Molecular consequences of NCED6 induction in imbibed seeds

The induction of NCED6 with the Gene Switch (GS) system, a chemically inducible gene expression system (Martinez-Andujar et al., Reference Martinez-Andujar, Ordiz, Huang, Nonogaki, Beachy and Nonogaki2011), is sufficient to cause germinating seeds to suspend or reverse the physiological processes proceeding towards radicle protrusion (Fig. 1A). While NCED is a rate-limiting enzyme for ABA biosynthesis, induction of an enzyme should not be effective if the substrates are not present. Therefore, the suppression of germination by NCED6 induction alone suggests that the pathways upstream of the rate-limiting reaction are operational and a substantial amount of the NCED substrates 9′-cis-neoxanthin and/or 9-cis-violaxanthin is continuously supplied even in non-dormant seeds. The steady supply of the substrates for the rate-limiting ABA biosynthesis enzyme even during germination is probably a critical strategy for seeds to respond to environmental changes promptly and cease the germination process immediately when they sense marginal conditions. The same mechanism may function in intact seeds before testa rupture to cause secondary dormancy, which could be induced by seed imbibition at high temperatures, in osmotica, or under darkness (Bewley et al., Reference Bewley, Bradford, Hilhorst and Nonogaki2013).

Figure 1.

The effects of nine-cis-epoxycarotenoid dioxygenase 6 (NCED6) induction on germination and gene expression. (A) Suppression of germination in the NCED6-induced Arabidopsis seeds. Germination of NCED6-inducible seeds (three independent lines 5-125, 8-181 and 15-132) was suppressed in the presence of the chemical ligand Intrepid2F (+IP) while those seeds germinated normally in the absence of the chemical ligand (–IP). (B) Differential gene expression between NCED6-induced and -uninduced seeds. Smear plot shows gene switch up-regulated (GS-UP) and down-regulated (GS-DOWN) genes (grey dots) upon NCED6 induction. CPM, count per million; FC, fold change. (C) Venn diagrams showing genes specifically up- (left, NCED6-UP) or down-regulated (right, NCED6-DOWN) in the NCED6-inducible seeds, excluding the GS-UP (59) and GS-DOWN (2) genes that were also differentially expressed in wild-type (WT) upon ligand application (WT-UP, WT-DOWN).

The ecological significance of seed responses to the environment is clear, however the molecular and biochemical mechanisms underlying the suppression of germination by ABA is not fully understood. To obtain more insights into ABA-regulated molecular events in seeds, RNA-seq was performed using the NCED6-induced and -uninduced Arabidopsis seeds. Induction of NCED6 using GS caused differential expression of a number of genes (Fig. 1B). The RNA-seq analysis identified 437 genes up-regulated (GS-UP) and 168 genes down-regulated (GS-DOWN) in the NCED6-induced seeds with statistical significance (q < 0.05). The genes that were also differentially expressed between ligand-treated or -untreated WT seeds were subtracted from the original GS-UP and GS-DOWN groups and the remainders were termed NCED6-UP (378) and NCED6-DOWN (166) (Fig. 1C, Table S2). NCED6 was one of the highly differentially expressed genes, which verified successful induction of gene expression during sample preparation for the RNA-seq experiments and the propriety of the computational and statistical analyses (Table S2). NCED6-UP genes included known ABA signalling components (e.g. ABA INSENSITIVE 5), annotated yet uncharacterized genes (e.g. DELAY OF GERMINATION1-LIKE4) and unknown proteins (e.g. At3g48510). Differential expression of the representative genes was confirmed by RT-PCR (Fig. 2).

Figure 2.

Differential expression of representative NCED6-UP genes. (A) Differential expression detected by RNA-seq analysis between induced (+, filled bar) and uninduced (–, open bar) seeds. FPKM, fragments per kilobase of exon per million fragments mapped. ABI5, ABA INSENSITIVE5; ACT2, ACTIN 2, DOGL4, DELAY OF GERMINATION1-LIKE 4. **P < 0.01 (Student's t-test compared with control). (B) RT-PCR of the same group of genes. Representative gel images are shown.

