The wild-type (WT) and mutant Arabidopsis plants used in this study were of the Columbia-0 (Col-0) background. The lfr-1 and lfr-2/+ mutants are described elsewhere (Wang et al., 2009). ice1-2 (SALK_003155; Kanaoka et al., 2008) was a gift from Prof. Shuhua Yang’s laboratory. Seeds were surface-sterilized in 75% ethanol and 1% NaClO, washed three times with sterile water, plated on half-strength Murashige and Skoog (1/2 MS) medium with 1% sucrose and 0.36% agar, and stratified at 4°C in the dark for 3 days. All plants were grown under long-day conditions (16 h of light/8 h of dark) at 22°C.

To map the domains of ICE1 (AT3G26744) that are involved in its interaction with LFR, full-length and truncated versions of ICE1 were fused to the GAL4 binding domain (BD) in pGBKT7. Meanwhile, LFR was fused to the GAL4 activation domain (AD) in pGADT7 (Lin et al., 2018). For our Bimolecular Fluorescence Complementation (BiFC) and firefly luciferase (LUC) complementation imaging (LCI) experiments, the full-length coding sequence (CDS) of ICE1 and LFR was amplified by PCR from Col-0 cDNA, cloned into pENTRY/D/SD-TOPO (Takara Bio Inc., Shiga, Japan), and then fused with the C-terminal half of cyan fluorescent protein (CFP) (cCFP), the C-terminal half of LUC (cLUC), the N-terminal half of yellow fluorescent protein (YFP) (nYFP), or the N-terminal half of LUC (nLUC). The full-length CDS of LFR was fused to nYFP (Lin et al., 2018).

To obtain the construct of pLFR : LFR-enhancedYFP (eYFP) and pLFR : LFR-3FLAG, the native promoter of LFR (1600 bp), the genomic sequence of LFR, and the coding sequences of eYFP and 3×FLAG were amplified and digested using appropriate restriction endonucleases and subsequently cloned into pCAMBIA1300. After being verified by DNA sequencing, the plasmid was introduced into lfr-1 and/or lfr-2/+ by Agrobacterium tumefaciens-mediated floral dip transformation (Clough and Bent, 1998). The homologous transgenic plants of pLFR : LFR-eYFP in lfr-2 background (pLFR : LFR-eYFP/lfr-2) and pLFR : LFR-3FLAG in lfr-1 and lfr-2 background (pLFR : LFR-3FLAG/lfr-1 and pLFR : LFR-3FLAG/lfr-2) were identified in the T4 generation and used for further phenotypic observation.

Total RNA was extracted from WT and mutant plants using Trizol reagent (Takara). First-strand cDNA was prepared from 500 ng of total RNA using a PrimeScipt RT Reagent Kit (Takara) and quantified using a CFX96™ Connect Real-Time PCR System (Bio-Rad, Hercules, CA, USA) with a SYBR Green Kit (Takara). The eIF4A1 gene was used as an internal control. The primers used are listed in Supplementary Table S1 .

A freezing tolerance assay was performed as described previously (Kim et al., 2015). Approximately 36 plants grown on 1/2 MS medium for 10 days were subjected to freezing conditions; the temperature was set to decrease 1°C/h until it reached -6°C (seedlings did not undergo cold acclimation before freezing) or -9°C (seedlings underwent cold acclimation at 4°C for 4 days before freezing). The plants were then incubated at 4°C overnight in the dark. Next, the plants were allowed to recover for 2 days at 22°C.

The chlorophyll content measurement was performed as described previously (Arnon, 1949; Mostofa et al., 2015). After freezing treatment, the seedlings were allowed to recover for three days at 22°C. Approximately 0.5 g of seedlings from a pool of plants from different plates were dipped in 50 mL aqueous acetone (80%) at 4°C in the dark for 48 h. The light absorption by acetone extracts of chlorophyll at 663 nm and 645 nm are detected in a spectrophotometer. The chlorophyll content was calculated as (20.2 × D645 + 8.02 × D663) × V/(1000 × W); where V = volume of 80% (v/v) acetone (mL), W = weight of sample (g).

