Splicing-mediated activation of SHAGGY-like kinases underpinning carbon partitioning in Arabidopsis seeds

Abstract

Glycogen synthase kinase 3 (GSK3) family members serve as signaling hubs for plant development and stress responses, yet the underlying mechanism of their transcriptional regulation remains a long-standing mystery. Here we show that the transcription of SHAGGY-like kinase 11/12 (SK11/12), two members of the GSK3 gene family, is promoted by the splicing factor SmD1b, which is essential for distributing carbon sources into storage and protective components in Arabidopsis seeds. The chromatin recruitment of SmD1b at the SK11/12 loci promotes their transcription associated with co-transcriptional splicing of the first introns in the 5′-untranslated region of SK11/12. The loss of SmD1b function generates transcripts with unspliced introns that create disruptive R-loops to hamper the transcriptional elongation of SK11/12, in addition to compromising the recruitment of RNA polymerase II to the SK11/12 genomic regions. These effects imposed by SmD1b determine the transcription of SK11/12 to confer a key switch of carbon flow among metabolic pathways in zygotic and maternal tissues in seeds.

SmD1b binds to SK11/12 loci to prevent disruptive R-loop formation caused by intron retention and facilitate RNA polymerase II recruitment, thus promoting SK11/12 transcription to regulate carbon partitioning in seeds.

IN A NUTSHELL.

Background: Maternal carbon source supports seed development and is eventually transformed into seed reserves, such as storage oil in the filial embryo and protective compounds in the maternal seed coat. During seed development of Arabidopsis (Arabidopsis thaliana), a regulatory module, including SHAGGY-like kinases 11/12 (SK11/12), determines carbon partitioning into different metabolites in distinct tissues. Although it has been known that SK11/12 participate in this process by modulating the transcriptional regulators they interact with, how these two kinases are regulated during seed development is currently unknown. SK11/12 are two members of the glycogen synthase kinase 3 (GSK3) family, which consists of highly conserved serine/threonine protein kinases across eukaryotes. In contrast to the well-known regulation of GSK3s at the posttranslational level, the mechanisms underlying the transcriptional regulation of their encoding genes remain a long-standing mystery.

Question: We asked how SK11/12 are transcriptionally regulated during seed development to determine the distribution of carbon sources into storage and protective components in Arabidopsis seeds.

Findings: Here we report that transcriptional regulation of SK11/12 is mediated by a splicing factor, SmD1b, conferring a new regulatory layer for carbon flow between metabolic pathways in zygotic and maternal tissues in seeds. SmD1b binds to the SK11/12 chromatin region to prevent disruptive R-loop formation caused by retention of the first intron and also to facilitate the recruitment of RNA polymerase II, which promotes SK11/12 transcription to regulate carbon partitioning in seeds. These findings not only reveal a key molecular event that constitutes a master regulatory module for determining carbon flow in seeds, but also provide new insights into the transcriptional regulation of GSK3s.

Next steps: We are further characterizing other SmD1b interacting partners, which may provide more insights into the role of SmD1b in transcriptome reprogramming at specific loci in different developmental contexts, including seed development. 

Introduction

Seeds contain dormant juveniles covered by the protective seed coat established by mother plants. In the model plant Arabidopsis (Arabidopsis thaliana), seeds store oil in the filial embryo, while the maternal seed coat covering the embryo contains nonoil and protective components including mucilage and pigments. Since all these seed reserves share the same carbon source from the parent plant, the amount of seed reserves in zygotic tissues, such as seed oil in the embryo, is restricted via their competition with the reserves constituting the maternal seed coat components (Shi et al., 2012; Li et al., 2018). The control of carbon distribution across various seed tissues is called seed carbon partitioning, which is crucial for coordinating the development of zygotic and maternal tissues in seeds to prepare them for subsequent seedling establishment and survival in changing environments.

Notably, carbon partitioning is not only essential for plant survival, but also practically determines the quality of oilseed crops as high-value agricultural commodities. For example, the yellow-seed trait of rapeseed (Brassica napus) is a desirable trait in breeding, because lower pigmentation in the seed coat of yellow-seed varieties is often associated with higher storage reserves in embryos, such as seed oil (Jiang et al., 2019; Zhai et al., 2020). As seed oils are important for human and animal consumption as well as industrial uses, a mechanistic understanding of carbon partitioning sheds light on the manipulation of seed oil production in those agriculturally important oilseed crops. We recently found that two glycogen synthase kinase 3 (GSK3) family members, SHAGGY-like kinase 11 (SK11) and SK12, in Arabidopsis act as key regulators of carbon partitioning to redundantly determine seed oil levels by phosphorylating TRANSPARENT TESTA GLABRA 1 and reducing its interaction with TRANSPARENT TESTA 2, leading to the downregulation of GLABRA2 (GL2) transcription (Li et al., 2018). Despite these known mechanisms downstream of SK11 and SK12, how these two kinases are regulated to exert their roles in regulating carbon partitioning in seeds remains elusive.

GSK3 family members, including SK11/12, are conserved serine/threonine protein kinases that integrate signals of plant development and stress responses (Li et al., 2018, 2021; Chen et al., 2020). They are mainly regulated at the posttranslational level, as evidenced by the best-studied plant GSK3, BRASSINOSTEROID INSENSITIVE 2 (BIN2) in Arabidopsis (Hao et al., 2016; Zhu et al., 2017; Houbaert et al., 2018; Li et al., 2020b). GSK3s share some conserved regulatory features among various homologs in different plant species (Li et al., 2021). For example, a gain-of-function variant of BIN2, harboring the E263K mutation in its C-terminal TREE motif, stabilizes BIN2 as a result of decreased degradation via the proteasome pathway (Li and Nam, 2002; Peng et al., 2008). The equivalent mutations in the corresponding sites of SK11 (E292K) and SK12 (E297K) also stabilize these two kinases (Kim et al., 2009; Li et al., 2018). Although posttranslational regulation of plant GSK3s pertaining to their stability, activity, and localization has been well studied (Li et al., 2021), the mechanisms underlying their transcriptional regulation and the implications for functional specificity of plant GSK3 are so far unknown.

