor species N. sylvestris (strict LD plant) and N. tomentosiformis (facultative SD plant) (Aoki and Ito, 2000; Murad et al., 2002; Harig et al., 2012; Sierro et al., 2014; Beinecke et al., 2018). Nicotiana FTs act antagonistically to regulate flowering, and for simplicity hereafter we identify the floral repressors (rep) and activators (act) using superscript notation. NtFT1^rep, NtFT2^rep, NtFT3^rep and NtFT4^act are primarily SD-specific floral regulators, whereas NtFT5^act induces flowering under SD and agriculturally-relevant LD conditions. Interestingly, NtFT5^act originates from N. tomentosiformis (NtomFTγ^act). Therefore, LD flowering in N. tabacum seems to be based largely on the facultative SD N. tomentosiformis flowering pathway rather than the strict LD-dependent flowering pathway in N. sylvestris. Furthermore, under SD conditions, N. tomentosiformis flowering is strongly promoted by NtomFTb^act (homologous to NtFT4^act). Under inductive LD conditions, N. sylvestris expresses NsFTc^act (homologous to NtFT6^act, carrying a premature stop codon in SR1) and NsFTd^act (homologous to NtFT7^act, not present in SR1), thereby promoting flowering, whereas flowering under SD conditions might be suppressed by the strong expression of NsFTa^rep (homologous to NtFT2^rep) (Beinecke et al., 2018). In this study, the overexpression of NtCOL2a and NtCOL2b did not substantially affect the flowering time in N. tabacum and, at least under LD conditions, 3xc-myc-tagged NtCOL2a and NtCOL2b proteins were present in transgenic plants and followed a diurnal pattern (i.e., largely following the diurnal expression profile of the endogenous NtCOL2a and NtCOL2b genes despite constitutive expression under the control of the 35S promoter). NtCOL2a and/or NtCOL2b knockout only marginally affected the flowering time under LD conditions in N. tabacum. These results suggest, that, despite their observed abundance, the NtCOL2 proteins appear to have little or no function in terms of day-neutral floral transition and do not act as floral key regulators in N. tabacum.

Although we introduced the P_35S:NtCOL2a construct into N. tomentosiformis, we were unable to produce transgenic lines. However, we introduced the P_35S:NtCOL2b construct (encoding NtCOL2b 100% identical to NsCOL2) into the strict LD plant N. sylvestris and recovered 12 independent transgenic T₀ lines. In the T₁ generation, seeds of all lines germinated under SD conditions on selective medium (with N. sylvestris wild-type seeds as controls on non-selective medium) and four seedlings per line were transferred to a phytochamber for phenotyping under SD conditions 6 weeks after germination. Seedling material was also harvested for qPCR analysis. N. sylvestris wild-type plants and all plants of five transgenic lines (L5, L6, L7, L9 and L13) grew solely in a vegetative manner and remained at the rosette growth stage, reflecting normal growth under these typically non-inductive SD conditions. Interestingly, the remaining seven lines (L2, L3, one of four L4 plants, L6, L8, L10, L11 and L12) started bolting and flowering 10–13 weeks after transfer to the phytochamber, resulting in only 24–27 leaves ( Figures 7A, B ).

Flowering in Nicotiana species strongly depends on antagonistically acting FTs, so we determined the levels of NsCOL2, the floral repressors NsFTa^rep and NsFTb^rep , and the floral activators NsFTc^act and NsFTd^act in seedlings and leaves of mature transgenic plants in the T₁ generation ( Figures 7C–E ; Supplementary Figure S5 ). This revealed 12–30-fold higher NsCOL2 expression levels in the leaves of flowering transgenic lines compared to the N. sylvestris wild-type, while non-flowering transgenic lines showed comparable or even lower NsCOL2 expression levels ( Figure 7C ). NsCOL2 overexpression thus correlated with the induction of flowering in the transgenic lines. NsCOL2 levels in seedlings were lower than in mature plants but showed the same trend, with higher levels in lines that started flowering after cultivation under SD conditions. Furthermore, although we could not detect NsFTd^act expression in seedlings from any transgenic or wild-type lines, NsFTd^act expression increased in the leaves of flowering plants, despite being typically very low in wild-type N. sylvestris plants under these condition ( Figure 7D ; Beinecke et al., 2018). Non-flowering transgenic plants from the other lines accumulated even less NsFTd^act mRNA than wild-type N. sylvestris plants. We also observed lower expression (near qPCR detection limit) of the floral repressor NsFTa^rep in seedlings from lines that started flowering, ranging from 3-fold (L10) and 11-fold (L3) to more than 260-fold (L12) lower than wild-type levels ( Figure 7E ). In the seedlings of non-flowering lines, NsFTa^rep remained at wild-type expression levels as expected (with the exception of non-flowering L9, where expression was ~50-fold lower than wild-type plants). In the leaves of mature plants, NsFTa^rep expression levels increased for all plants compared to the seedling stage, with no correlation between NsFTa^rep expression levels and a flowering phenotype (except L8, where expression is more than 10-fold lower than in other lines or WT; Figure 7E ). We also checked the expression of NsFTb^rep and NsFTc^act ( Supplementary Figures S6A, B ) but found no correlation with the flowering or non-flowering phenotype.

