Introduction The biological significance of non-protein-coding genomic sequences has been an issue for decades (Britten and Davidson, 1969; Mattick, 2004). This has recently been reinforced by the finding that most of the human genome is represented in primary transcripts (Djebali et al., 2012). The majority of these are long, spliced, and polyadenylated RNA Polymerase II (RNA Pol II) transcripts, and a large number are antisense transcripts to annotated genes (Derrien et al., 2012; Osato et al., 2007; Lehner et al., 2002; Lu et al., 2012; Wang et al., 2005; Yamada et al., 2003). Many of the long (>200 nt) noncoding RNAs show no evolutionary conservation, adding to the debate of whether they serve any function (Gerstein et al., 2012; Graur et al., 2013). Several in-depth studies in yeast have shown that noncoding transcripts have the potential to regulate gene expression through transcriptional interference or recruitment of chromatin modifiers (Camblong et al., 2007; Hongay et al., 2006; Castelnuovo et al., 2013). However, roles of noncoding transcripts in higher eukaryotes are less well understood. Some have been shown to play roles in chromatin regulation (Wang and Chang, 2011), although it can be the transcriptional overlap rather than the antisense transcript itself that is important for the functional consequence (Latos et al., 2012). We have focused on the functional consequences of antisense transcription through our study of the regulation of Arabidopsis FLOWERING LOCUS C (FLC) gene, a developmental regulator that controls the timing of the switch to reproductive development. FLC encodes a MADS box transcriptional regulator that represses flowering, and FLC expression quantitatively correlates with flowering time (Sheldon et al., 1999; Michaels and Amasino, 1999). There are several regulatory pathways that converge to regulate FLC: two that antagonistically regulate FLC in ambient temperatures—the FRIGIDA pathway, which activates FLC expression, and the autonomous pathway, which downregulates FLC—and one more, vernalization, which epigenetically silences FLC in response to prolonged cold (Figure 1A). All of these pathways involve a set of antisense transcripts, collectively named as COOLAIR, that fully encompass the FLC gene, initiating immediately downstream of the sense strand polyadenylation site and terminating beyond the sense transcription start site (Hornyik et al., 2010; Liu et al., 2010; Swiezewski et al., 2009). COOLAIR transcripts are polyadenylated at multiple sites with proximal polyadenylation promoted by components of the autonomous promotion pathway. These include the RNA-binding proteins FCA and FPA, the 3′ processing factors Cstf64, Cstf77 and FY, the CPSF component and homolog of yeast Pfs2p and mammalian WDR33 (Liu et al., 2010; Ohnacker et al., 2000; Simpson et al., 2003). Use of the proximal poly(A) site results in quantitative downregulation of FLC expression in a process requiring FLD, an H3K4me2 demethylase (Liu et al., 2010). FLD activity results in H3K4me2 demethylation in the gene body of FLC and transcriptional downregulation of FLC (Liu et al., 2007, 2010). Loss of any of the autonomous pathway components reduces usage of the proximal polyadenylation site, which leads to increased FLC transcription. Analysis of the regulation of COOLAIR transcription has recently identified an RNA-DNA heteroduplex, or R-loop, covering the COOLAIR promoter (Sun et al., 2013). Stabilization of this R-loop by a novel homeodomain protein limits COOLAIR transcription, adding another layer of regulation within the autonomous pathway. We have continued to investigate the transcriptional circuitry at FLC and how COOLAIR is linked to changes in FLC expression. Here, through identification of a hypomorphic mutation in the core spliceosome component PRP8, we reveal how COOLAIR functionally modulates FLC gene expression through a cotranscriptional coupling mechanism. The prp8 mutation reduces splicing efficiency of COOLAIR introns and usage of the proximal poly(A) site, increasing histone methylation in the gene body and upregulating FLC transcription. We also show a positive feedback mechanism between gene body histone methylation and COOLAIR processing. The involvement of COOLAIR splicing in this mechanism was supported through both disruption of COOLAIR production and cis mutation of the antisense proximal splice acceptor site. Cotranscriptional coupling mechanisms such as this may be of widespread importance in the quantitative regulation of gene expression. Results A Hypomorphic Mutation in the Core Splicing Factor