Y2H assays were carried out using the yeast GAL4 system following a standard protocol (Clontech Laboratories, Inc., Mountain View, CA, USA). To verify its interaction with LFR, truncated versions of ICE1 [ICE1-2 (the C-terminal region of ICE1, corresponding to amino acids 298–494) and ICE1-3 (the C-terminal region of ICE1, corresponding to amino acids 358–494)] were cloned into pGBKT7 and LFR was fused to the AD in pGADT7. Following verification of the constructs by DNA sequencing, pairs of the plasmids were co-transformed into yeast strain AH109, followed by incubation at 30°C for 3 days. The co-transformed yeast cells were grown on solid medium lacking leucine (Leu) and tryptophan (Trp) (SD/-L-W). Only when the two proteins interacted with each other could the co-transformed yeast cells grow on solid medium lacking Leu, Trp, histidine (His), and adenine (Ade) (SD/-L-W-H-A).

BiFC and LCI assays were performed using transiently transformed tobacco leaves (Chen et al., 2008; Ou et al., 2011). For the BiFC assay, cCFP-ICE1 and nYFP-LFR constructs were co-transformed and co-expressed for about 48 h in tobacco epidermal cells. Fluorescent green fluorescent protein (GFP) (excitation 488 nm; emission 500–550 nm) and 4’,6-diamidino-2-phenylindole (DAPI) (excitation 405 nm; emission 430–480) signals were then detected using a Zeiss 710 Confocal Microscope (Carl Zeiss AG, Jena, Germany). For the LCI assay, tobacco leaves co-expressing the cLUC-ICE1 and LFR-nLUC constructs were incubated at 22°C for about 48 h. Then the luciferase substrate (1 mM D-luciferin) was sprayed onto the surface of the leaves and images were taken using a low-light cooled imaging apparatus (Fusion FX7; Vilber, Marne-la-Valle, France). The LUC activity was analyzed at 48 hpi using the NEWTONZ.0 Bio.Plant Imaging System (Vilber Bio lmaging, France).

ChIP was performed as described previously (Yamaguchi et al., 2014). Briefly, 600 mg of 8-day-old seedlings were cross-linked using 1% formaldehyde; the reaction was stopped by the addition of 0.125 M glycine. The cross-linked chromatin was then sheared to an average size of ~500 bp, and 20 μL of the supernatant was used as the input control. Anti-FLAG antibodies (F3165; Sigma-Aldrich, St. Louis, MO, USA) were added to the remaining chromatin for immunoprecipitation followed by qPCR. Data are presented as the fold change relative to the wild type. The primers used are shown in Supplementary Table S1 .

Primary roots from 8-day-old pLFR : LFR-enhancedYFP (eYFP)/lfr-2 transgenic seedlings with or without cold treatment were stained with 10 μg/ml of propidium iodide (PI) for 5 min and then imaged using an Olympus FV 3000 Confocal Microscope (Olympus Corp., Tokyo, Japan) with the following settings: equal laser power, gain, and pinhole. Line scan analyses of the YFP and PI signals in the root tip were measured using Image J. For the relative eYFP fluorescence signals quantification, all the eYFP fluorescence signals in each picture were quantified by image J. Twenty roots were measured for each genotype. The signals for cold-treated roots are analyzed relative to the average fluorescence intensity of at time 0 h. Data were analyzed using Prism 8.0.

To determine the function of the SWI/SNF component LFR in freezing tolerance, we evaluate the freezing tolerance phenotype of 10-day-old wild-type (WT, Col-0) and lfr-1 seedlings grown on 1/2 MS medium under non-acclimated (NA) or cold-acclimated (CA) conditions. The NA seedlings were subjected to freezing stress at -6°C for 1 h. Under this condition, the lfr-1 mutant showed reduced freezing tolerance with a significantly lower survival rate and chlorophyll content compared to WT after freezing treatment ( Figures 1A–C ). This defect was successfully rescued in pLFR : LFR-3FLAG/lfr-1 transgenic lines, which showed no obvious difference compared to the WT control ( Figures 1A–C ). The CA seedlings were acclimated at 4°C for 4 days and then subjected to freezing stress at -9°C for 1 h. The lfr-1 mutant also displayed freezing-sensitive defects under CA conditions ( Figures 1D–F ), and the defects were complemented in pLFR : LFR-3FLAG/lfr-1 transgenic lines ( Figures 1D–F ). Thus, LFR is essential for freezing tolerance at the seedling stage under both CA and NA conditions.

