Light regulates widespread plant alternative polyadenylation through the chloroplast

Significance

Light is not only a source of energy but also a regulator of plant physiological adaptations. We previously showed that light affects alternative splicing in Arabidopsis thaliana. We report now that light also controls alternative cleavage/polyadenylation in approximately 30% of Arabidopsis genes. Similarly to alternative splicing, this effect requires functional chloroplasts, is not affected in photoreceptor mutants, and is observed in roots only if communication with shoots is not interrupted. We identified sucrose as the signaling molecule traveling from shoots to roots and confirm that mitochondrial activity is necessary for its action in the roots. However, unlike alternative splicing, changes in transcriptional elongation are not involved. Instead, the effect correlates with chloroplast-elicited changes in the abundance of constitutive cleavage/polyadenylation factors.

Abstract

Transcription of eukaryotic protein-coding genes generates immature mRNAs that are subjected to a series of processing events, including capping, splicing, cleavage, and polyadenylation (CPA), and chemical modifications of bases. Alternative polyadenylation (APA) greatly contributes to mRNA diversity in the cell. By determining the length of the 3’ untranslated region, APA generates transcripts with different regulatory elements, such as miRNA and RBP binding sites, which can influence mRNA stability, turnover, and translation. In the model plant Arabidopsis thaliana, APA is involved in the control of seed dormancy and flowering. In view of the physiological importance of APA in plants, we decided to investigate the effects of light/dark conditions and compare the underlying mechanisms to those elucidated for alternative splicing (AS). We found that light controls APA in approximately 30% of Arabidopsis genes. Similar to AS, the effect of light on APA requires functional chloroplasts, is not affected in mutants of the phytochrome and cryptochrome photoreceptor pathways, and is observed in roots only when the communication with the photosynthetic tissues is not interrupted. Furthermore, mitochondrial and TOR kinase activities are necessary for the effect of light. However, unlike AS, coupling with transcriptional elongation does not seem to be involved since light-dependent APA regulation is neither abolished in mutants of the TFIIS transcript elongation factor nor universally affected by chromatin relaxation caused by histone deacetylase inhibition. Instead, regulation seems to correlate with changes in the abundance of constitutive CPA factors, also mediated by the chloroplast.

Transcription of eukaryotic protein-coding genes generates immature mRNAs that are subjected to a series of cotranscriptional processing events, including capping, splicing, cleavage, and polyadenylation (CPA) and chemical modifications of bases. CPA is responsible for cleavage of the nascent RNA and addition of a poly(A) tail to cleaved RNA. Most of the core CPA factors found in yeast and mammalian cells, about 20 in total, have homologous counterparts in the plant Arabidopsis thaliana, in particular, within the complexes CPSF (cleavage and polyadenylation specificity factor), CstF (cleavage stimulation factor), and CFII (cleavage factor II). Equivalent subunits of mammalian CFI (cleavage factor I), however, have not been identified in plants (1). In all studied organisms, core CPA factors act together with the poly(A) polymerase and other RNA-binding proteins to yield poly(A)-tailed mature mRNAs. A key step in this process is the recognition of RNA motifs that define the polyA site (PAS). The AAUAAA hexamer is the most prominent PAS motif in metazoans, which is recognized by CPSF. However, this sequence has only been reported in approximately 10% of the Arabidopsis transcripts, which opens the yet-unsolved question of how the PAS is specified in plants (1).

Similar to alternative splicing (AS), alternative polyadenylation (APA) greatly contributes to mRNA diversity in the cell (2). By affecting the length and sequence of the 3’ untranslated region (3’UTR), APA generates transcripts with different architectures of target sites for microRNAs, which affects mRNA turnover and translation. When coupled with AS, APA can also alter C-terminal segments of the polypeptides encoded by a single gene. For example, this mechanism is involved in the control of seed dormancy in plants. The Arabidopsis DOG1 (delay of germination 1) gene gives rise to two mRNA isoforms that are generated by the use of proximal and distal (with respect to the promoter) PASs, respectively (3). The short and long mRNAs encode polypeptides of 30 and 32 kDa, respectively, differing in a nonconserved C-terminal stretch. However, only the protein encoded by the short mRNA is functionally active in promoting dormancy. Mutations in CPA factors that lead to reduced use of the proximal PAS cause weakened seed dormancy (3). APA of noncoding RNAs also has fundamental biological roles in plants. The best-studied example is the role of the long noncoding RNA COOLAIR in the control of flowering in Arabidopsis. COOLAIR is an antisense transcript of the flowering locus C gene (FLC) that displays two APA sites. When the distal site is used, a transcription-permissive histone mark is deployed along the FLC gene which allows for FLC expression and repression of flowering. Instead, if the proximal site is used, the short isoform of COOLAIR base-pairs with complementary FLC sequences, forming a so-called R-loop, which promotes the deployment of a transcription-repressive histone mark, which subsequently inhibits FLC expression and promotes flowering (4, 5).

