GIGANTEA accelerates wheat heading time through gene interactions converging on FLOWERING LOCUS T1

Precise regulation of flowering time is critical for cereal crops to synchronize reproductive development with optimum environmental conditions, thereby maximizing grain yield. The plant specific gene GIGANTEA (GI) plays an important role in the control of flowering time, with additional functions on the circadian clock and plant stress responses. In this study, we show that GI loss-of-function mutants in a photoperiod sensitive tetraploid wheat background exhibit significant delays in heading time under both long-day (LD) and short-day (SD) photoperiods, with stronger effects under LD. However, this interaction between GI and photoperiod is no longer observed in isogenic lines carrying either a photoperiod insensitive allele in the PHOTOPERIOD1 (PPD1) gene or a loss-of-function allele in EARLY FLOWERING 3 (ELF3), a known repressor of PPD1. These results suggest that the normal circadian regulation of PPD1 is required for the differential effect of GI on heading time in different photoperiods. Using crosses between mutants or transgenic of GI and those of critical genes in the flowering regulation pathway, we show that GI accelerates wheat heading time by promoting FLOWERING LOCUS T1 (FT1) expression via interactions with ELF3, VERNALIZATION 2 (VRN2), CONSTANS (CO), and the age-dependent microRNA172-APETALA2 (AP2) pathway, at both transcriptional and protein levels. Our study reveals conserved GI mechanisms between wheat and Arabidopsis, but also identifies specific interactions of GI with the distinctive photoperiod and vernalization pathways of the temperate grasses. These results provide valuable knowledge for modulating wheat heading time and engineering new varieties better adapted to a changing environment.