To confirm the function of LFR in freezing tolerance, we subjected 3-week-old lfr-1 and lfr-2 (a strong sterile allele of LFR, segregated from lfr-2+/-) mutant plants to freezing stress in soil. The mutants were subjected to freezing stress at -4 or -5°C for 1 h and then returned to a growth chamber set at 22°C for 7 days. In terms of survival, lfr-1 and lfr-2 both displayed a freezing-sensitive phenotype ( Figures 1G–J ).

Taken together, these results indicate that LFR is a positive regulator of freezing tolerance in Arabidopsis.

ICE1, CBFs, and downstream COR genes are important players in the response of plants to cold stress; thus, we next assessed the transcript levels of ICE1, CBF1, CBF2, CBF3, and COR15A in plants following low-temperature (4°C) treatment for 3 hours (3 h). Real time RT-qPCR showed that the expression levels of CBF1, CBF2, CBF3, and COR15A were significantly increased by cold treatment ( Figures 2A–D ). However, the expression levels of CBF1, CBF2, and CBF3 were significantly reduced in lfr-1 compared to Col-0 after cold treatment ( Figures 2A–C ). Similarly, the transcript level of COR15A was lower in lfr-1 than in Col-0 after low-temperature treatment ( Figure 2D ). We also tested the expression levels of these four genes in pLFR : LFR-3FLAG/lfr-1 complemented transgenic line (2-2), and we found that their expression levels were fully or partly rescued in the lfr-1 mutant ( Figures 2A–D ). We also examined the expression level of ICE1, and found that it was not regulated by cold stress as reported previously ( Figure 2E ; Chinnusamy et al., 2003; Shi et al., 2018) ICE1 was downregulated in lfr-1 following cold treatment, and the downregulation of ICE1 was partially rescued in complemented transgenic lines ( Figure 2E ). These results suggest that LFR is involved in cold acclimation via the direct or indirect transcriptional regulation of ICE1, CBFs, and COR15A.

To uncover the molecular mechanism underlying the role of LFR in cold acclimation, we examined Yeast two-hybrid (Y2H) data from a previous study (Lin et al., 2021). Interestingly, we noticed that ICE1 was isolated as an interacting partner of LFR. We confirmed this interaction by Y2H, Bimolecular Fluorescence Complementation (BiFC), and firefly luciferase (LUC) complementation imaging (LCI) assays. Because full-length LFR and ICE1 fused to the GAL4 BD showed obvious autoactivation in yeast (Ohta et al., 2018; Lin et al., 2021), truncated ICE1-2 (including the bHLH and ZIP domains) and ICE1-3 (including the ZIP domain) were fused to the BD to produce BD-ICE1-2 and BD-ICE1-3 ( Figure 3A ). We then co-transformed yeast strain AH109 with the AD-LFR and BD-ICE1-2 or BD-ICE1-3 plasmids. Yeast colonies co-transformed with AD-LFR/BD-ICE1-2 or AD-LFR/BD-ICE1-3 grew well on selective medium (SD/-L-W-H-A) ( Figure 3B ). However, negative control cells did not grow on the selective medium under the same conditions ( Figure 3B ). Thus, ICE1 probably interacts with LFR through its C-terminal ZIP domain.

We next conducted a BiFC assay using transiently transformed tobacco (Nicotiana benthamiana) leaves. GFP signals were detected in epidermal cell nuclei co-expressing nYFP-LFR and cCFP-ICE1; however, no GFP signal was detected in negative control cells co-expressing nYFP-LFR/cCFP or nYFP/cCFP-ICE1 ( Figure 3C ). These results indicate that LFR interacts with ICE1 in plant.

We next verified the LFR–ICE1 interaction using an LCI assay. Tobacco leaf areas co-transformed with LFR-nLUC and cLUC-ICE1 emitted light in the presence of luciferin. However, tobacco leaf areas co-transformed with the negative controls (LFR-nLUC/cLUC or cLUC-ICE1/nLUC or cLUC/nLUC) showed no light emission in the same condition ( Figures 3D, E ). These results confirm that LFR interacts with ICE1 in plant.