As anticipated, many of the differentially expressed genes were associated with abiotic stresses, including cold, high temperature and drought (Table S2). The ABA-induced protein phosphatases 2C (PP2C) HIGHLY ABA-INDUCED1 (HAI1), HAI2 and HAI3 (Bhaskara et al., Reference Bhaskara, Nguyen and Verslues2012) were detected as NCED-UP genes. In addition to ABI5, a major regulator of seed maturation, typical seed maturation genes, such as LATE EMBRYOGENESIS ABUNDANT (LEA) and storage proteins were also detected as NCED6-UP genes (Table S2, Fig. 3). These results suggest that the seed maturation programmes can, at least in part, be re-introduced to the testa-ruptured seeds if NCED expression reaches a certain threshold, even though they are at the very final step of germination.

Figure 3.

Schematic representation of the major NCED-UP (black) and NCED-DOWN (grey) genes found in the NCED6-induced seeds and their mechanistic association. See text for details. ELIP, EARLY LIGHT-INDUCIBLE PROTEIN; HAI, HIGHLY ABA-INDUCED; JMJ, JUMONJI; LEA, LATE EMBRYOGENESIS ABUNDANT; MFT, MOTHER OF FT AND TFL1; OLE, OLEOSIN; PP2C, PROTEIN PHOSPHATASE 2C; SOM, SOMNUS.

The misguided expression of NCED6 at the last stage of germination also triggered expression of the major regulators of seed dormancy, such as SOMNUS (SOM), a seed-specific nucleus-localized CCCH-type zinc finger protein (Kim et al., Reference Kim, Yamaguchi, Lim, Oh, Park, Hanada, Kamiya and Choi2008; Park et al., Reference Park, Lee, Kim, Lim and Choi2011) and MOTHER OF FT AND TFL1 (MFT), a phosphatidylethanolamine-binding protein (Yoo et al., Reference Yoo, Kardailsky, Lee, Weigel and Ahn2004; Xi et al., Reference Xi, Liu, Hou and Yu2010). SOM is activated by PHYTOCHROME-INTERACTING FACTOR3-LIKE5 (PIL5) and ABI3 (Park et al., Reference Park, Lee, Kim, Lim and Choi2011), which directly bind to the SOM promoter while ABI3, ABI5 and DELLA proteins can also interact with each other to bind the SOM promoter at high temperature (Lim et al., Reference Lim, Park, Lee, Jeong, Toh, Watanabe, Kim, Kang, Kim, Kawakami and Choi2013). It has been demonstrated that expression of the GA biosynthesis genes GA3ox1 and GA3ox2 are reduced by SOM expression, suggesting that SOM suppresses seed germination by reducing GA levels in seeds (Kim et al., Reference Kim, Yamaguchi, Lim, Oh, Park, Hanada, Kamiya and Choi2008). In fact, GA3ox1 and GA3ox2 were detected as NCED-DOWN genes in our analysis, together with SOM as an NCED-UP gene (Fig. 3).

Down-regulation of GA3ox1 and GA3ox2 can be mediated by suppression of the Jumonji domain-containing proteins 20 and 22 (JMJ20/22), which are histone arginine demethylases. JMJ20/22 activate GA3ox1 and GA3ox2 by removing the inhibitory methyl marks from these genes (Cho et al., Reference Cho, Ryu, Jeong, Park, Song, Amasino, Noh and Noh2012). It is known that the transcript levels of JMJ20/22 are increased in the pil5 and som mutants, suggesting that SOM suppresses JMJ20/22 (Cho et al., Reference Cho, Ryu, Jeong, Park, Song, Amasino, Noh and Noh2012). Therefore, it is conceivable that the down-regulation of GA3ox1 and GA3ox2 in the NCED6-induced seeds was the consequence of the up-regulation of SOM (Fig. 3). However, JMJ20 and JMJ22 were not detected as NCED-DOWN genes in our analysis. Thus the mechanism of GA3ox regulation by SOM in the NCED6-induced seeds is not clear.

MFT is a gene that is specifically induced in the radicle–hypocotyl transition zone of the embryo in response to ABA (Xi et al., Reference Xi, Liu, Hou and Yu2010). Up-regulation of MFT in the induced seeds also verifies the strong ABA response occurring in the NCED6-induced seeds, which must be the consequence of increased expression of ABI5, because MFT is directly regulated by ABI5 (Xi et al., Reference Xi, Liu, Hou and Yu2010) (Fig. 3). MFT exerts a negative feedback effect to ABI5 (Xi et al., Reference Xi, Liu, Hou and Yu2010) (Fig. 3). The extremely high levels of ABA production in the NCED6-induced GS seeds (Martinez-Andujar et al., Reference Martinez-Andujar, Ordiz, Huang, Nonogaki, Beachy and Nonogaki2011) probably triggered the negative feedback regulation to counteract the excessive ABA levels and maintain a certain level of ABA–GA balance in seeds. However, as ABI5 was still detected as an NCED-UP gene, the robustness of the GS system most likely overrode and masked the effects of the negative feedback function by MFT.

