TRANSPARENT TESTA GLABRA 1 participates in flowering time regulation in Arabidopsis thaliana

Plant material und growth conditions

The used A. thaliana mutants ttg1-9, ttg1-11, ttg1-21, ttg1-22, glabra3 (gl3)-3, enhancer of gl3 (egl3)-19114, tansparent testa 8 (tt8)-SALK, myc1-1, cop1-4 (all Col-0), ttg1-1 (Ler), ttg1-10 (Ws) (Table S7) have been described before (Alonso et al., 2003; Appelhagen et al., 2014; Jakoby et al., 2008; Koornneef, 1981; Larkin et al., 1994; Larkin et al., 1999; McNellis et al., 1994; Pesch et al., 2013; Rosso et al., 2003; Walker et al., 1999; Wester et al., 2009). Primers used for genotyping (including dCAPS primers) are listed in Table S8 (Appelhagen et al., 2011; Jaegle et al., 2016; Neff, Turk & Kalishman, 2002; Schrader et al., 2013). The floral dip method (Clough & Bent, 1998) was used to generate overexpression lines in Col-0 and cop1-4 background. T₁ plants were BASTA selected and resistant plants were screened for YFP fluorescence using a Leica stereomicroscope (MZ FLIII) (Leica Microsystems, http://www.leica-microsystems.com). This analysis was repeated for plants homozygous for the insert being at least in T₃ generation and overexpressing YFP-TTG1 (Table S4). Two walk-in plant chambers were used. Detailed conditions at the respective used areas are listed in Table S1. For flowering time experiments, single seeds were placed in parallelly prepared pots with soil and stratified for 7 d at 5 °C before being transferred to the respective growth condition.

Seedlings for qRT-PCR experiments were sterilized with 70% (v/v) ethanol, 2% NaOCl, stratified at 5 °C and grown on Murashige and Skoog medium containing 1% sucrose at “cold” LD conditions (see Table S1). Seedlings for circadian transcript profiles were snap frozen in liquid nitrogen (∼100 mg) on day eight starting at ZT0 in 4 h intervals until ZT20. Samples used comparing transcript levels in the overexpression lines OE01-03 and OE19-21 were similarly harvested at ZT11 and ZT13 as well as parallelly grown seedlings for comparing protein levels in the same lines at ZT11 and ZT10, respectively. Three biological replicates were analyzed for circadian profiles and comparisons of transcript and protein levels among the TTG1 overexpression lines.

Phenotyping

Flowering time was recorded as the number of post-stratification days until bolting and the total number of leaves at the time point of bolting. Bolting was defined as the time at which the first bud was visible. Two (bHLH overexpression lines’ experiment) or three experiments (TTG1 overexpression lines, all ttg1 mutant sets) were conducted for each set of analyzed genotypes and for each condition with at least six individual plants per genotype and experiment. See Table S2 for details.

Constructs

Gateway™ (Invitrogen™, http://www.invitrogen.com) entry clones containing the coding DNA sequence (CDS) for the respective protein were generated using BP reaction with the previously published vectors TTG1 pAS2.1 and GL3 pcACT2 (Pesch et al., 2015), EGL3 pcACT2, TT8 pAS2.1, MYC1 pAS2.1 (vectors provided by M Pesch) or PCR products using primers for TOC1, PRR5, PRR7, PRR9, bHLH92, LWD1 and LWD2 CDS are listed in Table S8 and cDNA from Col-0 seedlings as a template or using the vectors 35S::PRR7:CFP and 35S::PRR9:CFP (Hayama et al., 2017) as a template, respectively. The used entry vector in all cases was pDONR207 (Invitrogen). All entry vectors were sequenced. LWDs were amplified from the cDNA using GGGGACAAGTTTGTACAAAAAAGCAGGCTTAATGGGAACGAGCAGCGATCC and GGGGACCACTTTGTACAAGAAAGCTGGGTTTCAAACCCTGAGAATT for LWD1 and GGGGACAAGTTTGTACAAAAAAGCAGGCTTAATGGTTACGAGCAGCGATCA and GGGGACCACTTTGTACAAGAAAGCTGGGTTTCAGACCCGGAGAATC for LWD2. See Table S9 for the TTG1, LWD1 and LWD2 CDS which were also used in the alignment in Fig. S2. To generate the construct used for the overexpression lines OE01-03 in Col-0 and OE19-OE21 in cop1-4 background, TTG1 pDONR207 was recombined using Gateway™ LR Clonase™ (Invitrogen™) into pENSG-YFP (N Medina-Escobar, a version for C-terminal fusions was published before Feys et al., 2005). In analogy, the CDS in pDONR207 for GL3 (OE04-OE06), EGL3 (OE07-OE09), TT8 (OE10-OE12) and MYC1 (OE13-OE15) were recombined into pENSG-YFP and used to generate the overexpression lines numbered as given in brackets and expressing YFP-bHLH fusion proteins driven by the Pro35s.

For tobacco co-localization experiments, TTG1 pDONR207 and PRR5 pDONR207 were recombined into pNmR (Schrader et al., 2013) and PRR5, PRR7, PRR9, TOC1, bHLH92 (all in pDONR207) were recombined into pENSG-CFP (N. Medina-Escobar, a version for C-terminal fusions was published before Feys et al., 2005) and for the Y2H experiments, also in pAS2.1-attR and pACT-attR (Clontech, http://www.clontech.com, modified J.F. Uhrig). LWD1 pDONR207 and LWD2 pDONR207 were similarly used in combination with pAS2.1-attR and pACT-attR.

