Background

In eukaryotes, chromatin structure, composition, and dynamics determine the three-dimensional configuration of the genome and are critical for gene regulation, cell fate, and function [1, 2]. Chromatin conformation and related transcriptional states depend on coordinated shifts in DNA methylation, post-translational modifications of histone tails, and RNA interference (RNAi) pathways [3,4,5]. In mammalian genomes, DNA methylation is primarily observed at CpG dinucleotides and is estimated to occur at ~70–80% of CpG sites throughout the genome [6]. A quarter of non-CG methylation is found in embryonic stem cells, while the remaining unmethylated CpG sites are mostly found in dense clusters, near gene promoters, referred to as CpG islands [7,8,9]. Unlike in mammals, DNA methylation in plants predominantly occurs on transposons and other repetitive DNA elements, and exists in all possible sequence contexts: symmetric CpG and CpHpG—where H is any base except G—or asymmetric CpHpH [10, 11]. Non-CpG methylation is mainly distributed at heterochromatin regions, nevertheless, some euchromatic genes exhibit cytosine methylation in their promoter, a feature likely correlated with tissue specificity [10].

In Arabidopsis thaliana, the establishment of de novo methylation in all sequence contexts is catalyzed by DOMAINS REARRANGED METHYLTRANSFERASE 2 (DRM2), a plant homolog of mammalian DNA (CYTOSINE-5)-METHYLTRANSFERASE 3 (DNMT3) a and b [12, 13]. DRM2 is guided to chromatin by 24-nucleotide small interfering RNAs (24nt siRNAs) as part of a pathway known as RNA-directed DNA methylation (RdDM; [14]). RdDM involves two non-redundant plant specific RNA polymerases, Pol IV and Pol V, in addition to the canonical RNA interference machinery, which requires the activity of RNA-dependent RNA polymerase 2 (RDR2) and members of the Dicer and Argonaute families [15,16,17,18]. Once established, DNA methylation is maintained by three different pathways depending on the sequence context. CpG methylation depends on DNA METHYLTRANSFERASE 1 (MET1), a homolog of mammalian DNA METHYLTRANSFERASE 1 (DNMT1; [19, 20]). CpHpG methylation relies on CHROMOMETHYLASE 3 (CMT3), a plant specific DNA methyltransferase that associates with SU(VAR)3–9 HOMOLOG (SUVH) histone methyltransferases and recognizes dimethylated histone 3 tails at lysine 9 (H3K9me2; [11, 21,22,23,24]). Finally, CpHpH methylation is maintained through persistent de novo methylation by DRM2 and by the CMT3 homolog CHROMOMETHYLASE 2 (CMT2) that specifically reads the H3K9me2 mark [24,25,26].

Beside the contribution of methyltransferases, the DNA methylation process involves methylcytosine-binding proteins of the VARIANT IN METHYLATION (VIM/ORTH) family [27, 28]. VIM/ORTH proteins are homologous to the mammalian UBIQUITIN-LIKE CONTAINING PHD AND RING FINGER DOMAINS (UHRF) proteins known to regulate cytosine methylation through the recruitment of DNMT1 to target loci [29,30,31,32]. In particular, UHRF1 functions as an epigenetic regulator maintaining DNA methylation and histone modifications [33] and is stabilized by direct interaction with the long noncoding RNA (lncRNA) UPAT (UHRF1 Protein Associated Transcript; [34]). VIM proteins are characterized by the presence of a PHD domain recognizing trimethylated histone 3 tails at lysine 4 (H3K4me3; [35,36,37,38]), a SRA (SET [Su(var), Enhancer of Zeste, Trithorax], and RING [Really Interesting New Gene] Associated) domain that can associate with methylated DNA [11, 27, 39], and two RING domains conferring ubiquitin E3 ligase activity [31]. The Arabidopsis genome encodes five highly similar VIM proteins—named VIM1 to 5—and a related protein ORTH-LIKE1 (ORL1/VIM6; [31]). VIM1, VIM2, and VIM3 maintain MET1-mediated cytosine methylation at CpG dinucleotides throughout the genome and have been reported to function entirely redundantly to mediate epigenetic transcriptional silencing in collaboration with MET1 [25, 28, 40].

Interestingly, aberrant changes in active and repressive histone modifications have been observed in vim1/2/3 and met1 mutants, supporting that VIM proteins coordinate DNA methylation and histone modification [41,42,43,44]. In Arabidopsis, the transcriptionally repressive mark histone H3 lysine 27 trimethylation (H3K27me3) is largely restricted to the transcribed regions of single genes, exhibiting a global anti-correlated distribution with centromeric-enriched DNA methylation [11, 43, 45]. However, a loss of H3K27me3 was also reported at gene bodies in met1 mutants and at specific VIM1 targets located in euchromatic regions in the vim1/2/3 triple mutant [43]. Notably, the correlation between DNA hypomethylation and H3K27me3 reduction in the vim1/2/3 mutation was more prevalent in promoter regions than in transcribed regions [44]. Collectively, these observations hint at non-canonical mechanisms linking DNA methylation and repressive histone modifications over specific transcriptionally active loci.

In A. thaliana, H3K27me3 is deposited by Polycomb group (PcG) proteins in euchromatic regions containing protein-coding genes and is maintained by LIKE HETEROCHROMATIN PROTEIN 1 (LHP1), a component of the plant Polycomb Repressive Complex 1 (PRC1) and a homolog of mammalian HETEROCHROMATIN PROTEIN 1 (HP1; [46,47,48,49]). LHP1 capacity to localize and mediate epigenetic repression of PcG target genes was shown to rely on its RNA-binding hinge region, suggesting that LHP1 activity could be modulated by interacting RNAs [50]. Consistently, it was shown that the lncRNA APOLO (AUXIN-REGULATED PROMOTER LOOP) can regulate local chromatin conformation by decoying LHP1 away from target loci. APOLO directly recognizes multiple distant and independent auxin-related loci across the A. thaliana genome by short sequence complementarity and the formation of DNA-RNA duplexes (R-loops) [51, 52].

Here, we demonstrated that in addition to the PRC1 component LHP1, APOLO lncRNA interacts in vivo with VIM1, the plant homolog of UHRF1, linking Polycomb and DNA methylation machineries. RNA sequencing analyses of APOLO over-expression and vim1 mutant lines revealed that APOLO and VIM1 control a large common set of genes related to the thermomorphogenic response. In particular, the APOLO-VIM1-LHP1 complex directly targets the heat-responsive auxin-biosynthetic gene YUCCA2 and conjointly mediates cytosine methylation and H3K27me3 deposition at its promoter, representing a new epigenetic mechanism regulating the plant response to warm temperatures. Strikingly, we also demonstrate that the lncRNA UPAT, known to recognize UHRF1 in humans, can interact with VIM1 and LHP1 in plant cells despite the lack of sequence homology between UPAT and APOLO. Furthermore, the over-expression of APOLO or UPAT trigger a similar transcriptional reprogramming in Arabidopsis, and UPAT constitutive transcription in plants precludes LHP1 and VIM1 binding to the YUCCA2 promoter region. Hence, sequence-unrelated lncRNAs with potentially similar structures may exert similar molecular functions across kingdoms, integrating the epigenetic regulation of gene expression.

Results

Long noncoding RNA APOLO associates with the methylcytosine-binding protein VIM1 in vivo

In order to investigate the composition of the ribonucleoprotein complexes integrated by the lncRNA APOLO, we performed an exploratory Chromatin isolation by RNA purification (ChIRP) followed by protein precipitation and mass spectrometry. We used two independent sets of biotinylated DNA probes against APOLO (ODD and EVEN sets) and LacZ probes as a negative control [53]. Among the proteins identified by at least two positive hits of unique peptides in ODD and EVEN, excluded from LacZ ChIRPs, we found the protein VARIANT IN METHYLATION 1 (VIM1, AT1G57820). This interaction was confirmed by APOLO-ChIRP-dot blot in stably transformed Arabidopsis thaliana seedlings over-expressing GFP-VIM1 (OE VIM1-1) (Fig. 1A and Additional file 1: Fig. S1A). Transient expression of GFP-VIM1 in Nicotiana benthamiana and A. thaliana leaves revealed that VIM1 is located in the cell nucleus, in agreement with previous observations [27]. Moreover, this localization was observed regardless of the co-expression with nuclear-enriched APOLO transcripts (Additional file 1: Fig. S1B-C). Anti-GFP RNA immunoprecipitation (RIP) in N. benthamiana leaves transiently co-transformed with GFP-VIM1 and APOLO, or in Arabidopsis OE VIM1-1 (Fig. 1B; VIM1 levels shown in Additional file 1: Fig. S2A), revealed a high enrichment of APOLO transcripts in VIM1 immunoprecipitated samples. These results indicate that VIM1 is part of a novel ribonucleoprotein complex integrated by APOLO lncRNA.