The NCED GS system was originally created to gain a proof of concept for technology development for prevention of pre-harvest sprouting (PHS) in cereals (Martinez-Andujar et al., Reference Martinez-Andujar, Ordiz, Huang, Nonogaki, Beachy and Nonogaki2011; Nonogaki et al., Reference Nonogaki, Sall, Nambara and Nonogaki2014). In wheat, potential of altered expression of a MFT orthologue for dormancy regulation and PHS control has been demonstrated (Nakamura et al., Reference Nakamura, Abe, Kawahigashi, Nakazono, Tagiri, Matsumoto, Utsugi, Ogawa, Handa, Ishida, Mori, Kawaura, Ogihara and Miura2011). Therefore, the GS system (Martinez-Andujar et al., Reference Martinez-Andujar, Ordiz, Huang, Nonogaki, Beachy and Nonogaki2011), which can increase both ABA biosynthesis (NCED) and response (MFT), is expected to be effective for PHS prevention in cereals. A more advanced system of PHS prevention, which is devoid of potential pleiotropic effects with the GS system, has also been developed based on the knowledge obtained for seed responses to ABA (Nonogaki et al., Reference Nonogaki, Sall, Nambara and Nonogaki2014). Therefore, the basic mechanisms analysed here in NCED6-induced seeds will also be important for translational biology for the future (Nonogaki and Nonogaki, Reference Nonogaki and Nonogaki2017).

Other negative feedback responses, besides MFT-ABI5, were also observed in the NCED6-induced seeds. HAI PP2Cs mentioned above are negative regulators of ABA signal transduction (Fig. 3). CYP707A1, an ABA deactivation gene, was also an NCED-UP gene, presumably as another counteracting response to ABA increase in seeds (Fig. 3). NCED6 induction seems to cause a number of changes in hormone metabolism genes themselves (discussed below).

Positive feedback mechanisms in ABA biosynthesis

The initial analysis using the Tophat2 program had identified NCED2 and NCED9 as NCED6-UP genes, which suggests the presence of positive-feedback regulation in the ABA biosynthesis pathway. The positive-feedback regulation has been suggested for NCED5 (Okamoto et al., Reference Okamoto, Tatematsu, Matsui, Morosawa, Ishida, Tanaka, Endo, Mochizuki, Toyoda, Kamiya, Shinozaki, Nambara and Seki2010; Nonogaki et al., Reference Nonogaki, Sall, Nambara and Nonogaki2014) and NCED9 (Okamoto et al., Reference Okamoto, Tatematsu, Matsui, Morosawa, Ishida, Tanaka, Endo, Mochizuki, Toyoda, Kamiya, Shinozaki, Nambara and Seki2010) in the previous studies. Although the list of NCED6-UP by the HISAT2 analysis did not include these NCEDs (Table S2), NCED2, NCED5 and NCED9 seemed to be enhanced by NCED6 induction in the same dataset (Fig. 4, histograms). Experiments by RT-PCR confirmed that NCED2, NCED5 and NCED9 are also subject to positive feedback (Fig. 4, top right inset). In a separate project, we prepared DOGL4-inducible lines (authors’ unpublished observations) and found that induction of DOGL4, which was one of the NCED6-UP genes (Fig. 2), increased NCED9 expression (Fig. 4, bottom right inset). This observation also confirms that NCED9 is subject to positive feedback regulation in the NCED6-inducible system (NCED6→ABA→DOGL4→NCED9). The positive feedback mechanism seems to play a central role in the regulation of the rate-limiting step of (and hence the entire) ABA biosynthesis pathway (Fig. 4, pathway scheme).

Figure 4.