In this study, we show that a splicing factor, SmD1b, acts as a transcriptional activator of SK11/12 and consequently regulates seed carbon partitioning among the pathways leading to the production of seed oil, pigments, and seed coat mucilage in zygotic and maternal tissues during seed development. SmD1b is recruited to the SK11/12 loci to mediate co-transcriptional splicing of their first introns. This SmD1b-mediated splicing promotes the recruitment of RNA Polymerase II (RNA Pol II) and decreases the formation of DNA:RNA hybrids (R-loops) to facilitate RNA Pol II-mediated transcription elongation. The resulting SK11/12 upregulation in turn participates in a master regulatory module to determine the distribution of carbon sources into storage and protective components in Arabidopsis seeds.

Results

Isolation of trans-acting factors associated with the 5′-region of SK11

To understand how SK11/12 are transcriptionally regulated, we developed a genome targeting chromatin purification system called clustered regularly interspaced short palindromic repeats (CRISPR)/dead CRISPR associated protein 9 (dCas9)-mediated native chromatin affinity recovery (Cas9-ChAR) for an exploratory identification of trans-acting proteins associated with the 5′-region of SK11 (Supplemental Figures S1 and S2; see details in the “Materials and methods”). We isolated chromatin from wild-type Arabidopsis siliques 4-day after pollination (DAP), at a time when SK11 expression levels are the highest after fertilization (Li et al., 2018). Since the 5′-region of SK11 containing a 2.6-kb fragment upstream of the translation start codon was sufficient to recapitulate its native expression pattern in siliques (Li et al., 2018), we analyzed the chromatin-bound proteins over this 5′-region by liquid chromatography coupled with tandem mass spectrometry (LC–MS/MS). We carried out the same assay targeting the 3′-region of SK11 as the negative control (Supplemental Figure S1C). We identified 97 or 37 proteins associated with the 5′- or 3′-region of SK11, respectively, among which we selected 66 proteins associated only with the 5′-region as candidates (Supplemental Data Set 1). After filtering out some background proteins, we selected a peptide corresponding to SmD1 (Figure 1, A and B), which localizes to the nucleus according to the Bio-Analytic Resource for Plant Biology database (https://bar.utoronto.ca/eplant/) for further investigation.

Figure 1.

SmD1b functions upstream of SK11/12 to control carbon partitioning in seeds. A, Identification of SmD1 as a potential upstream regulator of SK11 through Cas9-ChAR. The identified SmD1 peptide is shown above the corresponding mass spectra. B, Sequence alignment of SmD1a and SmD1b. The framed sequences indicate the same peptide identified by LC–MS/MS. C–H, Examination of seed coat phenotypes of mature seeds in various genetic backgrounds. Scanning electron microscopy (SEM) of the seed coat (first column), seed coat mucilage stained with ruthenium red (second column), and seed pigmentation stained with DMACA (third column) are shown from left to right. In sem panels, the columellae are highlighted in red. Scale bars, 25 μm, 200 μm, and 1 mm (from left to right). I, Measurement of columella area in various genetic backgrounds. Over 50 columellae from five individual seeds were measured for each genotype. Samples with different lowercase letters are significantly different. P-values were determined by global one-way analysis of variance (ANOVA) with Tukey’s test (α = 0.05). J and K, Measurement of total fatty acid contents of mature seeds in smd1b-601, sk11 sk12, and smd1b-601 sk11 sk12 (J), and smd1b-601, SK11:SK11^E292K-GFP, and smd1b-601 SK11:SK11^E292K-GFP (K). Values are mean ± standard deviation (sd) of three biological replicates. Asterisks indicate significant differences between wild-type plants and the other genotypes (two-tailed paired Student’s t test, *P < 0.05, **P < 0.01).

SmD1b affects carbon partitioning in seeds

SmD1 is one of seven Smith (Sm) protein constituents that assemble into a ring-shaped heteroheptamer as a core ribonucleoprotein (RNP) component of the eukaryotic spliceosome (He and Parker, 2000). In Arabidopsis, SmD1 is encoded by two genes, SmD1a (At3g07590) and SmD1b (At4g02840), which produce almost identical proteins (Figure 1B). To understand the biological roles of SmD1a and SmD1b in Arabidopsis development, we isolated their respective T-DNA insertion mutants, smd1a-380 (named according to the seed stock GK-380A07-017232; CS436391) and smd1b-601 (named according to the seed stock SALKseq_047601) (Supplemental Figure S3A), in which SmD1a and SmD1b expression was dramatically reduced compared to wild-type plants (Supplemental Figure S3B). Both mutants exhibited an earlier flowering phenotype than wild-type plants (Supplemental Figure S3C), as previously reported for another smd1b mutant allele (suppressor of gene silencing14; Elvira-Matelot et al., 2016). To assess the extent of redundancy between the two genes, we crossed the two single mutants to generate the smd1a-380 smd1b-601 double mutant. Among their segregating F₂ progeny, smd1a-380/+ smd1b-601 and smd1a-380 smd1b-601/+ plants exhibited an extremely dwarf and slightly small stature, respectively, compared to wild-type plants (Supplemental Figure S3C), indicating that SmD1b plays a more dominant role than SmD1a in seedling development. This observation was also in agreement with a previous study showing that SmD1b is expressed at higher levels than SmD1a in various tissues and may play more important roles in Arabidopsis (Elvira-Matelot et al., 2016). Notably, we failed to identify smd1a-380 smd1b-601 homozygous double mutants (Supplemental Figure S3D), suggesting that both genes may be essential for either reproductive development or seed development.

To determine whether SmD1a and SmD1b affect carbon partitioning in seeds, we performed a detailed phenotypic analysis of seed reserves, including testa-deposited proanthocyanidins (PAs), mucilage, and embryo-stored fatty acids, in smd1a-380 and smd1b-601 seeds. PAs and mucilage can be visualized by p-dimethylaminocinnamaldehyde (DMACA) staining and ruthenium red staining, respectively. In addition, as a volcano-shaped structure called the columella at the seed coat surface enables compaction of the mucilage during seed dehydration (Francoz et al., 2015), we also examined the columellae associated with mucilage deposition. Interestingly, only smd1b-601 seeds phenocopied the abnormal seed characteristics of the sk11 sk12 double mutant compared to wild-type seeds, including larger columellae, more condensed mucilage, increased PA deposition, and lower fatty acid contents (Li et al., 2018) (Figure 1, C–E, I, and J), whereas smd1a-380 seeds exhibited similar characteristics as wild-type seeds (Supplemental Figure S4, A–C). Moreover, the expression of GL2, which is a major regulator of seed carbon partitioning and acts downstream of SK11/12 (Li et al., 2018), increased by 79% in smd1b-601 over wild-type levels, but only decreased by 10% in smd1a-380 (Supplemental Figure S4D). These observations suggest that SmD1b is a major regulator that affects carbon partitioning in seeds via SK11/12. Interestingly, like the sk11 sk12 double mutant, the loss of SmD1b function did not affect seed weight (Supplemental Figure S4E). Therefore, the decrease in seed oil for smd1b-601 measured in unit weight (Figure 1J) reflected a change in oil levels per seed, suggesting that the reduction of seed oil in smd1b-601 may result from altered carbon partitioning in seeds.