Fig. 1
figure 1

The lncRNA APOLO and the methylcytosine-binding protein VIM1 are thermomorphogenesis regulators. A Chromatin isolation by RNA purification (ChIRP)-Dot blot analysis of APOLO interaction with GFP-VIM1. ChIRP was performed using ODD and EVEN sets of probes against APOLO or using LacZ probes as a negative control. Dot blots are revealed with an anti-GFP antibody and an HRP-conjugated secondary antibody. Diluted INPUT (*1/50) were used as loading control. B RNA immunoprecipitation (RIP) assay in Nicotiana benthamiana leaves transiently co-transformed with APOLO and GFP-VIM1 translational fusion expressed under the control of the 35S-CaMV promoter, or in Arabidopsis thaliana 2-week-old VIM1 over-expression (OE VIM1-1) seedlings. Results are expressed as a percentage of the INPUT fraction. Anti-IgG antibodies were used as a negative control, and the asterisks indicate Student’s t test ≤ 0.05 (n = 3) between anti-GFP and anti-IgG RIPs. The non-specific background level of RNA precipitation (PP2A) is also shown in Arabidopsis. C Fold change of APOLO and VIM1 expression levels in relation to 23 °C control conditions in 4-day-old wild-type (WT) seedlings treated with heat (29 °C) for 6 h. Asterisks indicate Student’s t test ≤ 0.05 (n = 3) for levels of each corresponding gene between 23 and 29 °C. D, E Boxplots showing hypocotyl length quantification ratio at 29 °C over 23 °C of 4-day-old OE APOLO-1, OE APOLO-2 (D) or vim1-2, vim1-3 (E) seedlings and their associated WT. Values are represented by colored points. F Venn diagram of differentially expressed transcripts in WT treated with heat (29 °C) for 6 h, and in untreated OE APOLO-1 and vim1-3 seedlings. G Heatmap showing log2(fold change) compared to WT 23 °C. A correlation of up- and downregulated genes in WT in response to 29 °C is observed for OE APOLO-1 and vim1-3 at 23 °C. H Gene Ontology (GO) enrichment analysis of upregulated transcripts in WT treated with heat. Top ten GO categories with more significant p-values are shown. Violet bars: auxin-related pathways. The ShinyGO Browser is located at bioinformatics.sdstate.edu and published in [54]. In A, each sample was serially diluted as indicated in the plot, and two additional biological replicates are shown in Additional file 1: Fig. S1A. In B, bars represent average ± SD (n = 3 biological replicates). In C, transcript levels are normalized relatively to the untreated control to show fold changes. Bars represent average ± SD (n = 3 biological replicates). In D, E, results are the mean of three biological replicates and letters indicate significant differences compared to WT, based on a Kruskal-Wallis test (α = 0.05; n ≥ 134)

Full size image

APOLO and VIM1 are important thermomorphogenesis regulators in Arabidopsis thaliana

Abnormal morphological phenotypes, including DNA methylation-dependent late flowering, were previously reported for the vim1/2/3 triple mutant in contrast to the vim1 single mutant exhibiting no evident phenotype [28]. We thus wondered in what developmental context the APOLO-VIM1 interaction may occur and exert a regulatory role over target genes. By exploring the eFP Arabidopsis transcriptomic database [55], we found that VIM1, VIM2, and VIM3 genes showed different transcriptional dynamics following heat stress (38 °C followed by recovery at 25 °C; Additional file 1: Fig. S3A, blue box). Therefore, we evaluated the expression of these three genes and APOLO at 29 °C, a warm temperature usually faced by Arabidopsis in the wild [56]. After 6 h of exposure to 29 °C, transcript levels of APOLO decreased while those of VIM1, as well as the warmth-response markers PIF4 and YUC8, increased (Fig. 1C; Additional file 1: Fig. S3B-C). Unlike VIM1, expression of VIM2 and VIM3 were repressed or unaffected, respectively (Additional file 1: Fig. S3C). Interestingly, anti-GFP RIP in N. benthamiana leaves transiently co-transformed with GFP-VIM2 and APOLO revealed that APOLO can also interact in vivo with VIM2 (although showing a weaker interaction than VIM1; Additional file 1: Fig. S3D for VIM2 nuclear localization and Additional file 1: Fig. S3E for RIPs), suggesting that this complex may also form during Arabidopsis development or in response to the environment.

To assess whether the ribonucleoprotein complex integrated by APOLO and VIM1 is involved in the response to warmth, we tested the effect of their deregulation on the thermomorphogenesis response. Two independent APOLO over-expression (OE) lines (OE APOLO-1 and OE APOLO-2; [52]), together with two independent vim1 insertional mutants resulting in a knockdown of VIM1 (vim1-2 and vim1-3, respectively; Additional file 1: Fig. S2B-C), exhibited an impaired hypocotyl elongation after 4 days at 29 °C (Fig. 1D, E; Additional file 1: Fig. S4A-B; Additional file 2: Table S1). vim1-3 mutant plants transformed with proVIM1:GFP:VIM1 exhibited a partially rescued hypocotyl elongation phenotype (Additional file 1: Fig. S4C-D; Additional file 2: Table S1). Additionally, we observed a slight but significant reduction of hypocotyl elongation at 29 °C in one out of two independent RNAi APOLO lines (RNAi APOLO-1 and RNAi APOLO-2; [51]) as well as in the knockout line obtained by CRISPR-Cas9-mediated deletion of the full APOLO locus (CRISPR-APOLO; Additional file 1: Fig. S2D, S4E-H; Additional file 2: Table S1), and a minor but significant induction of hypocotyl elongation in OE VIM1 lines (OE VIM1-1 and OE VIM1-2; Additional file 1: Fig. S2A, Fig. S4F, S4I; Additional file 2: Table S1). Altogether, these results suggest that APOLO and VIM1 regulate thermomorphogenesis and incidentally indicate that VIM1 is not redundant with VIM2 or VIM3 in this developmental context.

To pinpoint molecular mechanisms that could explain impaired thermomorphogenesis upon APOLO or VIM1 deregulation, we profiled the transcriptomes of 4-day-old OE APOLO-1 and vim1-3 seedlings grown at 23 °C with RNA-Seq of 4-day-old wild-type (WT) seedlings grown at 23 °C and subjected or not to heat (29 °C) for 6 h. In WT seedlings, warmth caused the downregulation of over 3400 transcripts and the upregulation of approximately 2650 transcripts (Additional file 2: Table S2). Remarkably, 56% of the total set of up- and downregulated transcripts at 29 °C in WT were already deregulated at 23 °C in OE APOLO-1 and/or vim1-3 (3403 genes over 6073; Fig. 1F), whereas 52% of these transcripts (1782 out of 3403) were deregulated in both OE APOLO-1 and vim1-3 at 23 °C (Fig. 1F, central intersection). Moreover, a subset of up- and downregulated genes in response to warmth in WT showed similar expression profiles in OE APOLO-1 and vim1-3 at ambient temperature (Fig. 1G). Upregulated genes in WT in response to warmth were enriched in “Response to auxin,” “Cellular response to auxin stimulus,” and “Auxin-activated signaling pathway” GO categories (Fig. 1H; Additional file 1: Fig. S5), consistent with the well-established role of auxin in thermomorphogenesis [57]. Altogether, these transcriptomic analyses hint at a critical role for the APOLO-VIM1 complex in the transcriptional reprogramming of gene expression in response to warm temperatures.

We then aimed at identifying the potential direct targets of the APOLO-VIM1 complex during the thermomorphogenic response. Given the well-known role of VIM1 in DNA methylation, we compared the list of hypomethylated genes in vim1 vs. WT identified by bisulfite sequencing analyses (BiS-Seq; [25]), with potential APOLO direct targets identified by APOLO-ChIRP-Seq [52]. Significant enrichment of hypomethylated genes in the CpG and CpHpH contexts was observed among APOLO potential targets (Additional file 2: Table S3). Interestingly, among the eleven potential APOLO targets exhibiting a differential CpHpH methylation pattern in the vim1 mutant, we found YUCCA2 (YUC2, AT4G13260; Additional file 2: Table S3), a heat-responsive gene involved in auxin biosynthesis [58, 59]. By using a proYUC2:GUS transcriptional fusion [60], we observed that the YUC2 promoter region is activated in the hypocotyl of 4-day-old seedlings grown at 29°C (Fig. 2A). In addition, yuc2 loss-of-function mutant seedlings [60] exhibited a reduced hypocotyl elongation after 4 days at 29°C (Fig. 2B; Additional file 2: Table S1), indicating that YUC2 is required for a proper thermomorphogenic response. Consistently, we observed that the transcriptional activation of YUC2 after 6 h at 29 °C was impaired in OE APOLO-1 and vim1-3 showing an altered hypocotyl elongation (Fig. 2C) and a generally altered transcriptional behavior in APOLO- and VIM1-deregulated lines (Additional file 1: Fig. S6A). Among the other six genes assessed together with YUC2, three were induced by warmth in WT, whereas all of them exhibited an altered expression upon deregulation of APOLO and/or VIM1 (Additional file 1: Fig. S6A). Altogether, our results further suggest that the APOLO-VIM1 complex directly regulates a common subset of genes including YUC2 to trigger the plant auxin-related thermomorphogenic response at warm temperatures.