In analogy to YFPattB1-pBat-TL-B-p35s and RFP-HAattB1-pBat-TL-B-p35s previously created as negative controls (Schrader et al., 2013) CFPattB1 was amplified from pENSG-CFP with primers ANS393 and ANS235 and recombined in pBat-TL-B-p35s (Schrader et al., 2013) to obtain CFPattB1-pBat-TL-B-p35s.

qRT-PCR experiments

About 100 mg of seedlings was harvested for each RNA extraction. RNA extractions were done according to the manufacturer’s instructions (RNeasy Plant Mini Kit, Qiagen, http://www.qiagen.com) using a Tissue Lyser (Qiagen) and followed by DNase I (ThermoFisher, https://www.thermofisher.com) treatment. RNA integrity was tested on a gel prior to cDNA synthesis (SuperScript™ III First-Strand Synthesis System, Invitrogen, or the RevertAid First Strand cDNA Synthesis Kit, ThermoFisher) and RNaseH treatment as suggested before (Martel, Grundemann & Schomig, 2002). A PCR using Elongation factor 1-alpha 1 (EF1ALPHA) primers (Kirik et al., 2007) spanning an intron served as a control to ensure that there was no genomic DNA in the cDNA synthesis (Primers in Table S8).

qRT-PCR was performed using the QuantStudio 5 Real-Time PCR Systems (ThermoFisher) with POWER SYBR Green PCR-Master Mix (Applied Biosystems), the respective cDNA and gene-specific primers. UBQ10 (UBIQUITIN10) was used as a reference gene (Harari-Steinberg, Ohad & Chamovitz, 2001; Sun & Callis, 1997). Three biological replicates with three technical replicates each were performed. Calculations are described in detail in Tables S3 and Table S5. All used primers are listed in Table S8. Most of these were described before (Grigorova et al., 2011; Hayama et al., 2017; Li et al., 2008; Maier et al., 2013; Nakamichi et al., 2007; Shin et al., 2017; Wang et al., 2014; Wang et al., 2011; Wenden et al., 2011; Yu et al., 2012; Zhang et al., 2015; Zou et al., 2013). For TTG1 endo, TTG1 both and TTG1 no LWD see Fig. 1G and Fig. S2.

Comparison of protein levels

Samples were homogenized under liquid nitrogen to compare YFP-tagged TTG1 in the overexpression lines OE01-03 and OE19-21. 150 µl of lysis buffer (Kirik et al., 2007) were added to the powder and incubated for 30 min at 4 °C (rotating). 100 µl of the supernatant following centrifugation were mixed with 100 µl 2x Laemmli, boiled for 10 min at 95 °C and centrifuged for 1 min at 10 600 g. Samples were separated on SDS-PAGE gels, subsequently blotted and immunodetected (α-GFP (IgG1K, Roche), α-mouse (Jackson ImmunoResearch, http://www.jacksonimmuno.com)). After GFP detection using the SuperSignal® West Femto Maximum Sensitivity Substrate (ThermoFisher) and a LAS-4000 Mini bioimager (GE Healthcare Life Sciences (formerly Fuji), http://www.gelifesciences.com), blots were stripped as suggested by Abcam (http://www.abcam.com/ps/pdf/protocols/stripping%20for%20reprobing.pdf) using mild stripping buffer (1L: 15 g glycine, 1 g SDS, 10 ml Tween20, pH to 2.2) and re-probed with α-histone H3 (ab1791, Abcam, http://www.abcam.com) and α-rabbit (A6154, Sigma-Aldrich, http://www.sigmaaldrich.com).

Y2H experiments and Y2H screening

The TTG1pAS2.1-attR construct (Pesch et al., 2015) was used as bait to screen an A. thaliana root cDNA library in yeast (Klopffleisch et al., 2011). Y2H screening was performed as described before using 5mM of 3-AT (Soellick & Uhrig, 2001). Y2H assays were done by co-transformation of pAS2.1-attR/pACT-attR (or TTG1-pcACT2 Pesch et al., 2015) vector combination as described previously (Gietz & Schiestl, 2007). GFPpAS2.1-attR and GFPpACT (Schrader et al., 2013) served as a negative control. At least three replicates were conducted for each Y2H co-transformation experiment and 6 or 8 individual colonies per transformation were resolved in water in 96-well plates and transferred to SD-LW or SD-LWH plates supplemented with different 3-AT concentrations (3, 5, 10, 15, 20, 30 mM) using a 96-well replica plater. Plates were scanned after one (only SD-LW controlling for successful double transformation and providing a relative comparison of transferred yeast amounts), three and seven days.

Co-localization, microscopy, phenotypic characterization of ttg1-21 and ttg1-22

Nicothiana benthamiana leaves were infiltrated as described before (Yang, Li & Qi, 2000) but using the Agrobacterium tumefaciens strain GV3101 pMP90RK harboring the respective constructs and Agrobacteria expressing the silencing suppressor TBSV19K (Voinnet, Pinto & Baulcombe, 1999). Infiltrated plants were analyzed three days post-infiltration. CFP-attB1-pBat-TL-B-p35s (this study), YFP-attB1-pBat-TL-B-p35s and RFP-HA-attB1-pBat-TL-B-p35s (Schrader et al., 2013) were used as controls for co-expression with single fluorescent tag fusion protein. The experiment was conducted at least three times for each combination. Infiltrations for the co-expression of CFP and RFP-TTG1 or YFP-TTG1 and RFP, respectively, were included in all experiments. CLSM was performed using a Leica SP8 confocal microscope (Leica Microsystems). Z-stacks were acquired by sequential scanning starting with the laser with the higher wavelength. The LAS Application Suite X (Leica Microsystems) was used to extract and merge images for co-localization figures.

