Regulation of Flowering Time by the RNA-Binding Proteins AtGRP7 and AtGRP8

Abstract

The timing of floral initiation is a tightly controlled process in plants. The circadian clock regulated glycine-rich RNA-binding protein (RBP) AtGRP7, a known regulator of splicing, was previously shown to regulate flowering time mainly by affecting the MADS-box repressor FLOWERING LOCUS C (FLC). Loss of AtGRP7 leads to elevated FLC expression and late flowering in the atgrp7-1 mutant. Here, we analyze genetic interactions of AtGRP7 with key regulators of the autonomous and the thermosensory pathway of floral induction. RNA interference- mediated reduction of the level of the paralogous AtGRP8 in atgrp7-1 further delays floral transition compared of with atgrp7-1. AtGRP7 acts in parallel to FCA, FPA and FLK in the branch of the autonomous pathway (AP) comprised of RBPs. It acts in the same branch as FLOWERING LOCUS D, and AtGRP7 loss-of-function mutants show elevated levels of dimethylated lysine 4 of histone H3, a mark for active transcription. In addition to its role in the AP, AtGRP7 acts in the thermosensory pathway of flowering time control by regulating alternative splicing of the floral repressor FLOWERING LOCUS M (FLM). Overexpression of AtGRP7 selectively favors the formation of the repressive isoform FLM-β. Our results suggest that the RBPs AtGRP7 and AtGRP8 influence MADS-Box transcription factors in at least two different pathways of flowering time control. This highlights the importance of RBPs to fine-tune the integration of varying cues into flowering time control and further strengthens the view that the different pathways, although genetically separable, constitute a tightly interwoven network to ensure plant reproductive success under changing environmental conditions.

Introduction

Higher plants need to precisely synchronize the transition from vegetative to reproductive growth with favorable environmental conditions to ensure reproductive success. This has led to the evolution of an intricate network of interwoven signaling pathways that mediate the response to environmental and developmental factors (Wang et al. 2009, Srikanth and Schmid 2011, Andrés and Coupland 2012, Johansson and Staiger 2015). Arabidopsis thaliana is a facultative long-day (LD) plant. A long period of light promotes flowering via the photoperiodic pathway (Shim et al. 2017). Day length is sensed in the leaves by the circadian clock, an endogenous timekeeper that generates a 24-h rhythm in the expression of numerous genes. The oscillatory pattern of the CONSTANS (CO) transcript encoding the key regulator of the photoperiodic pathway determines the response to LDs (Putterill et al. 1995, Suárez-López et al. 2001). When the CO peak coincides with the light phase in LDs, CO protein accumulates and activates the floral integrator FLOWERING LOCUS T (FT). The FT protein then moves through the phloem to the shoot apex to induce flower formation (Corbesier et al. 2007, Jaeger and Wigge 2007, Mathieu et al. 2007). Vernalization, the exposure to an extended period of cold, promotes flowering by downregulating the floral repressor FLOWERING LOCUS C (FLC) that represses FT (Bastow et al. 2004). FLC is also downregulated by endogenous regulators, collectively referred to as the autonomous pathway (AP) (Rataj and Simpson 2014). RNA processing appears to play a crucial role in the AP, as FLOWERING LOCUS CA (FCA), FLOWERING LOCUS PA (FPA) and FLOWERING LOCUS KH DOMAIN (FLK) encode predicted RNA-binding proteins (RBPs) and FLOWERING LOCUS Y (FY) codes for an mRNA 3′ end processing factor (Macknight et al. 1997, Schomburg et al. 2001, Simpson et al. 2003, Lim et al. 2004). Because both FCA and FPA undergo negative autoregulation through promoting polyadenylation of their own transcripts which ultimately leads to altered protein activity (Quesada et al. 2003, Simpson et al. 2003, Hornyik et al. 2010), it was hypothesized that alternative polyadenylation of FLC may be a means of FLC regulation in the AP. However, the FLC transcript is not alternatively polyadenylated in fpa mutants (Duc et al. 2013). Rather, the region downstream of FLC serves as a promoter for antisense transcripts termed COOLAIR that correlate with downregulation of FLC (Liu et al. 2010). These antisense transcripts are alternatively polyadenylated either at a distal site opposite of the FLC transcription start or at a promoter-proximal site. FCA and FPA favor polyadenylation at the proximal site, and this correlates with reduced histone H3 lysine 4 dimethylation and reduced FLC sense expression (Liu et al. 2007, Hornyik et al. 2010, Liu et al. 2010). Alternative splicing (AS) of the antisense transcripts contributes to reduce FLC transcription (Marquardt et al. 2014). Overall, it appears that the RBPs of the AP control FLC levels through affecting its transcription rather than directly affecting the FLC mRNA (Liu et al. 2007, Rataj and Simpson 2014). Indeed, the other members of the AP, FLOWERING LOCUS VE (FVE), a component of a histone deacetylase complex, and FLOWERING LOCUS D (FLD), a homolog of human lysine-specific demethylase 1, inhibit FLC expression through affecting histone modification at the FLC locus (He et al. 2003, Ausín et al. 2004, Kim et al. 2004, Shi et al. 2004). In contrast to the photoperiodic or vernalization pathways, the hierarchical interaction of the AP components, if any, is not known (Rataj and Simpson 2014). FCA and FPA act independently of each other (Hornyik et al. 2010) but the full activity of FCA and FPA may require FLD (Liu et al. 2007).