Fig. 2
figure 2

The thermomorphogenesis-related gene YUCCA2 is directly co-regulated by APOLO and VIM1. A Histochemical localization of GUS activity in 4-day-old seedlings containing the proYUCCA2:GUS construct, grown at 23 °C or 29 °C. Scale bars, 0.1 cm. B Boxplots showing hypocotyl length quantification ratio at 29 °C over 23 °C of 4-day-old yuc2 seedlings and their associated wild-type (WT). Values are represented by colored points. Representative morphological phenotypes are shown on the right. Scale bars, 1 cm. CYUCCA2 (YUC2) transcript levels in 4-day-old WT, OE APOLO-1, and vim1-3 seedlings treated or not with heat (29 °C) for 6 h. Asterisks indicate Student’s t test ≤ 0.05; n = 3 between each corresponding genotype and WT. D Epigenetic profile at the YUC2 locus. Tracks 1 to 3 [52]: APOLO recognition by chromatin isolation by RNA purification (ChIRP) sequencing, using ODD (Track 1) and EVEN (Track 2) sets of probes against APOLO. ChIRP negative control using LacZ probes is shown in Track 3. Tracks 4 to 7 [61]: R-loop formation by DNA:RNA immunoprecipitation (DRIP) sequencing, on Watson (Track 4), Crick strand (Track 5), or unstranded sequencing (Track 6). DRIP negative control after RNAseH treatment is shown in Track 7. Gene annotation is shown at the bottom. On the YUCCA2 schematic representation in the bottom, red dots indicate the presence of six GAAGAA/TTCTTC boxes which may mediate APOLO recognition according to Ariel et al. [52]. EAPOLO association to DNA of the YUC2 locus by ChIRP-qPCR in WT, RNAi APOLO-1, and CRISPR-APOLO plants. The background level was determined using a set of probes against LacZ RNA. F RNA-DNA hybrid (R-loop) formation at the YUC2 locus by DRIP-qPCR in WT and RNAi APOLO-1 plants. G R-loop formation at the YUC2 locus by DRIP-qPCR in WT and OE APOLO-1 plants at 23 °C or 29 °C. Asterisks indicate a significant reduction (Student’s t test ≤ 0.05; n = 3) between R-loop levels at 23 °C or 29 °C in WT plants. H Chromatin immunoprecipitation (ChIP)-qPCR analysis of VIM1 binding at the YUC2 promoter in 4-day-old VIM1 over-expression (OE VIM1-1) seedlings treated or not with heat (29 °C) for 6 h. I Methylated DNA immunoprecipitation (MeDIP)-qPCR analysis at the YUC2 promoter in 4-day-old APOLO over-expression (OE APOLO-1) or vim1-3 mutant seedlings treated or not with heat (29 °C) for 6 h. In A, one representative picture out of ten stained seedlings is shown. In B, results are the mean of three biological replicates and letters indicate significant differences compared to WT, based on a Mann-Whitney test (α = 0.05; n ≥ 110). In C, transcript levels are normalized relatively to the untreated control to show fold changes. Bars represent average ± SD (n = 3 biological replicates). In E, F, bars represent average ± SD (n = 3 biological replicates). In G, H, results are expressed as a percentage of the INPUT fraction. Anti-IgG antibodies were used as a negative control. Bars represent SD (n = 3 biological replicates), except for H (n = 2 closest biological replicates out of 3 performed, all showing the same trend) and the asterisk indicates the Student’s t test ≤ 0.05, between 23 and 29 °C. In E and G, asterisks indicate the Student’s t test ≤ 0.05; n = 3, between WT and the corresponding genotype. In I asterisks indicate the Student’s t test ≤ 0.05; n = 3, between 23 and 29 °C for each genotype

Full size image

APOLO and VIM1 directly mediate heat-dependent methylation at the YUCCA2 promoter

As APOLO recognizes its target loci by sequence complementarity and R-loop formation [52], we first investigated the APOLO binding profile over the YUC2 locus by ChIRP-Seq. Compared to the LacZ probes used as a negative control, ODD and EVEN probes showed specific binding of APOLO across the YUC2 locus, notably to the exon 1, displaying six APOLO-binding motifs “GAAGAA/TTCTTC” (Fig. 2D, first three tracks in blue and light blue, binding motifs indicated as red dots over the YUC2 locus). APOLO-ChIRP-qPCR over YUC2 gene body in WT, RNAi APOLO-1, and CRISPR-APOLO seedlings confirmed the specificity of the interaction (Fig. 2E). Furthermore, DNA-RNA immunoprecipitation followed by sequencing (DRIP-Seq; [61]) revealed that R-loop formation occurs in the exon 1 of YUC2, in correlation with APOLO binding (Fig. 2D, red box), suggesting that APOLO directly recognizes YUC2 through the formation of R-loops. Consistently, DRIP-qPCR in WT, RNAi APOLO-1 and CRISPR-APOLO seedlings confirmed that R-loop formation over the YUC2 locus depends on APOLO expression (Fig. 2F). Consistently, DRIP-qPCR revealed that R-loop formation over YUC2 diminishes significantly more in WT than in OE APOLO-1 at 29°C, further supporting that YUC2 R-loop at 23°C is mediated by APOLO (Fig. 2G).

Interestingly, a NRPE1 Pol V-subunit chromatin immunoprecipitation (ChIP)-Seq [62] indicated that the YUC2 promoter region, located 560 to 1010 bp upstream of YUC2 TSS, is directly regulated by Pol V in a MET1-dependent manner (Fig. 2D, black box indicating the region for results in Additional file 1: Fig. S6B). In addition, a small RNA-Seq profiling reported a temperature-dependent accumulation of RdDM-related 24nt siRNAs over this locus ( [63]; Additional file 1: Fig. S6C), suggesting that DNA methylation in this regulatory region may contribute to the temperature-induced transcriptional regulation of YUC2. We thus assessed VIM1 binding and DNA methylation at the YUC2 promoter region in the same developmental stage as our phenotypic characterization. GFP-VIM1 ChIP performed in 4-day-old seedlings grown at 23 °C and treated or not at 29 °C for 6 h revealed that VIM1 binds to the region located 357 to 494 bp upstream of YUC2 TSS and that this binding is reduced at 29 °C compared to 23 °C (Fig. 2H). Methylated DNA immunoprecipitation (MeDIP) demonstrated that DNA methylation is reduced in this region in the WT after 6 h at 29 °C (Fig. 2I). Furthermore, in OE APOLO-1 and vim1-3, DNA methylation levels resulted to be lower at 23 °C and increased at 29 °C, exhibiting the opposite behavior to the WT (Fig. 2I).

To further support the relevance of RdDM on YUC2 regulation in the context of thermomorphogenesis, we additionally characterized the physiological response to warm temperatures of RdDM mutants nrpd2a (a common subunit of Pol IV and Pol V), rdr2-5, dcl3-1 and ago4-8, as well as CpG methyltransferase mutant met1-2 and non-CpG methyltransferase mutants cmt2-7 and cmt3-11 (for a visual summary of their respective roles in DNA methylation pathways, see Additional file 1: Fig. S7A). Hypocotyl elongation at 29 °C was reduced in nrpd2a, met1-2, rdr2-5, dcl3-1, and ago4-8 backgrounds, whereas it was unaffected in cmt3-11 and slightly enhanced in cmt2-7 backgrounds (Additional file 1: S7B-J; Additional file 2: Table S1). In agreement, BiS-seq analyses [25] revealed that RdDM and met1-2 mutants showing impaired thermomorphogenesis also displayed reduced CpHpH methylation levels in the YUC2 regulatory region (Additional file 1: Fig. S6D-E, blue box), further supporting that epigenetic modifications regulating DNA methylation at YUC2 promoter are critical to modulate thermomorphogenesis. Taken together, our results indicate that APOLO and VIM1 directly mediate DNA methylation at YUC2 promoter in response to warm temperatures, in a process likely modulated by the RdDM pathway.

VIM1 and LHP1 cooperate to regulate temperature-dependent histone and DNA methylation at the YUCCA2 promoter

Given that the plant PRC1 component LHP1 recognizes APOLO in vivo and co-regulates common target loci across the Arabidopsis genome [51, 52], we explored whether the YUC2 locus was regulated by LHP1 and the related transcriptionally repressive mark H3K27me3. ChIP-Seq analyses [49] showed that the YUC2 locus is enriched in LHP1 and H3K27me3 in standard growing conditions (Fig. 3A), whereas both marks are drastically reduced at 29 °C over the promoter region (Fig. 3A, black box; Fig. 3B, C). Moreover, LHP1 binding to the YUC2 gene body was impaired by APOLO over-expression (Fig. 3A, blue box; Fig. 3D). Consistent with an involvement of LHP1 in the thermomorphogenic response, 4-day-old lhp1 loss-of-function mutant seedlings [49] exhibited a reduced hypocotyl elongation at 29 °C and no induction of YUC2 after 6 h at 29 °C (Fig. 3E, F; Additional file 2: Table S1). Taken together, these results suggest that in addition to VIM1-dependent DNA methylation, the PcG-dependent histone methylation machinery also directly regulates YUC2 expression in response to warm temperatures.