Introduction
GIGANTEA (GI) is a plant-specific gene that encodes a large protein (1,173 amino acids in Arabidopsis) with no domains of known biochemical function (Fowler et al. 1999). The GI protein is predominantly localized to the nuclei where it forms nuclear bodies (Huq et al. 2000).
Both GI transcription and GI protein abundance are regulated by the circadian clock (Fowler et al. 1999;Park et al. 1999;David et al. 2006) and can be detected throughout plant development (Fowler et al. 1999). In Arabidopsis plants grown under long-day photoperiod (16 h light/8 h dark, henceforth LD), the GI transcripts peak at Zeitgeber time 10 (ZT10); while under short-day (8 h light/16 h dark, henceforth SD), the peak is at ZT8 (Fowler et al. 1999;Park et al. 1999).
Defects in several components of the circadian clock have been shown to affect the GI transcription profile (Fowler et al. 1999;Lu et al. 2012). Chromatin Immunoprecipitation (ChIP) analysis showed that the evening complex (EC), including EARLY FLOWERING 3 (ELF3), EARLY FLOWERING 4 (ELF4) and LUX ARRYTHMO (LUX), regulates GI transcription by direct binding to its promoter (Ezer et al. 2017). ELF3 can also regulate GI protein degradation through its interaction with CONSTITUTIVELY PHOTOMORPHOGENIC1 (COP1) in an ECindependent manner . The Arabidopsis elf3 gi double mutant is unable to synchronize the circadian oscillator to the light-dark cycles, indicating that these two genes are essential for light entrainment of the clock (Anwer et al. 2020).
The multiple gi mutants identified in Arabidopsis (Rédei, 1962;Koornneef et al. 1991) show different effects on the length of the circadian clock period, but all of them exhibit a late flowering phenotype (Fowler et al. 1999;Park et al. 1999). The effect of GI on Arabidopsis flowering time is stronger under LD than under SD, indicating an interaction between GI and the photoperiod pathway (Araki and Komeda, 1993;Mizoguchi et al. 2005). In addition to its role in the circadian clock regulation and photoperiodic flowering response, GI has been shown to play important roles in multiple other physiological processes such as light signaling, chlorophyll accumulation, starch accumulation and stress tolerance (reviewed in Mishra & Panigrahi, 2015;Brandoli et al. 2020). Because GI can physically interact with many proteins of the circadian clock and photoperiodic flowering pathways, GI has been suggested to function as a scaffold or hub protein and orchestrates other protein interactions (Imaizumi and Kay, 2006).
In Arabidopsis, GI regulates the expression of FT and flowering time through both CONSTANS (CO)-dependent and CO-independent mechanisms. In the CO-dependent pathway, GI acts between the circadian oscillator and the main photoperiod gene CO to accelerate flowering by promoting both CO and FT mRNA abundance (Mizoguchi et al. 2005). GI interacts with the FLAVIN-BINDING, KELCH REPEAR, F-BOX1 (FKF1) protein to form a complex that regulates the protein abundance of CYCLING DOF FACTOR 1 (CDF1), a transcriptional repressor of CO. The degradation of CDF1 protein elevates the transcription of CO, thereby promoting FT expression and flowering (Sawa et al. 2007). In addition, GI can interact with CO directly and indirectly through interactions with FKF1 and ZEITLUPE (ZTL) to regulate CO protein stability (Song et al. 2014;Hwang et al. 2019). Finally, the interaction between GI and SPINDLY (SPY), which is a negative regulator of gibberellin signaling, can inhibit SPY activity and promote FT transcription in a CO-dependent manner (Tseng et al. 2004).
GI also regulates FT expression in Arabidopsis through two CO-independent pathways. The first one involves microRNA172 (miR172), which inhibits the expression of flowering repressors TARGET OF EAT 1 (TOE1) and APETALA 2 (AP2) (Jung et al. 2007). In the second COindependent pathway, GI activates FT transcription through direct binding to the FT promoter alone or in a protein complex with FT repressors such as SHORT VEGETATIVE PHASE (SVP), TEMPRANILLO 1 (TEM1) and TEM2 (Sawa and Kay, 2011).
In Arabidopsis, CO plays a central role in the photoperiodic response by inducing flowering under LD conditions (Valverde et al. 2004), and CO homologs from rice and sorghum have also been shown to play major roles in the photoperiodic control of flowering (Yano et al. 2000;Yang et al. 2014). However, in the temperate grasses, the duplicated CO1 and CO2 genes show In the temperate grasses, PPD1 is the dominant gene in the photoperiodic pathway, showing strong differences in heading time between LD and SD even in the absence of functional copies of both CO1 and CO2 (Shaw et al. 2020). Although the interactions between GI and CO have been extensively studied in Arabidopsis (Mizoguchi et al. 2005;Sawa et al. 2007) and rice (Hayama et al. 2003), the interactions between GI and PPD1 remain largely unexplored.

5
Another understudied aspect of the role of GI in the temperate cereals is its interaction with VRN2, a LD repressor of FT1 with no known homologs in eudicots (Yan et al. 2004). VRN2 encodes a protein comprising a zinc finger motif at the N-terminus and a CCT domain at the Cterminus (Yan et al. 2004), which is orthologous to the rice protein GRAIN NUMBER, PLANT HEIGHT, AND HEADING DATE 7 (GHD7) (Xue et al. 2008;Woods et al. 2016). In rice, the GI protein has been shown to interact in vivo with GHD7 and to regulate its stability (Zheng et al. 2019). Loss-of-function mutations in GHD7 result in earlier flowering under LD in rice (Xue et al. 2008), and mutations in VRN2 result in accelerated heading under LD and a spring growth habit in wheat (Yan et al. 2004;Distelfeld et al. 2009).
In this study, we show that GI loss-of-function mutants in wheat exhibit significant delays in heading time under both LD and SD. This delay is larger under LD than under SD in photoperiod sensitive wheats (PS), but the differences disappear in photoperiod insensitive (PI) wheats and elf3 mutants. Moreover, we show that GI regulates FT1 expression and heading time through its interactions with ELF3, VRN2, CO, and the age-dependent miR172-AP2 pathway, revealing both conserved and specific mechanisms for the temperate grasses. Kronos carries the photoperiod insensitive allele Ppd-A1a, which has a 1,027 bp deletion in the promoter region and is associated with misexpression of PPD1, accelerated induction of FT1 expression, and early flowering under SD (Wilhelm et al. 2009). To study the effect of the GI mutations in a photoperiod sensitive background, we crossed the four gi mutants twice with a 6 near-isogenic Kronos line carrying the photoperiod sensitive Ppd-A1b allele (henceforth Kronos-PS, Pearce et al. 2017), and intercrossed them to generate three different BC 1 F 2 lines homozygous for gi mutants in both the A-and B-genome homoeologs. The gi-1 line, generated from the cross between A-2019 and B-2205, headed 12 d later than Kronos-PS, whereas the gi-2 (A-401 x B-2205) and gi-3 (A-401 x B-3825) mutants headed 16 d later (Fig. 1B). These results indicate that GI functions as a promoter of heading time in wheat under LD.