Apart from extended periods of low temperature, ambient temperature impacts floral transition. At 16°C Arabidopsis flowers later than at 20°C (Halliday et al. 2003). A temperature rise from 23°C to 27°C accelerates flowering in noninductive short days (SDs) to the same extent as extension of the photoperiod to LDs (Blázquez et al. 2003, Balasubramanian et al. 2006). Mutants lacking SHORT VEGETATIVE PHASE (SVP) function flowered early regardless of the ambient temperature (Lee et al. 2007) suggesting that the temperature information is mediated via SVP. SVP represses FT expression via direct binding to the FT promoter (Lee et al. 2007).

AS is a mechanism by which different transcript variants are generated from one pre-mRNA containing multiple exons and introns. This greatly enhances the coding capacity of the transcriptome and enables the cell to rapidly react to endogenous or exogenous cues. In plants, AS has been particularly associated with various stress responses (Reddy et al. 2013, Staiger and Brown 2013, Meyer et al. 2015, Laloum et al. 2018) and was also shown to play a role in the regulation of seed development (Sugliani et al. 2010, Zuo et al. 2019). Recently, AS has also gained interest in flowering time regulation. FLOWERING LOCUS M (FLM)/MADS AFFECTING FLOWERING 1 (MAF1) encoding a MADS-Box transcription factor (Scortecci et al. 2001) that acts in the thermosensory pathway has been shown to be subject to temperature-dependent AS (Balasubramanian et al. 2006, Severing et al. 2012). The splice product FLM-β is preferentially generated at cooler temperatures like 16°C and forms a repressor complex with SVP to prevent flowering. At higher temperatures, expression of FLM-β is reduced and another splice isoform termed FLM-δ is generated (Posé et al. 2013). The SVP-FLM-δ complex does not interact with the FT promoter. Moreover, the SVP protein is preferentially degraded at higher temperatures (Lee et al. 2013). This combined regulation at the post-transcriptional and post-translational level allows activation of FT to promote flowering preferentially at higher temperatures (Posé et al. 2013). Recent publications have questioned this model and the role of FLM-δ in the regulation of flowering and instead have highlighted the occurrence of a suite of noncanonical splice isoforms of FLM at higher temperatures. These mostly contain premature termination codons and are subject to degradation by the nonsense-mediated decay (NMD) pathway (Lutz et al. 2015, Sureshkumar et al. 2016, Capovilla et al. 2017, Lutz et al. 2017). This might serve as a way to reduce the pool of active FLM-β and thus enables flowering.

The RBP A. thaliana GLYCINE-RICH RNA-BINDING PROTEIN 7 (AtGRP7) is under control of the circadian clock with maximal expression in the evening. It autoregulates its own expression by AS and subsequent degradation via NMD (Heintzen et al. 1994, Staiger et al. 2003, Schöning et al. 2007). A reverse genetic approach has shown that it contributes to transduce timing information from the circadian clock to downstream transcripts (Streitner et al. 2010). It also promotes resistance to pathogenic Pseudomonas syringae both via salicylic acid-dependent and salicylic acid-independent pathways (Fu et al. 2007, Jeong et al. 2011, Nicaise et al. 2013, Hackmann et al. 2014) and influences AS of numerous target transcripts (Streitner et al. 2012, Meyer et al. 2017). Furthermore, AtGRP7 impacts flowering time (Streitner et al. 2008). The atgrp7-1 T-DNA mutant and RNA interference (RNAi) lines with reduced levels of AtGRP7 flower later than wild-type (wt) plants, particularly in SDs. The late-flowering phenotype correlates with elevated FLC levels and late flowering is suppressed by introducing the flc-3 loss-of-function mutation (Michaels and Amasino 1999) into atgrp7-1. Nuclear localization of AtGRP7 is regulated by JACALIN-LECTIN LIKE 1 (Xiao et al. 2015). The closely related AtGRP8 undergoes negative autoregulation by the same mechanism as AtGRP7 and both proteins reciprocally regulate their expression (Heintzen et al. 1997, Schöning et al. 2008, Schmal et al. 2013). AtGRP8 is rapidly upregulated under oxidative stress and plays a role in the regulation of root hair development (Schmidt et al. 2010, Foley et al. 2017) but apart from that, very little is known about its physiological functions. To unravel how AtGRP7 and the paralogous AtGRP8 regulate flowering time, we tested the genetic interactions with key regulators of the AP in the control of FLC expression and investigated the influence of both RBPs on the thermosensory pathway of flowering time control.

Results

Reduced levels of AtGRP8 further delay flowering of atgrp7-1

Loss of AtGRP7 delays flowering in the atgrp7-1 T-DNA mutant (Streitner et al. 2008). In the absence of a true loss-of-function line of AtGRP8, the effect of this AtGRP7 paralog on flowering time is not known. In particular, it is unknown whether the elevated AtGRP8 level in atgrp7-1 may alleviate the loss of AtGRP7 for flowering time control. Therefore, we aimed at counteracting the elevated AtGRP8 level in atgrp7-1 by an RNAi construct against AtGRP8. The resulting line atgrp7-1 8i displayed wt AtGRP8 level in the absence of AtGRP7 (Streitner et al. 2012). We assayed two independent atgrp7-1 8i lines for flowering time in SDs. Both flowered with significantly more leaves than atgrp7-1 but the number of days to bolting was increased only for atgrp7-1 8i #6 (Fig. 1A, B). FLC levels were slightly but not significantly elevated compared with atgrp7-1 (Fig. 1C). The protein level of AtGRP8 was elevated in atgrp7-1 and reduced below the level of wt in both atgrp7-1 8i lines (Fig. 1D). Thus, the reduction of the AtGRP8 level in atgrp7-1 8i compared with atgrp7-1 leads to an additional small delay in floral transition, suggesting that the paralogous AtGRP8 also is involved in flowering time control.