Positive feedback mechanisms in the ABA biosynthesis pathway found in the NCED6-induced seeds. Histograms show differential expression of the ABA biosynthesis genes in the NCED6-induced (+, filled bar) and -uninduced (–, open bar) seeds. **P < 0.01 (Student's t-test compared with control). Schematic representation of the major ABA biosynthesis pathway is shown in the center. Top right inset, RT-PCR for expression of other NCEDs in NCED6-induced (+) or uninduced (–) seeds. Bottom right inset, RT-PCR for NCED9 expression in DOGL4 (DELAY OF GERMINAITON1-LIKE4)-induced (+) or -uninduced (–) seeds (see text for DOGL4). Induction of NCED6 caused up-regulation of NCED2, NCED5 and NCED9, suggesting positive feedback regulation of NCEDs in the native system (dashed black arrow in the centre). Representative gel images are shown. ZEAXANTHIN EPOXIDASE (ZEP) and ABSCISIC ALDEHYDE OXIDASE 3 (AAO3), which function up- and downstream of NCEDs, were also subject to positive feedback (top and bottom dashed arrows), which suggests coordinated enhancement of ABA biosynthesis in seeds. FPKM, fragments per kilobase of exon per million fragments mapped.

The high magnitude of NCED expression, which leads to over-production of the rate-limiting enzymes, could deplete its substrate 9′-cis-neoxanthin and 9-cis-violaxanthin from the NCED6-induced seeds (Fig. 4). Interestingly, ZEAXANTHIN EPOXIDASE (ZEP), which catalyses the upstream reaction and subsequently provides the substrates for NCED, was an NCED-UP gene (Table S2, Fig. 4), indicating that there is another positive-feedback loop upstream of ABA biosynthesis pathway (Fig. 4).

While the steady supply of the substrates could meet the demands of enhanced NCED activities, the products of enzyme reaction could over-accumulate if the downstream pathways are not operating accordingly. Therefore, another positive feedback loop may be present downstream as well. Indeed, Arabidopsis ABSCISIC ALDEHYDE OXIDASE 3 (AAO3), which catalyses the last step of abscisic aldehyde conversion to ABA, was also detected an NCED-UP gene (Table S2) (Fig. 4). At least the major steps of the ABA biosynthesis pathway in seeds seem to function coordinately to enhance ABA biosynthesis (Fig. 4). These results explain the robustness of the induction of NCED6, a single gene, for the suspension of seed germination.

Positive feedback mechanisms were suggested for ZEP and AAO3 in non-seed tissues (Xiong et al., Reference Xiong, Lee, Ishitani and Zhu2002) and for NCEDs in seeds (Okamoto et al., Reference Okamoto, Tatematsu, Matsui, Morosawa, Ishida, Tanaka, Endo, Mochizuki, Toyoda, Kamiya, Shinozaki, Nambara and Seki2010; Nonogaki et al., Reference Nonogaki, Sall, Nambara and Nonogaki2014) previously. Our data provided strong evidence of multiple layers of positive-feedback loops present in the NCED6-induced seeds. The mechanistic details of the positive feedback regulation in ABA metabolism in seeds are not clear, except for the DOGL4–NCED9 interaction. However, some of the NCED-UP genes described above, such as SOM and MFT (Fig. 3), probably contribute to the positive feedback mechanism in ABA metabolism in the NCED6-induced seeds. ZEP expression is reduced in the som mutant, suggesting that SOM up-regulates ZEP (Kim et al., Reference Kim, Yamaguchi, Lim, Oh, Park, Hanada, Kamiya and Choi2008). Therefore, it is possible that the positive feedback to ZEP, which was observed in the NCED6-induced seeds (Fig. 4), was mediated through the enhancement of SOM expression by NCED6 induction (Fig. 3). SOM also up-regulates NCED6 (Kim et al., Reference Kim, Yamaguchi, Lim, Oh, Park, Hanada, Kamiya and Choi2008). Therefore, it is possible that expression from the native NCED6 gene in the genome was stimulated by SOM, in addition to the ligand-dependent expression of the switchable NCED6 transgene. Probably, the expression of the native NCED6, which was enhanced by the increased SOM activity, also contributed to the large fold changes of NCED6 expression between induced and -uninduced seeds (Table S2). The major ABA deactivation gene CYP707A2 in Arabidopsis seeds (Kushiro et al., Reference Kushiro, Okamoto, Nakabayashi, Yamagishi, Kitamura, Asami, Hirai, Koshiba, Kamiya and Nambara2004) is negatively regulated by SOM (Kim et al., Reference Kim, Yamaguchi, Lim, Oh, Park, Hanada, Kamiya and Choi2008), which could also contribute to maintaining high levels of ABA in seeds, although this gene was not detected as an NCED-DOWN (i.e. SOM-DOWN) gene in our data set. All of these events by SOM that enhance ABA accumulation in seeds, together with the down-regulation of the GA biosynthesis genes by SOM (possibly through JMJ20/22 described above), are thought to have increased the ABA–GA ratio in seeds and prevented seed germination. Thus SOM is also a potential target of modification for future applications.