GI loss-of-function mutants delay heading time in wheat under LD photoperiod
To further explore the function of GI, we generated wheat transgenic lines constitutively expressing a fusion of the full-length coding region of the GI gene and a C-terminal 4xMYC tag under the control of the maize UBIQUITIN promoter in a Kronos-PI background (hereafter referred to as UBI-GI-MYC). Among the six independent transgenic events, transgenic line #13 showed the highest transcript levels of GI relative to the non-transgenic sister line ( Fig. S1A, P < 0.001) and was selected for further characterization under both LD and SD conditions. Transgenic T 1 plants of the UBI-GI-MYC line #13 headed 3.3 d earlier than the non-transgenic Kronos-PI controls when grown under LD (Fig. S1B), but showed no difference in heading time relative to the controls under SD (Fig. S1C). These results confirmed that GI is a LD promoter of wheat heading time.
We next crossed the UBI-GI-MYC #13 transgenic plant (PI) with the gi-2 mutant (PS) to test the ability of the transgene to rescue the delayed heading time of the gi-2 mutant. In the F 2 progeny, we selected sister lines homozygous for the wild-type control (WT), the non-transgenic gi-2 mutant (gi-2), and the combined UBI-GI-MYC / gi-2 (all homozygous for the PPD1-A1b allele, PS). The UBI-GI-MYC / gi-2 plants headed 12.4 d earlier than the gi-2 mutant, but still 2.4 d later than the wild type under LD (Fig. 1C). These results indicate that the UBI-GI-MYC #13 transgene can largely, but not completely, rescue the delayed heading time of the gi-2 mutant ( Fig. 1C).
In addition to the late heading under LD, the gi-2 mutants also showed an increased number of spikelets per spike ( Fig.1D-E) and tillers ( Fig. 1F-G). Compared with the wild-type control plants, the gi-2 mutant produced on average 5.8 more spikelets per spike (Fig.1E) and had an average of 5.6 more spike-bearing tillers (Fig. 1G), likely associated with its extended vegetative phase.