Fig. 1

Flowering time of the atgrp7-1 8i lines. Col wt, atgrp7-1, and two atgrp7-1 8i lines were grown in SDs. The number of rosette leaves (A) and the days until bolting (B) (n = 20) are shown as mean ± SD. ANOVA followed by a Dunnett’s test was performed to determine statistical significance (**P ≤ 0.01, ***P ≤ 0.001). (C) The FLC level was determined in 5-week-old plants. A t-test was performed to determine statistical significance based on atgrp7-1 as reference line. Bars show mean ± SD of three biological replicates (n.s., not significant). (D) Immunoblot analysis of the AtGRP7 and AtGRP8 levels. Amidoblack staining of the membrane is shown as loading control. Samples were quantified relative to Col.

AtGRP7 acts in parallel to the AP components FCA, FPA and FLK

The late-flowering phenotype of atgrp7-1 plants and their elevated FLC level can be overcome by vernalization, a trait shared with mutants in the AP (Streitner et al. 2008). Presently it is not entirely clear how the RBPs and the chromatin modification factors of the AP interact in the control of flowering time (Rataj and Simpson 2014). To test a genetic interaction of AtGRP7 with other AP components, we generated double mutants of atgrp7-1 or atgrp7-1 8i #9 with fca-9, flk-1, fpa-3 and fld-3 and assayed the flowering time (Supplementary Fig. S1, detailed statistics in Supplementary Table S2). We used atgrp7-1 8i #9 for most of the crosses to avoid attenuating effects of the elevated AtGRP8 as much as possible. Whereas fca-9 flowered after 60 d with 46 leaves in LDs, four homozygous fca-9 atgrp7-1 8i lines derived from independent crosses in reciprocal directions flowered after around 100 d with a large number of leaves that could not be determined reliably (Supplementary Fig. S1A, B). The high FLC level in the fca-9 mutant was somewhat further elevated in fca-9 atgrp7-1 8i. However, this was not significant, likely due to the inhomogeneous plant material causing a high variability (Supplementary Fig. S1C). flk-1 flowered after 43 d with 28 leaves, whereas flk-1 atgrp7-1 flowered after around 60 d with 45 leaves and flk-1 atgrp7-1 8i flowered between 74 and 81 d with 55 leaves. Again, somewhat higher FLC levels were seen in the double mutant compared with flk-1 although due to the large variation this failed to pass significance testing. fpa-3 flowered after 62 d with 55 leaves. The flowering time of the double mutants depended on the direction of the cross: When fpa-3 was used as the pollen donor (lines 29 and 54), fpa-3 atgrp7-1 8i flowered after around 84 d with 65 leaves. When fpa-3 was used as the pollen recipient (lines 17 and 19), fpa-3 atgrp7-1 8i flowered at 110 d with a large number of leaves that could not be determined reliably. The FLC level was not significantly different from the fpa-3 single mutant. In turn, steady-state abundance of FCA, FLK and FPA was not affected in atgrp7-1 loss-of-function lines or in AtGRP7-ox lines (Supplementary Fig. S1D). Together, these data indicate that AtGRP7 acts in parallel to these three RBPs in the AP.

Phenotypic appearance of atgrp7-1 8i fca-9 lines changed from that of a normal rosette plant with a central meristem to a shrub-like structure with multiple additional rosettes emerging from lateral meristems which made reliable counting of rosette leaves impractical (Supplementary Fig. S1E). Interestingly, this was not the case for atgrp7-1 flk-1 and atgrp7-1 8i flk-1 lines. These lines maintained the normal structure of an Arabidopsis plant (Supplementary Fig. S1F). In the case of atgrp7-1 8i fpa-3 double mutants, the growth phenotype depended on the direction of the cross. The atgrp7-1 8i fpa-3 plants where fpa-3 was the pollen donor showed normal plant body structure, whereas fpa-3 atgrp7-1 8i plants where fpa-3 was the pollen recipient showed a shrub-like structure much like fca-9 atgrp7-1 8i (Supplementary Fig. S1G). Together, the function of both AtGRP7 and FCA seems to be essential for normal plant development. This is also the case for AtGRP7 and FPA but synergistic effects on the plant phenotype are only visible in a background where the loss of FPA function is inherited maternally. Loss of both AtGRP7 and FLK can be tolerated by the plant without dramatic effects on the phenotype.

AtGRP7 genetically interacts with the AP component FLD

The fld-3 atgrp7-1 double mutants did not flower significantly later than fld-3 and with about the same leaf number (Fig. 2A, B). No additive effect on FLC levels was observed in these plants (Fig. 2C). We also generated atgrp7-1 8i fld-3 mutants that did not flower later than the fld-3 mutant (Supplementary Fig. S2). This indicates that AtGRP7 and FLD act in the same branch of the AP. FLD transcript levels were not altered in plants misexpressing AtGRP7 (Fig. 2D).