Stacks of small leaves acquired as described before (Failmezger et al., 2013) were merged using Combine ZP (by Alan Hadley, https://combinezp.software.informer.com/) for Fig. S1. Pictures of seeds, seedlings and older leaves (14d-old soil and LD grown) were acquired using a stereo microscope Leica stereomicroscope (MZ FLIII) with the MultiFocus and Montage option of the Leica Application Suite V3 (Leica Microsystems) The step-size was 20 µm (seeds, seedlings) and 50µm (leaves). Seedlings were sterilized as described above and grown on MS (4% sucrose) at constant light at 21 °C.

Data analysis and statistics

All statistics (Tables S2, S3 and S5), most data analysis, all box plots and plots for qRT-PCR results were generated using R version 3.4.1 (R Core Team, 2017) with the following packages: dplyr (Wickham et al., 2019), extrafont (Chang, 2014), ggplot2 (Wickham, 2016), plyr (Wickham, 2011), scales (Wickham, 2017), tidyr (Wickham & Henry, 2018). Schematics for Fig. 1G, Figs. S1A and S2 were extracted from CLC DNA Workbench (CLC bio A/S, http://www.clcbio.com).

Relative protein amounts were determined using fiji (imageJ 1.52 h, http://imagej.nih.gov/ij). ROIs of the same size for all bands analyzed within both images of one blot detected with both antibody combinations were measured for their mean grey value intensity. The background close to each band was subtracted, GFP values were set relative to the respective Histone H3 values and values obtained for one blot were normalized to OE01 and, in one case for which OE01 was not evaluated, OE20 was used (for OE19-OE21 analysis). Results are shown in Fig. S3.

TTG1 has an effect on flowering time

To date, TTG1 has been analyzed in detail for early traits while little is known about its role in the regulation of late developmental traits. When growing ttg1 mutants, we observed that these flowered slightly earlier than the wild type. Therefore, we selected flowering time as a key late developmental trait. We analyzed the classical ttg1-9 and ttg1-11 mutants at the same “warm” long-day conditions (chamber set to 22 °C, on average 23.7 °C at the plant’s level, Table S1). We also analyzed flowering time at an about 2 °C reduced temperature which we named “cold” long-day conditions (chamber set to 20 °C, on average 21.4 °C at the plant’s level, Table S1). Both mutants flowered significantly earlier at both conditions (Figs. 1A–1B, Table S2). ttg1-9 exhibited the strongest flowering time phenotype and was only slightly responsive to the difference in temperature in contrast to ttg1-11 and the wild type. Here, the “cold” condition proofed to beneficial in order to reveal differences in flowering time regulation among the ttg1 mutants.

In both used EMS mutants, an amino acid in the TTG1 protein is changed. These mutants are known to be no null-mutants at least with respect to their effect on trichome patterning. Therefore, we obtained the recently described additional mutants ttg1-21 and ttg1-22 in Col-0 background to extend the flowering time analysis (Appelhagen et al., 2014; Rosso et al., 2003). In both mutants, a T-DNA insertion causes a premature stop codon within the inserted T-DNA. The insertion in ttg1-21 is close to the start and in front of the WD40 domain (Fig. S1A). Therefore, it can be expected that this mutant is a null mutant or at least a comparably strong mutant. In ttg1-22, the T-DNA inserts in proximity to the stop codon. This might also causes a strong phenotype as seen for the premature stop codon mutant ttg1-1 in Landsberg erecta (Ler) background. We tested these mutants for some of the early TTG1-dependent developmental traits. Compared to the wild type, all used Col-0 mutants showed the analyzed aspects of the ttg1 syndrome (Koornneef, 1981) (Fig. S1). Moreover, we wondered, why the flowering time phenotype of ttg1 mutants was not reported before. Therefore, we added an often used mutant in Ler background, ttg1-1, a point mutant with a premature stop codon close to the end of TTG1 (Larkin et al., 1994; Walker et al., 1999), and a mutant in Wassilewskija (Ws) background, ttg1-10, carrying a point mutation in the TTG1 promoter (Larkin et al., 1994; Larkin et al., 1999).

The different mutants and variants showed different patterns of flowering time phenotypes relative to the respective backgrounds (Figs. 1C–1F). In Col-0 background, similar to ttg1-9 and ttg1-11, ttg1-21 and ttg1-22 flowered significantly earlier in terms of time and leave number at both tested conditions with one exception. For ttg1-21 grown in the warm condition, the number of days only deviated significantly from the wild type in one out of three repeats. Therefore, early flowering cannot be concluded in this case. For ttg1-21 and ttg1-22 it can be summarized that the mutants responded to temperature in the same way as the wild type in regard to time (days) and number of leaves produced until flowering. Again, there was one exception, only in one of three repeats for ttg1-21, a significantly different number of leaves was recorded when comparing between the warm and cold condition at flowering time. This suggests that these mutants are less responsive to temperature affecting its number of leaves at flowering time than the wild type and ttg1-22.

As compared to the Ler wild type, flowering time was not significantly reduced in ttg1-1 at both conditions. Therefore, it is not surprising that the flowering time phenotype was not reported before for this heavily used ttg1 mutant. In contrast to the other analyzed mutants, ttg1-10 carries its mutation as a point mutation in the TTG1 promoter. Interestingly, ttg1-10 mutants flowered significantly later at both conditions for both recorded flowering phenotypes as compared to its wild type. Ler and ttg1-1 did not respond to the difference in temperature, when it comes to the number of leaves produced at flowering time. ttg1-1 and Ws responded only in one out of two experiments to the difference in temperature for the number of days (ttg1-1) and leaves (Ws), respectively. In these cases, a reduced response to temperature cannot be concluded. In all other cases, the reduced temperature caused a delay in flowering time as also observed for Col-0 in the other experiments.