Fig. 3
figure 3

VIM1 and LHP1 co-regulate histone and DNA methylation at the YUCCA2 promoter. A Epigenetic landscape at the YUCCA2 (YUC2) locus. Track 1: H3K27me3 deposition by ChIP-sequencing. Track 2: LHP1 binding by ChIP-sequencing [49]. The black signal represents the input sequencing for each respective track. Gene annotation is shown in the bottom. B ChIP-qPCR analysis of H3K27me3 deposition at the YUC2 promoter in 4-day-old wild-type (WT) seedlings treated or not with heat (29 °C) for 6 h. The asterisk indicates the Student’s t test ≤ 0.05, n=3, between 23 and 29 °C. C ChIP-qPCR analysis of LHP1 binding at the YUC2 promoter in 4-day-old WT seedlings treated or not with heat (29 °C) for 6 h. D ChIP-qPCR analysis of LHP1 binding at the YUC2 gene body in WT and APOLO over-expression (OE APOLO-1) plants. The negative control corresponds to an APOLO-independent LHP1 target gene AT4G23720 [49]. E Boxplots showing hypocotyl length quantification ratio at 29 °C over 23 °C of 4-day-old lhp1 seedlings and their associated WT. Values are represented by colored points. Representative morphological phenotypes are shown on the right. Scale bars, 1 cm. FYUC2 transcript levels in 4-day-old WT and lhp1 seedlings treated or not with heat (29 °C) for 6 h. G Methylated DNA immunoprecipitation (MeDIP)-qPCR analysis at the YUC2 promoter in 4-day-old lhp1 mutant seedlings treated or not with heat (29 °C) for 6 h. H, I Chromatin immunoprecipitation (ChIP)-qPCR analyses of LHP1 binding (in H) and H3K27me3 deposition (in I) at the YUC2 promoter in 4-day-old WT and vim1-3 mutant seedlings. The negative control corresponds to an APOLO-independent LHP1 target gene AT4G23720 [49]. J Bimolecular fluorescence complementation (BiFC) assay in transiently transformed Nicotiana benthamiana leaves. CYFP was fused to VIM1 or SRA/RING2, and NYFP was fused to LHP1. In both panels, bright-field images (left), YFP fluorescence (middle), and merged images (right) are shown. A zoom-in including mCherry constitutive expression for nuclei and membrane visualization is shown in Additional file 1: Fig. S8A. Scale bars, 50 μm. In B–D and G–I, results are expressed as a percentage of the INPUT fraction. Anti-IgG antibodies were used as a negative control. Bars represent standard deviation (n = 3 biological replicates). In E, results are the mean of three biological replicates and letters indicate significant differences compared to WT, based on a Mann–Whitney test (α = 0.05; n ≥ 111). In F, transcript levels are normalized relatively to the untreated control to show fold changes. Bars represent average ± SD (n = 3 biological replicates) except for G (n = 2 closest biological replicates out of 3 performed, all showing the same trend). In J, one representative picture out of six biological replicates is shown

Full size image

Considering the similarities between VIM1 and LHP1 behaviors at the YUC2 locus, we wondered if VIM1-associated DNA methylation and LHP1-related H3K27me3 transcriptional regulations were dependent on each other. We observed that YUC2 regulatory region in the proximal promoter displayed lower levels of cytosine methylation in the lhp1 mutant at 23 °C, which resulted to be restored at 29 °C in lhp1, exhibiting opposite behavior to the WT (Fig. 3G) and the same trend as OE APOLO-1 and vim1-3 (Fig. 2H). To further elucidate the link between LHP1 and VIM1, we assessed the levels of H3K27me3 and LHP1 capacity to bind to the YUC2 promoter region in 4-day-old vim1-3 seedlings. Notably, LHP1 binding and H3K27me3 levels were impaired in vim1-3 (Fig. 3H, I). In combination, these results suggest that while lhp1 mutation affects DNA methylation of a VIM1-target region, vim1 mutation impairs LHP1 binding and H3K27me3 deposition on the same locus, hinting at a cooperative interaction between these two epigenetic factors. To test any direct cooperative interaction, we performed bimolecular fluorescence complementation (BiFC) assays in N. benthamiana leaves transiently transformed with CYPF-VIM1 and NYFP-LHP1, which demonstrated that both proteins can interact in vivo (Fig. 3J; Additional file 1: Fig. S8). Altogether, our results established that the lncRNA APOLO associates with both a regulator of DNA methylation (VIM1) and a PcG component (LHP1), which further cooperate to mediate DNA methylation and H3K27me3 deposition at a specific APOLO target locus.

Human lncRNA UPAT can exert similar regulatory functions as APOLO in planta

Arabidopsis VIM1 is a homolog of the mammalian methylcytosine-binding protein UHRF1 [31, 64]. Interestingly, UHRF1 was shown to directly recognize in vivo the human lncRNA UPAT, through its flexible spacer region positioned between the SRA and RING domains [34, 65, 66]. This interaction stabilizes UHRF1 by interfering with its ubiquitination and subsequent degradation [34]. In order to determine if the capacity of VIM1 and UHRF1 to interact with lncRNAs is conserved between plants and animals, we first delimited the minimal region of VIM1 that interacts with APOLO. We generated independent GFP fusion constructs bearing different VIM1 coding regions, according to Woo et al. [27] (Fig. 4A right panel). All the construct-encoded proteins accumulated exclusively or partially in the nucleus of N. benthamiana cells, with or without APOLO co-expression (Additional file 1: Fig. S9A-B). Similar to the reported interaction between UHRF1 and UPAT [34], an anti-GFP RIP performed in N. benthamiana leaves transiently co-transformed with APOLO and GFP-VIM1 derivatives revealed that the two VIM1 portions containing the full spacer region (GFP-SRA/RING2 and GFP-Spacer) bind to APOLO in vivo, although with lower efficiency than the full-length VIM1 (Fig. 4A). Thus, our results indicated that the lncRNAs-binding capacity of the SRA/RING-spacer regions of VIM1 and UHRF1 is evolutionary conserved between plants and animals.

Fig. 4
figure 4

The UHRHF1-interacting lncRNA UPAT binds to VIM1 and LHP1 in plant cells and decoys the complex away from chromatin. A RNA immunoprecipitation (RIP) assay in Nicotiana benthamiana leaves transiently co-transformed with APOLO and translational fusions expressing GFP-VIM1 or derivatives (GFP-PHD/RING1, GFP-SRA/RING2, GFP-SRA, GFP-RING2, GFP-SPACER) under the control of the 35S-CaMV promoter. Schematic representation of VIM1 protein and derivatives tested for RIP is shown on the right. Amino acid coordinates are indicated between brackets. B, C RIP assay in N. benthamiana leaves transiently co-transformed with UPAT and GFP-UHRF1, GFP-VIM1B, or LHP1-GFPC translational fusions expressed under the control of the 35S-CaMV promoter. D–F In vitro interaction of recombinant VIM1:GFP, LHP1: and the canonical RNA-binding protein NSRa, with APOLO, UPAT, GFP mRNA, and the ASCO-interacting lncRNA ASCO. Equimolar concentrations of each RNA (for 100 ng of APOLO as a reference) were incubated with purified proteins before proceeding with a regular RIP. G–H Chromatin immunoprecipitation (ChIP)-qPCR analysis of LHP1 (in D) and VIM1 (in E) binding at the YUC2 promoter in wild-type (WT), APOLO over-expression (OE APOLO-1), and UPAT over-expression (OE UPAT) plants transiently transformed or not with GFP-VIM1. In A–H, results are expressed as a percentage of the INPUT fraction. Anti-IgG antibodies were used as a negative control. Bars represent average ± SD (n = 3 biological replicates). I Venn diagrams showing the overlap of upregulated and downregulated genes in plants constitutively expressing APOLO or UPAT vs. WT 4-day-old plants grown in standard conditions (23 °C). The p-value indicated below was calculated by hypergeometric test considering all genes annotated in Araport11. Fifty-three percent of common DEG are responsive to warmth in WT plants (Additional file 2: Table S5b)

Full size image

Considering the common ability of the homologous VIM1 and UHRF1 to interact with lncRNAs, we wondered whether the lncRNA UPAT, exhibiting no apparent sequence homology with APOLO (Additional file 1: Fig. S9C) but potentially sharing functional secondary structures (predicted by using RNAfold; [67, 68]; Additional file 1: Fig. S9D), could also interact with VIM1 and its partner LHP1 in planta. Strikingly, an anti-GFP RIP in N. benthamiana leaves transiently co-transformed with UPAT and GFP-UHRF1, GFP-VIM1, or LHP1-GFP demonstrated that UPAT is able to interact with UHRF1, as well as with VIM1 and LHP1 in plant cells (Fig. 4B, C). In order to further characterize the interaction between VIM1 and LHP1 with APOLO and UPAT, we purified recombinant GFP-tagged proteins and incubated them with different combinations of in vitro-transcribed APOLO, UPAT, and GFP mRNA, and with the unrelated lncRNA ASCO, previously linked to alternative splicing regulation [69]. We compared VIM1 and LHP1 binding to each transcript with the ASCO-partner NSRa, a canonical RNA-binding protein [70]. VIM1 and LHP1 exhibited specific recognition of APOLO and UPAT over GFP mRNA and ASCO. Interestingly, in the presence of equivalent molar concentrations of APOLO and UPAT, LHP1 showed a preferential binding to APOLO, in contrast to VIM1 (Fig. 4D–F).