To confirm that the seed phenotypes of smd1b-601 are attributed to SmD1b activity, we introduced two SmD1b-3FLAG complementation constructs in the smd1b-601 background, gSmD1b-3FLAG and 35S:SmD1b-3FLAG, which contained three copies of the FLAG tag (3FLAG) translationally fused to either a 3.2-kb SmD1b genomic fragment or the 1.5-kb SmD1b coding sequence driven by two copies of the cauliflower mosaic virus 35S promoter, respectively (Supplemental Figure S4, F and G). Among the representative transgenic lines tested, SmD1b expression was restored to wild-type levels in developing seeds of smd1b-601 gSmD1b-3FLAG complementation lines, while SmD1b expression in smd1b-601 35S:SmD1b-3FLAG overexpression lines reached much higher than in wild-type plants (Supplemental Figure S4H). Consequently, fatty acid levels in smd1b-601 were fully rescued to wild-type levels in the complementation lines, or accumulated to higher levels in overexpression lines (Supplemental Figure S4I), indicating that the seed phenotypes of smd1b-601 are indeed causally linked to SmD1b activity.

SmD1b regulates seed carbon partitioning via SK11/12

We proceeded to investigate the genetic relationship between SmD1b and SK11/12 pertaining to carbon partitioning in seeds. The smd1b-601 sk11 sk12 triple mutant displayed similar seed phenotypes as smd1b-601 and sk11 sk12 (Figure 1, C–F, I, and J). This nonadditive genetic effect indicated that SmD1b, SK11, and SK12 may function in the same regulatory pathway. To test their epistatic relationship, we further crossed smd1b-601 with an established SK11 gain-of-function line (SK11:SK11^E292K-GFP), which accumulates the variant SK11^E292K with high protein stability and showed phenotypes opposite to those of sk11 sk12 (Li et al., 2018). SK11:SK11^E292K-GFP suppressed the seed phenotypes of smd1b-601 in smd1b-601 SK11:SK11^E292K-GFP seeds, which displayed smaller columellae, sparser mucilage extrusion, lower PA deposition, and higher fatty acid contents compared to wild-type seeds (Figure 1, E, G–I, and K). These results substantiated the idea that SmD1b acts upstream of SK11.

To test if the other splicing factors in the Sm complex play similar roles in seed carbon partitioning as SmD1, we obtained two T-DNA insertion lines, sme1-2 (SALK_089521C; Huertas et al., 2019) and smd3b-410 (named according to the seed stock SALK_006410C), where the respective expression of SmE1 and SmD3b was much lower than in wild-type plants (Supplemental Figure S5, A and B). Although smd1b-601, sme1-2, and smd3b-410 seeds accumulated less seed oil than wild-type seeds (Figure 1J;  Supplemental Figure S5C), their seed coat phenotypes varied to different extents (Figure 1, E and I; Supplemental Figure S5, D and E). While all three genotypes produced larger columellae than wild-type seeds, they showed different phenotypes in terms of mucilage and PA deposition. In contrast to smd1b-601, smd3b-410 showed less mucilage and PA deposition than wild-type seeds, whereas sme1-2 displayed similar mucilage and PA deposition to wild-type seeds (Supplemental Figure S5, D and E). These observations imply that SmD1b could control seed characteristics upstream of SK11/12 via the mechanism(s) that may differ from those of other components in the Sm-complex.

SmD1b shares overlapping accumulation patterns with SK11/12

In agreement with the fact that SmD1 was isolated from chromatin containing the 5′-region of SK11, immunolocalization of SmD1b-3FLAG in smd1b-601 gSmD1b-3FLAG lines revealed its dominant localization in the nucleus (Elvira-Matelot et al., 2016), although it was also detectable in the cytoplasm (Figure 2A). During seed development, the relative SmD1b, SK11, and SK12 transcript levels displayed a similar downward trend (Figure 2B), which paralleled the dynamic change observed for SmD1b-3FLAG protein levels in developing seeds (Figure 2C). We also constructed a β-glucuronidase (GUS) fusion reporter (gSmD1b-GUS), in which we translationally fused the same SmD1b genomic fragment used in gSmD1b-3FLAG to the GUS coding sequence (Supplemental Figure S4F). Representative GUS staining of gSmD1b-GUS also exhibited a decreasing trend during seed development with a strong signal in the integument (Figure 2D). We further monitored SmD1b protein abundance during seed development by immunostaining of smd1b-601 gSmD1b-3FLAG seeds using anti-FLAG antibody. Before fertilization, SmD1b accumulated in the entire ovule including the maternal integument and gametophytic cells (Figure 2, D and E), while SmD1b became highly abundant in the maternal integument of developing seeds after fertilization (Figure 2, D and E). The overall accumulation pattern of SmD1b protein was consistent with that of SK11/12 mRNA, which exhibited high expression levels in the integument and gradually decreased during seed development (Li et al., 2018). In agreement with SmD1b accumulation in the maternal integument, the seed oil contents of F₁ hybrids from reciprocal crosses between wild-type and smd1b-601 plants was mainly determined by the genotype of the female parent (Figure 2F), which was similar to the maternal effect observed in reciprocal crosses between wild-type and sk11 sk12 plants (Li et al., 2018). These results, together with the findings that SmD1b was isolated as a potential trans-acting factor of SK11 and acted genetically upstream of SK11/12, prompted us to investigate whether SmD1b directly regulates SK11/12 transcription during seed development.

Figure 2.