Fig. 2

Analysis of the fld-3 atgrp7-1 mutants. Leaf number (A) and days to bolting (B) of fld-3 atgrp7-1 double mutants. Col wt, atgrp7-1, fld-3 and four fld-3 atgrp7-1 lines were grown in LDs. Bars show mean ± SD (n = 19–20). A Kruskal–Wallis test was performed to show statistical significance between lines Col and atgrp7-1, statistical significance between fld-3 and fld-3 atgrp7-1 was assayed with an ANOVA followed by a Dunnett’s test (***P ≤ 0.001, n.s., not significant). (C) The FLC level was determined in 3-week-old plants grown in LDs. A t-test was performed to determine statistical significance. Bars show mean ± SD of three biological replicates. (D) Transcript levels of FLD were determined in 5-week-old plants grown in SDs. Bars show mean ± SD of three biological replicates.

In fld-3, increased histone H3 lysine 4 dimethylation (H3K4me2) is observed in a region of the FLC locus comprising intron 1 and the subsequent exons and correlates with increased levels of FLC transcript (Liu et al. 2007). As AtGRP7 appears to genetically act together with FLD, we tested whether AtGRP7 affects FLD-dependent histone modification at the FLC locus (Fig. 3A). We monitored H3K4me2 levels in atgrp7-1 8i, fld-3 and atgrp7-1 8i fld-3. In fld-3, we observed wt levels of H3K4me2 in the region spanning the part of the 5′-UTR and the first exon (region B) and increased levels of the chromatin mark in the first intron (region D) and in the genomic region comprised of exon 2 till exon 4 (region G), as previously reported (Liu et al. 2007, Marquardt et al. 2014). For atgrp7-1 8i, H3K4me2 levels were elevated compared with wt in all regions tested. In atgrp7-1 8i fld-3, the pattern of H3K4me2 was very similar to the fld-3 single mutant (Fig. 3B). Taken together, the results suggest that an increase in H3K4me2 could be the cause of the elevated FLC transcript levels in atgrp7-1 8i. The double mutant atgrp7-1 8i fld-3 completely mimics the methylation pattern of the fld-3 single mutant, underpinning the genetic interaction of AtGRP7 and FLD and pointing toward an epistasis of FLD over AtGRP7. In contrast to fld-3, atgrp7-1 8i also shows an increase in H3K4me2 in region B, raising the possibility that FLD is not the only factor controlling this chromatin mark at the FLC locus in atgrp7-1 8i.

Fig. 3

Analysis of H3K4me2 enrichment and splicing efficiency of FLC. Schematic representation of the FLC locus (A). Solid boxes represent exons, lines represent introns. Genomic regions analyzed by ChIP are underlined. Arrows and dotted lines represent primers used for the determination of splicing efficiency. (B) H3K4me2 enrichment in the regions depicted in (A) in fld-3, atgrp7-1 8i and fld-3 atgrp7-1 8i relative to Col wt. Data represent mean ± SEM of three biological replicates. (C) Transcript abundance of unspliced (intron 1 and 6) and spliced FLC variants was determined in atgrp7-1 and two atgrp7-1 8i lines relative to Col wt using the primer combinations specified in (A). The splicing efficiency (D) was determined as the ratio of spliced vs. unspliced transcript relative to wt. Data represent mean ± SD for three biological replicates. A t-test was performed to show statistical significance (n.s., not significant).

AtGRP7 does not regulate splicing of the FLC sense mRNA

Increased transcriptional activity at the FLC locus should lead to higher levels of unspliced nascent RNA in atgrp7 mutants. In addition, AtGRP7 is a known regulator of splicing and it could be possible that the protein affects the processing of the FLC pre-mRNA to the mature transcript. To address this, we estimated the unspliced nascent FLC transcript by measuring variants containing intron 1 or intron 6 in comparison to the fully spliced variants (Fig. 3C) as previously done (Liu et al. 2007). For the region containing intron 6, we found elevated levels of spliced FLC in atgrp7-1 and both atgrp7-1 8i lines. Transcripts containing unspliced intron 6 were likewise affected in these lines leading to a calculated splicing efficiency that was not different from wt (Fig. 3D). We, therefore, concluded that AtGRP7 is not involved in the regulation of intron 6 splicing. For the region containing intron 1, we also found elevated levels of the spliced FLC mRNA in the AtGRP7 loss-of-function lines compared with wt. In atgrp7-1 and atgrp7-1 8i #9, the intron 1 containing variants accumulated to the same level, leading to an unchanged splicing efficiency of intron 1 compared with wt. In atgrp7-1 8i #6, we measured a higher accumulation of variants containing intron 1 than of the fully spliced variant (Fig. 3C). This leads to a significant decrease in the calculated splicing efficiency for this particular line (Fig. 3D). Taking into consideration that atgrp7-1 8i #6 shows the strongest phenotype of all three loss-of-function lines, a possible role for AtGRP7 in the regulation of FLC intron 1 splicing cannot be entirely ruled out but might only be obvious under strong repression of the possibly counteracting AtGRP8.

Taken together, our data suggest a possible role for AtGRP7 in the regulation of FLC transcription rather than a prominent role in splicing regulation of this particular transcript. Even if a reduced splicing efficiency was seen as for intron 1 in atgrp7-1 8i #6, it did not result in reduced levels of mature FLC mRNA, possibly due to compensation by the higher transcription rate. Recently, this mode of action in FLC regulation has also been reported for two other RBPs, RZ1-B and RZ-1C (Wu et al. 2016).

Late flowering of atgrp7 mutant plants is absent at cooler temperatures

A modest rise in ambient temperature promotes flowering in Arabidopsis whereas cooler temperatures substantially delay floral induction (Blázquez et al. 2003, Balasubramanian et al. 2006). Therefore, we tested whether small changes in ambient temperature may have an effect on AtGRP7-dependent regulation of flowering time by growing plants at 16°C, 20°C and 27°C. To estimate the changes in the onset of flowering among the different temperatures, we assayed the leaf number at bolting and the days from sowing to bolting for each genotype.