When grown under SD (8 h light), Kronos-PI plants head between 80 and 100 days, whereas
Kronos-PS plants fail to head (Shaw et al. 2020;Alvarez et al. 2023). Under these conditions Kronos-PI plants produce normal spikes and viable grains, whereas Kronos-PS plants undergo spike and stem elongation arrest, and spikes fail to emergence from the leaf sheaths (Pearce et al. 2017;Shaw et al. 2020). Similarly, the gi-2 mutants failed to head under SD before the experiment was terminated at 150 d ( Fig. 2A), but showed fully developed young spikes wrapped inside the leaf sheath (Fig. 2B).
Although it is not possible to determine heading time in Kronos-PS under SD, the final number of leaves can be used to estimate the timing of the transition from the vegetative shoot apical meristem (SAM) to the reproductive phase (flowering time). In two separate SD experiments, one performed under 8 h of light (SD) and the other under 10 h of light (SD-10h), Kronos-PS primary shoots produced 13.9 and 13.7 leaves in SD and SD-10h, respectively (including leaves still wrapped inside the sheaths). The gi-2-PS mutant produced ~1 more leaf than the Kronos-PS control under both conditions: SD (14.9 leaves) and SD-10h (14.5 leaves, Fig. 2C). However, when the same lines were grown under LD, the difference between wild type and gi-2 increased to 3.4 leaves, indicating a highly significant interaction (P< 0.001) between the effects of GI and photoperiod on leaf number (Supplemental Information S2).
This interaction between GI and photoperiod was no longer significant when we tested the effect of the gi-2 mutant on heading time in a Kronos-PI background (henceforth gi-2-PI). The gi-2-PI mutant showed similar delays in heading time relative to Kronos-PI under both LD (15.4 d) and SD conditions (16.5 d, Fig. 2D). A factorial ANOVA for heading time using these lines, showed highly significant effects for GI and photoperiod (P < 0.001), but no significant interaction between them (P = 0.4347, Supplemental Information S2).
To explore the interactions between GI and PPD1 under LD, we crossed Kronos-PS gi-2 with a loss-of-function ppd1 mutant (Pearce et al. 2017) and compared heading times among the four possible homozygous classes. A factorial ANOVA showed highly significant effects of PPD1 and GI on heading times (P < 0.001), but no significant interaction between the two genes ( Fig.   2E, P = 0.3807, Supplemental Information S2). We did not study these lines under SD because gi-2 and ppd1 mutants, as well as Kronos-PS controls do not head under SD. Taken together, 8 these results suggest that the PPD1 PS-allele is important for the GI differential flowering responses in LD and SD.
Lastly, we compared PPD1 expression profiles in Kronos-PS and the gi-2-PS mutant under LD ( Fig. 2F). In Kronos-PS, PPD1 transcripts showed diurnal rhythms with a peak expression 12 hours after the lights were turned on (zeitgeber time 12 = ZT12). In the gi-2-PS mutant, PPD1 transcript levels were significantly downregulated relative to Kronos-PS from the ZT12 expression peak to ZT20 (Fig. 2F). in the elf3 mutant background. In the F 2 progeny of a cross between gi-2 and elf3 (both in Kronos-PS), we selected homozygous elf3 and elf3 gi-2 mutants and evaluated them under LD and SD. Both GI and photoperiod showed highly significant effects on heading time (P < 0.001), however, no interaction was observed between them (P = 0.7608, Supplemental Information S3).
The delayed heading caused by the gi-2 mutant in the elf3 background was similar under LD (14.7 d) and SD (14.9 d, Fig. 3A), indicating that in the absence of a functional ELF3 the effect of GI is no longer modulated by photoperiod.
We further explored the interaction between GI and ELF3 in an additional LD experiment including Kronos-PS, gi-2, elf3, and elf3 gi-2 mutants in the same PS genetic background.
Consistent with previous findings (Alverez et al. 2016;2023), the elf3 allele was associated with a highly significant acceleration in heading time and a reduced number of leaves relative to Kronos-PS, but the differences were larger in the gi-2 than in the wild-type background ( Fig. 3B and C). By contrast, the gi-2 allele was associated with significant increases in leaf number and delayed heading, with larger differences in the presence of the functional Elf3 allele than in the elf3 mutant background ( Fig. 3B and C). These differential responses were reflected in highly significant interactions between GI and ELF3 for both traits (P < 0.001, Supplemental Information S3).

9
We next examined the diurnal expression pattern of FT1 transcripts in the same four genotypes.
RNA samples were extracted from the 4 th fully expanded leaves (L4, before the SAM transition to the reproductive phase) collected at 4 h intervals for a 24 h period. As expected, the highest transcript levels of FT1 were observed in the elf3 mutant (earliest heading) and the lowest in the gi-2 mutant (latest heading, Fig. 3D). Factorial ANOVAs at each ZT point revealed highly significant effects of ELF3 and GI and highly significant interactions on the regulation of FT1 (P < 0.001, Supplemental Information S3). These results suggest that the differences in heading Finally, because ELF3 was previously shown to physically interact with GI and modulate the cyclic accumulation of the GI protein in Arabidopsis , we explored the physical interactions between the wheat GI and ELF3 proteins. Indeed, we observed a strong interaction between the two wheat proteins in yeast two-hybrid (Y2H) assays (Fig. S2). The GI and ELF3 interactions at the protein level, together with the interactions at the transcriptional level, likely contribute to the observed genetic interactions between GI and ELF3 on heading time observed in this study.