TTG1-9 and TTG1-11 encode TTG1 protein variants. Therefore, the observed early flowering might be due to a gain- or loss-of-function of the TTG1 variant or TTG1 gene in the respective mutants. To specify this, we generated overexpression lines of TTG1 (Col-0) driven by the constitutively active 35s promoter in Col-0 background (Pro35S::YFP-TTG1 (Col-0), the three lines are subsequently named OE01-OE03, Fig. 1G). As a regulatory hub in light signaling, COP1 is known to regulate protein stability of relevant flowering time regulators (Jang et al., 2008; Liu et al., 2008; Yu et al., 2008) and COP1 can interact with a TTG1 gene regulatory network component at the protein level (Maier et al., 2013). Therefore, we included cop1-4 as a background in the flowering time analysis (Pro35S::YFP-TTG1 (cop1-4), the three lines are subsequently named OE19-OE21). All overexpression lines in wild-type background flowered later than the wild type at both temperatures and in respect to time and number of leaves (Figs. 1H–1I). This suggests a loss-of-function in the mutant scenario. Compared to Col-0, cop1-4 produced significantly less leaves at the time point of flowering. An increase in number of leaves and days in this background was only observed for the overexpression line OE20 as compared to its background (Figs. 1H–1I). This might be due to different transcript or protein levels in the overexpressors.

The overexpression constructs did not affect the endogenous TTG1 transcript levels despite for OE19 in cop1-4 background. In this line, a significant reduction of endogenous TTG1 transcript was observed (Fig. 1J, Table S3, see Fig. 1G and Fig. S2 for the selective primer design). Interestingly, in cop1-4 mutants, TTG1 CDS transcript levels were significantly increased (3-4-fold) as compared to Col-0 (Figs. 1J–1L). All overexpression lines showed a significant overexpression of the construct (Figs. 1K–1L). Highest expression and protein levels were reached by line OE01 in Col-0 background and by line OE20 in cop1-4 background (Figs. 1K–1M, Fig. S3). Both, expression and protein levels, were comparable for line OE02 (Col-0) and line OE20 (cop1-4). For OE19 and OE21 (cop1-4), in different repeats, YFP-TTG1 levels were either comparable (Fig. 1M) or close to or even below the detection limit (Fig. S3). In Col-0 background, line OE03 exhibited the highest variance of YFP-TTG1 levels compared to other lines in between the repeats. This is in agreement with the observation at the fluorescence stereo microscope using older plants. Here, in several OE03 plants YFP-fluorescence was absent in areas of the leaves or in the center of the rosette. This patchiness of YFP fluorescence was also observed in OE21 (cop1-4) and sometimes in OE01 (Col-0) plants but only in one out of 50 OE02 (Col-0) plants and in none of the OE20 (cop1-4) plants (Table S4). Therefore, and due to the similar transcript and protein level, OE02 (Col-0) and OE20 (cop1-4) were chosen for subsequent quantitative RT-PCR (qRT-PCR) experiments.

Together, we revealed that TTG1 has an effect on flowering time. Subsequently, we used q-RT-PCR experiments for an initial embedding of TTG1 in the transcriptional regulation of the flowering time pathway.

TTG1 can reduce FT and SOC1 transcript levels

TTG1 acts early in cell fate determination and is a pleiotropic regulator of transcription. The known mechanisms of TTG1 molecular activity are at the protein level at which it acts in differing complex composition with transcription factors that act as the direct modulators of transcription (e.g., Zhang & Schrader, 2017). Therefore, overexpression lines are most informative to initially reveal if the TTG1 protein can have an impact on the transcriptional regulation of specific targets within the individual branches of the flowering time regulatory pathway. Moreover, on the one hand, by using the cop1-4 mutant, we added a sensitized background with a significant modulation in protein composition. On the other hand, COP1 interacts with at least one TTG1-complex component (PAP2), therefore, this background also allows for a conclusion if TTG1 activity requires a functional COP1 protein. Results in this background provide insides if TTG1 at elevated protein levels is able to even overwrite the transcriptional scenario in the light-signaling and LD flowering time mutant cop1-4. This would underline even more than in the wild-type scenario the potential adaptive value of TTG1 and relevance as a valuable target for flowering time modulation in various environmental settings. All selected targets were analyzed with the most suitable overexpression line as characterized and described above. In addition, seed material of the two EMS mutants ttg1-9 and ttg1-11 was available.

First, we assessed the circadian expression profile of endogenous and overexpressed TTG1 in OE02. As for all circadian qRT-PCR experiments in this study, we used 8-day-old LD grown seedlings harvested first at ZT0 and thereafter in intervals of 4 h with the last sample at ZT20. A similar overexpression level of TTG1 was seen throughout the day. The endogenous expression was not affected (see Fig. S4, Table S5).