In wild-type N. sylvestris plants, the floral repressor NsFTa^rep is strongly expressed during vegetative growth under SD conditions, probably repressing flowering, although NsFTc^act is also expressed (Beinecke et al., 2018). The overexpression of NsCOL2 in transgenic N. sylvestris plants under SD conditions potentially suppressed NsFTa^rep expression in transgenic seedlings, enabling flowering after a certain period of vegetative growth. In contrast, NsFTa^rep expression levels in the leaves of mature transgenic lines were comparable to wild-type levels regardless of the flowering/non-flowering phenotype ( Figure 7E ). However, although NsFTd^act expression was near the qPCR detection level in wild-type N. sylvestris and non-flowering transgenic lines (Beinecke et al., 2018; Figure 7D ), it increased in flowering NsCOL2 overexpression lines, suggesting that NsFTd^act expression is activated by NsCOL2 overexpression, which overcomes the relatively high expression of NsFTa^rep detected in all lines. In flowering transgenic lines, the floral activator/repressor ratio is higher than in non-flowering lines or wild-type N. sylvestris plants and is thus shifted toward floral induction. We have already proposed the importance of the activator/repressor ratio in the transition to flowering in N. tabacum, where the floral promoter NtFT4^act is expressed at a lower level than NtFT1^rep and NtFT2^rep under SD conditions, but the fold increase in abundance during development is much higher than the two floral repressors (Harig et al., 2012). Likewise, others have discussed local ratios of FT-like and TFL1-like proteins that control the balance between determinate and indeterminate growth in tomato (Shalit et al., 2009; McGarry and Ayre, 2012).

NsCOL2 may have a dual role, acting as a repressor of NsFTa^rep while inducing the expression of NsFTd^act . A dual role has also been described for OsHd1: although this protein acts as a floral repressor under LD conditions by repressing the expression of Hd3a (FT homolog and floral activator), it induces the expression of the FT-like genes Hd3a and RFT1 under SD conditions (Yano et al., 2000; Izawa et al., 2002; Kojima et al., 2002; Hayama et al., 2003; Ishikawa et al., 2011). NsCOL2 may also influence NsFT expression indirectly or in concert with other factors that are only present at a certain developmental stage, given that the downregulation of NsFTa^rep was only detected in seedlings and the upregulation of NsFTd^act was only observed in flowering plants. The potential indirect effect of NsCOL2 on NsFTd^act expression is supported by the absence of CO-specific regulatory elements (COREs) in the NsFTd^act promoter, 5 kb of which we checked in silico using PLANTPAN3.0 (Chow et al., 2019) for the COREs TGTG(N_2–3) ATG (Tiwari et al., 2010) and TGTGGT (Abelenda et al., 2016). However, our experiments did not reveal any major influence of NtCOL2a and NtCOL2b on the regulation of flowering time in N. tabacum, whereas NsCOL2 was able to induce flowering under otherwise non-inductive SD conditions in N. sylvestris. The assumed inability of NtCOL2 to regulate flowering time in N. tabacum cv. SR1 may reflect the absence of the NsFTd^act homolog NtFT7^act in the SR1 genome (Beinecke et al., 2018). Although the two N. tabacum COL2 homologs are unlikely to be key regulators of the floral transition, the observed abundance of 3xc-myc-tagged NtCOL2a and NtCOL2b (at least during long days) suggests that their abundance is controlled by light and/or the circadian clock and that the proteins are not completely functionless, but may regulate other clock/photoperiod-dependent processes. This hypothesis is supported by the circadian expression profile of the genes. The involvement of StCO and StCOL1 in the control of photoperiod-dependent tuberization in potato demonstrates that such scenarios are possible (Navarro et al., 2011; González-Schain et al., 2012; Abelenda et al., 2016). Furthermore, the B-box (BBX) protein family is known to be involved in diverse developmental processes influenced by light, including shade avoidance, seedling de-etiolation, and photomorphogenesis (Crocco and Botto, 2013). The role of the other NtBBX genes remains elusive, Song et al. (2022) suggested a role in the response to multiple stresses and we assume that some of them may help to control the photoperiod-dependent expression of FTs in day-neutral N. tabacum. Preliminary computational analysis of the remaining Nicotiana FT promoter regions supports this by revealing the presence of putative COREs ( Supplementary Table S13 ; Accession numbers in Supplementary Table S11 ). Recently, the expression of SlCOL, SlCOL4a and SlCOL4b in day-neutral tomato was negatively associated with flowering time (Yang et al., 2020). Thus, future studies should elucidate the role of the remaining uncharacterized tobacco BBX genes (e.g., NtBBX3–NtBBX7) to determine which other processes they might control in N. tabacum.