Interactions between GI and the LD flowering repressor VRN2 contribute to the regulation of wheat heading time
To determine the relationship between GI and VRN2, we crossed gi-2 with a Kronos vrn2 lossof-function mutant (Distelfeld et al. 2009) and studied the effect of the four different homozygous genotypes on heading time. When grown under LD, the vrn2 mutant headed 5.3 d earlier than the wild type, and this effect increased to 12.4 d when the same comparison was performed in the gi-2 mutant background (Fig. 5A). On the contrary, the 21.6 d delay in heading time associated with the gi-2 mutant in plants homozygous for the functional Vrn2 allele was reduced to 14.5 d in plants homozygous for the vrn2 mutant allele (Fig. 5A). These differential effects were reflected in a highly significant interaction between GI and VRN2 on heading time (P < 0.001, Supplemental Information S5).
When we analyzed the same plants for total leaf number, we also found highly significant differences associated with VRN2 and GI (Fig. 5B), but the interaction was not significant (P = 0.2665, Supplemental Information S5). These results suggest that the interaction between GI and VRN2 has a limited effect on the initial transition of the SAM from the vegetative to the reproductive phase, but a stronger effect on the subsequent phase of spike development and stem elongation.
To characterize better the genetic interaction between GI and VRN2 on heading time, we determined the effect of the gi mutation on the VRN2 transcription profile by qRT-PCR (Fig. 5C).
In both the Kronos-PS control and the gi-2 mutant, VRN2 transcripts increased during the day to reach a peak at ZT12, with slightly lower levels in gi-2 at ZT0 and ZT8 (P< 0.05). (Fig. 5C).
However, after the lights were turned off, VRN2 transcript levels decreased rapidly in Kronos-PS but remained high in the gi-2 mutant (Fig. 5C), suggesting a role of GI in the downregulation of VRN2 at night.
Since VRN1 is a known repressor of VRN2 (Chen & Dubcovsky, 2012), we also analyzed the transcript levels of VRN1 using the same leaf samples. VRN1 transcript levels were significantly higher in the wild type than in the gi-2 mutant in most time points, but at ZT20 the two genotypes showed no significant differences (Fig. 5D). This result suggests that the role of GI in the downregulation of VRN2 during the night is not mediated by VRN1, but we cannot completely rule out the possibility of a residual effect of the higher transcript levels of VRN1 at dusk (ZT16) in Kronos-PS relative to gi-2.
In addition to their interactions at the transcriptional level, GI and VRN2 proteins show strong interaction in Y2H assays (Fig. S2). A similar interaction has been reported between the two orthologous rice proteins by Y2H and further validated by BiFC assays in Nicotiana benthamiana and by Co-IP experiments in rice protoplasts (Zheng et al. 2019). In summary, the observed interactions between GI and VRN2 at the protein and transcriptional levels likely contribute to their significant genetic interactions on heading time.