Due to the strong late flowering phenotype of the overexpression lines, we expected that the transcript levels of the floral integrators (e.g., Blumel, Dally & Jung, 2015) differed from the respective backgrounds. Towards this end, we continued with analyzing the CO-FT module’s and SOC1 transcript levels. In line with the flowering time phenotype, for CO, an activator of FT (Suarez-Lopez et al., 2001), a slight tendency to lower transcript levels in the overexpression lines was observed. Nevertheless, this was neither significant nor sufficient to explain the strong phenotype especially in the overexpression in wild-type background (Fig. 2, Table S5). GI is an FT regulator that can increase CO transcript levels but can also activate FT in a CO-independent way. It can directly bind to the FT promoter and interacts with FT suppressors (Sawa & Kay, 2011; Sawa et al., 2007). Also, GI transcript levels in OE02 were not significantly changed (Fig. S5). In both overexpression lines, for FT, the transcript levels almost dropped to the detection limit and exhibited a general reduction at all timepoints which was significant at ZT12 and 16 in OE02. The overexpressors’ SOC1 transcript profiles were very similar in their circadian pattern and both exhibited reductions throughout the day, which was significant at ZT0, 4, 8 and 16 for OE20. In the mutant scenario, we expected to find more subtle effects. We found only a trend to elevated SOC1 levels in ttg1-9 mutants which might explain the early flowering time phenotype but was not significant. In summary, we found that TTG1 can reduce FT and SOC1 transcript levels.

Early and late effects on transcript levels of AP2-domain containing factors in ttg1 mutants

We moved our focus on factors that can suppress FT transcript levels, the AP2 domain containing factors of the flowering time regulatory pathway: TEM1, TEM2 (RAV transcription factors with AP2/ERF and B3 DNA-binding domain) and AP2, SMZ, SNZ, TOE1-3 e.g., (Song, Ito & Imaizumi, 2013; Wang, 2014). In addition, we tested the transcript level of SVP which acts as an activator of the aforementioned AP2-like factors and as a suppressor of FT (Lee et al., 2007; Tao et al., 2012). Only at night, a significant increase of the TEM2 transcript level was observed for OE02 (Fig. 3, Table S5). No significant change was found for OE20 suggesting that elevated TTG1 levels do not have a strong impact on these genes. However, in the mutants’ case a trend for reduced transcript levels was observed for TEM2 (ttg1-9), AP2, TOE1 and TOE3 (all ttg1-11) and a slight reduction for SNZ (ttg1-11, at night also for ttg1-9) and SMZ (both mutants). Significance analysis (P < 0.05) supported the reduction of TOE1 transcript levels at ZT4 and ZT16 for ttg1-11 and SNZ transcript levels at ZT20 for ttg1-9. Interestingly, although not supported by the significance analysis, the trend of reduced transcript levels for SVP seemed to be opposed by elevated SVP levels at ZT0 and ZT20 for OE02 pointing to a possibly flattened circadian amplitude of SVP transcript levels upon TTG1 overexpression. Therefore, we had a closer look on circadian clock components for OE02.

TTG1 can regulate circadian clock components

LWD1 and LWD2, the closest homologs of TTG1, regulate flowering through transcriptional modulations within the circadian clock (Wang et al., 2011; Wu et al., 2016; Wu, Wang & Wu, 2008). LWD1 was shown to bind to the promoter of PRR5, PRR9 and PRR1/TOC1 (Wang et al., 2011). Therefore, we analyzed the transcript levels of the core clock components LHY, CCA1 and TOC1/PRR1 and also those of its feed-back regulators PRR5, PRR7 and PRR9 (Shim, Kubota & Imaizumi, 2017) in the TTG1 overexpression line OE02. Figure 4 shows the results sorted by the time point of the maximal peak in the wild type. In general, a flattened circadian amplitude is to be seen despite for TOC1 which is not affected by TTG1 overexpression. For LHY, the minimum seems to be shifted to ZT8 instead of ZT12 in the wild type. A significant increase (P < 0.5) in transcript levels was observed for CCA1 at ZT12 and ZT16, for PRR9 at ZT20, for PRR7 at ZT20 and ZT0 and for PRR5 at ZT0 and ZT4. Thus, we found that TTG1 can regulate the transcript levels of circadian clock components and modulates their transcriptional profiles mainly through flattening the amplitude as a consequence of elevated transcript levels at distinct timepoints on the day under investigation. Similar as its homologs LWD1 and LWD2, TTG1 can change the transcript levels of PRRs but did not modulate TOC1 transcript levels.

The transcript levels of the so far analyzed branches of the flowering time regulatory pathway did not provide a convincing explanation for the strong suppression of FT and SOC1 transcript levels in the overexpression lines. FLC represents and integrates additional branches of this pathway. The FLC protein can bind directly to the promoters of FT and SOC1 and suppresses the respective transcript levels (Helliwell et al., 2006; Searle et al., 2006). Therefore, we decided to complete our initial embedding of TTG1 in the transcriptional control of the flowering time pathway by analyzing the transcript levels of FLC. We found that overexpression of TTG1 resulted in elevated FLC transcript levels throughout the day (Fig. 4). These were significantly and more than 2-fold increased as compared to the wild type at ZT8. We conclude that TTG1 can act as an activator of FLC.

One well selected and characterized overexpression line was analyzed in three biological replicates based on three independent seed batches for this initial embedding of TTG1 in the flowering time regulatory pathway. The different and specific time points of transcript modulation spread throughout the day based on the overexpression line OE02 suggest the general ability of TTG1 to elevate transcript levels of clock components and FLC and identify both branches as targets for specific and more detailed follow up studies.

Together, we found that TTG1 acts as a transcriptional regulator in various parts of the flowering time regulatory pathway.