PRP8 Affects FLC Expression We pursued a suppressor mutagenesis strategy to identify additional factors contributing to flowering time regulation through FLC repression by FCA (Figure 1A). We mutagenized a line that is suitable to identify factors required for FCA-mediated FLC repression (also referred to as C2; Liu et al., 2010). It relies on FCA overexpression (35S-FCAγ transgene) to enhance FCA activity and establish low levels of FLC, a FLC-LUCIFERASE (FLC-LUC) reporter to efficiently monitor FLC levels, and a functional FRIGIDA (FRI) allele to amplify changes in FLC expression to increase sensitivity of detection (Johanson et al., 2000). Interestingly, the commonly used Arabidopsis accessions such as Landsberg erecta (Ler) and Columbia (Col) contain loss-of-function fri alleles, and the functional FRI we added originated from a Swedish accession (Johanson et al., 2000). We screened for mutants with increased luciferase activity of FLC-LUC and identified suppressor of overexpressed FCA (sof) 81 (Figure 1B). sof81 was a weaker suppressor than fld, the first mutant identified as a sof (Liu et al., 2007), and was found to be recessive in crosses to the C2 progenitor (Figure 1B). The mutation was mapped to At1g80070 (Figures S1A and S1B available online), a gene that has previously been identified as essential for plant development, as null mutations lead to embryonic lethality and abnormal suspensor development (sus phenotype) (Schwartz et al., 1994). At1g80070 encodes PRP8, the conserved and central component of the spliceosome (Grainger and Beggs, 2005). The sof81 mutation changes a glycine to glutamic acid at amino acid position 1,891 (Figure 1C) within the RNase H domain of PRP8 (Figure 1D). The mutation did not change PRP8 protein levels in the plant (Figure S1C). The RNase H domain of PRP8 is thought to be an integral part of the spliceosome (Pena et al., 2008; Galej et al., 2013) that prevents premature U4/U6 unwinding and acts as a platform for exchange of U6 snRNA for U1 at the 5′ splice site (Mozaffari-Jovin et al., 2012). The five available null alleles of PRP8 (sus2) plants are embryonic lethal, indicating the mutation in sof81 (referred to from now as prp8-6) is hypomorphic. The prp8-6 mutant phenotype was rescued by a genomic PRP8 clone (Figure S2A), and heteroallelic combinations between one copy of a prp8-sus2 allele (either sus2-4 or sus2-5) and one copy of the prp8-6 allele showed no complementation based on FLC-LUC bioluminescence and flowering-time analyses (Figures S2B and S2C). We therefore conclude that prp8-6 is a recessive, hypomorphic mutation that increases FLC expression in sof81. Unlike yeast and human, Arabidopsis thaliana carries a second copy of PRP8 (At4g38780) transcribed at low levels (Figure S1D) (Liu et al., 2009); however, given the mutant phenotype, this cannot completely cover the function of At1g80070 in FLC regulation. The prp8 Mutation Also Affects Endogenous FLC Expression and Flowering Time A similar forward mutagenesis screen had led to identification of DCL4 as a regulator of FCA expression with reduction in FCA expression resulting in elevated levels of FLC (Liu et al., 2012). Therefore, we first tested whether there was any change in the expression or functionality of FCA in prp8-6. We found no change in expression of the transgene 35S-FCAγ by western and northern blot analysis (Figures S3A and S3B). Additionally, the autoregulatory feedback limiting FCA levels was unaffected (Figures S3B and S3C) (Quesada et al., 2003). Previous data had shown that FCA associates with FLC chromatin (Liu et al., 2007). We found no reduction of FCA binding to the FLC locus in prp8-6; if anything, there was an elevated level (Figure S3D). A similar lack of effect of prp8-6 was observed on expression of other autonomous pathway components (Figure S3E); thus, we concluded that the increase of FLC expression by prp8-6 is unlikely to be due to an indirect effect on autonomous pathway function. Various polymorphisms have been reported between the FLC alleles of the Col and Ler laboratory strains (Col-FLC and Ler-FLC), including the presence of a Mutator transposon at the 3′ end of intron 1 (Liu et al., 2004). As FLC-LUC is based on Col-FLC, we tested the effect of prp8-6 on both alleles in the same samples by northern blotting using an FLC probe that discriminates by size. We detected only two transcript species reflecting Ler-FLC and FLC (Col)-LUC in prp8-6, both of which were increased compared to the progenitor (Figure 2A). We therefore concluded that the prp8-6-induced increase in expression is independent