GI shows strong interactions with CO1 and CO2 on the regulation of heading time in wheat
The regulation of flowering time by GI in Arabidopsis is mainly mediated through the COdependent pathway (Sawa et al. 2007), so we studied the interactions between GI and the two CO paralogs present in wheat: CO1 and CO2 (Shaw et al. 2020). We first determined the effect of the Kronos-PS gi-2 mutant on the diurnal expression patterns of CO1 and CO2 in the 4 th leaves. The transcript levels of CO1 were significantly up-regulated in the gi-2 mutant relative to the Kronos-PS control at ZT0 and ZT4, but were mostly indistinguishable between the two genotypes at other time points (Fig. 6A). Transcript levels of CO2 were significantly higher in the gi-2 mutant relative to Kronos-PS at three (ZT4, ZT8 and ZT12) of the six analyzed time points (Fig. 6B).
Interestingly, GI also showed strong physical interactions with both CO1 and CO2 proteins in Y2H assays (Fig. S2). These results suggest that wheat GI may also regulate CO protein stability through protein interaction, as previously described in Arabidopsis GI (Song et al. 2014;Hwang et al. 2019). Interactions at both the transcriptional and protein levels may contribute to the genetic interaction between GI and CO1/CO2 on heading time described in the next section.
From the cross between gi-2 and co1co2 double mutant in the Kronos-PS background we generated the eight possible homozygous classes and studied their effect on heading time (all genotypes were grown in the same LD growth chamber). Among the four genotypes carrying the Gi wild-type allele, co1 flowered 3.8 d earlier, co2 2.7 d later and the co1co2 combined mutant 4.3 d earlier than Kronos-PS (Fig. 6C), similarly to what was reported in a previous study (Shaw et al. 2020). Among the four genotypes carrying the gi-2 mutation, the differences between the individual mutants and the wild type (co1 4.5 d earlier and co2 5.7 d later) were in the same direction as in the corresponding lines with the Gi allele (Fig. 6C). However, the co1 co2 combined mutant was 6.1 d later than the wild type in the gi-2 background and 4.3 d earlier in the Gi background (Fig. 6C). These opposite results were reflected in highly significant interactions among these three genes in a 3-way ANOVA (P < 0.001, Supplemental Information S6). The effect of GI on heading time was stronger in the presence of the co1 allele than in the presence of the Co1 allele, whereas the effect of CO1 was stronger in Gi than in gi-2 (Fig. 6C).
The effects of both co2 and gi-2 on heading time were stronger in the presence of the mutant allele of the other gene than in the presence of the wild-type allele (Fig. 6C).  (Fig. 7D), consistent with a previous study (Debernardi et al. 2022). In the gi mutant background, the acceleration of heading time by UBI-miR172 (6.3 d) was greater than in the wild type (2.3 d), and both were highly significant (Fig. 7D). On the contrary, the delayed heading associated with gi-2 was smaller in the UBI-miR172 transgenic plants (15.6 d) than in the non-transgenic sister lines (19.6 d, Fig. 7D). A factorial ANOVA for heading time revealed a 13 highly significant interaction (P < 0.001) between GI and UBI-miR172 (Supplemental Information S8). Taken together, these results confirmed that interactions between GI and the miR172-AP2L1 pathway contribute to the regulation of FT1 expression and to the GI effects on wheat heading time.

Integration of GI into a working model of the regulation of wheat heading time
To summarize the multiple pathways through which GI regulates wheat heading time, and its interactions with other flowering genes, we present a working model in Fig. 8, and a  This model shows that GI promotes FT1 expression and heading time through multiple pathways, including the grass-specific VRN2 gene, as well as through complex interactions with CO1/CO2 and the miR172-AP2 age-dependent pathway. The model also shows that the interactions between GI and photoperiod are dependent on the presence of a wild-type PPD1-PS allele. Some of the interactions are shown by bidirectional grey arrows because they are too complex to be summarized by a promotion arrow or repression T-bar. For example, VRN2 expression is upregulated in the gi-2 mutant at ZT20 but downregulated at ZT0-ZT8 (Fig. 5); and GI is downregulated in the elf3 mutant at ZT12 but upregulated at ZT20-ZT4 (Fig. 4).
This model presents only the strongest effects, but additional weaker effects are known to exist.
For example, PHYB affects heading time and expression of multiple flowering genes even in the absence of ELF3, whereas ELF3 affects heading time even in the absence of PPD1 (Alvarez et al.  15