We have previously shown that light regulates Arabidopsis AS through the chloroplast. We demonstrated that the photosynthetic electronic transport chain initiates a chloroplast retrograde signaling that regulates nuclear splicing (6). Later, we found that light promotes RNAPII elongation while in darkness, elongation is lower (7). These changes in transcriptional elongation are causative of the observed changes in AS, indicating that the chloroplast control responds to the kinetic coupling between transcription and RNA processing found in mammalian cells [for reviews see ref. 8 and (9)], and providing unique evidence that coupling is important for a whole organism to respond to environmental signals.

In view of the physiological importance of APA in plants, we decided to investigate the effects of light/dark conditions and the underlying mechanisms and to compare them to the ones elucidated for AS. We report here that light deeply affects Arabidopsis 3’UTR APA in 28.7% (3,467) of 12,070 genes assessed by genome-wide analyses. Three gene groups are clearly defined: those in which light promotes preferred usage of proximal PASs, those in which distal PASs are preferred, and those in which light has not effect on APA at all. Similar to AS, the effect of light on APA requires functional chloroplasts, is not affected in mutants of the phytochrome and cryptochrome photoreceptor pathways, is observed in roots when the communication with the photosynthetic tissues is not interrupted. Furthermore, like in AS, mitochondrial and TOR kinase activities are necessary for the APA effect of light (10). However, unlike AS, coupling with transcriptional elongation does not seem to be involved in changes in APA triggered by light. Alternatively, regulation seems to correlate with light-elicited changes in the abundance of constitutive CPA factors, also mediated by the chloroplast.

Results

Arabidopsis Genes Display Substantial 3’UTR APA Isoform Expression.

In order to investigate the role of light/dark exposure in Arabidopsis APA, we applied a light regime similar to the one used in our previous AS studies (6, 7, 10). Briefly, seedlings were grown for two weeks in constant white light to minimize interference from the circadian clock and then transferred to light or dark conditions for different time periods (Fig. 1A). To obtain APA regulation information genome-wide, we subjected total RNA to RNA sequencing by using the 3’ region extraction and deep sequencing (3’READS) method (11, detailed in Materials and Methods). We generated ~82.5 million PAS-supporting reads from the two sample groups (SI Appendix, Table S1). After clustering of adjacent cleavage sites that resulted from heterogenous cleavage for each PAS (SI Appendix, Fig. S1A, see Materials and Methods for detail), we identified 160,881 PASs in the Arabidopsis genome. Consistent with previous studies (11, 12), plant PASs identified by 3’READS are surrounded with A/T-rich sequences, with a prominent T-rich peak around −10 nt and a modest A-rich peak around −20 nt (cleavage site is set to position 0, SI Appendix, Fig. S1B). By analyzing hexamers in neighboring regions of the PAS, we found substantially enriched TA-rich motifs surrounding Arabidopsis PASs (SI Appendix, Fig. S1C), which is similar to PASs in S. cerevisiae but distinct from those in metazoans (13–16).

Fig. 1.

Light/dark conditions elicit widespread APA changes in Arabidopsis thaliana. (A) Protocol scheme of the light/dark regime used in this study. Total RNA from seedlings was subject to 3’READS for gene expression and APA analyses. (B) Distribution of identified PASs (Left) and PAS reads (Right) in different regions of the plant genome. (C) Different types of APA genes identified in this study. (D) Top, schematic of 3’UTR APA isoform; bottom, 3’UTR sizes for mRNAs of genes without 3’UTR APA (single 3’UTR) and mRNAs with the longest or shortest 3’UTRs of genes with 3’UTR APA. The median value for each group is indicated. (E) Diagram showing 3′UTR APA analysis. The two most abundant APA isoforms per gene were selected for comparison, which are named proximal PAS (pPAS) and distal PAS (dPAS) isoforms, respectively. The distance between the two PASs is considered alternative 3′UTR (aUTR). (F) Scatterplot showing genes with pPAS and dPAS isoform abundance differences between light- and dark-treated seedlings. Results represent analyses of 12,873 genes in three biological replicates. Genes with significantly (FDR < 0.05, DEXseq analysis) higher or lower abundance of pPAS isoforms in light vs. dark conditions are shown in blue and red, respectively.

Overall, 59% of all PASs were mapped to the last exon of mRNA genes, mostly in 3’UTRs. In addition, 10% were in introns, 6% in coding exons, and 22% in intergenic regions (Fig. 1B, left). Notably, the vast majority of the PAS-supporting reads (96%) were mapped to the last exon (Fig. 1B, right), indicating that transcripts using last exon PASs are much more abundant than those using PASs in other regions.