At 20°C, atgrp7-1 and both atgrp7-1 8i lines flowered later and developed more leaves at bolting than wt plants, as previously observed (Fig. 4A, B; cf. Fig. 1). Conversely, three independent AtGRP7-ox lines flowered with significantly fewer leaves than wt (Fig. 4A). However, the time the AtGRP7-ox lines needed from sowing to bolting was similar as for the wt plants (Fig. 4B).

Fig. 4

Flowering phenotype of AtGRP7 loss-of-function and gain-of-function mutants at different ambient temperatures. Col wt, three independent AtGRP7-ox lines, atgrp7-1, and two atgrp7-1 8i lines were grown in SDs at 16°C, 20°C and 27°C, respectively. (A) The number of rosette leaves at bolting is shown as mean ± SD (n = 17–36). ANOVA followed by a Dunnett’s test was performed to show statistical significance (**P ≤ 0.01, ***P ≤ 0.001, n.s., not significant). (B) The number of days until bolting at 16°C, 20°C and 27°C. A Kruskal–Wallis test was performed to show statistical significance.

At 16°C, wt plants flowered later than at 20°C, both in terms of leaf number and days to flowering. In contrast, no delay in flowering was observed for the atgrp7-1 8i lines at 16°C and atgrp7-1 even flowered with fewer leaves than wt (Fig. 4A). The three independent AtGRP7-ox lines D, G and M flowered with fewer leaves than wt at 16°C (Fig. 4A). Again, time to flowering was not dramatically different from wt. Only AtGRP7-ox G showed a moderate but significant decrease in time to flowering (Fig. 4B).

At 27°C, flowering of wt plants was strongly accelerated, resulting in plants flowering with approximately 20 leaves (Fig. 4A). The atgrp7-1 and atgrp7-1 8i lines flowered significantly later than wt plants both in terms of leaf number and days to flowering. In contrast, the three AtGRP7-ox lines flowered with fewer leaves and after a significantly shorter time than the wt (Fig. 4A, B). Taken together, these results show that the misexpression of AtGRP7 has stronger effects on flowering time at temperatures above 20°C.

AS of FLM in plants with altered levels of AtGRP7 and AtGRP8

Recent studies have shown that temperature-dependent AS of the MADS-box transcription factor FLM and complex formation of FLM and SVP contribute to ambient temperature dependent regulation of flowering (Lee et al. 2013, Posé et al. 2013). Since late flowering of atgrp7-1 and atgrp7-1 8i is lost at 16°C and AtGRP7 affects transcript levels of the FLM paralog FLC as well as it regulates AS of a series of other transcripts (Streitner et al. 2012), we tested whether FLM abundance or splicing were affected in atgrp7-1, atgrp7-1 8i or AtGRP7-ox plants. At 20°C the total FLM transcript level was more or less similar in all lines compared with wt (Fig. 5A). However, the ratio of the alternative splice forms was changed. In AtGRP7-ox, the abundance of the FLM-δ isoform was strongly reduced compared with wt, while the amount of the FLM-β isoform was increased. In atgrp7-1 no obvious change in the ratio of isoforms was visible compared with wt, but both atgrp7-1 8i lines showed a higher amount of FLM-δ than wt whereas FLM-β levels were unchanged or somewhat lower (Fig. 5B). This points toward an additional effect of AtGRP8 in the regulation of these transcripts. For all lines, the splicing pattern was also observed in 16°C and 27°C whereas the abundance of total FLM transcript was mostly unchanged (Supplementary Fig. S3A–D). The reduction of the isoform FLM-δ in AtGRP7-ox was most pronounced at 16°C.

Fig. 5

AS of FLM. Col wt, atgrp7-1, two atgrp7-1 8i lines, and three AtGRP7-ox lines were grown in SDs at 20°C until they developed 10 leaves. Transcript levels of total FLM (A), FLM isoforms FLM-β and FLM-δ (B) and FLC (C) were determined using qRT-PCR. Data represent mean ± SD of three biological replicates.

Taken together, these results indicate that the FLM transcript pool in AtGRP7-ox plants consists mainly of the FLM-β isoform, irrespective of the ambient temperature. This is contradictory to the observed early flowering phenotype of these plants, as FLM-β was reported to be the floral repressing isoform of FLM and AtGRP7-ox flowered with less leaves than wt. One possible explanation is that transcript steady-state abundance of the floral repressor FLC is reduced in AtGRP7-ox (Fig. 5C). This leads to a floral promotive effect that may eventually overcome the repressive effect of FLM-β and is in line with the finding that FLC has a strong effect on flowering time at 27°C (Balasubramanian et al. 2006).

In contrast, FLM-δ was elevated in atgrp7-1 8i. Assuming that this isoform has floral promotive effects, this would contradict the observed late-flowering phenotype of atgrp7-1 8i at 20°C. However, these plants also showed elevated transcript levels of FLC (Fig. 5C), which might overcome a promotive effect of FLM-δ and, as opposed to the phenotype of AtGRP7-ox, lead to later flowering of atgrp7-1 8i. Additionally, the isoform FLM-δ is reported to have only little effects on flowering time and instead FLM-β levels are sufficient to explain the flowering phenotype (Lutz et al. 2015, Capovilla et al. 2017, Lutz et al. 2017). At 16°C, atgrp7-1 8i was flowering with about the same number of leaves as wt, suggesting that at this temperature the effects of the elevated floral repressor FLC were probably compensated by FLM-β.