Finally, we assessed if the constitutive expression of UPAT or APOLO could modulate VIM1 and LHP1 binding to YUC2 promoter. ChIP-qPCR analyses performed in WT, OE APOLO-1, and OE UPAT Arabidopsis stable lines transiently transformed or not with GFP-VIM1 (Additional file 1: Fig. S1B) revealed that both APOLO and UPAT over-expression impairs the interaction of VIM1 and LHP1 with YUC2 promoter (Fig. 4G, H). Furthermore, UPAT constitutive expression in stable Arabidopsis seedlings resulted in decreased transcript levels of YUC2 (Additional file 1: Fig. S10A-B). Moreover, by performing a new RNA-Seq comparing the transcriptional profile of WT vs. OE APOLO-1 and OE UPAT-1 plants, a significant overlap between APOLO and UPAT-deregulated genes emerged (both for up and downregulated genes), including 53% of common differentially expressed genes (DEG) which are modulated by warmth, strongly suggesting that they exert similar molecular mechanisms (Fig. 4I; Additional file 2: Table S5a and b). Altogether, our results revealed a functional interaction between sequence-unrelated long noncoding transcripts and the key epigenetic regulators LHP1 and VIM1, which are remarkably conserved across kingdoms (Fig. 5 for a model in plants and a comparison of UPAT and APOLO integrated complexes across kingdoms).

Fig. 5
figure 5

Long noncoding RNA-mediated regulation of DNA and histone methylation in plants and animals uncovers mechanistic commonalities across kingdoms. A Model for the regulation of YUC2 expression in response to heat by the APOLO-VIM1-LHP1 complex: at 23 °C, APOLO lncRNA recognizes the YUCCA2 (YUC2) locus by sequence complementarity and R-loop formation. APOLO interacts with VIM1 and LHP1 over the YUC2 promoter region, which exhibits RdDM and H3K27me3 deposition, blocking YUC2 transcription. meDNA (light brown balls) and H3K27me3 (dark brown balls) are cooperatively maintained by the VIM1-LHP1 complex. At 29 °C, APOLO transcript levels decrease and VIM1-LHP1 binding to the YUC2 promoter region is reduced. Conjointly, 24nt siRNA accumulation is decreased, impairing RdDM over the YUC2 promoter. As a result, YUC2 transcriptional activity increases. YUC2 participates in auxin synthesis, promoting hypocotyl elongation in response to heat. B Proposed explanation for the molecular impact of APOLO or UPAT constitutive expression. Although the ability of UPAT to form R-loops in animals (or when expressed in plants) remains unknown, high levels of both lncRNAs in Arabidopsis achieve to decoy VIM1 and LHP1 away from chromatin, hinting at a stoichiometrically modulated mechanism. C Comparison between conserved lncRNA-integrated epigenetic machineries across kingdoms. Homolog proteins between Homo sapiens and Arabidopsis thaliana are shown in the same color. References of the articles demonstrating direct interactions or co-localizations are indicated. Question marks “?” mean that the interaction has not been proved yet

Full size image

Discussion

Epigenetic transcriptional reprogramming in the thermomorphogenic response

When plants are exposed to warmer nonstressful temperatures, some organs grow and develop at a faster rate without affecting their final dimensions, whereas others suffer morphological or developmental changes. The latest response is known as thermomorphogenesis and includes petiole elongation, leaf hyponasty, and auxin-dependent hypocotyl elongation [71,72,73]. These modifications of the plant architecture rely on a deep transcriptional reprogramming determined by dynamic changes of chromatin organization [74,75,76,77,78,79].

A strong correlation between temperature and DNA methylation was first established in Swedish A. thaliana accessions, where higher levels of CpHpH methylation were observed in plants grown at 16 °C compared to 10 °C [80, 81]. Interestingly, CpHpH methylation is conjointly regulated by CMT2 and the RdDM pathway [24,25,26] and both have been involved in the heat response. It was shown that cmt2 mutants are more tolerant to heat stress [82] and that the heat-stress response relies on the integrity of the RdDM pathway [83, 84]. In agreement, we observed that cmt2 mutants displayed a slight but significantly increased response to warm temperatures in contrast to RdDM mutants exhibiting hypocotyl elongation defects. We also observed that met1 mutants exhibited the same developmental phenotype as RdDM mutants. Accordingly, it was demonstrated that met1 knockout causes a loss of DNA methylation, a global loss of Pol V at its normal locations and a Pol V redistribution to sites that become hypermethylated [62]. Furthermore, we observed a reduction in CpHpH methylation at the YUC2 promoter in the vim1-3 mutant, linking the canonical CpG-related machinery with the RdDM pathway.

Importantly, met1 was also shown to suffer H3K9 hypermethylation at PcG target genes and a redistribution of H3K27me3 to transposons [43], hinting at a relation between DNA methylation and PRC-dependent histone modifications. Consistently, we uncovered here a new molecular link between the DNA methylation-associated protein VIM1 and the H3K27me3-related protein LHP1, which can interact together and with the lncRNA APOLO. An indirect relation between LHP1 and VIM1 had already been described through the histone methyltransferase NtSET1, a common interactor. It was first shown in Nicotiana tabacum that NtSET1 directly interacts with LHP1 [85], and it was then reported that NtSET1 can bind to AtVIM1 in planta [86], although no link was established between LHP1 and DNA methylation.

Although VIM1, VIM2, and VIM3 have been reported to function redundantly to mediate epigenetic transcriptional silencing [28], we showed here that vim1 single mutants exhibit particular molecular and related physiological phenotypes in response to warmth. The expression of VIM1 is induced by warmth, whereas VIM2 is repressed. Interestingly, we showed that although only VIM1 was first identified by ChIRP-MS as a potential APOLO partner, VIM2 is also able to weakly but significantly recognize APOLO in N. benthamiana leaves, suggesting that the VIM2-APOLO interaction may also occur during Arabidopsis development and that lncRNA-mediated regulation of VIM protein activity is a general feature in plants. Further research will be needed to uncover if VIM2 and VIM3 are also involved in thermomorphogenesis.

Here, we related the hypocotyl elongation defect observed in vim1 to the transcriptional activation of the YUC2 gene [58, 59], likely among other auxin-related genes emerging as potential APOLO and VIM1 common targets (Additional file 2: Table S3). The control of APOLO over auxin synthesis and transport may affect global auxin homeostasis exceeding YUC2. Therefore, as expected, we cannot mimic all APOLO-related phenotypes in the yuc2 single mutant. YUC2 regulation in response to heat was first observed in flowers, where heat shock (33 °C) repressed its expression and caused plant male sterility [59]. At milder temperatures (27 °C), YUC2 transcript levels were reported to increase, in correlation with the demethylation of YUC2 proximal promoter region targeted by RdDM [63], suggesting that cytosine methylation at the YUC2 regulatory region is critical for the control of its expression in response to heat. In addition to DNA methylation, a correlation was established between the deposition of H3K27me3 and genes exhibiting either high or low transcription rates under warm temperatures [87]. Consistently, we observed that the lhp1 mutant displayed impaired hypocotyl elongation and showed no activation of YUC2 at 29 °C. Interestingly, LHP1 was previously reported to exert a positive role in the regulation of YUCCA genes under standard growth conditions. YUC5, YUC8, and YUC9 displayed an abnormal transcriptional activity in the lhp1 mutant [88], correlating with a reduction of H3K27me3, as revealed by genome-wide approaches [49]. In addition, LHP1 was found to localize to YUC2, whose expression was not affected by the lhp1 mutation at this developmental stage, in an auxin-dependent fashion [88].

It was recently shown in Arabidopsis that the 5′ untranslated region of several mRNAs may adopt alternative hairpin conformations under warm cycling daytime temperatures, modulating translation efficiency. It was proposed that mRNA thermoswitches enhancing protein synthesis may constitute a conserved mechanism enabling plants to respond rapidly to high temperatures [89]. In contrast, the potential role of lncRNAs as thermosensors in higher organisms remains unexplored. In this work, we showed that APOLO participates in the thermomorphogenic response in Arabidopsis. Interestingly, APOLO transcriptional accumulation increases in response to cold [53], whereas we showed here that it decreases under warm temperatures. Further research will be needed to uncover the effect of temperature on the structure of APOLO and other lncRNAs that determines their molecular role and contributes to the plant adaptation to environmental constraints.