Using transcript abundance of 5% as the cutoff to call an isoform (SI Appendix, Fig. S2A), we found that 81.6% of plant mRNA genes display APA in our samples. On average, an APA gene expresses 2.9 APA isoforms (SI Appendix, Fig. S2B). Of all APA genes, 91.9% have APA sites exclusively in the last exon and 4.7% have APA sites in both the last exon and upstream introns/exons (Fig. 1C). Therefore, most plant genes display 3’UTR APA through alternative usage of PASs in the last exon.

For genes that do not display APA, the median value of their 3’UTR size is 151 nt (Fig. 1D). By contrast, for genes that express 3’UTR APA isoforms, the median values for the shortest 3’UTR and the longest 3’UTR are 125 nt and 214 nt, respectively (Fig. 1D). The alternative 3’UTR size has a median value of 78 nt across genes (SI Appendix, Fig. S2C). Therefore, 3’UTR APA in plants can potentially alter a substantial portion (30 to 60%) of the 3’UTR sequence, impacting posttranscriptional control of gene expression. We also found that the longest 3’UTR isoform is typically expressed at a higher level than the shortest 3’UTR isoform (SI Appendix, Fig. S2D). In addition, the PASs of the shortest and longest isoforms are surrounded with distinct mRNA motifs (SI Appendix, Fig. S2E), with the AATAAA motif highly enriched for the last PAS (longest 3’UTR isoform) and TGT motif for the first PAS (shortest 3’UTR isoform).

Light Elicits Widespread APA Isoform Changes.

We next examined APA isoform changes in light vs. dark conditions. Because most genes display 3’UTR APA, we focused on the top two most expressed 3’UTR APA isoforms for each gene. For simplicity, they are named proximal PAS (pPAS) isoform and distal PAS (dPAS) isoform, respectively (Fig. 1E). We calculated their relative expression (RE, dPAS vs. pPAS) in each sample group and then compared RE between the groups (light vs. dark), yielding the value RED (relative expression difference).

Using a p-value cutoff of 0.05 (DEXSeq), we identified 3,467 genes that showed significant differences in APA isoform expression levels between light and dark conditions (Fig. 1F), which represents 28.7% of the genes displaying APA in Arabidopsis. Strikingly, genes showing 3’UTR lengthening, i.e., upregulation of dPAS isoform relative to pPAS isoform, outnumbered those showing 3’UTR shortening by 3.3-fold (Fig. 1F). By using Gene Ontology analysis, we found that genes showing 3’UTR lengthening tend to have functions in various metabolic pathways, such as ATP, cellular amide, and hexose metabolisms. (SI Appendix, Table S2).

Representative genes showing 3’UTR lengthening, 3’UTR shortening, or no change are shown in SI Appendix, Fig. S3. HTA9 and RKH are two paradigmatic examples of 3’UTR lengthening and 3’UTR shortening genes, respectively (Fig. 2A). Using PCR primers targeting alternative 3’UTR sequences of their isoforms and upstream common regions (SI Appendix, Table S3), we validated our 3’READS data by RT-qPCR.

Fig. 2.

The light effect on APA is sensed by the chloroplast (genetic and biochemical evidence). (A) 3’READS data for two Arabidopsis genes with opposite APA changes. Top: a representative event [HTA9 gene (AT1G52740), reads in red] with higher usage of dPAS in the light. Bottom: a representative event [RKH gene (AT5G15270), reads in blue] with higher usage of pPAS in the light. For each APA event, two pairs of primers were designed to validate the APA changes using RT-qPCR: amplicon dPAS (dark orange) only exists if the dPAS is used (long isoform); amplicon cod is common to all isoforms in the upstream coding region. Changes in APA are quantified as ratios of dPAS/cod amplicons relative mRNA expression levels for every gene. (B–D) Light is not sensed by photoreceptors. APA response to light/dark in different Arabidopsis phytochrome and cryptochrome mutant genotypes in a Landsberg erecta background (wt, Ler). Three selected genes are shown: HTA9, whose APA events increases its dPAS usage in the light (A), RKH, whose APA event diminishes its dPAS usage in the light (B) and FAD6, whose APA event is not affected by the light/dark conditions and serves as a negative control. (E–G) Effect of the photosynthetic electron transfer chain inhibitor DCMU on the light/dark effect on APA events of the HTA9 (E), RKH (F), and FAD6 (G) genes. Seedlings were grown in constant light, transferred to darkness for 48 h. and then treated with 20 µM DCMU during a 6-h. light/dark further incubation. RT-qPCR experiments were quantified with n ≥ 3, where n = ~25 to 30 Arabidopsis seedlings growing in one Petri dish. White and black bars represent light and dark treatments respectively. Changes considered significant show differences with a P value < 0.05 (two-tailed Student’s t test). ***P < 0.001; **P < 0.01; *P < 0.05; NS (not significant) = P > 0.05.

A Functional Chloroplast Electron Transport Chain is Necessary for the Light Effect on APA.