of the cis polymorphism between these two FLC alleles. We also analyzed flowering time and established that prp8-6 delays flowering (Figure 2B), suggesting that prp8-6 elevates biologically relevant levels of FLC. We then undertook an extensive genetic study analyzing combinations of fca-1, prp8-6, and FRIGIDA to investigate how PRP8 influences the autonomous and FRIGIDA pathway (Figure 2C). prp8-6 delayed the early flowering of the progenitor line (carrying the 35S-FCAγ and FRIGIDA transgenes) and delayed flowering much more extensively when the 35S-FCAγ transgene was crossed out, but was epistatic (nonadditive) with the loss-of-function mutation of FCA, fca-1. Consistent with this, when prp8-6 was combined with just FRI, the expression of FLC was significantly higher (Figure 2D). The effect of prp8-6 in fri genotypes increased when in combination with a sus2 null allele suggesting stronger alleles than prp8-6 would confer later flowering if they were viable (Figures 2C and S2C). The epistasis (nonadditivity) of prp8-6 with fca-1 indicates that PRP8 works in the same genetic pathway as FCA in wild-type plants. Overall, these results suggest the prp8-6 mutation causes a small reduction in PRP8 activity, which functions in the same genetic pathway as FCA to oppose FRIGIDA activation of FLC. PRP8 Influences Sense FLC Expression through Effects on COOLAIR Splicing PRP8 has a central role in splicing in most eukaryotes, but since the null phenotype is embryonic lethality (Schwartz et al., 1994), a role in FLC regulation had not previously been detected. Interestingly, mutations in two other Arabidopsis splicing factors, SR45 and PRP39-1, have been shown to increase FLC expression and cause late flowering (Ali et al., 2007; Wang et al., 2007). We therefore analyzed the effect of prp8-6 on splicing of FLC transcripts and also more generally. Four alternatively spliced gene models are annotated for sense FLC (Figure 3A). Measuring splicing efficiency of these alternative FLC introns by quantitative RT-PCR (qRT-PCR) indicated no effect of prp8-6 on these splicing events (Figure 3B). No disruption of the size of the sense transcript was detected by northern blot analysis, even though a cis mutation in an FLC sense splice site in the flc-5 mutant reveals this species (Figure S4A). The splicing efficiency of two control genes was also unaffected in prp8-6 (UBC9 and EF1a, Figure 3C). As our analysis provided no evidence of FLC sense splicing defects we analyzed splicing of COOLAIR introns. COOLAIR is alternatively spliced in different environmental conditions and different genotypes (Hornyik et al., 2010; Liu et al., 2010; Swiezewski et al., 2009). We assayed the efficiency of splicing of introns present in the most abundant COOLAIR transcripts (schematically summarized in Figure 3A) and found it was reduced in prp8-6 (Figures 3D). The class IIiii and IIiv forms represent 10). The activity of the FLC-LUC reporter of the transformants was compared to untransformed sof81 mutant controls. Cloning of FLC, COOLAIR AA , and COOLAIR TEX FLC was cloned as a genomic SacI fragment (∼12 kb) into the Arabidopsis binary vector pCambia-1300, which confers hygromycin resistance in plants. To generate COOLAIR AA , fragments F1 (1,325 bp) and F2 (311 bp) were amplified from FLC with primers for F1 (FLC3ss_F1-forward and FLC3ss_F1-reverse) and F2 (FLC3ss_F2-forward and FLC3ss_F2- reverse) containing a mutated sequence for the 3′ splice site of FLC antisense class Ii intron (AA instead of AG). PCR amplification was performed with Phusion polymerase (NEB). Resulting fragments F1 and F2 with overlapping ends were fused together in 1:1 molar ratio by PCR amplification with Phusion polymerase (NEB) employing the forward primer for F1 and the reverse primer for F2. The resulting fragment was digested with NheI and BglII, gel purified, and subsequently cloned into an SphI fragment of FLC, replacing the wild-type NheI-BglII fragment. The resulting SphI fragment with the mutated class Ii antisense 3′ splice site was inserted into FLC-pCambia-1300. This mutation creates a recognition site for DraI (TTTAAA), which has been used for genotyping the hygromycin resistant transformants to verify presence of the COOLAIR AA mutation. F2 homozygotes of the following genotypes: prp8-6/flc-2/FRI and PRP8/flc-2/FRI were obtained from crosses of prp8-6 and flc-2/FRI. The F2 homozygotes were transformed using Agrobacterium-mediated transformation of floral buds with the either FLC-pCambia-1300 or COOLAIR AA -pCambia-1300. The seeds from a total of 49 T1 (first generation) transformants (13 plants of COOLAIR AA /PRP8/flc-2/FRI, 11 plants of COOLAIR AA /prp8-6/flc-2/FRI, 15 plants of FLC/PRP8/flc-2/FRI and 10 plants of FLC/prp8-6/flc-2/FRI) were sown on GM medium without glucose and selected for hygromycin resistance (T2 generation). RNA for analysis was extracted from 4-week old seedlings. For cloning COOLAIR TEX , the sequence TAGCCACC that contains FLC translational stop TAG codon was mutagenized to create EheI restriction site TGGCGCCC. A SspI-SspI fragment containing the strong RBCS terminator (706 bp) was PCR amplified and cloned in sense direction between EheI and SwaI restriction sites (SwaI is located 741 bp downstream of the FLC stop codon, therefore replacing the corresponding genomic sequence of 3′ UTR of FLC and flanking downstream region to create COOLAIR TEX ). To analyze the effect of FLC tex seeds were collected from four homozygous plants of FLC tex /flc-2/prp8-6/FRI and five homozygous plants of FLC tex /flc-2/PRP8/FRI. These plants were obtained from the three independent crosses of FLC tex /flc-2/FRI to prp8-6/Ler. As a control for the FLC tex analysis, the flc-2/FRI plants were transformed with pSLJ-FLC15 (10 kg clone of Columbia FLC gene) and crossed with prp8-6/Ler (two independent crosses). Three plants from either FLC/flc-2/prp8-6/FRI or FLC/flc-2/PRP8/FRI were obtained. The seedlings from FLC tex and corresponding pSLJ-FLC transgenic plants were grown on GM medium without glucose and BASTA resistant transformants were isolated for analysis. Measuring FLC Sense Transcript For the sense FLC mRNA analysis, reverse transcription was performed using FLC specific reverse primers with SuperScript®III Reverse Transcriptase (Invitrogen). qPCR analysis was performed on LightCycler480®II (ROCHE) with primers FLC Unspliced_LP and FLC Unspliced_RP for the unspliced sense FLC transcript and with primers FLC Spliced_LP and FLC Spliced_RP for the spliced sense FLC transcript. qPCR data was normalized to UBC (which was amplified with primers UBC-F and UBC-R). The primers are described in Table S1. Measuring COOLAIR Splicing Efficiency To measure the splicing efficiency of class Ii intron, 5 μg of total RNA isolated from seedlings were reverse-transcribed into cDNA, primed by Int1_RT, which is located in the exon 2 of class I and class II ii, (for locations of the primers, see also the illustration presented in Figure 3A). Resulting cDNA was used as template in qPCR reactions to amplify cDNA with the first small intron spliced by primers Int1_spliced_LP and Int1_spliced_RP, which covers the splicing junction. cDNA with the first small unspliced intron was amplified by primers Int1_unspliced_LP and Int1_unspliced_RP, which is located in the first small intron. Triplicates of all PCR reactions were performed and quantified against standard curves of cDNA dilutions. These data were then used to calculate the mean together with the spliced/unspliced ratio. RT− controls were always included to confirm absence of genomic DNA contamination. To measure COOLAIR class II intron splicing efficiency, 5 μg of total RNA isolated from seedlings was reverse-transcribed into cDNA, primed by Class II unspliced F, and located in the last exon of all the class II antisense RNA. The resulting cDNA was used as template in qPCR to amplify spliced class II i with primers Class II-1_LP and Class II-1_RP, which cover the splicing junction; Class II ii intron 2 spliced with primers Class II-2_LP and Class II-2_RP, which cover the splicing junction; and FLC antisense big introns unspliced with primers Class II unspliced F and Class II unspliced R. Triplicate PCR reactions were performed and quantified against standard curves of cDNA dilutions before calculating the mean and spliced/unspliced ratio. RT− controls were always included to confirm absence of genomic DNA contamination. Measuring Polyadenylated COOLAIR The following primers were employed for the analysis of the COOLAIR transcripts: (a) for proximal poly(A) site transcript oligo(dT) primer was used for the reverse transcription and forward primer, set1_RP, and reverse primer, LP_FLCin6polyA, used for the qPCR analysis (Liu et al., 2010), and (b) for the distal poly(A) site, oligo(dT) primer was used for the reverse transcription and forward primer Set4_RP and reverse primer Set4_LP used for the qPCR analysis. qPCR reactions were performed in triplicates for each sample. Average values of the triplicates were normalized to the expression of total COOLAIR (which was amplified with Total COOLAIR_LP and Total COOLAIR_RP primers). The primers are summarized in the Table S1.