Conserved eukaryotic factors XCT and COP1 work together to control circadian clock function and reproductive timing in plants

Mutations that affect the circadian clock often influence photoperiodic regulation of the transition from vegetative to reproductive growth; for example, short-period mutants often flower early in short days but have wild-type flowering phenotypes in long days^40–42. We therefore investigated whether mutation of XCT also affects flowering time regulation. Surprisingly, xct-2 mutants flower early in long days but do not have a phenotype in short-day conditions (Supplementary Fig. 1). This suggests that this flowering phenotype is not caused by the fast pace of the circadian clock in xct-2^43,44.

Other repressors of flowering previously reported to act downstream of CRY2 and upstream of CO and GI include the E3 ubiquitin ligase COP1 and the scaffold protein ELF3²³. We therefore examined the genetic interactions between XCT, COP1, and ELF3. Since null alleles of COP1 are lethal⁴⁵, we used the reduction-of-function allele cop1-4 for these studies. We found that in long days, xct-2, cop1-4, and elf3-1 single mutants all flowered significantly earlier than wild-type plants (Fig. 1b). xct-2 cop1-4, xct-2 elf3-1, and cop1-4 elf3-1 double mutants produced the same number of leaves as the cop1-4 and elf3-1 single mutants before transitioning to flowering (Fig. 1b). Although we cannot draw definitive conclusions from this epistasis analysis given that cop1-4 is not a null allele, the lack of an additive phenotype in these three types of double mutants suggests that XCT, COP1, and ELF3 function in the same pathway. Notably, a variety of light signaling and flowering time mutants have been reported to transition to flowering at an even earlier stage of development than elf3-1 and cop1-4^46,47, suggesting that the lack of additivity in the double mutants is biologically relevant. Together, our genetic analysis suggests that XCT acts near COP1 and ELF3 to directly or indirectly repress the floral-promoting functions of GI and CO.

CO is a key regulator of flowering that promotes expression of the florigen gene FT (Fig. 1e). CO expression is controlled by both transcriptional and post-transcriptional mechanisms; for example, GI promotes transcription of the CO gene while COP1 and ELF3 promote degradation of CO protein^19,22,25. To help determine where in the flowering time pathway XCT functions, we used quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR) analysis to examine expression levels of CO and FT, as well as CRY2 and COP1, in cry2-1, xct-2, and xct-2 cry2-1 mutants. We found expression of CRY2 was not significantly different between wild-type and xct-2 plants and that CO and COP1 mRNA levels were not significantly different in the four genotypes tested (Fig. 1c). However, and as expected based on previous studies⁴⁸, FT levels were significantly reduced in cry2-1 plants grown in long days (Fig. 1c). Consistent with the early-flowering phenotype of xct-2 mutants, FT levels were significantly elevated in these plants in the late afternoon and at the end of the day (Fig. 1c). FT levels were indistinguishable from wild type in xct-2 cry2-1 mutants, indicating that loss of the negative regulator XCT can compensate for mutation of CRY2 in the regulation of florigen expression.

To further investigate the relationship between COP1 and XCT in control of flowering time, we next examined FT mRNA levels in wild-type, xct-2, cop1-4, and xct-2 cop1-4 mutants grown in long-day conditions (Fig. 1d). We found that FT levels were elevated in cop1-4 mutants at all time points tested, with significant peaks at both ZT4 and ZT16. Similar to our previous experiment, FT levels were most increased in xct-2 mutants at ZT12 and ZT16 (at the ZT16 time point, FT expression in all three mutant genotypes was higher than in wild type with p < 0.1), with no upregulation of FT at ZT4 or ZT20. These data are consistent with a previous study demonstrating significant upregulation of FT in cop1 mutants grown in long days, likely due to a role for COP1 in a phyA-mediated signaling pathway^19,22,25. FT expression in the xct-2 cop1-4 double mutants was indistinguishable from that in cop1-4 mutants, even in the middle and late daytime when FT mRNA levels were elevated in both cop1-4 and xct-2 single mutants. These gene expression data suggest that XCT and COP1 influence flowering by a similar mechanism during part of the diel cycle but that COP1 has additional, XCT-independent roles in the control of flowering during the late night and early day. Altogether, our genetic and gene expression data suggest that XCT may act near COP1 and ELF3 to modulate CO protein function or stability (Fig. 1e).