PRR5 recruits TTG1 to subnuclear foci and bHLH92 nuclear enrichment is counteracted by TTG1

In its role as a regulator of early developmental traits, TTG1 acts through differential complex composition, it acts in competitive scenarios, trapping to the nucleus and mutual localization with respective interactors (Balkunde, Bouyer & Hulskamp, 2011; Bouyer et al., 2008; Pesch et al., 2015; Wester et al., 2009; Zhang et al., 2019; Zhang & Schrader, 2017; Zhao et al., 2008). All these scenarios occur at the protein level in dependence of its interactors. bHLH factors like GL3, EGL3, TT8 or MYC1 and R2R3-MYB factors form R2R3MYB-bHLH-WD40 (MBW) complexes with TTG1 to regulate the early TTG1-dependent traits (Balkunde, Pesch & Hulskamp, 2010; Broun, 2005; Koornneef, 1981; Lepiniec et al., 2006; Miller, Chezem & Clay, 2015; Ramsay & Glover, 2005; Tominaga-Wada, Ishida & Wada, 2011; Walker et al., 1999; Zhang et al., 2003). We wondered, which interactors might be relevant for TTG1 function towards flowering time regulation. As the classical bHLH interactors did not modulate flowering time in a similar way as TTG1 (Fig. S6), we conducted a Y2H screening to identify candidates which are related to the flowering time regulatory pathway. Among the results of this screen was one of the classical bHLH interactors, EGL3 (Table S6) (Zhang et al., 2003). In addition, and related to the flowering time regulatory pathway, we identified PRR5 and bHLH92, a bHLH factor shown to be expressed in a circadian pattern (Hanano et al., 2008).

Both selected candidates could be verified as interactors of TTG1 in Y2H experiments in which they were fused to the GAL4-binding as well as to the GAL4-activation domain (Figs. 5A–5B, Figs. S7A–S7B). We also tested PRR7, PRR9 and TOC1 in the GAL4 system (Figs. 5A–5B) with different 3-AT (3-amino-1,2,3triazole) concentrations and an adjusted optical density of the samples (Figs. S7A–S7B). All three PRRs interacted with TTG1 when fused to the activation domain of GAL4. TOC1 exhibited the weakest interaction, as indicated by growth of yeast on the respective plates, followed by PRR9, PRR7 and PRR5. A very weak possible interaction of LWD1 and LWD2 was seen with PRR5 in yeast (Figs. S7B–S7C). However, among the homologs, the interaction with all PRRs is TTG1-specific. bHLH92 did neither interact with LWD1, LWD2 nor with the PRRs in this assay (Figs. S7B–S7C). This suggests that the mechanisms of LWDs and of the TTG1 protein in flowering time regulation including the transcriptional modification of the circadian clock differ.

The TTG1 protein is cell-to-cell mobile (Bouyer et al., 2008). Knowing that nuclear trapping by GL3 is a relevant mechanism for TTG1 function in trichome patterning (Balkunde, Bouyer & Hulskamp, 2011; Bouyer et al., 2008), we analyzed the localization of TTG1 in presence and absence of PRR5 and bHLH92 and vice versa (Figs. 5C–5U, Figs. S8–S9). As reported before, tagged TTG1 is localized in the cytoplasm and in nucleus in epidermal cells of infiltrated tobacco leaves (e.g., Bouyer et al., 2008). PRR5 is only localized to the nucleus where it forms nuclear foci. When co-expressing RFP-tagged TTG1 and YFP-tagged PRR5, RFP-TTG1 localized predominantly in the nucleus where it co-localized with YFP-PRR5 (Figs. 5C–5K, Figs. S8–S9). In case of bHLH92, we obtained a different result. YFP-bHLH92 is enriched in the nucleus but also localizes to the cytoplasm when co-expressed with RFP alone. When co-expressed with RFP-TTG1, TTG1 localization did not change but the nuclear enrichment of bHLH92 did not occur (Figs. 5L–5U, Figs. S8–S9). We repeated the experiment with each of the PRRs (PRR5, PRR7, PRR9 and TOC1/PRR1) being fused to CFP in combination with RFP alone or RFP-TTG1 (Fig. 6, Figs. S8–S9). The PRRs did not only recruit TTG1 to the nucleus, TTG1 also changed the subnuclear localization for PRR9 and PRR7.

TTG1 variants, protein levels and the background

The Ler wild type exhibited the earliest flowering wild type in our experiment which could impair an early flowering of ttg1-1 relative to its wildtype with respect to time and number of leaves produced. Due to these results, it is not surprising that the flowering phenotype was not reported earlier as the ttg1-1 has been heavily used in previous studies since the 1980s. The late flowering phenotype of ttg1-10 in Ws background might be initially surprising but can be explained by the localization of the mutation within the TTG1 gene. Floral buds did not express the TTG1 transcript in this mutant (Larkin et al., 1999). This might deviate at different developmental stages and tissues as ttg1-10 is an EMS mutant with a point mutation in the TTG1 promoter. The mutation might change the expression pattern of TTG1, which in turn can suppress flowering. Also, a manifested second site mutation cannot be excluded as well as an effect of the Ws background.

Interestingly, the ttg1-21 mutant, the presumably strongest mutant analyzed in this study, showed the mildest effect on flowering time as compared to the wildtype. At the warm condition, a significant deviation from the wild type could not be concluded for this mutant in terms of time. This suggests that the properties of the TTG1 protein might have a stronger effect on flowering time regulation than the reduced level or absence of TTG1. Presence of the TTG1 protein is not required for the plant to flower but elevated levels and variants can modulate flowering time in A. thaliana. Therefore, an overexpression line seems to be more beneficial to initially identify targets and interaction partners for TTG1-dependent regulation within the flowering time regulatory pathway. Thereby, young plant material can be used to provide the basis for a broad embedding of TTG1 in this pathway. The phenotype of the mutants is in agreement with the phenotype of the overexpression lines in the same background. However, TTG1-dependent regulation through protein level and type of protein variant might follow overlapping as well as differing regulatory mechanisms which might be dependent on the respective background and variant-specific interaction patterns to be dissected in the future.