The interaction between GI and photoperiod is disrupted in Kronos PI and the elf3 mutant
The misexpression of PPD1 in Kronos-PI also resulted in the elimination of the significant interaction on leaf number observed in Kronos-PS between GI and photoperiod (Fig. 2C). These results indicate that the photoperiod sensitive PPD1 allele is critical for conveying the photoperiodic information to GI and for generating its differential flowering response in LD and SD (Fig. 2C).
This hypothesis is further supported by the absence of a significant interaction between GI and  (Fig. 3A). These results indicate that the differential effect of GI on heading time in different photoperiods is also dependent on a functional ELF3.
The altered PPD1 expression in elf3 may contribute to the lost interaction between GI and photoperiod, but additional post-translational mechanisms cannot be ruled out. In this study, we show that GI and ELF3 physically interact with each other by Y2H assay (Fig. S2), an interaction that has been also observed in Arabidopsis . In Arabidopsis, GI protein stability and photoperiodic flowering is also regulated by the interaction between ELF3 and COP1 . ELF3 is essential for the interaction between GI and COP1, and the Arabidopsis elf3 mutant exhibits a disruption in the cyclic accumulation of GI associated with early flowering and photoperiod-insensitivity (Zagotta et al. 1996;Yu et al. 2008). Similar protein interactions may also contribute to the disrupted photoperiodic response of GI in the elf3 wheat mutant background.

GI interacts with the grass-specific VRN2 repressor to regulate wheat heading time
The effects of VRN2 on heading time reported in this study are smaller than those reported in previous studies in winter wheat (Yan et al. 2004;Distelfeld et al. 2009). In winter cereals, VRN2 is a LD flowering repressor that prevents FT1 expression during the fall, and is repressed in the spring by VRN1, which is upregulated by vernalization (Chen & Dubcovsky, 2012). However, in spring wheats, such as the Kronos variety used in this study, VRN1 induction does not require vernalization and the earlier down-regulation of VRN2 results in smaller effects of VRN2 on heading time in spring than in winter wheat varieties.
In this study, we show that the effects of VRN2 on heading time are modulated by its interactions with GI (Fig. 5A-B) both at the transcriptional and protein levels. In the gi-2 mutant, the expression levels of VRN2 vary during the day, with significant increases at ZT20 and significant decreases between ZT0 and ZT8 (Fig. 5C). These diurnal changes likely reflect GI's own circadian oscillation (Fig. 4A) and/or the effects of the gi-2 mutants on the central oscillators ( Fig. 4B-C). GI and VRN2 also interact at the protein level (Fig. S2), a result that has been validated for the rice orthologs (Zheng et al. 2019). A recent study in rice has shown that coexpression of GI and GHD7 causes reduced accumulation of the GHD7 protein, and that GI and phytochromes function antagonistically in the regulation of GHD7 protein stability (Zheng et al. 2019).
The existence of a similar negative effect of GI on the stability of the VRN2 protein in wheat is indirectly supported by the stronger effect of VRN2 on heading time in the gi-2 mutant than in the presence of the functional Gi allele (Fig. 5A-B). This significant genetic interaction between GI and VRN2 on heading time also resulted in larger differences between the gi-2 mutant and the wild type in the presence of the functional Vrn2 allele than in the vrn2 mutant. This result indicates that part of the effect of GI in the acceleration of heading time in wheat is mediated by its interactions with VRN2.
In summary, these results show that in the grasses, GI interacts with VRN2, a gene that has no homologs in the eudicots plants. It will be interesting to determine if these genetic interactions are the result of direct interactions between these two genes or an indirect effect of the multiple interactions between GI and other flowering genes.