In addition to affecting the pace of the circadian clock, mutation of circadian genes often affects the rhythmic robustness of the circadian oscillator. This can be quantified by assessing the relative amplitude error (RAE) associated with luciferase activity rhythms, with an RAE of 0 indicating highly rhythmic expression patterns and an RAE of 1 indicating no statistically significant rhythmic element is present. For example, null mutations in either ELF3 or ELF4, another component of the Evening Complex (EC), result in very high RAE values (Fig. 2b), indicating these plants are nearly arrhythmic^51,52. We therefore examined genetic interactions between circadian clock gene mutants and xct-2 and cop1-4 on rhythmic robustness. Mutation of XCT or COP1 caused modest but statistically significant increases in RAE values. The xct-2 elf3-1 and xct-2 elf4-1 plants had similarly high RAE values as the elf3-1 and elf4-1 single mutants (Fig. 2b). This non-additivity is consistent with XCT acting in a similar position within the oscillator as these EC components.

GI is another clock component reported to interact with both COP1 and ELF3²³. We found that a null mutation in GI caused a significant reduction in rhythmicity and decreased luminescence levels relative to wild type (Fig. 2b, c), consistent with previous reports^53,54. To our surprise, we found that combining this gi-2 mutation with either xct-2 or cop1-4 significantly increased the robustness of CCR2::LUC+ rhythms and overall luminescence when compared to the gi-2 single mutant (Fig. 2b, d, Supplementary Fig. 2a–c). This demonstrates that XCT and COP1 are epistatic to GI in control of rhythmic robustness. xct-2 and cop1-4 were also largely epistatic to gi-2 in their effects on free-running circadian period (Supplementary Fig. 2a). The improved rhythmicity in xct-2 gi-2 and cop1-4 gi-2 relative to the null allele gi-2⁵³ suggests that XCT and COP1 may act antagonistically to GI in the regulation of a shared circadian regulatory component or process. Overall, these genetic data suggest that XCT and COP1 act in a similar manner to affect the circadian oscillator.

COP1 has previously been shown to regulate flowering time and circadian clock function via a physical interaction with ELF3, which leads to a COP1-dependent decrease in ELF3 protein abundance^23,55. Given the physical interaction between COP1 and XCT, we speculated that XCT might also modulate ELF3 protein levels. To test this, we first introduced a transgene with the Arabidopsis ELF3 promoter controlling transcription of a translational fusion between ELF3 and luciferase (ELF3::ELF3-LUC)⁵⁶ from wild type into cop1-4 and xct-2 plants by crossing. In light/dark cycles, wild-type plants displayed a complex pattern of luciferase activity that in constant light conditions resolved into a circadian pattern with peak expression during the subjective night (Fig. 3b). cop1-4 mutants displayed a similar overall pattern of luciferase activity, but with substantially higher luminescence than wild type. These results are consistent with previous reports that COP1 mediates proteasome-dependent degradation of ELF3^23,55. xct-2 mutants expressed significantly higher levels of ELF3-LUC than wild type and slightly more than cop1-4. This suggests that, like COP1, XCT may negatively regulate ELF3 protein levels.

To test the effects of the xct-2 mutation on the abundance of native ELF3 protein, we next performed immunoblotting experiments with anti-ELF3 antibody on nuclear extracts prepared from plants grown in 12 h light/12 h dark cycles. We found higher levels of ELF3 protein in xct-2 than in Col, particularly in the late afternoon and the first part of the night (Fig. 3c). Cosinor analysis of ELF3 protein abundance with CircaCompare⁵⁷ revealed a greater than two-fold higher amplitude of ELF3 protein level oscillations in xct-2 than in control plants (Fig. 3d; Supplementary Data 1). We next investigated whether this difference might be due to increased ELF3 transcript levels in xct-2 mutants. qRT-PCR analysis of ELF3 mRNA levels in plants maintained in light/dark cycles revealed a phase advance in the accumulation of ELF3 transcript in xct-2 mutants relative to wild type (Fig. 3e), presumably due to the short-period phenotype of these plants³³. However, we did not detect a significant difference in the amplitude of ELF3 mRNA level oscillations in xct-2 and Col (Supplementary Data 1). Together, these results indicate that XCT negatively regulates ELF3 protein abundance, most likely by collaborating with COP1 to promote ELF3 degradation.

We next examined genetic interactions between ELF3-OX, COP1, and XCT on the regulation of circadian clock pace. We found that the long-period phenotype of ELF3-OX plants⁵⁹ was partially suppressed by the cop1-4 mutation, with cop1-4 ELF3-OX plants having a period phenotype indistinguishable from wild type and intermediate between cop1-4 and ELF3-OX mutants (Fig. 4b). The suppressive effects of xct-2 were stronger, with xct-2 ELF3-OX plants having a short-period phenotype indistinguishable from xct-2 single mutants (Fig. 4b). This epistasis indicates that XCT and to a lesser extent COP1 are required for ELF3 control of clock pace.