At 27°C, the transcript abundance of FLM was very low compared with 16°C and 20°C but the differences in FLM splice isoforms were still visible in the different lines (Supplementary Fig. S3B, D). In comparison, the expression level of FLC was much stronger than that of FLM (Supplementary Fig. S3F) suggesting that the flowering phenotype of the AtGRP7 loss-of-function lines and the AtGRP7-ox lines at 27°C might be more dependent on FLC than on FLM.

FLM is a target transcript of AtGRP7

The inverse regulation of the FLM-δ isoform in atgrp7-1 8i and AtGRP7-ox prompted us to test whether FLM is a direct target of AtGRP7 in vivo. Plants expressing AtGRP7 fused to green fluorescent protein (GFP) under control of the endogenous promoter and all regulatory elements in the transcribed region were grown in SDs at 16°C. Using our established protocol for RNA immunoprecipitation (RIP) (Köster and Staiger 2014, Köster et al. 2014a), AtGRP7-GFP was precipitated with GFP-Trap® beads (IP+) and bound FLM isoforms were quantified by qRT-PCR. A mock precipitation using Sepharose beads without antibody (IP−) served as the control for unspecific binding. Transcript levels were expressed relative to the input (IN). As an additional control, RIP was also performed with plants expressing GFP only under the control of the same regulatory elements as AtGRP7-GFP. Using a primer combination that detects all isoforms, we found FLM to be enriched in the IP+ fraction of AtGRP7-GFP about 8-fold relative to the controls, namely the IP− fraction of AtGRP7-GFP and plants expressing GFP only (Fig. 6A). When the two major isoforms FLM-β and FLM-δ were assayed individually, we found an enrichment of FLM-β in the IP+ of AtGRP7-GFP whereas it was absent in the control line expressing GFP only (Fig. 6B). FLM-δ was not precipitated with AtGRP7-GFP at 16°C. The transcript was only found in one single replicate in the IP− fraction (Fig. 6C). This suggests that FLM is an in vivo target of AtGRP7 and that at 16°C the isoform FLM-β is preferentially bound, likely reflecting the higher abundance of FLM-β compared with FLM-δ at 16°C.

Fig. 6

FLM is an in vivo target of AtGRP7. RIP was performed in plants expressing AtGRP7-GFP in the atgrp7-1 background or plants expressing GFP only grown in SDs at 16°C. Levels of total FLM transcript (A), FLM-β (B) and FLM-δ (C) coprecipitated with GFP-Trap^® beads (IP+) or in a mock IP without antibody (IP−) are given relative to the input. Mean ± SEM of three biological replicates are presented (n.d., not detectable).

Discussion

Both AtGRP7 and its paralog AtGRP8 affect flowering time

Here, we show that the RBP AtGRP7 affects flowering time via at least two different pathways. In the AP, AtGRP7 regulates the MADS-box transcription factor FLC on the level of transcription. In the thermosensory pathway, AtGRP7 regulates AS of FLM and shifts the balance of the isoforms FLM-β and FLM-δ toward the floral repressive isoform FLM-β.

Reducing the level of the paralogous protein AtGRP8 in the atgrp7-1 mutant causes a small additional delay in floral transition, suggesting that AtGRP8 also impacts flowering time. The abundance of FLC in atgrp7-1 8i is not significantly higher compared with atgrp7-1 and additionally, atgrp7-1 already flowers later than wt although these plants show an elevated AtGRP8 level. This could be a hint that AtGRP7 possibly exerts stronger functions in the AP than AtGRP8. On the other hand, the splicing pattern of FLM is clearly changed in atgrp7-1 8i whereas in atgrp7-1 it resembles wt. This suggests that the higher level of AtGRP8 in atgrp7-1 masks the effect on AS of FLM and could mean that AtGRP8 plays a more prominent role in the ambient temperature responsive pathway than in the AP. Although we assume that both paralogous proteins are very similar in their mode of action, there might be differences under specific conditions. A similar effect of AtGRP8 was found for the regulation of pri-miR398b that was only seen in atgrp7-1 8i but not in atgrp7-1 (Köster et al. 2014b). To elucidate the role of AtGRP8 in more detail, further studies with plants lacking both AtGRP7 and AtGRP8 are needed.

AtGRP7 acts in the AP to regulate FLC

Because late flowering of the atgrp7 mutants correlates with increased FLC levels and is overcome by vernalization, AtGRP7 may be assigned to the AP of flowering time control that already comprises other well-known RBPs (Simpson 2004, Srikanth and Schmid 2011). In contrast to ‘classical’ AP mutants that show dramatically delayed flowering and high FLC expression compared with wt, late flowering of atgrp7-1 and even atgrp7-1 8i is only moderate and also FLC levels are only mildly elevated. FLC is a multi-exon gene but no reported target of prominent AS in wt (Mahrez et al. 2016). Instead, FLC seems to be regulated at the transcriptional level by long non-coding antisense RNAs termed COOLAIR. RBPs of the AP regulate alternative polyadenylation and AS of COOLAIR leading to the downregulation of FLC transcription in a process requiring FLD (Liu et al. 2007, Hornyik et al. 2010, Liu et al. 2010, Marquardt et al. 2014). Our analysis of double mutants revealed that AtGRP7 appears to act additively to the RBPs FCA, FPA and FLK of the AP. Interestingly, it appears to act in the same branch of the AP as FLD, further highlighting the pivotal role of this histone demethylase in the regulation of FLC. We find elevated levels of dimethylated H3K4 in the FLC gene body in atgrp7-1 8i which point to the direction that the increase in FLC expression is a consequence of increased transcription of this locus possibly caused by attenuated FLD function. How exactly AtGRP7 and FLD interact is still unclear. The FLD steady-state abundance is not altered in AtGRP7 loss-of-function or overexpressing lines, so an indirect mode of regulation involving another factor seems more likely. Interestingly, despite its mild effect on FLC sense expression, AtGRP7 is reported to have a comparably strong influence on polyA site selection of the antisense transcript COOLAIR by direct binding to its preRNA (Xiao et al. 2015). This could provide the missing connection between the RNA-binding ability of AtGRP7 and the interaction with FLD. However, due to the generally low abundance of these transcripts, we could not reliably detect changes in COOLAIR polyA site selection or changes in the AS pattern of COOLAIR transcripts neither in atgrp7-1 nor in atgrp7-1 8i under our growth conditions.