The developmental stage and age pathway

Our results acquired along the chronological and developmental axis point to an involvement of TTG1 in modulating the plastochron. Detailed meristem analysis at and around the time point of flowering are required in the future. The developmental stage and a possible tissue-specificity might also explain the FT and SOC1 transcript levels which were not significantly increased in ttg1 mutants. As TTG1 is a factor required for cell fate determination (Galway et al., 1994), developmental stage, tissue and cell specific effects might occur. Therefore, older plants than the used seedlings and the analysis of tissue specific expression might be required in a more detailed future analysis. This would be of particular relevance for the age pathway and GA signaling which were not covered in this study.

Overlaps with the age pathway might occur at the level of the SQUAMOSA BINDING PROMOTER BINDING PROTEIN-LIKE (SPL)s which are suppressed by microRNA156. SPLs are involved in regulating trichome density at later stages e.g., at the stem (Yu et al., 2010), SPL9 activates TRIPTYCHON and TRICHOMELESS1 and is thought to modulate trichome density thereby (Yu et al., 2010). Both R3-MYBs belong to the so-called inhibitors which compete in the MBW complex scenarios with R2R3-MYBs for bHLH factor binding (Balkunde, Pesch & Hulskamp, 2010; Esch et al., 2003; Wang et al., 2008; Wester et al., 2009). TTG1 itself was reported to interact with SPL4 and SPL5 in yeast (Ioannidi et al., 2016) and the mutant of another inhibitor, enhancer of try and cpc 3 (etc3), exhibits a differential FT and SOC1 regulation in 21d-old LD-grown plants as compared to the wild type which equals the time of its early flowering time phenotype (Wada & Tominaga-Wada, 2015).

WERWOLF (WER) and FT mRNA stability

If TTG1 acts at the FT gene, TTG1 can either act on the promoter or on other regulatory regions of FT (and SOC1) or affect the FT mRNA stability as reported for WER, a TTG1 network component (Seo et al., 2011). Seo and co-workers found that the mutant of WER flowers late and the R2R3-MYB factor WER is required for FT mRNA stability. The role of WER towards flowering time regulation was found to be independent of CO and FLC. Together with our results, the role of TTG1 and WER would be opposing which is not in line with a joined regulation following the classical MBW complexes regulatory mechanisms. However, we can not exclude that FT mRNA stability is changed in dependence of TTG1 complexes. By forming an MBW complex with WER, TTG1 could prevent WER from its function towards FT mRNA stability which would add to the late flowering time phenotype observed in TTG1 overexpressors.

AP2-domain containing factors and the GA signaling branch

Several AP2-domain containing factors exhibit the expected tendencies to reduced transcript levels in ttg1 mutants. However, the overexpressor lines did not show the respective opposing effect. Hence, an intact TTG1 seems to be required for normal circadian transcript profiles of these AP2-domain containing factors.

At the transcriptional level, TEM1 and TEM2 act as repressors of the TTG1-MBW complex components GL1, an R2R3-MYB factor, and the bHLH factors GL3 and EGL3 while TTG1 transcript levels are not affected (Matias-Hernandez et al., 2016). On the one hand, TEM1 and TEM2 act in a cell type-dependent manner (Matiaz-Hernandez, 2016). Therefore, an effect of deviating cell fate and differentiation in the ttg1 mutants can have additional effects. On the other hand, TEMs control GA accumulation and distribution in the leaf mesophyll. They also integrate the photoperiod and GA signaling pathway in LD and SD conditions (Castillejo & Pelaz, 2008; Matiaz-Hernandez, 2016; Osnato et al., 2012). At the molecular level, reduced TEM2 transcript levels might circumvent FT and SOC1 transcript levels and cause an early flowering of ttg1-9. Reduced TEM2 levels lead to elevated GA levels which promote flowering time. The bHLH factors GL3, EGL3 and the R2R3-MYB factor GL1, can interact with the DELLA proteins RGA (REPRESSOR OF GA1-3 (mutant of GA REQUIRING 1)) 1 and RGA2 which both repress the transcriptional activation properties of the MBW complex. This suppression is derepressed by GA through GA-induced degradation of the DELLA proteins (Qi et al., 2014). Therefore, different binding properties of the mutant protein variants might cause additional regulatory loops through the GA signaling pathway to play a role.

Competitive scenarios modulating CO protein levels

AP2-domain containing factors can bind directly to the FT promoter (Castillejo & Pelaz, 2008; Mathieu et al., 2009; Zhang et al., 2015). In addition, at the protein level, TOEs can interact with CO and thereby prevent CO from activating FT transcription (Zhang et al., 2015). In contrast, PRRs can stabilize the CO protein (Hayama et al., 2017). Although a reduction of CO transcript levels upon TTG1 overexpression at ZT12 is observed with similarities to the patterns of PRR overexpressors (Hayama et al., 2017), a leading role of CO in the TTG1-dependent regulation of FT transcript levels cannot be concluded based on the CO transcript levels. Nevertheless, at the protein level, TTG1 might either decrease CO protein levels or inactivate the CO protein and, thereby, reduce a CO-mediated FT activation. The interaction of TTG1 with PRRs and re-localization of PRRs, as suggested by our results, could have such an effect.