Long non-coding RNAs

Many unknown proteins that are induced by ABA have been detected by our RNA-seq analysis (Table S2). Some known but uncharacterized genes, such as DOGL4, were also found as NCED-UP or ABA-responsive genes. In addition, RNA-seq reads were assembled to the intergenic regions where protein coding genes were not annotated by the Salk T-DNA express Arabidopsis Gene Mapping Tools (http://signal.salk.edu/cgi-bin/tdnaexpress) (examples shown in Fig. 5A). These RNAs, which were expressed from intergenic regions, were considered as long intergenic non-coding RNAs (lincRNAs). We found five lincRNAs responding to NCED6 induction in seeds (termed N6LINCRs) (Fig. 5B). N6LINCR1-5, which are about 400–800 bases in size spread over different chromosomes and show distinct sequences while all of them were found between coding genes without overlapping them (Fig. 6) (except for some known non-coding RNAs discussed later).

Figure 5.

Expression of lncRNAs from intergenic regions (lincRNA) of the Arabidopsis genome upon NCED6 induction. (A) Example of assembly of RNA-seq reads at the coding genes (black boxes) and an intergenic region (boxed by dashed line). RNA-seq reads from three independent lines of the wild-type (WT1, WT2, WT3) and NCED6-inducible (N6-1, N6-2, N6-3) seeds with (+) or without (–) ligand treatment are assembled. (B–F) Expression of lincRNAs (N6LINCRs) in the NCED6-inducible seeds with (+, filled bar) or without (–, open bar) ligand application. **P < 0.01, *P < 0.05 (Student's t-test compared with control). FPKM, fragments per kilobase of exon per million fragments mapped.

Figure 6.

The N6LINCR loci in the Arabidopsis genome. (A) Schematic representation of the approximate positions of N6LINCR1-5 in the Arabidopsis chromosomes. (B) Schematic representation of the genome positions of N6LINCR1-5 relative to the neighbouring genes.

To examine a potential role of lincRNAs in seeds, we focused on and characterized N6LINCR1, which exhibited the most obvious response to NCED6 induction (Fig. 5B). RT-PCR analysis of N6LINCR1 using the NCED6-inducible seeds showed that it was expressed in seeds without induction but its expression was enhanced by NCED6 induction (Fig. 7A). N6LINCR1 expression was enhanced by exogenous ABA (10 µM) in WT seeds and also in leaves (Fig. 7B). Strand-specific RT-PCR showed that N6LINCR1 was expressed preferentially from one strand (Fig. 7C). In this case, enhancement of its expression by NCED6 induction was also observed, confirming the N6LINCR1 response to the increase in endogenous ABA. The other strand was also expressed at a detectable level (discussed below).

Figure 7.

Expression of N6LINCR1 and N6LINCR1-regulated genes. (A) Expression of N6LINCR1 in the NCED6-inducible seeds with (+) or without (–) ligand application. (B) Expression of N6LINCR1 in wild-type (WT) seeds (left) or leaf (right) in water (H₂O) or ABA. (C) Strand-specific RT-PCR using the forward (N6LINCR1-F0) or reverse (N6LINCR1-R0) primer (see Table S1). RNA was extracted from the NCED6-inducible seeds incubated with (+) or without (–) the ligand. (D) RT-PCR showing enhanced expression of N6LINCR1 in the N6LINCR1-inducible seeds upon ligand application (+) compared with control (–). (E) and (F), RT-PCR showing up-regulation (E, UP) and down-regulation (F, DOWN) of the representative genes that were identified by RNA-seq analysis of the N6LINCR1-inducible seeds. ACT2, ACTIN2; ALMT1, ALUMINUM-ACTIVATED MALATE TRANSPORTER 1; ATERF2, ETHYLENE RESPONSE FACTOR2; CYP94B3, CYTOCHROME P450, FAMILY 94; DIC2, DICARBOXYLATE CARRIER 2; LBD41, LOB DOMAIN-CONTAINING PROTEIN 41; MYB15, MYB DOMAIN PROTEIN 15; PUB22, PLANT U-BOX 22; STZ, SALT TOLERANCE ZINC FINGER. Representative gel images are shown.