Next, we wanted to address whether signaling through photoreceptors was involved in APA regulation in response to light. We exposed seedlings of Arabidopsis mutants for the red/far red-light photoreceptor phytochrome A [phyA-201, ref. 17] and for the blue light photoreceptor cryptochrome cry1 [cry1-1, ref. 18] and cry2 [fha-1, ref. 19] to our light regime protocol (Fig. 2 B–D), using the wild-type Landsberg erecta (Ler) background as control. APA was assessed by RT-qPCR in the HTA9 and RKH genes, as examples of 3’UTR lengthening and shortening by light, respectively (Fig. 2 B and C), and the FAD6 gene (Fig. 2D) as a gene with no change in APA. Both types of mutants behaved similarly to wild-type seedlings for all three genes, which refutes the notion that photoreceptors play a role on light-controlled APA.

Because retrograde signals from the chloroplast have been shown to modulate nuclear gene expression (6, 10, 20), we reasoned that chloroplast functions might be involved in light-elicited APA regulation. To this end, we used the herbicide DCMU [3-(3,4-dichlophenyl)-1,1-dimethylurea] (21) that blocks the photosynthetic electron transport from photosystem II to the plastoquinones. Interestingly, DCMU inhibited the effect of light on APA in both HTA9 (Fig. 2E) and RKH (Fig. 2F) genes but had no effect on FAD6 (Fig. 2G), suggesting that chloroplast function is necessary for modulating APA in response to light.

The Light Effect on APA is Sensed by the Photosynthetic Tissues.

To obtain further evidence of the involvement of the chloroplast in APA, we performed dissection experiments. Because roots have no chloroplasts, we reasoned that if seedlings were cut to separate roots from green photosynthetic tissues (shoots) (Fig. 3 A and B), APA differences upon light/dark treatments should only be observed in the green tissues. Similarly to what happens with AS (6), when dissection was performed 6 h after light/dark treatment, in the HTA9 gene, the effect was observed both in dissected leaves and roots (Fig. 3C, left). This led us to perform a dissection experiment in which shoots were separated from roots before the treatment (Fig. 3C, right). In these conditions, dissected shoots retain the same light response for HTA9 APA as undissected seedlings, but light has no effect on APA in dissected roots. In the case of RKH APA, the light effect was only observed in shoots independently of whether dissection was performed after or before treatment (Fig. 3D). The negative control FAD6 was unresponsive in neither shoot nor roots in both dissection protocols (Fig. 3E). These results strengthens the previous evidence of the involvement of the chloroplast electron transport chain (Fig. 2) and rules out any effects of other plastids present in the roots. The HTA9 dissection experiments suggest that a putative signal generated by light in the green tissues may move to the roots to modulate APA in a similar way as in the green parts of the plant.

Fig. 3.

The light effect on APA is sensed by the photosynthetic tissues (anatomical evidence). (A and B) Schemes for dissections of Arabidopsis seedlings performed after (post-) or before (pre-) the light/dark treatment. (C–E) APA isoform analysis in green tissue (shoots) and in the roots of post- (Left) and pre-dissection (Right) light/dark treatments of the HTA9 (C), RKH (D), and FAD6 (E) genes. Bar colors and RT-qPCR conditions were as in Fig. 2.

Sugars, Mitochondrial Activity, and TOR Kinase Modulate the Light Effect on APA.

In view of the evidence emerging from DCMU inhibition and the dissection experiments, we decided to further investigate the signaling from shoots to roots. Sugars are the main photosynthates in terrestrial plants. They are generated in the green tissues and either metabolized in their cells or loaded into the phloem to feed nonphotosynthetic tissues. In a previous study (10), it was shown that sucrose, the most important phloem-mobile sugar of Arabidopsis (22), is responsible for mediating AS changes in the roots triggered by light exposure of the shoots. More interestingly, externally applied sucrose solutions mimicked the effect of light on AS patterns in the roots but had no major effect in the shoots. In agreement with AS results (10), incubation with 100 mM sucrose of seedlings subjected to the light/dark regime that were later dissected (Fig. 4A, POST) did not alter the light effect on HTA9 APA in the shoots, but abolished it in the roots, in a way that mimics the effect of light in roots kept in the dark (note the height of the black bar that corresponds to roots treated with sucrose in Fig. 4A). In order to rule out any osmotic effect of sucrose we used equal concentrations of sorbitol as negative control. When dissection was performed before light/dark treatments (Fig. 4B, PRE) incubation with sucrose neither had effect on the change in HTA9 APA observed in the shoots nor on the lack of it observed in the roots. As far as the RKH APA event is concerned, since we found no effects of light in roots (Fig. 3D), experiments on shoots also revealed that sucrose mimics the effect of light on this APA event (SI Appendix, Fig. S4A)

Fig. 4.