Given that CBF3 is a direct target of ICE1 (Ding et al., 2015; Kim et al., 2015; Tang et al., 2020) and LFR interacts directly with ICE1, we speculated that LFR may also bind to the CBF3 gene. ChIP-qPCR assays of pLFR : LFR-3FLAG/lfr-2 complemented transgenic lines were conducted using anti-FLAG antibodies with Col-0 as a negative control. According to a previous report on the ICE1-binding site of CBF3 (Ding et al., 2015), primers for the amplification of three different fragments (P1–P3) of CBF3 chromatin were designed ( Figure 4A ). Under normal growth condition, the ChIP-qPCR result showed that LFR binds to CBF3 at the P1 and P2 loci ( Figure 4B ). After low-temperature treatment, the ChIP-qPCR result showed that LFR binds to CBF3 chromatin at the P2 locus ( Figure 4C ). The P1 and P2 loci of CBF3 are also binding sites of ICE1 (Ding et al., 2015; Kim et al., 2015; Tang et al., 2020). Thus, these results demonstrated that CBF3 is a direct target of LFR under normal and cold stress conditions.

Because we detected reduced expression of ICE1 in lfr-1 plants, we tested whether ICE1 is also a direct target of LFR. A ChIP-qPCR assay was performed using Col-0 and the pLFR : LFR-3FLAG/lfr-2 complemented transgenic line with anti-FLAG antibodies. Different primers for the amplification of different fragments of the ICE1 locus were designed ( Figure 5A ). The ChIP-qPCR results showed that LFR was able to bind directly to the P2 locus of ICE1 at 22°C ( Figure 5B ). After cold stress treatment, LFR was associated with the P3 locus of ICE1 ( Figure 5C ). Thus, these results demonstrated that ICE1 is also a direct target of LFR with or without low-temperature treatment.

To elucidate whether LFR is regulated by cold tress at the transcriptional level, we analyzed cis-regulatory elements in the LFR promoter using PLANTCARE. Five low temperature-responsive element (CANNTG) bound by ICE1 was found in the LFR promoter, indicating that LFR may be regulated by ICE1 or a cold-responsive gene ( Figure 6A ). And we indeed detected significant downregulation of LFR in ice1-2 mutant plant ( Supplementary Figure S1A ). And we also checked the transcript level of LFR in plants treated with or without cold exposure at different time points. RT-qPCR showed that LFR was downregulated by cold treatment ( Figure 6B ). We also studied the protein level of LFR using pLFR : LFR-eYFP/lfr-2 complemented lines. Furthermore, when the pLFR : LFR-eYFP/lfr-2 lines were held at 4°C, the eYFP signals were reduced, especially after 6 and 24 h of cold stress ( Figures 6C–E ). Taken together, these results indicate that LFR may be repressed by cold stress.

Plants have developed sophisticated mechanisms to survive fluctuating ambient temperatures (Shi et al., 2018; Chen et al., 2021). At chilling or freezing temperatures, plants utilize a cascade of TFs to adapt to low-temperature stress (Shi et al., 2018). Among these TFs, ICE1, a constitutively expressed MYC-type TF, activates the expression of CBFs, which subsequently induce the transcription of downstream cold-responsive genes such as CORs, which promote cold tolerance (i.e., the ICE1–CBF–COR cascade) (Shi et al., 2018). However, the interacting factors that are needed for ICE1 to activate CBFs are largely unknown. In this study, we demonstrated that ICE1 is an LFR-interacting partner using Y2H, BiFC, and LCI assays ( Figure 3 ). Disruption of LFR resulted in reduced cold tolerance under both NA and CA conditions, similar to ice1-2 ( Figure 1 ). In addition, the transcript levels of ICE1–CBF–COR cascade TF genes (e.g., ICE1, CBF1, CBF2, CBF3, and COR15A) were all downregulated after cold treatment in lfr plants ( Figure 2 ). Like ICE1, LFR was able to directly bind to the CBF3 locus ( Figure 4 ). LFR was also found to associate directly with ICE1 chromatin ( Figure 5 ). Together, our results demonstrate that LFR directly interacts with ICE1 and positively regulates the transcription of genes in the ICE1–CBF–COR signaling pathway.

Interestingly, we also detected significant downregulation of LFR in ice1-2 mutant plants, and we found ICE1-binding sites (i.e., CANNTG) in the proximal region of the LFR promoter ( Figure 6 and Supplementary Figure S1A ). Furthermore, using published ChIP-seq data for GFP-ICE1 (Tang et al., 2020), we found that ICE1 was enriched at the proximal promoter region and gene body of LFR and ICE1 chromatin ( Supplementary Figure S1 ; Tang et al., 2020). Together with our data showing that ICE1 is not only an interacting factor but also a target of LFR, which tends to bind target(s) via interaction with transcription factor (Lin et al., 2018), these results may indicate that LFR and ICE1 interact physically and form transcriptional feedback loops. It will be interesting to test this possibility in the future.