Transcription from part or the vicinity of the N6LINCR1 genomic region was also detected by previous EST (expressed sequence tag), tiling array or RNA-seq analysis (Matsui et al., Reference Matsui, Ishida, Morosawa, Mochizuki, Kaminuma, Endo, Okamoto, Nambara, Nakajima, Kawashima, Satou, Kim, Kobayashi, Toyoda, Shinozaki and Seki2008; Nakashima et al., Reference Nakashima, Fujita, Kanamori, Katagiri, Umezawa, Kidokoro, Maruyama, Yoshida, Ishiyama, Kobayashi, Shinozaki and Yamaguchi-Shinozaki2009; Okamoto et al., Reference Okamoto, Tatematsu, Matsui, Morosawa, Ishida, Tanaka, Endo, Mochizuki, Toyoda, Kamiya, Shinozaki, Nambara and Seki2010; Richter et al., Reference Richter, Behringer, Muller and Schwechheimer2010; Visscher et al., Reference Visscher, Paul, Kirst, Guy, Schuerger and Ferl2010; Qin et al., Reference Qin, Kodaira, Maruyama, Mizoi, Tran, Fujita, Morimoto, Shinozaki and Yamaguchi-Shinozaki2011; Jin et al., Reference Jin, Liu, Wang, Wong and Chua2013) (Fig. S1). Two non-coding RNAs (At1g06473 and At1g06483) are now annotated at this genomic region in the Salk T-DNA express Arabidopsis Gene Mapping Tools, although they are in opposite orientation to N6LINCR1. In any case, this genomic region, which does not code a protein, appears to be activated for transcription, and in the case of N6LINCR1 it is hormone-dependent transcription.

While N6LINCR1 was originally identified as an 819-bp lncRNA, the results of tiling RT-PCR using fine-mapped primers suggested that the broader genomic region of N6LINCR1 (1185 bp) can be transcribed (Fig. S2). This possible longest N6LINCR1 region (1185 bp) was used for further analyses. LincRNAs could regulate gene expression through sequence specificity to their targets (Turner et al., Reference Turner, Galloway and Vigorito2014). The 1185-bp N6LINCR1 sequence contained the oligonucleotides that completely or nearly completely matched part of the Arabidopsis genome (Fig. S3A and B). The duplicated sequence AAGAAATATATTAGTAATT of unknown function was found in N6LINCR1, which matched the vicinity of potential microRNA (miRNA) target genes in the Arabidopsis genome (Fig. S3C). The potential triplex forming oligonucleotides (TFO) (Buske et al., Reference Buske, Bauer, Mattick and Bailey2012) sequence TTTCCTCTTTCTCTTATCTCTC, through which a lncRNA could form a triplex with double-stranded DNA, was also found in N6LINCR1 (Fig. S3A). LncRNA could repress genes through TFO formation (Buske et al., Reference Buske, Bauer, Mattick and Bailey2012) or de-repress genes by serving as a miRNA sponge (Hansen et al., Reference Hansen, Jensen, Clausen, Bramsen, Finsen, Damgaard and Kjems2013; Turner et al., Reference Turner, Galloway and Vigorito2014) or target mimic (Liu et al., Reference Liu, Wang and Chua2015). However, any of the genes in the vicinity of the N6LINCR1-matched regions in the Arabidopsis genome, including the potential miRNA targets and TFO targets, did not show differential expression in the original RNA-seq data. The possibility that N6LINCR1 acts as a transregulator of genes located on other chromosomes (Galupa and Heard, Reference Galupa and Heard2015), through the TFO, miRNA or other mechanisms, should not be excluded yet because our examination was not comprehensive. However, it turned out to be difficult to speculate on possible targets of N6LINCR1 only through sequence specificity.