Sugars, mitochondrial activity, and TOR kinase modulate the light effect on APA. HTA9 gene APA isoform analysis in green tissues (shoots) and roots of Arabidopsis seedlings dissected post- (A, C and D) and pre- (B) light/dark treatments. (A and B) Incubations were performed with 100 mM sucrose or sorbitol (negative osmotic control) as indicated. (C) Incubations were performed with 20 µM of the mitochondrial uncoupler dinitrophenol (DNP) or vehicle (ctrl.) as indicated. (D) Seedling were incubated with TOR kinase inhibitor AZD or vehicle (ctrl.) as indicated. Bar colors and RT-qPCR conditions were as in Fig. 2.

Sugars have dual roles in plants, serving both as metabolic fuel and as signaling molecules (23). The signaling pathway in which sugars activate target of rapamycin (TOR) kinase and subsequently gene expression has been involved in the regulation of AS by sugars in Arabidopsis roots (10). Xiong et al. (24) revealed that glucose activation of TOR kinase in Arabidopsis meristems depended on glycolysis-mitochondria-mediated energy. This is consistent with findings that the respiratory chain/oxidative phosphorylation uncoupler 2,4-dinitrophenol (DNP) suppresses mammalian target of rapamycin (mTOR) activation in brain (25). Since DNP was also shown to abolish changes in root AS induced by light exposure of the green tissues (10), we investigated the effects of DNP on APA. In order to be able to assess DNP effects in the roots, experiments were carried out in the POST protocol, i.e., where excision was performed after the light/dark or drug treatments. Fig. 4C shows that DNP does not suppress the light effect on APA in the shoots, but completely abolishes the conspicuous light effect in the roots, in a way that imitates darkness. Again, looking only at the shoots in the case of the RKH APA event, we see that DNP abolishes the light effect, mimicking darkness (SI Appendix, Fig. S4B).

To link the need of mitochondrial activity to a putative role of the TOR kinase pathway, we decided to investigate the effects of a pharmacological inhibitor of this enzyme. Fig. 4D and SI Appendix, Fig. S4C show that the TOR kinase inhibitor AZD (AZD-8055) abolishes the effect of light on APA in the HTA9 and RKH respectively. Control experiments for sucrose, DNP, and AZD with the APA unresponsive FAD6 gene are shown in SI Appendix, Fig. S5.

Unlike AS, Light Control of APA is Not Affected by Treatments Modifying RNAPII Dynamics.

We used two experimental approaches to evaluate whether the effect of light on HTA9 APA was linked to changes in transcriptional elongation according to the kinetic coupling mechanism. To affect transcriptional elongation, we assessed an Arabidopsis mutant of the TFIIS transcription elongation factor. TFIIS is a factor required for RNAPII processivity that affects backtracking and stimulates RNAPII to reassume elongation after pausing (26, 27). The tfiis mutant was achieved by replacing the key amino acids D290 and E291 of the acidic loop responsible for TFIIS stimulatory activity by alanines (27) as described in yeast (28), giving rise to a dominant negative phenotype showing a range of developmental defects, such as defective growth and serrated leaves. We have previously reported that in the tfiis mutant, transcript elongation is reduced and the change in AtRS31 AS induced by light is abolished, mimicking the effects of darkness (7). While this is confirmed as a positive control in Fig. 5A, the effect of light on HTA9 APA is not affected in the tfiis mutant (Fig. 5B). The negative control FAD6 behaves similarly (Fig. 5C). On the activation side, we explored the use of a histone deacetylase inhibitor (HDAC) trichostatin A (TSA). By promoting higher histone acetylation and chromatin opening, HDACs have been shown to promote elongation in animal (29, 30) and plant cells (7) and, in the latter, to mimic the chloroplast-mediated light effect on AtRS31 AS. Again, while this is confirmed as positive control in Fig. 5D, the effect of light on HTA9 APA is not affected by TSA treatment (Fig. 5E) and the negative control FAD6 behaves similarly (Fig. 5F).

Fig. 5.

AS and APA respond differently to factors affecting transcript elongation. Effects of genetic disruption (tfiis mutant) of the transcription elongation factor TFIIS (A, B and C) and of treatment with the histone deacetylase inhibitor TSA (D, E and F) on AS of the Arabidopsis RS31 gene (A and D) and on APA of the HTA9 (B and E) and FAD6 (C and F) genes. Bar colors and RT-qPCR conditions were as in Fig. 2.

The Light Effect on APA does not Correlate with Changes in Gene Expression.