Expression pattern of SmD1b transcripts and encoded protein. A, Immunolocalization of SmD1b-3FLAG in smd1b-601 gSmD1b-3FLAG ovules using an anti-FLAG antibody. Merge, merge of differential interference contrast microscopy and fluorescence channels. Scale bar, 10 μm. B, Relative SK11, SK12, and SmD1b transcript levels in developing seeds of wild-type plants, as determined by RT-qPCR analysis. Results were normalized against the expression levels of U-BOX as an internal control. Values are means ± sd of three biological replicates. C, Examination of SmD1b-3FLAG protein levels in developing seeds of smd1b-601 gSmD1b-3FLAG. Siliques at different DAP were harvested for protein extraction and immunoblot analysis using anti-FLAG antibody (upper). The amount of rubisco large subunit (RbcL) in each sample is visualized by Ponceau S staining (lower). An equal amount of total protein based on the Bradford protein assay was loaded on each lane. D, Representative GUS staining pattern of the translational reporter construct gSmD1b-GUS. GUS signal is detectable in inflorescences (left), ovules (before fertilization) and seeds at an early developmental stage (2 and 4 DAP) (right). e, embryo; en, endosperm; ii, inner integument; oi, outer integument. Scale bars, 5 mm (left) and 50 μm (right). E, Immunolocalization of SmD1b-3FLAG in an ovule (before fertilization) and seeds at an early developmental stage (2 and 4 DAP) of smd1b-601 gSmD1b-3FLAG. en, endosperm; ii, inner integument; oi, outer integument. Scale bars, 50 μm. F, Measurement of total fatty acid contents of mature F₁ seeds from reciprocal crosses between wild-type and smd1b-601 plants. Crosses are indicated as “female × male,” as per convention. Values are means ± sd of three biological replicates. Samples with different lowercase letters are significantly different. P-values were determined by global one-way ANOVA with Tukey’s test (α = 0.05).

SmD1b regulates SK11/12 expression and binds to their genomic loci

We then compared SK11/12 expression in developing seeds in various genetic backgrounds, and observed that their expression is downregulated in smd1b-601, but restored to wild-type levels in the smd1b-601 gSmD1b-3FLAG complementation lines (Figure 3A). Furthermore, relative SK11/12 transcript levels increased in SmD1b overexpression lines (smd1b-601 35S:SmD1b-3FLAG) (Figure 3A). We subsequently introduced the established SK11pro:GUS and SK12pro:GUS reporter lines containing the 5′-upstream regions of SK11 and SK12 including their 5′-untranslated regions (UTRs) (Li et al., 2018) into smd1b-601, and detected decreased GUS signals for both reporters in the mutant background compared to wild-type plants (Figure 3B). These data strongly suggest that SmD1b promotes SK11/12 expression.

Figure 3.

SmD1b directly regulates SK11/12 transcription. A, Relative SK11 and SK12 transcript levels in siliques in various genetic backgrounds at 4 DAP, as determined by RT-qPCR. Results were normalized against the expression levels of U-BOX as an internal control. Values are means ± sd of three biological replicates. Asterisks indicate significant differences between wild-type plants and other genetic backgrounds (two-tailed paired Student’s t test, *P < 0.001). B, Representative GUS staining of SK11pro:GUS and SK12pro:GUS in developing seeds at 4 DAP in the wild-type (upper) and smd1b-601 backgrounds (lower). Scale bar, 200 μm. C and D, ChIP analysis of SmD1b-3FLAG binding to the SK11 (C) and SK12 (D) loci. Schematic diagrams of the SK11 and SK12 genomic regions are shown above the plots. Black boxes, exons; white boxes, UTRs; introns and other genomic regions, black lines. Twelve (1–12) and six (1–6) DNA fragments (1–12) spanning the SK11 and SK12 genomic regions, respectively, were designed for ChIP analysis, which was performed on smd1b-601 gSmD1b-3FLAG siliques harvested at 4 DAP. Chromatin from nuclear extracts served as the input, while the fraction immunoprecipitated by anti-FLAG antibody was used as the eluate. A TUB2 fragment was amplified as a negative control. Values are means ± sd of three biological replicates. Asterisks indicate significant changes in fold-enrichment compared to the TUB2 fragment (two-tailed paired Student’s t test, *P < 0.05). ND, not detectable.

Next, we carried out chromatin immunoprecipitation (ChIP) assays on smd1b-601 gSmD1b-3FLAG siliques at 4 DAP to examine whether SmD1b is associated with the SK11 and SK12 loci in vivo. We detected a clear enrichment of SmD1b in the first intron located within the 5′-UTR of SK11 and SK12 (Figure 3, C and D). This observation was consistent with our initial identification of SmD1 as a potential binding protein at the 5′-region of SK11 in the Cas9-ChAR assay (Supplemental Figure S1C), suggesting that SmD1b binds to the SK11/12 loci in vivo.

SmD1b promotes splicing of the first intron of SK11/12 and RNA Pol II recruitment

Interestingly, although the coding sequences of SK11 and SK12 are highly similar with ∼85% sequence identity, the sequence similarity of their 5′-UTR was relatively low (Supplemental Figure S6, A and B), suggesting that recruitment of SmD1b to the SK11 and SK12 loci may rely on other factors beyond simple sequence similarity. Notably, the first introns bound by SmD1b were the largest introns of SK11 and SK12 (Figure 4A), which were also uracil-rich (Supplemental Figure S6, C and D), supporting their specific recognition by Sm-proteins including SmD1b, as described in a previous study (Achsel et al., 2001). Indeed, we observed the occurrence of intron retention (IR) for the first intron of SK11 or SK12 only in smd1b-601, but not in wild-type plants (Figure 4B), suggesting that SmD1b binds to the first intron of SK11/12 and directly regulates the splicing of their transcripts.

Figure 4.