CO protein levels are elevated in cop1-4 background (Jang et al., 2008). Overexpression of TTG1 can also delay flowering time in cop1-4 background. This indicates that either the TTG1 protein levels were sufficient to counteract the CO protein function at the protein level or that the TTG1 function can be or is mainly independent of CO. However, the effects on flowering time are difficult to compare among wild type and cop1-4 background as we found increased TTG1 transcript levels in cop1-4.

LWDs, TTG1 and the clock

Further upstream in the photoperiodic pathway, LWD proteins act as activators within the loops of the circadian clock (Shim, Kubota & Imaizumi, 2017). It is conceivable that a partial overlap in function exists as these are the two closest homologs of TTG1 in A. thaliana but also differences are expected. Towards this end, a detailed evolutionarily focused analysis will be of interest.

While LWD1 transcript levels show a strong circadian response with highest levels late at night and in the long-day morning, LWD2 and, in our results, TTG1 do not show this pattern and remain at a similar level during the day and night (Wu, Wang & Wu, 2008). Interestingly, promoter-luciferase constructs showed rhythmic activity of both LWD promoters (Wang et al., 2011). With a focus on LWD1, its binding in a time dependent manner at the promoters of PRR5, PRR7, PRR9 and CCA1 was revealed (Wang et al., 2011). In lwd1lwd2 double mutants, CCA1 and LHY transcript levels are reduced, the period is shortened and shifted forward (Wu, Wang & Wu, 2008). In the late afternoon to early morning at LD condition, TTG1 overexpression increases the transcript levels of PRR5, PRR7, PRR9, CCA1 and LHY and potentially reduces the respective transcriptional amplitudes.

CCA1 expression can be activated by LWD1 which acts together with interacting TCP transcription factors (Wang et al., 2011; Wu et al., 2016). TTG1 might act similarly but, also, a mechanism as suggested for CO is possible for CCA1 and LHY. Through binding of TTG1 to PRRs, these might no longer be able to form a complex with the transcriptional repressors TOPLESS/TOPLESS RELATED PROTEINs at the CCA1 and LHY promoters (Wang, Kim & Somers, 2013). Consequently, CCA1 and LHY transcript levels would increase and the circadian period would be lengthened (Wang, Kim & Somers, 2013). We observed respective transcript level characteristics upon TTG1 overexpression apart of the period which was not tested.

Differential complex composition of TTG1 with interacting transcriptional activators and repressors might act in comparable regulatory mechanisms mediating target specific DNA-interaction. PRRs and bHLH92 described in this study are excellent candidates. As we did not find strong indications for a clear interaction with LWDs, mechanisms related to PRRs and bHLH92 at the protein level are likely to differ between the TTG1 and the LWDs. Being a bHLH factor, bHLH92 might fit well into the regulatory scheme known for TTG1. This would require the identification of an R2R3-MYB interacting with bHLH92.

TTG1 might act through TTG1-PRR and TTG1-FLC-FT modules (Fig. 7, Fig. S10)

The subnuclear localization patterns upon co-expression of PRRs and bHLH92 with TTG1 can provide a regulatory mechanism. Does this have an influence on transcriptional activation? What is the identity of the subnuclear foci? Are the interaction partners binding in concert to specific loci, stabilize each other, are they stored or deactivated within these foci or are competitive scenarios at work? These are pressing questions to be answered.

Through such mechanisms, known trait networks of the clock, like those of the PRRs, can be linked with the TTG1 trait network. For PRR5, target promoters were identified which comprise different transcription factors involved e.g., in auxin production, hypocotyl growth and cold-stress response. This might be intermingled with traits and temperature responsiveness in TTG1-dependence. TTG1 might suppress PRR target modulation on relevant downstream targets of both factors at a respective developmental stage. FT might be such a target as it was shown that a PRR-CDF-FT module can exists. PRR5 can directly suppress CDF expression (Nakamichi et al., 2012) and CDFs can suppress FT at its promoter (Song et al., 2012). A TTG1-PRR-CDF-FT module could bypass GI and CO and could link TTG1 effects on clock gene modulation with FT suppression in a competitive scenario with PRR-CDF-FT modules.

Our results suggest a possible, parallel TTG1-FLC-FT module. Elevated FLC transcript levels lead to an increase in FLC-mediated FT and SOC1 suppression and consequently to late flowering in the overexpression line. In line with this role of TTG1, the weak allele of FLC in Ler background could explain the absence of a flowering time phenotype in the strong and heavily used ttg1-1 mutant. Therefore, when anlalyzing this mutant, a role of TTG1 towards flowering time regulation was not detectable before. All used accessions in this study are rapid cycling accessions lacking a functional FRI allele and therefore immediately exposing modulations at FLC to potential phenotypic detection.

Overlapping regulatory network

The annual plant A. thaliana completes its life cycle with the production and ripening of seeds and enters the new life cycle following seed dormancy with seed germination. The reproductive success depends therefore on the appropriate timing of flowering and seed ripening as well as germination thereafter. Therefore, it is not surprising that a pleiotropic regulator like TTG1 which is strongly involved in the regulation of various relevant seed traits, is also involved in the regulation of flowering time. Here, it is noteworthy that TTG1-dependent gene regulatory network components including TTG1 have the potential to intervene in several sub-pathways of flowering time regulation. We found that TTG1 can even overwrite the transcriptional scenario in cop1-4 in regard to the floral integrators FT and SOC1. Moreover, TTG1 variants are likely to be of relevance in adaptation to temperature seasonality, minimum temperature and daylength (Hancock et al., 2011). This strongly suggests an adaptive value of the TTG1-dependent trait network which is strengthened by the overlapping gene regulatory networks of TTG1 and flowering time regulation as substantiated with this study.