Our 3’READS data also detected widespread gene expression changes in light/dark conditions (SI Appendix, Fig. S6A). A detailed analysis in SI Appendix, Fig. S6B, shows that of the 815 genes in which light shortens 3’UTRs, approximately 71% do not change expression, 14% upregulate their expression, and 15% downregulate it. In a similar way, of the 2,652 genes in which light lengthens 3’UTRs, approximately 72% do not change expression, 11% upregulate their expression, and 17% downregulate it. This distribution per se suggests that the APA changes caused by the light/dark conditions do not correlate with changes in gene expression. Moreover, we confirmed this conclusion by demonstrating that the FAD6 gene, whose APA does not change upon light and was therefore used as negative control in this report and displays big changes in gene expression that are not affected in phytochrome/cryptochrome (SI Appendix, Fig. S6C) and TFIIS mutants (SI Appendix, Fig. S6D) nor by DCMU treatment (SI Appendix, Fig. S6E).

Changes in CPA Factor mRNA Abundance.

In view of the fact that the light effect on APA does not seem to follow the kinetic coupling mechanism, we decided to investigate whether the changes in APA in the light could be explained by changes in the expression of CPA factors measured at the mRNA level. We assessed three subunits of CPSF and two of CstF in shoots and roots obtained from excision performed after the light/dark treatment. All three CPSF subunits (100, 160, and 30) exhibited reduction in their steady-state mRNA levels in the dark both in the shoots and in the roots (Fig. 6A). The effect is smaller in the roots (note smaller values for the white bars) suggesting that a signal generated in the green tissues reaches the roots but is not as effective as the signal in the shoots. In the case of CstF, while no changes in mRNA levels were observed for CstF64 neither in the shoots nor in the roots, CstF77 mRNA abundance drops in the dark in the shoots but is not affected in the roots (Fig. 6B). To investigate if, as shown for APA in Fig. 2 E–G, a functional chloroplast electron transport chain is needed for upregulation of CPSF100, 160, and 30 and CstF77 by light, we treated whole seedlings with DCMU. Consistently, the DCMU treatment abolished the light/dark effect on mRNA levels for these subunits (Fig. 6 C and D).

Fig. 6.

Light upregulates mRNA levels of cleavage/polyadenylation factors through the chloroplast. (A and B) RT-qPCR quantification of mRNA levels encoding subunits of CPSF (A) and CstF (B) in Arabidopsis shoots and roots obtained in a post-light/dark treatment excision experiment. (C and D) Effect of the photosynthetic electron transfer chain inhibitor DCMU on the light/dark effect on cleavage/polyadenylation factor mRNA levels in whole seedlings for CPSF (C) and CstF (D). Factor mRNA levels were relativized to mRNA levels of protein phosphatase 2A (PP2A). Bar colors and RT-qPCR conditions were as in Fig. 2.

Discussion

We report here an unforeseen, potent regulation of APA by light/dark conditions in plants. It should be noted that the effects here described occur in an intact living organism (Arabidopsis seedlings) and under a fundamental physiological external signal (light). Although this investigation was initially inspired by our previous work on the effect of light on plant AS (6, 7,10), our findings indicate that, despite some similarities, APA regulation has fundamental differences with respect to that of AS. In approximately 30% of Arabidopsis genes having APA, this alternative RNA processing event is affected by light/dark conditions, which highlights a global impact. In the regulated genes, light promotes the usage of both dPAS (75% of genes) and pPAS (the remaining 25%) (Fig. 1). It is important to point out that the light effect on APA does not correlate with changes in gene expression (SI Appendix, Fig. S6).

Because of the predominance of dPAS usage and the robustness of its response, we decided to use the APA event of a gene encoding a histone H2A variant (HTA9) as a model to investigate the mechanisms involved. An ideal event to study would have been APA in the long noncoding RNA COOLAIR (AT5G01675) that is key in the control of flowering (4, 5), but unfortunately, COOLAIR is not expressed in Col0 or Ler background seedlings (31), the paradigm system of our previous work on splicing and the present work on polyadenylation.

Similar to the regulation of AS, the light effect on APA is not mediated by phytochrome nor cryptochrome photoreceptors but depends on chloroplast function. The correct functioning of the photosynthetic electron transport chain is necessary since its inhibition by DCMU abolishes the effect (Fig. 2). This is consistent with the fact that the light effect is observed both in the green tissues and the roots as long as their connection is not interrupted by dissection (Fig. 3). Since the roots have no functional chloroplasts, the absence of effect in isolated roots indicates that a signaling molecule must travel through the phloem from shoots to roots. We identify this molecule as the main photosynthate sucrose because it fully mimics the effect of light in the roots when seedlings are kept in the dark (Fig. 4). Most interestingly, we found that root mitochondrial activity is necessary in the roots since the respiratory chain/oxidative phosphorylation uncoupler 2,4-dinitrophenol (DNP) abolishes the light effect fully mimicking the effect of darkness in the roots when seedlings are kept in the light. The opposite effects of sucrose and DNP allowed us to hypothesize that, like in AS regulation (10), the APA mechanism could involve sugar activation of TOR kinase, previously shown to depend on mitochondria-generated energy (24). Consistently, the TOR kinase inhibitor AZD behaves like DNP by abolishing the light/dark effect on HTA9 APA in the roots (Fig. 4). It is worth mentioning that both DNP and AZD abolish the change in APA by mimicking the effect of darkness, further supporting the idea that mitochondrial and TOR kinase activities are needed to interpret the light-generated signal (i.e., sucrose). Notably, 3’UTR size regulation through APA is increasingly associated with cell metabolism in mammalian cells (32). Whether plants and metazoans share similar mechanisms in energy-mediated APA regulation is to be examined in the future.