ELF3 and other EC components directly negatively regulate the expression of key clock genes such as PRR7, PRR9, and LUX^12,13,60. To investigate whether XCT might affect EC target gene expression in an environmentally-responsive manner, we introduced a PRR9::LUC2 reporter gene into xct-2 plants by crossing with wild type. This transgene confers early day-phased rhythms in luminescence in plants, consistent with the accumulation patterns of PRR9 mRNA⁶¹. In 12 h light: 12 h dark cycles, peak levels of PRR9::LUC2 activity were much higher in xct-2 than in control plants (Fig. 4c). However, after 24 h in constant light conditions, peak levels of luminescence were very similar between the two types of seedlings. qRT-PCR experiments also revealed higher peak levels of PRR9 expression in xct-2 compared to controls in plant grown in light/dark conditions (Supplementary Fig. 3). These data are consistent with our previous finding that PRR7 and PRR9 levels are elevated in xct-2 plants maintained in short-day conditions but not in constant light³³ and suggest the ability of the EC to repress gene expression may be reduced in the absence of XCT.

To investigate possible ways in which the ability of ELF3 to control gene expression might be altered in xct mutants, we carried out chromatin immunoprecipitation (ChIP) experiments. A transgene encoding epitope-tagged ELF3 expressed under the control of the 35S promoter was introduced into xct-2 by crossing with wild type. Plants were grown in 12 h light: 12 h dark cycles, and samples were harvested shortly before dawn. Immunoprecipitations were carried out using anti-Myc antibodies, with non-transgenic Col seedlings serving as a negative control. The abundance of genomic DNA in regions of the PRR7, PRR9, and LUX promoters known to serve as LUX binding sites (LBS) as well as regions within the coding sequences of these genes not expected to associate with the EC (Fig. 4d) was assessed by qRT-PCR. We found that regions containing known EC binding sites were considerably more abundant in both wild-type and xct-2 plants expressing Myc-tagged ELF3 than in non-transgenic controls (Fig. 4e). However, in xct-2 plants, the enrichment of PRR9 and PRR7 promoter sequences in ELF3 chromatin immunoprecipitations was significantly lower than in plants expressing functional XCT (Fig. 4e, Supplementary Data 2). Given that ELF3 protein levels are actually elevated in xct-2 relative to wild type (Fig. 3c, d), these data indicate that XCT promotes ELF3 association with chromatin despite its negative effects on ELF3 protein levels.

All plants used are in the Columbia (Col-0) background. The following genotypes have been previously described: co-9⁸⁰; cop1-4⁴⁵; cry1-301⁸¹; cry2-1²⁴; elf3-1⁵⁹; elf4-1⁵² (this mutation was originally isolated in the Ws accession but was introgressed into Col-0 by four generations of backcrossing); gi-2^82,83; xct-2³³; and ztl-103⁸⁴. The CCR2::LUC+, PRR9::LUC2, and ELF3::ELF3-LUC reporters have been previously described^41,56,85. Myc-tagged, ELF3-overexpressing plants were generated as follows. The ELF3 cDNA was amplified (see primer sequences in Supplementary Data 3) and cloned into the pEarleyGate 203 transformation vector⁸⁶ using LR Clonase reaction mix (Invitrogen). Transformation of Col-0 plants was achieved using the floral dip method⁸⁷ and transgenic T1 seedlings selected using the herbicide Basta. Late-flowering lines were selected, and ELF3 overexpression was verified by immunoblotting. Double-mutant genotypes were generated by crossing. All transgenes were introduced into mutant backgrounds by crossing with the corresponding wild-type controls.

Seeds were surface-sterilized with chlorine gas and stratified in the dark for 2–4 days at 4 °C. For luciferase imaging, seeds were plated on 1× Murashige and Skoog media, 0.7% agar, and 3% sucrose. Seedlings were entrained in light-dark cycles (12 h light, 12 h dark) with 50–60 μmol m⁻² s⁻¹ cool white fluorescent light at 22 °C for 6 days. For chromatin immunoprecipitation experiments, seeds were plated on 1× Murashige and Skoog media, 0.7% agar, 1% sucrose and grown in light–dark cycles (12 h light, 12 h dark) with 50–60 μmol m⁻² s⁻¹ cool white fluorescent light at 22 °C for 12 days and then at 17 °C for 2 additional days. Plants were harvested at ZT23.5.