SmD1b co-transcriptionally affects the expression and splicing of the first intron of SK11/12. A, Schematic diagram of the SK11 and SK12 gene structures and the primers used for expression and splicing analysis. Black boxes, exons; white boxes, UTRs; introns and other genomic regions, black lines. Black arrows above the diagram indicate the primers for the IR test presented in (B). Blue and red lines below the diagrams indicate the amplicons of unspliced and spliced forms, respectively, which were measured to quantify the splicing efficiency of nascent RNA. B, The first introns of SK11 and SK12 transcripts exhibit IR in smd1b-601. Siliques at 4 DAP were used to isolate total RNA. A TUB2 fragment was amplified as a control. SP, spliced form. C, Splicing efficiency in nascent RNA isolated from siliques in wild-type and smd1b-601 plants at 4 DAP. The ratios between unspliced and spliced forms at different introns of SK11 and SK12 were calculated, and presented as means ± sd of three biological replicates. Asterisks indicate significant differences between wild-type and smd1b-601 plants (two-tailed paired Student’s t test, *P < 0.05). D, Relative SK11/12 transcript levels in nascent RNA isolated from siliques in wild-type and smd1b-601 plants at 4 DAP, as determined by RT-qPCR. Results were normalized against the expression levels of U-BOX, with levels in wild-type plants set to 1.0. Values are means ± sd of three biological replicates. Asterisks indicate significant differences between wild-type and smd1b-601 plants (two-tailed paired Student’s t test, *P < 0.01). E, Co-IP assay showing the in vivo interaction between SmD1b and RNA Pol II. Nuclear protein extracts from smd1b-601 gSmD1b-3FLAG siliques at 4 DAP were immunoprecipitated with an antibody against the CTD of RNA Pol II (Pol II-CTD) or nonspecific IgG. The input and co-immunoprecipitated proteins were examined by immunoblot using anti-FLAG antibody (upper) or anti-Pol II-CTD antibody (lower). F, ChIP analysis of the enrichment of Pol II over the SK11 and SK12 genomic regions in wild-type and smd1b-601 plants. Chromatin was isolated from siliques at 4 DAP, and immunoprecipitated with anti-Pol II CTD antibody. Values are means ± sd of three biological replicates. The ChIP-PCR fragments were designed as shown in Figure 3, C and D. Asterisks indicate significant differences between wild-type and smd1b-601 plants (two-tailed Student’s t test, *P < 0.05).

Since SmD1b affects SK11/12 splicing (Figure 4B) and mRNA levels (Figure 3A), we further compared splicing efficiency and expression levels of SK11/12 in the pool of chromatin-bound nascent RNA extracted from wild-type and smd1b-601 siliques. In wild-type nascent RNA, the unspliced-to-spliced ratios of 3′-proximal introns including the 11th intron of SK11 and the 10th intron of SK12 were much higher than those of the first introns of SK11 and SK12 (Figure 4, A and C). This observation was consistent with a previous report (Wu et al., 2016) and the expectation of sequential splicing along the nascent transcripts following the forward orientation (5′–3′), indicating successful isolation of nascent RNA. It is worth noting here that in nascent RNA, the unspliced-to-spliced ratios of the first introns of SK11/12 were higher in smd1b-601 compared to wild-type plants, whereas the unspliced-to-spliced ratios of 3′-proximal introns (the 11th intron of SK11 and the 10th intron of SK12) remained unchanged in both genotypes (Figure 4C). These results, together with the IR occurrence for the first intron of SK11 or SK12 in smd1b-601 (Figure 4B), substantiated the notion that SmD1b specifically mediates splicing of the first intron of SK11/12. Since SK11/12 transcript levels were lower as both mature RNA (Figure 3, A and B) and as nascent RNA in smd1b-601 compared to wild-type plants (Figure 4D), SmD1b may regulate SK11/12 splicing and transcript levels co-transcriptionally rather than posttranscriptionally.

Furthermore, we detected the interaction between SmD1b and RNA Pol II in vivo, as evidenced by co-immunoprecipitation (Co-IP) of SmD1b in nuclear protein extracts from smd1b-601 gSmD1b-3FLAG siliques using an antibody recognizing the C-terminal domain (CTD) of RNA Pol II (anti-Pol II CTD) (Figure 4E), suggesting that the endogenous SmD1b function may be coupled to RNA Pol II-controlled transcription. ChIP assays of chromatin extracted from wild-type and smd1b-601 plants using the anti-Pol II CTD antibody further revealed that RNA Pol II recruitment at the entire SK11/12 loci is much lower in smd1b-601 compared to wild-type plants (Figure 4F). These findings, together with the altered levels of SK11/12 nascent RNA in smd1b-601, indicate that SmD1b promotes SK11/12 transcription at least partially by enhancing Pol II recruitment at their loci.

SmD1b affects R-loop formation in the first introns of SK11/12

To further understand the link between SmD1b-mediated splicing of SK11/12 and SK11/12 transcript levels in seeds, we performed a detailed analysis of SK11/12 introns. These introns were thymidylate (T)-rich, thus generating uridylate (U)-rich nascent transcripts and introducing an obvious AT-skew around the first introns of the SK11/12 loci (Supplemental Figure S6, C–E). As regions with an AT-skew were reported to form R-loops in Arabidopsis (Xu et al., 2017), we performed an in vitro transcribed (IVT) R-loop gel-shift assay to test whether the first introns of SK11/12 can form such R-loops. Accordingly, we cloned the first SK11/12 introns into the pGEM-T vector to drive their in vitro transcription by SP6 RNA polymerase. The formation of an R-loop upon in vitro transcription turns the supercoiled state of the circular plasmid harboring the putative R-loop structure into a relaxed state that moves slower than the supercoiled plasmid during electrophoresis (Figure 5A), which can be detected by an RNase H-sensitive gel-shift assay in which RNase H cleaves the RNA strand of an RNA/DNA hybrid and eliminates R-loops (Powell et al., 2013). Through this assay, we established that the transcripts bearing the first introns of SK11/12 indeed form R-loops in vitro, as illustrated by the RNAse H digestion-sensitive band-shift (Figure 5B).

Figure 5.