R-loop-associated long noncoding RNAs modulate DNA methylation and histone modifications

RNA-DNA hybrids (R-loops) have been identified as important regulators of chromatin conformation and gene transcriptional regulation [90, 91]. APOLO is known to recognize its multiple trans targets by sequence complementarity and R-loop formation. Upon interaction, it decoys LHP1, shaping local chromatin 3D conformation and subsequently modulating gene transcriptional activity [52]. Here, we showed that APOLO recognizes YUC2 gene body through the formation of R-loops. Recently, it was shown that the human lncRNA HOTTIP can form R-loops in trans at the base of CTCF-related topologically associated domains (TADs), suggesting that lncRNA-DNA duplexes may exert a general role in 3D chromatin conformation [92].

Interestingly, APOLO additionally interacts with VIM1 to co-regulate DNA methylation at the YUC2 promoter, mediating YUC2 transcriptional response to warm temperatures. Although the involvement of lncRNAs in the regulation of DNA methylation in plants remains largely unknown, previous reports have linked lncRNA activity to promoter methylation in mammals, through the interaction with protein partners [93] or the formation of R-loops [94, 95]. Notably, it was reported that DNMT1-interacting lncRNAs can promote or block local DNA methylation [93, 96, 97], whereas it was proposed that R-loops formed by nascent transcripts can preclude DNA methylation from promoter regions by repelling DNA methyltransferases [98, 99].

In addition to DNA methylation, a link between R-loops and the activity of PRC complexes has also been proposed. In mammals, it was shown that R-loops can form at a subset of PcG target genes, resulting in alternative regulatory outputs depending on the interplay between the DNA-RNA duplex and PRC components [100]. In this work, we showed that the over-accumulation of APOLO can titer VIM1 and LHP1 away from the YUC2 promoter region. It is thus tempting to hypothesize that R-loops formed by APOLO at the YUC2 locus stoichiometrically modulate VIM1 and LHP1 recruitment to DNA in response to changes in ambient temperature. At 23 °C, a precise amount of APOLO over the YUC2 locus may guide the LHP1-VIM1 complex to the target region. In contrast, at 29 °C APOLO transcript levels decrease, hindering VIM1 and LHP1 efficient recognition of YUC2. On the other hand, over-accumulation of APOLO can decoy LHP1 and VIM1 away from the YUC2 locus, explaining the similar hypocotyl elongation phenotypes of OE APOLO, vim1, and lhp1, as well as the similar DNA methylation levels observed at the YUC2 promoter in these three backgrounds (Fig. 5A, B). The re-methylation of the YUC2 promoter in OE APOLO, vim1, and lhp1 at warm temperatures, in contrast to WT, might be due to altered RdDM activity when any component of the regulatory complex is affected. Thereby, we propose that APOLO functions as a key mediator of the PRC and DNA methylation machineries, hinting at a stoichiometric factor fine-tuning the activity of R-loop-related lncRNAs.

Long noncoding RNAs as key mediators of the Polycomb and DNA methylation machineries

Connections between Polycomb and DNA methylation machineries have been reported in few studies in mammals. DNMT1 was shown to directly interact with HP1 to mediate silencing of euchromatic genes [101], and its enzymatic activity is also stimulated by direct interaction with UHRF1 [102,103,104,105,106]. Interestingly, HP1 and UHFR1 have been shown to co-exist on chromatin [107], although their direct interaction has not been proven. Here, we showed that LHP1 and VIM1 interact in planta, suggesting that a direct interaction between their homologs HP1 and UHRF1 may occur in mammals. As in mammals, it was shown in Arabidopsis that VIM1 recruits MET1 to hemi-methylated DNA [40, 44, 64, 108], hinting at the existence of analogous complexes formed by MET1/DNMT1, VIM1/UHRF1, and LHP1/HP1 in plants and animals (Fig. 5C).

Remarkably, DNMT1, UHRF1, and HP1 were all shown to interact with coding or long noncoding RNAs [34, 97, 109, 110], whereas only LHP1 was previously shown to bind to RNAs in vitro [50] and in vivo [51, 52, 111] in plants. Here, we uncovered the interaction between VIM1 and lncRNAs. Notably, UHRF1 can interact with the lncRNA UPAT whose scaffold function prevents UHRF1 from ubiquitination and proteasomal degradation [34]. Knockdown of UPAT results in a drastic decrease in the levels of DNMT1 protein and a reduction of hemimethylation at its target genes [34]. Here, we showed that UPAT can interact with VIM1 and LHP1 in plant cells, and high amounts of UPAT can decoy VIM1 and LHP1 from YUC2 promoter, as the sequence-unrelated transcript APOLO. Furthermore, the transcriptomics of plants constitutively expressing UPAT significantly resemble the output of APOLO over-expression, further supporting that common molecular mechanisms can be mediated by both noncoding transcripts.

Although a global anti-correlation is observed in the Arabidopsis genome between DNA methylation and H3K27me3 deposition, characterizing heterochromatic and euchromatic regions respectively [11, 43, 45, 112], DNA hypomethylation was associated to H3K27me3 reduction at specific genes in met1 and vim1/2/3 mutants [43, 44]. The increasing evidence that PRC2 also targets transposons across distantly related eukaryotic lineages has recently served to hypothesize that an ancestral role of Polycomb is linked to silencing repetitive genomic elements [113]. Consistently, our results indicate that APOLO lncRNA interacts with LHP1 and VIM1 to regulate YUC2, constituting a lncRNA-mediated non-canonical cooperative interaction between Polycomb and DNA methylation machineries. Here we showed that H3K27me3 deposition and LHP1 binding to YUC2 was impaired in vim1 mutants, whereas DNA methylation was altered in lhp1 mutants. We cannot exclude that indirect effects mediated by transcriptional activation of the YUC2 gene may be affecting the interplay between the two epigenetic marks. Although other lncRNAs have been linked to thermomorphogenesis or light-dependent hypocotyl elongation, their mechanisms of action or target recognition remain unknown [114, 115]. Considering that the sequence-unrelated lncRNA UPAT mimicked the action of APOLO over the chromatin-binding capacity of the VIM1-LHP1 complex and global transcriptional reprogramming, it is likely that additional yet-uncovered lncRNAs exert a similar role as APOLO in Arabidopsis.

Conclusions

Altogether, our results hint at a mechanism possibly conserved across kingdoms, which is likely based on the structure rather than the sequence of long noncoding transcripts. A deeper knowledge about the molecular basis behind lncRNA-related regulatory mechanisms will likely push back the frontiers of human therapeutics and will allow the design of innovative strategies for sustainable agriculture in a climate change context.

Methods

Plant material and growth conditions

All the Arabidopsis thaliana lines used were in the Columbia-0 (Col-0) background. vim1-2 (SALK_050903) and vim1-3 (SALK_149277c) seeds were obtained from the Arabidopsis Biological Resource Center (http://www.arabidopsis.org/). Mutant plants were genotyped by PCR using specific primers to amplify the endogenous locus and T-DNA borders (primers used are listed in Additional file 2: Table S4). VIM1 and UPAT over-expression (OE), as well as CRISPR-Cas9-APOLO transgenic plants, were generated through Agrobacterium tumefaciens (strain EHA105 or ASE containing the pSOUP helper plasmid respectively)-mediated transformation [116]. Two independent lines of transformants harboring GFP-VIM1 or UPAT were selected on MS/2 medium supplemented with kanamycin (40 μg/mL). VIM1 expression levels were measured by RT-qPCR (primers used are listed in Additional file 2: Table S4). Transgenic CRISPR-Cas9-APOLO lines were selected for Basta resistance. T2 plants were genotyped for heterozygous deletion of APOLO, and T3 plants were selected for homozygous deletion of APOLO (confirmed by PCR followed by Sanger sequencing) and for the absence of T-DNA (primers used are listed in Additional file 2: Table S4).

Seeds were surface-sterilized by treatments with 70% EtOH and 5% hypochloryte-1% SDS, washed, and stratified at 4 °C for 2 days to obtain homogeneous germination. Seedlings were grown at 23 °C or 29 °C on solid half-strength Murashige and Skoog (MS/2) medium (Duchefa), under long day conditions (16 h light, 95 μE m−2 s−1/8 h dark, ARALAB Fitoclima 600, LED light). For phenotypic characterization of hypocotyl elongation, seedlings were grown for 3 h at 23 °C and transferred to 29 °C or maintained at 23 °C for 4 days. Hypocotyl lengths were measured using the ImageJ software. For RNA sequencing, chromatin immunoprecipitation, and methylated DNA immunoprecipitation assays performed in Arabidopsis stable lines, seedlings were grown for 4 days at 23 °C and transferred to 29 °C or maintained at 23 °C for 6 h. For chromatin isolation by RNA purification and DNA-RNA immunoprecipitation, seedlings were grown for 11 days at 23 °C.