For flowering time, immunoblotting, and quantitative PCR experiments, seeds were sown directly in soil, and plants were grown under light–dark cycles of the specified photoperiod with 150–200 μmol m⁻² s⁻¹ cool white fluorescent light at 22 °C. For immunoblotting and quantitative PCR, plants were collected at the indicated times and immediately frozen in liquid nitrogen. For flowering assays, rosette leaves were counted when inflorescence stems were 1 cm long.

Sample preparation and qRT-PCR were performed as previously described⁸⁸ using a BioRad CFX96 thermocycler (Bio-Rad Laboratories). Relative expression and SEM values were obtained from the BioRad CFX96 software package. Primer sequences are in Supplementary Data 3.

ChIP assays were performed primarily as previously described⁸⁹. Briefly, Arabidopsis seedlings were harvested and fixed with 1% formaldehyde under vacuum for 10 min. For each immunoprecipitation, nuclei were isolated from 2.5 g of seedling tissue, lysed, and then chromatin was sheared into fragments with an average length of 100–500 bp using a Covaris E220 sonicator (Covaris). Chromatin was incubated with 4.5 μg of mouse monoclonal anti-Myc antibody (9E 10; Developmental Studies Hybridoma Bank) overnight at 4 °C. Ten micrograms of rabbit anti-mouse secondary antibody (#12-488, Millipore Sigma) and protein A Dynabeads (#10001, ThermoFisher) were then added. Input and purified, immunoprecipitated DNA were quantified by real-time PCR. The sequences of the primers used are listed in Supplementary Data 3.

For the co-immunoprecipitation experiments, XCT-YFP-HA and Myc-COP1 were transiently expressed in Nicotiana benthamiana. The XCT-YFP-HA plasmid has been previously described³³; the Myc-COP1 plasmid was generated by amplifying the COP1 cDNA (see primer sequences in Supplementary Data 3) and cloning it into the pEarleyGate 203 transformation vector⁸⁶ using LR clonase reaction mix (Invitrogen). Leaf infiltrations were carried out largely as described⁹⁰ except that Agrobacterium tumefaciens GV3101 strain was diluted to an OD = 0.1 in 10 mM MgCl₂ for infiltration and tissue was harvested 3 days after infiltration. Extraction of nuclear proteins from Nicotiana benthamiana and Arabidopsis and anti-HA immunoprecipitations were carried out as previously described³⁶. Nuclear lysates or immunoprecipitated samples were analyzed by immunoblotting. The primary antibodies used were: rabbit anti-ELF3¹⁴; mouse anti-Myc (#ZMS1032, Millipore Sigma); mouse anti-HA conjugated to HRP (#12013819001, Millipore Sigma); and mouse anti-actin (#A0480, Millipore Sigma). The secondary antibodies used were: goat anti-mouse HRP (#31430, ThermoFisher Scientific) and goat anti-rabbit HRP (#31460, ThermoFisher Scientific). HRP activity was determined using SuperSignal West Pico chemiluminescent substrate (#34580, ThermoFisher Scientific) and protein abundance quantified using Image Lab software (Bio-Rad Laboratories).

After entrainment in light-dark cycles, seedlings were sprayed with 3 mM D-luciferin (Gold Biotechnology) and moved to the specified light conditions using red and blue LEDs (35 μmol m⁻² s⁻¹ each color; XtremeLUX, Santa Clara, CA) to provide illumination. Plants were imaged for 5–6 days using a cooled CCD camera (DU434-BV, Andor Technology; or ORCA-R2, Hamamatsu Photonics). Bioluminescence was quantified using MetaMorph software (Molecular Devices) and circadian rhythmicity assessed using Biological Rhythm Analysis Software System (BRASS)⁹¹.

Protein sequences of the Arabidopsis thaliana, Homo sapiens, Chlamydomonas reinhardtii, and Schizosaccharomyces pombe XCT homologs were obtained from Uniprot (accession numbers Q8H110, Q14320, A0A159YK42, and Q7LKZ5). Protein alignments were generated using Clustal Omega⁹² and viewed and edited using Jalview 2⁹³. Predictions of disordered domains were obtained from the Database of Disordered Protein Predictions (d2p2)⁶⁷. d2p2 uses nine predictor algorithms (VLXT, VSL2b, PrDOS, PV2, IUPred-S, IUPred-L, Espritz-N, Espritz-X, and Espritz-D^94–99) and requires that 75% of the programs agree before defining a region as disordered.