SmD1b affects R-loop accumulation in the first intron of SK11/12. A, Schematic diagram of IVT R-loop gel-shift assay. R-loops created by the nascent transcript (red line) and its template can relax the supercoiled plasmid and reduce its electrophoretic mobility, whereas RNase H degrades the RNA in the DNA/RNA hybrid and converts the relaxed state of the plasmid to a supercoiled state. B, IVT R-loop gel-shift assays showing that the first intron of SK11 (left) and SK12 (right) forms R-loops during in vitro transcription. C and D, DRIP analysis at SK11 (C) and SK12 (D) genomic regions using siliques harvested at 4 DAP. Immunoprecipitation with the S9.6 antibody was used to detect R-loops. Values are means ± sd of three biological replicates. Asterisks indicate significant differences between wild-type and smd1b-601 plants (two-tailed Student’s t test, *P < 0.05, **P < 0.01, ***P < 0.001) The ChIP-PCR fragments were designed as shown in Figure 3, C and D. “ND” indicates barely detectable amplicons at the given sites. E, DRIP analysis of the reporter locus carrying either the normal or mutated first intron in the 5′-UTR of SK11. The 35S:INTRON:SK11^E292K-GFP (labeled as “INTRON”) or 35S:intron:SK11^E292K-GFP (labeled as “intron”) harbors the normal or mutated intron in the 5′-UTR of SK11, respectively (Supplemental Figure S7). The ChIP fragment (i) used in the DRIP assay specifically detects the intron region of the reporters. Chromatin isolated from siliques harvested at 4 DAP was analyzed in the DRIP assay using the S9.6 antibody. Values are means ± sd of three biological replicates. Asterisks indicate a significant difference between 35S:intron:SK11^E292K-GFP and 35S:INTRON:SK11^E292K-GFP plants (two-tailed Student’s t test, ***P < 0.001).

We then assessed the effect of SmD1b on R-loop levels in vivo at the SK11/12 loci by DNA:RNA hybrid immunoprecipitation (DRIP) using the S9.6 antibody, which specifically recognizes DNA:RNA hybrids. An RNase H-resistant DRIP signal shown as a high RNase H(–)/RNase H (+) ratio indicates the existence of R-loop(s) in vivo. Indeed, we detected higher levels of R-loops in the smd1b-601 background compared to wild-type plants, especially over the first intron of SK11 (ChIP fragment 8) and SK12 (ChIP fragment 4) (Figure 5, C and D), suggesting that SmD1b-mediated splicing of SK11/12 may compromise R-loop formation in their first introns. To test for a causal link between R-loop formation and splicing of the first intron of SK11, we created two types of transgenic reporter lines, 35S:INTRON:SK11^E292K-GFP (designated INTRON thereafter) and 35S:intron:SK11^E292K-GFP (intron), which harbored the constructs with the normal first intron and the variant with a mutated acceptor site (3′-splicing site), respectively (Supplemental Figure S7A). This mutation of the 3′-splicing site completely abolished intron splicing from the 35S:intron:SK11^E292K-GFP transgene, as expected (Supplemental Figure S7B). We further performed DRIP assays to specifically assess the intron regions of these two reporter loci, and determined much higher R-loop levels in 35S:intron:SK11^E292K-GFP than in 35S:INTRON:SK11^E292K-GFP (Figure 5E;  Supplemental Figure S7A), suggesting that the unspliced first intron of SK11 indeed results in higher R-loop levels in vivo.

Loss of SmD1b function inhibits RNA Pol II elongation

Despite being driven by the same constitutive 35S promoter, SK11 expression levels were consistently lower in independent 35S:intron:SK11^E292K-GFP lines than in 35S:INTRON:SK11^E292K-GFP lines (Figure 6A). Moreover, when we crossed a representative 35S:INTRON:SK11^E292K-GFP line (#1) into the smd1b-601 background, SK11 transcript levels were lower in smd1b-601 than in wild-type siliques (Figure 6B). Thus, although the 35S promoter recruited RNA Pol II similarly at the initial stage of SK11 transcription, the smd1b-601 mutation or retention of the first intron still resulted in lower SK11 expression, indicating that transcriptional elongation of SK11 might be affected in smd1b-601 or by IR.

Figure 6.

SmD1b-mediated splicing regulates Pol II elongation. A, Relative SK11 transcript levels in siliques of 35S:INTRON:SK11^E292K-GFP (INTRON) and 35S:intron:SK11^E292K-GFP (intron) reporter lines, as determined by RT-qPCR. Results were normalized against the expression levels of U-BOX, with the levels in wild-type plants set to 1.0. n, the number of independent transgenic lines examined for each genotype. Asterisk indicates a significant difference between “intron” and “INTRON” lines (two-tailed paired Student’s t test, *P < 0.001) B, Relative SK11 transcript levels in siliques of INTRON and smd1b-601 INTRON lines, as determined by RT-qPCR. A representative INTRON line (#1) was crossed to smd1b-601. SK11 expression levels in INTRON and its derived smd1b-601 INTRON lines were normalized against the expression levels of U-BOX, with the levels in wild-type plants set to 1.0. Values are means ± sd of three biological replicates. Asterisk indicates a significant difference between INTRON and smd1b-601 INTRON (two-tailed paired Student’s t test, *P < 0.001). C, Co-IP assay showing the in vivo interaction between SmD1b and RNA Pol II in different phosphorylation statuses. Nuclear protein extracts from smd1b-601 gSmD1b-3FLAG siliques at 4 DAP were immunoprecipitated with an antibody against the CTD of RNA Pol II (anti-Pol II CTD), anti-S2-phosphorylated CTD antibody (anti-Pol II CTD [S2P]), anti-S5-phosphorylated CTD antibody (anti-Pol II CTD [S5P]), or nonspecific IgG. The input and co-immunoprecipitated proteins were examined by immunoblot using anti-Pol II-CTD antibody (upper) or anti-FLAG antibody (lower panel). D, Schematic diagrams of structures of the reporter constructs and the primers designed for ChIP analysis. For the “INTRON” reporter, the first intron in the 5′-UTR of SK11 is indicated as a black fold line. For the “intron” reporter, the 3′-splicing site-mutated intron replaces the wild-type intron, which is indicated as a black straight line. Blue lines below the diagram indicate the fragments designed for ChIP analyses shown in (E–G). These fragments spanning the intron region (i), middle region (m), and 3′-end region (e) are amplified only from the reporter constructs, but not from the native SK11 locus. E, Measurement of Pol II pausing index in different genetic backgrounds. The index indicates the relative Pol II enrichment at the 5′-proximal region to that in the gene body region, shown as the relative ChIP signal at fragment (i) versus fragment (m) or fragment (e). F and G, ChIP analysis of the enrichment of S2-phosphorylated Pol II (F) and S5-phosphorylated Pol II (G) in different genetic backgrounds. Values are means ± sd of three biological replicates. In (E–G), chromatin was isolated from the siliques harvested at 4 DAP, and immunoprecipitated by anti-Pol II CTD (E), anti-Pol II CTD (S2P) (F), or anti-Pol II CTD (S5P) antibodies (G). Asterisks indicate significant differences between intron or smd1b-601 INTRON lines and INTRON lines (two-tailed Student’s t test, *P < 0.05, **P < 0.01, ***P < 0.001).