Two experiments of different nature indicate that, unlike AS, the effect on APA is not linked to the control of RNAPII elongation. The light effect on APA is neither abolished in a mutant of the transcription elongation factor TFIIS nor in seedling treated with the histone deacetylase inhibitor TSA, previously shown to cause higher transcript elongation due to histone acetylation and chromatin relaxation (Fig. 5). Important controls show that both the TFIIS mutant and TSA treatment suppress the light effect on AS, the former mimicking darkness and the latter mimicking light as published (7). In search for an alternative mechanism we wondered whether light would be affecting the expression of cleavage/polyadenylation factors. We found indeed that light upregulates mRNA levels for CPSF100, 160, and 30 and CstF77 and that this increase is suppressed by treatment with DCMU (Fig. 6). In particular, Arabidopsis CPSF30 has been reported to interact with other polyadenylation factors like FIPS5, CPSF100, and CstF77 and can be considered as a central hub in the protein–protein interaction network of plant polyadenylation complex subunits (33–35). Indeed, Arabidopsis PAS selection is determined by the presence or absence of this factor (36). The precise mechanism by which CPSF30 controls Arabidopsis APA remains elusive. However, the current model for canonical plant PA signaling involves three discreet cis-elements collaborating to the effective 3’ end formation of mRNAs: i) the far upstream element (FUE), consisting of an extended U and G-rich region situated more than 50 nucleotides upstream from the PAS; ii) the near-upstream element (NUE), a 6 to 10 nucleotide-A-rich region situated 10 to 30 nucleotides upstream from the PAS; and iii) the cleavage element (CE), a U-rich region centered around the PAS. Although most functioning aspects of these elements have not yet been characterized, CPSF30 was shown to play a role in the functioning of the NUE because PAS that are only used in the wild type, but not the CPSF30-deficient mutant, possess the characteristic A-rich NUE signature, while PAS used only in the CPSF30-deficient mutant lack this signature (36). In mammalian models, there is accumulated evidence that a variety of core CPA factors regulate APA (37–42). Moreover, a model was proposed in which the choice of PASs is dependent on the strengths of the cis-elements present in the PAS and the relative usage dependent on the competition between PASs for the available CPA factors (43, 44). Upon loss or diminishment of core CPA factor(s) the relative strengths of all PASs decrease. However, any factor that would increase the window of opportunity for the CPA factors to recognize pPAS would lead to a shift toward the pPAS, such as RNAPII pausing, slowing RNAPII elongation, or increasing the distance between pPAS and dPAS. Even though there is still no evidence of this model in plants, it is an interesting putative scenario to see the light as a regulator of CPA factors abundance and PAS selection. Interestingly, we found that the distance between two APA sites is important for both types of APA isoform regulation, with 3’UTR lengthening and those with 3’UTR shortening (SI Appendix, Fig. S6).

As 3’UTRs are very rich in regulatory elements, such as miRNA target sites, we performed a preliminary bioinformatics analysis revealing that the average number of miRNA target sites per 3’UTR is 1.16 (SI Appendix, Fig. S8A). However, alternative 3’UTRs regulated by light are not more enriched in miRNA target sites compared to constitutive ones (SI Appendix, Fig. S8B). Consistently, we found that the percentage of genes with miRNA target sites is similar in the three categories assessed in this study in response to light (longer 3’UTRs, shorter 3’UTR, and no change).

In particular, the gene with the event chosen as output, HTA9, encodes a histone H2A variant (H2A.Z). This variant has been associated with environmental responses to temperature and stress (45, 46). On the other hand, it was shown that H2A.Z-containing nucleosomes wrap DNA more tightly than canonical H2A nucleosomes, which may affect RNAPII elongation and, in turn, AS (7). In any case, further work will be necessary to better understand the biological roles of the unique link between plant APA and light reported here.

Materials and Methods

Plant Material, Growth Conditions, and Drug Treatments.