Cloning procedures

The entire coding region of VIM1 (AT1G57820), VIM1 derivatives (PHD/RING1, SRA/RING2, SRA, RING2; [27]), LHP1 (AT5G17690), and the lncRNA UPAT [34] were amplified by PCR on genomic DNA or pBluescript:UPAT vector [34] respectively and cloned into the Gateway entry vector pENTR/D-TOPO (Invitrogen). The entire coding region of VIM2 (AT1G66050) was amplified by PCR on cDNA and cloned into the pDONOR207 vector by Gateway technology (BP reaction) (Invitrogen). Entry clones were recombined by Gateway technology (LR reaction) into the pK7WGF2, pK7FWG2, or pK7WG2 vectors containing respectively a p35S-GFP-GW, a p35S-GW-GFP, or a p35S-GW cassette (http://www.psb.ugent.be/gateway/index.php). pENTR/VIM1, pENTR/SRA/RING2, and pENTR223/LHP1 were recombined by LR reaction into the pGPTVII.Bar-C/NYFP vectors containing respectively a p35S-CYFP or a p35S-NYFP cassette (http://www.psb.ugent.be/gateway/index.php). The spacer of VIM1 and the entire coding region of UHFR1 were amplified by PCR on cDNA and pGE-HIS-MBP UHRF1 plasmid respectively and cloned into the GreenGate entry vector pGGC000. Destination vectors were constructed associating the modules: 35S promoter, GFP, Ubiquitin 10 terminator, and pNOS:BastaR, tNOS, according to Lampropoulos et al. [117]. p35S-mCherry was constructed associating the modules: 35S promoter, mCherry, and NOS terminator [117]. All Gateway constructs were subsequently transformed into the A. tumefaciens strain EHA105 and GreenGate constructs into the A. tumefaciens strain ASE containing a pSOUP helper plasmid.

To generate the CRISPR/Cas9 APOLO system, modified Greengate vectors [117] were used. A fragment containing two BpiI recognition sites between the AtU6-26 promotor and the sgRNA scaffold was synthetize by the Eurofins genomics service and further subcloned into pGGA000, pGGB000, pGGC000, and pGGD000. The obtained pGGA010, pGGB010, pGGC010, or pGGD010 plasmids were further used to clone specific guide RNA sequences (20 nt in length-NGG) in a one-step digestion-ligation reaction with the BpiI restriction enzyme. Four specific guide RNAs targeting APOLO genomic sequence were designed using the TEFOR website (http://crispor.tefor.net/crispor.py). For each sgRNA, forward and reverse primers were designed with complementary overhangs to BpiI digested plasmids, annealed, and subcloned into pGGA/B/C/D010 vectors (primers used are listed in Additional file 2: Table S4). Cas9 sequence [118, 119] from Streptococcus pyogenes (SpCas9) was subcloned in pGGE000 under the control of an AtUbi10 promotor using the pDe-CAS9 vector [118] as a matrix and 5′-TGTGAAGACAACCATGGATAAGAAGTACTCTATC-3′ and 5′-TGTGAAGACTTTAGTAAGCCTATACTGTACTTAAC-3′ as forward and reverse primers to create the pGGE011 entry plasmid. The four pGGA/B/C/D-sgRNA specific entry vectors, pGGE011, and pGGF008 (containing a Basta resistance cassette) were subsequently assembled into the pGGZ001 destination vector in a one-step digestion-ligation reaction using BsaI (Additional file 1: Fig. S11).

Transient expression and confocal microscopy

Leaves of Nicotiana benthamiana were transiently transformed as described in Waadt and Kudla [120]. For RNA immunoprecipitation, GFP-VIM1 or derivatives (GFP-PHD/RING1, GFP-SRA/RING2, GFP-SRA, GFP-RING2, and GFP-Spacer) were co-transformed with p35S-APOLO [52], or GFP-UHRF1, GFP-VIM1, or LHP1-GFP were co-transformed with p35S-UPAT. For sub-cellular localization, VIM1 protein or derivatives were co-transformed or not with p35S-APOLO. For bimolecular fluorescence complementation, YFPC-VIM1 or YFPC-SRA/RING2 were co-transformed with YFPN-LHP1. mcherry was co-transformed and used as a membrane and nuclear marker. Leaves were harvested or cells were analyzed 2 days after infiltration using a Zeiss confocal microscope equipped with Plan-Apochromat 10x/NA 0,45/M27 or Plan-Apochromat 20x/NA 0,8/r M27 US-VIS-IR dry lens. eGFP, eYFP, and mCherry fluorescence were excited with 488- and 514-nm argon laser lines and 561-nm diode, emission recorded between 490 and 580 nm, 520 and 600 nm, and 580 and 640 nm respectively. Image acquisitions and analyses were performed on the IPS2 Imaging Facility.

Leaves of 3-week-old Arabidopsis thaliana were transiently transformed as described in Zhang et al. [121]. Transformation with GFP-VIM1 was carried out in leaves > 0.5 cm in length, in 10 to 15 plants per genotype (WT, OE APOLO-1, and OE UPAT). Leaves were crosslinked and harvested (for ChiP) or cells were analyzed 3 days after infiltration, using a Leica TCS SP8 confocal laser scanning microscope. eGFP was excited at 488 nm (intensity=8%) and emission recorded between 495 and 530 nm for GFP and 610 and 670 nm (gain 650) for chlorophyll fluorescence. Images were processed using the Fiji software [122].

Chromatin isolation by RNA purification and mass spectrometry or dot blot

ChIRP-qPCR was performed as previously described [51, 52]. Formaldehyde-crosslinked seedlings were ground, and nuclei were isolated and sonicated using a water bath Bioruptor Pico (Diagenode; 30 s on/ 30 s off pulses, at high intensity for 10 cycles). APOLO-associated chromatin was purified using 100 pmol of two independent sets of primers (ODD and EVEN) and a negative control (primers matching the LacZ RNA), respectively [51, 52]. Primers used are listed in Additional file 2: Table S4.

A method adapted from the ChIRP protocol [123] was developed to allow identification of plant nuclear proteins bound to specific lncRNAs, as described in Rigo et al. [69]. Briefly, plants were in vivo crosslinked, and cell nuclei were purified and extracted through sonication. The resulting supernatant was hybridized against biotinylated complementary oligonucleotides that tile the lncRNA of interest, and putative lncRNA-containing protein complexes were isolated using magnetic streptavidin beads. Co-purified ribonucleoprotein complexes were eluted and used to purify RNA or proteins, which were later subject to downstream assays for identification and quantification.

Crosslinking and ribonucleoprotein complex purification

For protein extraction, approximately 250 g of 7-day-old Col-0 plants grown on solid half-strength MS medium was irradiated three times with UV using a CROSSLINKERCL-508 (Uvitec) at 0.400 J/ cm2. For RNA extraction, 10 g of 7-day-old Col-0 plants grown on solid MS/2 medium was crosslinked under vacuum for 15 min with 37 ml of 1% (v/v) formaldehyde. The reaction was stopped by adding 2.5 ml of 2M glycine, and seedlings were rinsed with Milli-Q purified water. For both crosslinking methods, 6 g of the fixed material was ground in liquid nitrogen (representing 15 ml of plant material ground to fine dust) and added to 50-ml tubes with 25 ml of extraction buffer 1– (the nuclei were prepared starting with 30 tubes; buffer 1: 10 mM Tris–HCl pH 8, 0.4 M sucrose, 10 mM MgCl2, 5mM β-mercaptoethanol, 1 ml/30 g of sample powder Plant Protease Inhibitor Sigma P9599). The solution was then filtered through Miracloth membrane (Sefar) into a new tube and 5 ml of extraction buffer 2 (10mM Tris–HCl pH8, 0.25 M sucrose, 10 mM MgCl2, 5 mM β-mercaptoethanol, 1% Triton X-100, 10 μl 100 μl PMSF) was added. The solution was then centrifuged, the supernatant discarded, and the pellet was resuspended in 500 μl of extraction buffer 3 (10 mM Tris–HCl pH 8, 1.7 M sucrose, 2 mM MgCl2, 5 mM β-mercaptoethanol, 0.15% Triton X-100, 50 μl protease inhibitor) and layered on top of 600 μl of fresh extraction buffer 3 in a new tube. After centrifugation at 13000 rpm for 2 min at 4 °C to pellet the nuclei, the supernatant was discarded and the pellet resuspended in 300 μl of nuclei lysis buffer (50 mM Tris–HCl pH7, 1% SDS, 10 mM EDTA, 1 mM DTT, 50 μl protease inhibitor, 10 μl RNAse inhibitor per tube) to degrade nuclear membranes. Samples were sonicated three times in refrigerated Bioruptor Plus (Diagenode), 30 cycles 30 s ON/30 s OFF in a Diagenode TPX microtube M-50001. After centrifugation, the supernatant was transferred to a new tube and diluted two times volume in hybridization buffer (50 mM Tris–HCl pH 7, 750 mM NaCl, 1% SDS, 15% formamide, 1 mM DTT, 50 μl protease inhibitor, 10 μl RNAse inhibitor). One hundred picpmoles of probes against APOLO (ODD and EVEN set of probes, [51, 52]) and the corresponding negative set against LacZ were added to samples and incubated 4 h at 50°C in a thermocycler. Samples were transferred to tubes containing Dynabeads- Streptavidin C1 (Thermo Fisher Scientific) and incubated 1 h at 50 °C. Then, samples were placed on a magnetic field and washed three times with 1 ml of wash buffer (2× SSC, 0.5% SDS, 1 mMDTT, 100 μl protease inhibitor). Protein purification samples for protein extraction were DNase-treated according to the manufacturer (Thermo Scientific). After addition of 1.8 ml of TCA-acetone (5 ml 6.1N TCA + 45 ml acetone + 35 μl β-mercaptoethanol), samples were incubated overnight at −80 °C. After centrifugation at 20,000 rpm for 20 min at 4 °C, the supernatant was discarded and 1.8 ml of acetone wash buffer (120 ml acetone, 84 μl β-mercaptoethanol) was added to the samples. Then, samples were incubated 1 h at −20 °C and centrifuged again at 20,000 rpm for 20 min at 4 °C. The supernatant was discarded, and the dry pellet was used for mass spectrometry analyses.