RNA Pol II activity in vivo is correlated with the phosphorylation status of its CTD. For instance, while phosphorylation of both serine 2 and serine 5 residues within the CTD (S2P/S5P-CTD) is associated with productive Pol II elongation, S5P-CTD and S2P-CTD are closely related to Pol II pausing-relieving and transcriptional termination, respectively (Phatnani and Greenleaf, 2006). In particular, the release of Pol II pausing is crucial for Pol II elongation, thus being a major regulatory step of transcription (Adelman and Lis, 2012). Our Co-IP assays revealed that SmD1b is associated with both nonphosphorylated and phosphorylated Pol II, but with a relatively strong interaction with Pol II bearing S5P-CTD (Figure 6C), indicating that SmD1b may be functionally related to productively elongating Pol II in addition to its role in Pol II recruitment.

To further examine the effects of SmD1b and IR on RNA Pol II elongation at SK11, we measured the Pol II pausing index, defined as the relative ratio of Pol II occupation in the 5′-proximal region (pausing status) and over the gene body region (elongating status; Zhang et al., 2017; Zhu et al., 2018) using three pairs of primers that specifically amplify the 5′-intron region [fragment (i)], the middle part [fragment (m)], and the end part [fragment (e)] of the reporter locus (Figure 6D). Based on the ratio of Pol II enrichment at fragment (i) relative to fragments (m) or (e), the Pol II pausing indexes of 35S:INTRON:SK11^E292K-GFP were lower than those of 35S:intron:SK11^E292K-GFP and smd1b-601 35S:INTRON:SK11^E292K-GFP (Figure 6, D and E). Furthermore, examination of S2P/S5P-CTD enrichment in these reporter lines revealed that the active Pol II with S2P-CTD or S5P-CTD is less associated with the reporter locus in 35S:intron:SK11^E292K-GFP or smd1b-601 35S:INTRON:SK11^E292K-GFP than in 35S:INTRON:SK11^E292K-GFP lines (Figure 6, D, F, and G). These observations all suggest that Pol II elongation is compromised when IR occurs in 35S:intron:SK11^E292K-GFP or in the absence of SmD1b.

Discussion

Plant GSK3s, including SK11/12, are highly conserved serine/threonine protein kinases that serve as key regulatory hubs for integrating various signaling pathways in plant development and responses to biotic and abiotic stresses (Li et al., 2021). Although the functional modules involving GSK3s are well elucidated at the posttranslational level, such as protein stability (Zhu et al., 2017), localization (Houbaert et al., 2018; Li et al., 2020b), and modifications (Kim et al., 2009; Hao et al., 2016; Li et al., 2020c), how transcriptional regulation of GSK3s is modulated to confer their functional specificity in plants is largely elusive. In this study, we revealed that the splicing factor SmD1b plays an essential role in mediating transcriptional regulation of SK11/12 by facilitating the recruitment of RNA Pol II and regulating the splicing of the first intron in the 5′-UTR of SK11/12 to prevent the formation of disruptive R-loops and promote Pol II productive elongation at SK11/12 chromatin regions. These promotive effects conferred by SmD1b upregulate SK11/12 transcript levels, which in turn constitute a master regulatory module for determining a key switch of carbon flow among metabolic pathways in zygotic and maternal tissues in Arabidopsis seeds (Figure 7).

Figure 7.

A model depicting the regulation of carbon partitioning in seeds by splicing-mediated activation of SK11/12 transcription. In wild-type seeds, the splicing factor SmD1b binds to the SK11/12 chromatin regions over the first intron in their 5′-UTR, which activates their transcription by promoting RNA Pol II recruitment and productive transcriptional elongation associated with phosphorylation at serines 2 and 5 (S2P and S5P) of the Pol II CTD. SK11/12 transcription promotes the biosynthesis of fatty acids in the seed storage component (embryo). In smd1b-601 seeds, loss of SmD1b function results in the retention of the first intron in SK11/12 transcripts, which generates disruptive R-loops that hamper transcriptional elongation of SK11/12, and also compromises the recruitment of RNA Pol II to the SK11/12 loci. These effects downregulate SK11/12 transcription and promote the formation of seed mucilage and flavonoid pigments in the seed protective component (seed coat). Therefore, SmD1b-mediated splicing of SK11/12 plays a key role in transcriptional regulation of SK11/12, which underpins the regulatory module for distributing carbon sources into storage and protective components in Arabidopsis seeds. The figure was created with BioRender (BioRender.com).

The findings herein provide several pieces of evidence suggesting that SmD1b activates the transcription of SK11/12 via promoting both RNA Pol II recruitment and its productive elongation at the SK11/12 loci (Figure 7). First, SmD1b promotes the expression of SK11/12 nascent transcripts, and is physically associated with the first intron of the SK11/12 loci and RNA Pol II, whereas loss of SmD1b compromises the recruitment of RNA Pol II with the entire SK11/12 loci, including the transcriptional start sites, suggesting that SmD1b directly facilitates RNA Pol II recruitment to the SK11/12 loci. Second, SmD1b specifically mediates splicing of the first intron in the 5′-UTR of SK11/12. Loss of SmD1b function results in retention of the first intron in the 5′-UTR of SK11/12, which creates disruptive R-loops and inhibits RNA Pol II elongation, as evidenced by the higher Pol II pausing index and reduced association of active RNA Pol II with S2P-CTD and S5P-CTD at the SK11/12 loci. Moreover, as R-loop formation and compromised recruitment of RNA Pol II also occur concurrently at the SK11/12 loci in smd1b-601, it is possible that R-loops exert another effect on disrupting Pol II recruitment at these two loci, particularly given that these two events happen in close physical proximity in the 5′-UTR regions.