The importance of the secondary structure of lncRNAs, rather than their sequence homology to the target genes, has been suggested (Glazko et al., Reference Glazko, Zybailov and Rogozin2012; Turner et al., Reference Turner, Galloway and Vigorito2014). It is possible that N6LINCR1 regulates gene expression through its secondary structure (Fig. S4), which hinders identification of its targets only by computational sequence analysis. Therefore, we decided to create N6LINCR1-inducible lines to see whether its induction could affect gene expression profiles. We used the GS system to generate N6LINCR1-inducible lines. Resulting transgenic plants increased N6LINCR1 expression upon ligand application (Fig. 7D). Unlike NCED6 induction, apparent phenotypic changes were not observed in ruptured seeds by N6LINCR1 induction. However, when we ran RNA-seq for ligand-treated and -untreated seeds of the three independent N6LINCR1 lines, we observed obvious changes in gene expression profiles by N6LINCR1 induction (Table S3), some of which have been confirmed by RT-PCR (Fig. 7E and F). These results suggest that N6LINCR1 might function as a regulatory RNA, which potentially mediates ABA responses in seeds through transcriptional control. N6LINCR1 induction caused more down-regulation (66 N6LINCR1-DOWN) than up-regulation (26 N6LINCR1-UP) (Table S3).

Regulatory roles of lncRNAs in gene expression, particularly their suppressive roles, have been well documented. LncRNAs could interact with Polycomb Repressive Complex 2 (PRC2) through a secondary structure (Zhao et al., Reference Zhao, Ohsumi, Kung, Ogawa, Grau, Sarma, Song, Kingston, Borowsky and Lee2010), which triggers histone H3 lysine 27 (H3K27) trimethylation to cause gene silencing (Simon and Kingston, Reference Simon and Kingston2009). Possible regulation of the dormancy gene DOG1 through the PRC2 pathway has been suggested (Bouyer et al., Reference Bouyer, Roudier, Heese, Andersen, Gey, Nowack, Goodrich, Renou, Grini, Colot and Schnittger2011). Antisense DOG1 lncRNA (asDOG1), which suppresses DOG1 expression and negatively affects seed dormancy, has also been found (Fedak et al., Reference Fedak, Palusinska, Krzyczmonik, Brzezniak, Yatusevich, Pietras, Kaczanowski and Swiezewski2016). Seed dormancy mechanisms appear to be regulated by gene repression through chromatin remodelling, part of which could be mediated by lncRNA (Liu et al., Reference Liu, Koornneef and Soppe2007; Wang et al., Reference Wang, Cao, Sun, Li, Chen, Carles, Li, Ding, Zhang, Deng, Soppe and Liu2013; Nonogaki, Reference Nonogaki2014). Therefore, it is possible that N6LINCR1 plays a suppressive role in gene expression through epigenetic mechanisms.

While the major role of N6LINCR1 could be suppression of gene expression, 26 genes were detected as N6LINCR1-UP (Table S3). LncRNAs could promote gene expression by binding to transcription factors through a secondary structure, which in turn binds to the promoter regions of target genes (Turner et al., Reference Turner, Galloway and Vigorito2014). We do not know which genes are directly or indirectly regulated by N6LINCR1. However, it is interesting that some genes are up-regulated by N6LINCR1 induction.

In terms of individual genes, EARLY LIGHT-INDUCED PROTEIN2 (ELIP2), which has been characterized for seed germination mechanisms (Rizza et al., Reference Rizza, Boccaccini, Lopez-Vidriero, Costantino and Vittorioso2011), was found as a N6LINCR1-DOWN gene (Table S3). ETHYLENE RESPONSE FACTOR105 (ERF105), another gene identified as N6LINCR1-DOWN (Table S3), is known to be repressed through chromatin remodelling in seeds (Wang et al., Reference Wang, Cao, Sun, Li, Chen, Carles, Li, Ding, Zhang, Deng, Soppe and Liu2013), which can be regulated by a lncRNA. While these are interesting targets and possibilities, it probably does not make much sense to speculate on a possible role of N6LINCR1 in germination mechanisms just by extracting individual genes out of the lists and discussing available information. Understanding the mechanisms underlying how N6LINCR1 regulates those genes and seed germination will require a number of experiments in the future. Nonetheless, as demonstrated by this study, N6LINCR1 response to endogenous and exogenous ABA is unequivocal and its induction obviously alters gene expression profiles in seeds. Therefore, it is possible that N6LINCR1 mediates, at least in part, the suppressive effects of NCED6 induction on germination, through regulation of gene expression.

Thus, the transcriptome analysis in this study identified multiple layers of positive feedback regulation for coding genes and possible involvement of lncRNA in seed transcriptome.