Double mutants with fve that are defective in the other chromatin modification factor of the AP could not be obtained due to the close proximity to AtGRP7 on the chromosome. Employing a CRISPR/Cas-based approach to generate a double mutant would help to unravel whether AtGRP7 acts solely via FLD or whether it has a more widespread role in the epigenetic regulation of FLC.

AtGRP7 and AtGRP8 influence AS of FLM

The regulation of flowering time in response to varying ambient temperatures has gained a lot of interest in the recent years. The existence of mutual exclusive isoforms FLM-β and FLM-δ generated by AS and the observation that only FLM-β was able to form functional complexes with the floral repressor SVP led to a model where flowering was repressed by FLM-β at cooler temperatures but promoted by FLM-δ at warmer temperatures (Lee et al. 2013, Posé et al. 2013). This model has since then been controversially discussed and more recent findings show that FLM-β seems to be the isoform predominantly responsible for the control of flowering time and that higher temperatures lead to unproductive splicing of the FLM transcript. The resulting noncanonical isoforms contain premature termination codons and are degraded by the NMD pathway. This reduces the amount of functional FLM-β and enables flowering at higher temperatures (Lutz et al. 2015, Sureshkumar et al. 2016, Capovilla et al. 2017, Lutz et al. 2017).

We show here that overexpression of AtGRP7 impacts AS of FLM and strongly shifts the ratio of FLM-β/FLM-δ toward the floral repressive isoform FLM-β. In contrast, loss of AtGRP7 together with a reduction of AtGRP8 leads to higher levels of FLM-δ. The primer combination used here to quantify FLM-δ also detects numerous of the noncanonical isoforms that harbor intron 3 and are subsequently degraded by NMD (Sureshkumar et al. 2016). Thus, it appears likely that AtGRP7 and AtGRP8 are required for the correct splicing of FLM and that loss of both results in unproductive splicing. This effect seems to be mediated through direct binding of FLM by either one or both proteins. So far, we could verify direct binding for AtGRP7, since we find FLM coprecipitating with AtGRP7-GFP in our RIP experiments. FLM-β is still present in the atgrp7-1 8i lines, so both RBPs are likely not the only factors responsible for the correct splicing of FLM.

In contrast to the observed effects on AS of FLM, overexpression of AtGRP7 reduces FLC levels while the loss-of-function lines show a higher expression of this floral repressor. This complex mode of regulating two distinct floral repressors in different pathways makes it difficult to explain the flowering phenotypes of plants with altered AtGRP7 levels under the temperatures tested. One possible explanation could be that FLC and FLM exert their functions at different temperatures, which results in the observed differences in flowering time. Indeed, it was reported that FLM has a strong repressive effect on flowering time at 15°C (Lutz et al. 2015, Lutz et al. 2017) while FLC represses flowering also at warmer temperatures like 27°C (Balasubramanian et al. 2006).

The late-flowering phenotype of the atgrp7-1 mutant is lost at 16°C and the plants do not show a change in FLM-β/FLM-δ levels compared with wt but a higher expression of FLC. Also, the atgrp7-1 8i lines do flower like wt at this temperature despite higher expression of FLM-δ and FLC. These findings are in line with the observations that at cooler temperatures flowering time is more dependent on FLM than on FLC (Lutz et al. 2015) and that it is the FLM-β level that is sufficient to explain the flowering phenotype whereas FLM-δ seems to have no effect on floral timing (Lutz et al. 2015, Capovilla et al. 2017, Lutz et al. 2017). At 20°C and 27°C, the loss-of-function lines show late flowering that can be explained by their higher FLC expression and the minor impact of FLM on flowering time under these temperatures. In contrast, the AtGRP7-ox lines flower early at 27°C due to their lower FLC expression. At 20°C and 16°C, however, these plants flower with less leaves than wt but still need the same amount of time to flower as wt plants (cf. Fig. 5). Apparently, overexpression of AtGRP7 also affects the leaf initiation rate leading to the production of fewer leaves per day than wt. However, when using days until flowering as a measurement, AtGRP7-ox plants nearly need the same amount of time to flower as wt plants at 16°C and 20°C. This could be due to a trade-off between the reduced FLC level in these plants and the fact that the FLM transcript pool consists mainly of the repressive FLM-β isoform. This in summary could lead to an unaltered flowering time compared with wt. Finally, it should also be noted that AtGRP7 probably affects other factors contributing to flowering time as well. For example, it was reported that the protein regulates AS of the miRNA172b precursor (Köster et al. 2014b) which likely has an effect on flowering time regulation via the age-dependent pathway.