For ChIRP-dot blot analyses, three independent samples of 10 g each of OE VIM1-1 seedlings were used for APOLO-ChIRP. Co-purified proteins without precipitation were deposed on PVDF membrane (Roche, Basel, Switzerland) pre-activated with methanol. Twenty microliters of ODD, EVEN, and LacZ samples was deposed, and the same amount of a 1/50 dilution of the input fraction, to avoid saturation of the signal. Blots were hybridized with polyclonal rabbit antibodies against GFP (Abcam ab290; dilution 1:1,000) and developed with anti-rabbit immunoglobulin (Agrisera, dilution 1:15,000) conjugated with horseradish peroxidase using the AgriseraECL SuperBright western blot detection reagent.

DNA-RNA immunoprecipitation

DNA-RNA immunoprecipitation (DRIP) was performed as described in Ariel et al. [52]. Non-crosslinked seedlings were used for nuclei purification, and samples were sonicated using a water bath Bioruptor Pico (Diagenode; 30 s on/ 30 s off pulses, at high intensity for 4 cycles). Chromatin samples were incubated with 50 ml of washed Protein G Dynabeads pre-coated with 1 mg of S9.6 antibody (Millipore MABE1095) for 16 h at 4 °C. Samples treated with RNAseH (Invitrogen) for 2 h at 37 °C were used for DRIP in parallel as a negative control. After washing, DNA was recovered for qPCR over R-loop-related loci (primers used are listed in Additional file 2: Table S4).

RNA sequencing analysis

Total RNA was prepared from 4-day-old A. thaliana wild-type (WT), OE APOLO-1 [51], and vim1-3 seedlings (grown in ARALAB Fitoclima 600, 95 μE m−2 s−1 of LED light) using the QIAGEN RNeasy plant mini kit and treated with DNase (Fermentas). Three independent biological replicates were produced per genotype. For the RNA-Seq of WT, OE APOLO-1, and OE UPAT-1 (Fig. 4, Additional file 2: Table S5), plants were grown in ARALAB Fitoclima 600, 95 μE m−2 s−1 of fluorescent light. After RNA extraction, libraries were constructed using the Tru-Seq Stranded mRNA Sample Prep kit (Illumina®). Sequencing was carried out at the IPS2. The Illumina NextSeq500 technology was used to perform 75-bp sequencing. A minimum of 15 million of single end reads by sample was generated. RNA-seq preprocessing included trimming library adapters, and quality controls were performed with Trimmomatic [124]. Paired end reads with Phred Quality Score Qscore > 20 and read length > 30 bases were kept, and ribosomal RNA sequences were removed with SortMeRNA [125]. Processed reads were aligned using STAR with the following arguments: --alignIntronMin 20 --alignIntronMax 3000 --outSAMtype BAM SortedByCoordinate --alignIntronMax 3000 --outSAMtype BAM SortedByCoordinate --alignIntronMax 3000 --outSAMtype BAM SortedByCoordinate. Read overlapping exons per genes were counted using the featureCounts of the subreads package using the GTF annotation files from the Araport11 project [126]. Significance of differential gene expression was estimated using DEseq2 [127] and the FDR correction of the p-value was used during pairwise comparison between genotypes. A gene was declared differentially expressed if its adjusted p-value (FDR) was < 0.01. Heatmap was generated using log2 transformed fold change values compared WT 23 °C and computed with the pheatmap R [128].

RNA immunoprecipitation assay

RNA immunoprecipitation (RIP) assays were performed on transiently transformed N. benthamiana leaves as described in Sorenson and Bailey-Serres [129], or in 2-week-old A. thaliana OE VIM1.1 seedlings as described in Bardou et al. [70], using anti-GFP (Abcam ab290) and anti-IgG (Abcam ab6702). RIP was performed using Invitrogen Protein A Dynabeads. Precipitated RNAs were prepared using TRI Reagent (Sigma-Aldrich), treated with DNaseI (Fermentas) and subjected to RT-qPCR (High Capacity cDNA Reverse Transcription Kit (Thermo); primers used are listed in Additional file 2: Table S4). Total RNAs were processed in parallel and considered as the input sample.

Chromatin immunoprecipitation assay

Chromatin immunoprecipitation (ChIP) assays were performed using anti-GFP (Abcam ab290), anti-LHP1 (Covalab pab0923-P), anti-H3K27me3 (Diagenode pab-195-050), and anti-IgG (Abcam ab6702), as described in Ariel et al. [52]. Crosslinked chromatin was sonicated using a water bath Bioruptor Pico (Diagenode; 30 s ON/30 s OFF pulses; 10 cycles; high intensity). ChIP was performed using Invitrogen Protein A Dynabeads. Precipitated DNA was recovered using Phenol:Chloroform:Isoamilic Acid (25:24:1; Sigma) and subjected to RT-qPCR (primers used are listed in Additional file 2: Table S4). Untreated sonicated chromatin was processed in parallel and considered as the input sample.

In vitro protein-RNA interaction assay

Recombinant VIM1:GFP, LHP1:GFP, and NSRa:GFP were purified from 21-day-old stably transformed Arabidopsis plant leaves without crosslinking. Nuclei were isolated as previously described and immunoprecipitation was performed using the anti-GFP Abcam ab290 antibody and 20 μg of RNAseA (Thermo) and 10 U of DNAseI (Fermentas) per sample. Protein purification included astringent washes as for histone ChIP protocol [52]. For in vitro transcription of the APOLO, UPAT, ASCO, and GFP RNAs, 1 μg of purified DNA of each template including the T7 promoter at the 5′ end was used following the manufacturer instructions (HiScribe T7 High Yield RNA Synthesis kit, NEB). One hundred nanograms of each of the corresponding RNAs was incubated with protein A Dynabeads (Thermo) containing the recombinant purified proteins, as previously described [70]. After 1 h of incubation at 4 °C in binding buffer 1X (10 mM HEPES, pH 7.0; 50 mM KCl; 10% glycerol; 1 mM EDTA; 1 mM dithiothreitol; 0.5% Triton X-100), three 5-min washes were done with the same buffer in rotation at 4 °C. Precipitated RNAs were purified using TRI Reagent (Sigma-Aldrich), treated with DNaseI (Fermentas) and subjected to RT-qPCR (High Capacity cDNA Reverse Transcription Kit (Thermo); primers used are listed in Additional file 2: Table S4).

Methylated DNA immunopreciptation

Methylated DNA immunoprecipitation (MedIP) assays were performed using anti 5-mC (Diagenode Mab-081-100) and anti-IgG (Abcam ab6702), as described in Nagymihály et al. [130]. Genomic DNA (1 μg) was sonicated using a water bath Bioruptor Pico (Diagenode; 30 s on/30 s off pulses; 4 cycles; high intensity). MedIP was performed using Invitrogen Protein G Dynabeads. Precipitated DNA was recovered using Phenol:Chloroform:Isoamilic Acid (25:24:1; Sigma) and subjected to RT-qPCR (primers used are listed in Additional file 2: Table S4). Untreated sonicated genomic DNA was processed in parallel and considered as the input sample.

RT-qPCR

RT-qPCR were performed using the LightCycler 480 SYBR Green I Master Kit on a StepOne Plus apparatus (Applied Biosystems) using standard protocols (40 to 45 cycles, 60 °C annealing). PP2A (AT1G13320) was used for normalization (primers used are listed in Additional file 2: Table S4) for ∆∆Ct quantification method. The efficiency of all primers was verified by consecutive dilutions of standardized samples. PP2A exhibited a homogenous behavior in all the RNA-Seq approaches included in this work (WT 23 or 29 °C, OE APOLO, vim1-3, and OE UPAT plants).

Histochemical staining

Four-day-old pYUCCA2:GUS transgenic seedlings [60] grown at 23 or 29°C were infiltrated with GUS staining solution (0.1 M phosphate buffer pH 7.5; 100 mM K3Fe(CN)6; 100 mM K4Fe(CN)6; 20% Triton X-100; 50 mM X-Gluc) under vacuum and subsequently incubated at 37 °C for 12 h. Stained tissues were cleared in 70% EtOH for 24 h at 37 °C and observed under a light microscope.

Statistical analyses

Statistical analyses were performed with non-parametric tests, Mann–Whitney when n=2 independent groups, and Kruskal-Wallis test when n>2 independent groups.