The Arabidopsis Columbia ecotype (Col 0) and Landsberg erecta (Ler) were used as wild type, according to the mutants assessed. Seeds were stratified for three days in the dark at 4 °C and then germinated on Murashige and Skoog 0.5x (MS) medium containing 1% (w/v) agar. Standard treatment protocol: A. thaliana seedlings were grown in Petri dishes with MS medium at a constant temperature of 22 °C under constant white light provided by fluorescent tubes with an intensity of irradiance between 70 and 100 μmol/m²sec (20 seeds per dish) for a period of 2 wk and then transferred for 48 h to darkness. After this period, seedlings were kept in the dark or transferred to the light for 6 h. At the end of this light/dark treatment, seedlings were harvested in liquid nitrogen. For all pharmacological treatments, drugs were added 1 h prior to the light/dark treatment. For the DCMU [3-(3,4-dichlophenyl)-1,1-dimethylurea; Sigma] treatment, seedlings were transferred to 6-well plates and incubated in 20 µM DCMU. For the sucrose treatment, plants on agar plates were covered with 10 mL of 100 mM sucrose or sorbitol, used as osmotic control. A vacuum was applied for five minutes to facilitate drug uptake. For the DNP (2,4-dinitrophenol; Sigma) and AZD (AZD-8055; Chemdea LLC) treatments, the drugs were added up to 20 µM, using ethanol and DMSO, respectively, as vehicle control.

3′READS and PAS Identification.

Total RNA extraction of seedlings was carried out by using the RNeasy Plant Mini Kit (Qiagen) following the manufacturer’s instructions. The 3′ region extraction and deep sequencing (3′READS) method were described previously (47) Libraries were sequenced on an lllumina HiSeq machine (2 × 150 bases) at Admera Health (New Jersey). 3’READS data were analyzed as previously described (16, 48, 49). Briefly, after adapter sequence removal, reads were mapped to the A. thaliana reference genome (TAIR10) by using the Bowtie2 program (50). Sequences with a mapping quality (MAPQ) score ≥10 and ≥2 nongenomic Ts at the 5′ end after alignment were considered as PAS-supporting (PASS) reads and were used for subsequence analysis. Identified cleavage sites within 24 nt from one another were clustered into PAS clusters (51). 3’READS data statistics are shown in SI Appendix, Table S1. Gene annotation was based on the TAIR10 and the Ensembl databases.

Analysis of APA Isoforms.

Analysis of 3’UTR APA was based on the top two most expressed APA isoforms of a gene, which were named proximal and distal PAS isoforms. To eliminate spurious APA isoforms, we further required that the number of PASS reads for the minor APA isoform (the second-most expressed) to be above 5% of all isoforms combined. For 3’UTR APA analysis, only the APA sites in the last exon were used. The relative expression (RE) of two PAS isoforms, e.g., pPAS and dPAS, was calculated by log2(RPM) of dPAS vs. pPAS, where RPM was reads per million PASS reads. Relative expression difference (RED) of two isoforms in two comparing samples was based on the difference in RE for the two isoforms between the two samples. DEXSeq was used to derive statistically significant APA events (FDR < 0.05) (52).

Gene Expression Analysis.

The DESeq method (R Bioconductor) (53) was used to analyze gene expression changes. Significantly regulated genes are those with adjusted P < 0.05 and fold change > 1.2. PAS reads of each gene were summed to represent gene expression.

PAS Motif Analysis.

The PROBE program was used to examine sequence motifs around the PASs (54). The genomic region surrounding each PAS was divided into four subregions: −60 to −31 nt, −30 to −1 nt, +1 to +30 nt, and +31 to +60 nt. The observed frequency of each k-mer in a subregion was enumerated and compared with the expected frequency based on randomized sequences of the region. Randomization was carried out by using the first-order Markov chain model (54). The enrichment score (Z-score) was calculated based on the difference between the observed and expected frequencies. The Fisher’s exact test was used to determine significance.

Gene Ontology Analysis.

The GOstats hypergeometric test (R Bioconductor) was used to test for significant association of genes with gene ontology (GO) terms. GO annotation for A. thaliana was obtained from org.At.tair.db (R Bioconductor). GO terms associated with more than 1,000 genes were considered too generic and were discard. To remove redundant GO terms, each reported GO term was required to have at least 25% of the genes that were not associated with another term with a more significant p-value.

APA Analysis by RT-qPCR.

Seedlings were grown following specifications given in Plant material, growth conditions, and drug treatments. Samples were harvested and total RNA was purified using TRIzol (Invitrogen). Then, 500 ng of RNA was further used to synthesize cDNA with MMLV-RT enzyme (Invitrogen) and oligo-dT as primer following the manufacturer’s instructions. Synthesized cDNAs were amplified with 1.5 U of Taq DNA polymerase (Invitrogen) and SYBR Green (Roche) using the Eppendorf Mastercycler Realplex. Primer sequences for qRT-PCR are available in SI Appendix, Table S3. RT-qPCR experiments were quantified with n ≥ 3, where n = about 25 to 30 Arabidopsis seedlings growing in one Petri dish. Changes considered significant show differences with a P value < 0.05 (two-tailed Student’s t test).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Data sets have been deposited in GEO NCBI (55).