In summary, we could show that the RBPs AtGRP7 and AtGRP8 influence flowering time via affecting the expression of FLC and regulating AS of FLM. The phenotypes of plants misexpressing both proteins suggest that their flowering time is the result of a complex interplay of the underlying molecular pathways. In contrast to the other RBPs assigned to the AP, where a loss of function leads to dramatic late-flowering phenotypes, a loss of AtGRP7 and a concomitant downregulation of AtGRP8 have weaker effects. This could mean that rather than being key players that regulate fundamental steps of flower initiation, both RBPs act at an intersection of the thermosensory pathway and the AP to fine-tune floral induction by integrating an external cue like temperature into flowering time control. It remains challenging to find conditions under which AtGRP7 and AtGRP8 might have a greater impact on flowering time and how they would act on flowering time under the fluctuating conditions observed in nature.

Materials and Methods

Plant material

The atgrp7-1 T-DNA line and the atgrp7-1 8i line carrying an RNAi construct against AtGRP8 (Fu et al. 2007, Streitner et al. 2012) as well as transgenic lines ectopically overexpressing AtGRP7 under control of the Cauliflower Mosaic Virus (CaMV) 35S promoter in the Col-2 ecotype (AtGRP7-ox) (Heintzen et al. 1997, Schöning et al. 2007, Streitner et al. 2008) have been described. AtGRP7-GFP fusion and GFP under control of 1.4 kb of the AtGRP7 promoter, the AtGRP7 5′-UTR and 3′-UTR have been described (Staiger and Apel 1999, Streitner et al. 2012). Double mutants of atgrp7-1 and atgrp7-1 8i were generated with fca-9 (Page et al. 1999), fpa-3 (Meier et al. 2001), flk-1 (SALK_112850) (Lim et al. 2004) and fld-3 (He et al. 2003), respectively. F2 plants were genotyped by PCR using primers indicated in Supplementary Table S1. The RNAi construct against AtGRP8 was detected based on resistance to BASTA® and reduced levels of AtGRP8 protein compared with atgrp7-1. Homozygous F3 plants were used for all subsequent experiments. All mutant lines were created in the Columbia background.

Determination of the flowering time

Seeds were sown on soil, stratified at 4°C for 2 d, and germinated and grown in SDs (8-h light/16-h dark cycles) or LDs (16-h light/8-h dark cycles). Plants were grown in a randomized fashion at 16°C, 20°C, or 27°C in Percival incubators AR66-L3 (CLF Plant Climatics, Wertingen, Germany). Flowering time was determined by counting the rosette leaves once the bolt was 0.5 cm tall. Mean values ± SD were calculated (Steffen et al. 2014). For statistical analysis, ANOVA followed by a Dunnett’s test and a factorial ANOVA were performed in case of a normal distribution of the data. Otherwise, a Kruskal–Wallis test was performed.

Quantitative real-time RT-PCR

Quantitative real-time RT-PCR (qRT-PCR) was performed as described (Streitner et al. 2012). Primers used are listed in Supplementary Table S1.

Protein analysis

Protein extracts were prepared as previously described (Heintzen et al. 1994). Western blot analysis with anti-peptide antibodies against AtGRP7 and AtGRP8 was done as described (Lummer et al. 2011). Protein levels were quantified using ImageJ (Schneider et al. 2012). Amidoblack staining of the membrane served as loading control.

Chromatin immunoprecipitation

Chromatin immunoprecipitation (ChIP) was essentially performed as described (Kaufmann et al. 2010). Chromatin was sheared to approximately 300-bp-long fragments using a Bioruptor (Diagenode Inc., Denville, NJ) with 10 cycles of 30 s on/off in high mode. An antibody against dimethylated lysine 4 of histone H3 (Merck-Millipore, Darmstadt, Germany; Catalog no. 07–030) was used. The antibody was precipitated using recombinant Protein A Sepharose beads (GE Healthcare, Freiburg, Germany). Cross-links between DNA and proteins were reversed by Proteinase K and heat treatment and DNA was recovered using phenol/chloroform extraction. FLC fragments were analyzed by qRT-PCR using the primers listed in Supplementary Table S1.

RNA immunoprecipitation

RNA immunoprecipitation (RIP) was performed as described (Köster and Staiger 2014, Köster et al. 2014a) using plants that express AtGRP7 fused to GFP under control of the AtGRP7 promoter and cis-regulatory sequences in the 5′-UTR, 3′-UTR and intron in atgrp7-1. Plants expressing GFP only under control of the same regulatory elements were used as a control. For immunoprecipitation (IP+), GFP-Trap® beads (ChromoTek, Planegg-Martinsried, Germany) were used. For the mock IP (IP−), Sepharose beads were used. To determine the input (IN), RNA was isolated from 100 µl cell extract and the concentration was determined spectrophotometrically. Fractions of the input reflecting the individual differences between the replicates but comprising not >3 µg of RNA were used for retrotranscription in order to not inhibit the reverse transcriptase (AMV native reverse transcriptase, Roboklon, Berlin, Germany). Coprecipitated transcripts were quantified relative to the input.

Acknowledgments

We acknowledge Prof. R. Amasino for providing the fld-3 mutant, Prof. C. Dean for providing seeds of fca-9 and Prof. G. Simpson for kindly providing seeds of fpa-3. We thank K. Neudorf and E. Klemme for expert technical assistance.

Funding

Deutsche Forschungsgemeinschaft (DFG) [STA653/9–1, STA653/5–2, priority program SPP1530].

Disclosures

The authors have no conflicts of interest to declare.