Although LFR may act as a positive regulator of cold signaling, we found that LFR was repressed after prolonged cold treatment ( Figures 6 ). This is similar to ICE1, which is degraded during long-term cold treatment (Dong et al., 2006; Ding et al., 2015; Zhao et al., 2017). The cold-induced downregulation of LFR and ICE1 may help balance plant cold responses and growth by preventing detrimental effects on plant growth resulting from the over-stimulation of critical freezing tolerance factors (e.g., CBFs and CORs) (Zhao et al., 2017).

In addition, we found that the lfr-1 ice1-2 double mutant showed obvious defects in seed germination and greening ( Supplementary Figure S2 ), which may indicate that LFR and ICE1 may also participate in these developmental processes.

Epigenetic factors play critical roles in long-term adaptation to the ambient environment (Miryeganeh, 2021; Chung et al., 2022). However, the exact functions of different CRC components in cold adaptation are unknown. Recently, Gratkowska-Zmuda et al. (2020) provided solid evidence showing that SWI3C of the SWI/SNF complex can directly bind, change the nucleosome occupancy and influence the transcription of ICE1, MYB15, and CBF1 in a temperature-dependent manner. SWI3C was required to activate ICE1 expression at 22°C but not at 14°C, and it repressed the transcription of all three CBF genes. Consistent with the upregulation of CBFs, the swi3c mutant displayed increased freezing tolerance compared to wild type (Gratkowska-Zmuda et al., 2020). However, in this study, we found that ICE1–CBF–COR cascade genes were downregulated in the lfr mutant, which displayed hypersensitivity to freezing. Our previous study showed that LFR interacts with SWI3A/SWI3B, but not SWI3C (Lin et al., 2021). Consistently, LFR together with SWI3A/SWI3B, not SWI3C, were identified in the immunoprecipitation mass spectrometry (IP-MS) data of a recent study of the MINU-associated SWI/SNF (MAS) complex (Diego-Martin et al., 2022; Hernández-García et al., 2022). In addition, a comprehensive characterization of the SWI/SNF complexes by proteomic approaches revealed that SWI3C and LFR are involved in different complexes with specific chromatin accessibility and binding ability for active histone modification (Guo et al., 2022). Thus, the opposite phenotype and differential expression pattern of CBF genes may result from that SWI3C and LFR participate in distinct SWI/SNF complexes with different transcriptional properties. So, it will be interesting to decipher the detailed molecular and genetic interactions between LFR, SWI3C, and the core ATPases of the SWI/SNF complex (BRM, SYD, MINU1, and MINU2), which is of great importance to elucidate the molecular regulatory mechanisms of SWI/SNF complexes in controlling plant response to cold treatment.

Arabidopsis LFR encodes an ARM-repeat domain-containing protein (Wang et al., 2009). Phylogenetic analysis of the ARM domain revealed that plant LFR has a common origin with the signature subunits ARID1/2 of the animal SWI/SNF complex (Hernández-García et al., 2022). Recently, the structural approaches demonstrated that the ARM domain in ARID1 serves as a rigid core in the “Base/Core module” organization and stabilization by directly interacting with the core ATPase and all other Base subunits (He et al., 2020; Mashtalir et al., 2020). Besides, ARID1 is also essential for the efficient nucleosome sliding activity of the SWI/SNF complex (He et al., 2020). However, the molecular weight of Arabidopsis LFR is much lower than its mammalian orthologs, because of the loss of the DNA-interacting AT-rich interaction domain (ARID) (Wang et al., 2009; Hernández-García et al., 2022). Thus, it will be interesting to decipher the biological contribution of plant LFR, to the complex organization, architecture, stability, and nucleosome sliding activity of target genes in plant development and stress in the future.

Taken together, our results demonstrate that the SWI/SNF complex component LFR interacts physically with ICE1 to act as a positive regulator of the ICE1–CBF–COR cascade ( Figure 7 ). This study will aid our understanding of plant-specific epigenetic changes through chromatin remodeling in response to global climate change.