The conserved GI-CO pathway contributes to the regulation of wheat heading time
The mechanistic characterization of GI's role in photoperiodic flowering regulation is mostly from Arabidopsis, where GI promotes CO transcription and acts as a major mediator between the circadian clock and CO. Under LD, the peak of GI protein accumulation coincides with FKF1 in the middle of the day, resulting in the repression of CDF1 and, consequently, in increased transcript levels of CO and FT (Sawa et al. 2007). Under SD, GI accumulation peaks three hours earlier than that of FKF1, precluding the repression of CDF1, and resulting in reduced levels of CO transcripts (Sawa et al. 2007). Therefore, the formation of the GI-FKF1 complex is required for day-length measurement in Arabidopsis. In addition, GI can regulate CO temporal stability through direct physical interactions and through its interactions with FKF1 and ZTL (Song et al. 2014;Hwang et al. 2019).
In this study, we show a direct physical interaction between GI and both CO1 and CO2 wheat proteins in Y2H assays (Fig. S2). A similar interaction between GI and CO has been reported in  (Fig. 6A-B). Therefore, although the presence of the GI-CO pathway seems to be conserved in Arabidopsis and wheat, the mechanisms by which they control photoperiodic flowering likely differ between these two species.
These differences are also reflected in the more limited role of CO1 and CO2 in the regulation of the photoperiodic response in the temperate grasses, where PPD1 is the main photoperiod gene

GI can regulate wheat heading time by CO-independent regulatory pathways
In addition to the CO-mediated pathway, we show that in wheat GI can also regulate FT1 expression through the age-dependent miR172-AP2 module. The interaction between GI and miR172-AP2 seems to be conserved between Arabidopsis and wheat, with GI promoting miR172 and repressing AP2 expression in both species (Jung et al. 2007; Fig. 7).
In Arabidopsis, GI has been shown to promote FT transcription through direct binding to its promoter regions (Sawa & Kay, 2011). However, as SVP, TEM1 and TEM2 also bind to the same FT promoter regions, it is also possible that the binding of GI to the FT promoter is mediated by a complex with other proteins (Sawa & Kay, 2011). In wheat, we also observed a strong downregulation of FT1 in the gi-2 mutant (Fig. 7C), which was observed even in the presence of the elf3 mutation (Fig. 3D). Although these results may reflect a direct interaction between GI and the FT1 promoter, we cannot rule out indirect effects mediated by the multiple pathways through which GI regulates FT1.

Plant materials
The reference sequences of GI (TraesCS3A02G116300 and TraesCS3B02G135400)  To test genetic interactions between GI and ELF3, CO1, CO2, PPD1 and VRN2, we combined gi-2 with the respective loss-of-function mutants of these genes and generated elf3 gi, co1 gi, co2 gi, co1 co2 gi, ppd1 gi and vrn2 gi in the same Kronos-PS background. Primers used to genotype and verify the presence of individual gene mutations are listed in Table S1. We also crossed the gi-2 mutant with a Ubi-miR172 transgenic line (Debernardi et al. 2022) to investigate the interaction between miR172 and GI in the Kronos-PS background. Primers used to differentiate the Ppd-A1b allele (PS) from Ppd-A1a allele (PI) are also included in Table S1.

Growth conditions
All experiments were performed in controlled environment conditions using the Conviron PGR15 growth chambers. During the lights-on period, the growth chambers were set at 22°C, but the first and last hour of this lights-on period were set at 20°C to provide a more gradual change between temperatures. Night temperatures were set at 17°C. All PGR15 chambers used similar metal halide and high-pressure sodium light configurations, and lights were set to the same intensity in all experiments (~330µM m -2 s -1 ). Lights were on for 8 hours for SD-8h, 10 hours for SD-10h experiments and 16 hours for LD experiments.  Table S2.  Table S2. ACTIN was used as endogenous control except for miR172, the small nucleolar RNA 101 (SnoR) was used as internal reference. Primers used in qRT-PCR analyses are included in Table S2.

Acknowledgments and Funding
This project was supported by the Howard Hughes Medical Institute (

Conflict of interest
The authors declare no conflict of interest

Data and materials availability
All Kronos mutants and transgenic lines are available from the authors upon requests without any restrictions for use. Kronos single mutants are also available from the Germplasm Resources Unit (GRU) at the John Innes Centre.   Table S1. Primers used for mutant genotyping. Table S2. Primers used in the qRT-PCR analysis.

Supplemental Materials
Supplemental Information S1-S8: raw data and statistics supporting figures.  for differences with the wild-type control using two-tailed t-test in B, E and G. Raw data and statistical analyses are available in Supplemental Data S1.