Previous studies have revealed that 5′-intronic regions, including T-rich introns, play important roles in activating gene expression mostly independently of splicing in plants (Rose and Beliakoff, 2000; Clancy and Hannah, 2002), suggesting that 5′-intronic regions may accommodate important trans-acting factors to enhance transcription (Rose et al., 2008). In yeast (Saccharomyces cerevisiae), 5′-intronic sequences enhance gene expression by counteracting R-loop formation, in which intron-mediated mRNP assembly on the nascent transcripts rather than the splicing event per se is critical for preventing R-loop formation (Bonnet et al., 2017). Our results showed that SmD1b is recruited specifically to the T-rich intronic region in the 5′-UTR of SK11/12 to activate SK11/12 mRNA. Despite the similar importance of 5′-intronic regions in promoting gene expression revealed in this study, our findings demonstrate two unique characteristics of SmD1b as an intron-associated RNP component. First, the effect of SmD1b on preventing the formation of disruptive R-loops at SK11/12 should be dependent on its splicing activity, as IR by blocked splicing in 35S:intron:SK11^E292K-GFP is sufficient to accumulate R-loops even in the presence of SmD1b (Figure 5E). This mechanism is different from the negligible role of splicing in R-loop formation proposed in yeast (Bonnet et al., 2017), suggesting that SmD1b function in splicing is fundamental to R-loop formation at the SK11/12 loci. Second, in addition to affecting RNA Pol II elongation, SmD1b directly or indirectly affects RNA Pol II recruitment through either its physical interaction with Pol II or R-loop formation. The new features revealed in this study suggest that the mechanisms underlying the effect of 5′-intronic regions on gene expression are highly attributed to the associated trans-acting factors in a context-specific manner.

Investigations of R-loops in yeast and human cells have proposed a genome-wide interdependence between inhibition of Pol II elongation and accumulation of R-loops (El Hage et al., 2010; Chen et al., 2017; Zhang et al., 2017; Shivji et al., 2018; Edwards et al., 2020). In plants, the existence of R-loops and their interrelations with other cellular processes, such as transcriptional regulation and genome instability, have been described on a genome-wide scale in Arabidopsis (Xu et al., 2017, 2020), rice (Oryza sativa; Fang et al., 2019), and maize (Zea mays; Liu et al., 2021). Meanwhile, ongoing efforts to clarify the regulatory pathways involving R-loops at specific gene loci are providing a better understanding of R-loop roles in plant development and response to stresses. For example, in Arabidopsis, the long noncoding RNA (lncRNA) APOLO acts in trans by generating R-loops to target other genes (Ariel et al., 2020; Moison et al., 2021). R-loops at FLOWERING LOCUS C (FLC) also modulate FLC expression and flowering time by directly repressing the transcription of its natural antisense transcript COOLAIR (Sun et al., 2013), while the intragenic tRNA-promoted R-loops inhibit the transcription of the overlapping host gene NUDIX HYDROLASE1, thus affecting oxidative stress responses (Liu and Sun, 2021). In both scenarios, local R-loops affect the interplay between two overlapping transcripts. In contrast, this study showed an inhibitory effect of IR-derived R-loops on the expression of their source transcripts SK11/12, demonstrating a different regulatory module for R-loops acting in cis to affect their target gene expression.

In summary, the molecular framework of SmD1b-mediated transcriptional regulation of SK11/12 in seed carbon partitioning provides an emerging paradigm shift in understanding both the transcriptional regulation of GSK3s and splicing-modulated R-loop formation as well as their effects on plant development. Further characterization of SmD1b interacting partners, including lncRNAs such as ALTERNATIVE SPLICING COMPETITOR (Rigo et al., 2020), will provide more insights into the role of SmD1b in transcriptome reprogramming at specific loci in different developmental contexts.

Materials and methods

Plant materials and growth conditions

Arabidopsis thaliana plants were grown under long-day conditions (16-h light/8-h dark; illuminated by white light-emitting diodes with a light intensity of 135–150 μmol m⁻² s⁻¹) at 22°C. The smd1b-601 (SALKseq_047601), smd1a-380 (CS436391), sme1-2 (SALK_089521C), and smd3b-410 (SALK_006410C) mutants were obtained from the Arabidopsis Biological Resource Centre. sk11 sk12, SK11pro:GUS, SK12pro:GUS, and SK11:SK11^E292K-GFP were described previously (Li et al., 2018).

Plasmid construction for plant transformation

To construct gSmD1b-3FLAG, a 3.2-kb genomic fragment of SmD1b (–1,718 bp to +1,525 bp, taking the A of the translation start codon as +1 bp) was cloned into the pC1305 vector to obtain an in-frame fusion of the SmD1b sequence with that of three copies of the FLAG tag (3FLAG). To construct SmD1b overexpression lines, the coding sequence of SmD1b was cloned into pC1306, where SmD1b was driven by a 2 × 35S promoter and cloned in-frame upstream of the sequence for a 3FLAG tag. These two constructs were both transformed into smd1b-601. To construct gSmD1b-GUS, the same 3.2-kb genomic fragment of SmD1b was cloned into pGWB3 to obtain an in-frame fusion between SmD1b and GUS. To generate 35S:INTRON:SK11^E292K-GFP and 35S:intron:SK11^E292K-GFP constructs, the sequence encoding the SK11^E292K fragment was amplified from the SK11:SK11^E292K-GFP construct (Li et al., 2018) and cloned into pC1302E, where the wild-type or mutated form of the intron sequence was inserted between the 2 × 35S promoter and SK11^E292K coding sequence. These two constructs were both transformed into wild-type plants. All transgenic plants were generated through Agrobacterium (Agrobacterium tumefaciens; strain GV3101)-mediated transformation. The primers used are listed in Supplemental Data Set 2.

sgRNA cloning and probe preparation

The T7-sgRNA cassette containing the sgRNA backbone (Feng et al., 2013) was cloned into pUC19 to generate pUCT7SG (Supplemental Figure S1B). The putative sgRNA targeting regions of interest were screened by CRISPRdirect online (Naito et al., 2014) based on target specificity, PAM (NGG), and GC ratio (30%–75%) (Supplemental Figure S1C). To generate the pUCT7SG-sgRNAs, two types of primers, 5′-TACGACTCACTATA-(20-nt sgRNA targeting sequence)-GTTTTAGAGCTAGAAATAG-3′ and 5′-CTATTTCTAGCTCTAAAAC-(reverse complementary 20 nt sgRNA targeting sequence)-TATAGTGAGTCGTA-3′ (Supplemental Data Set 2), were annealed and inserted into the polymerase chain reaction (PCR)-linearized pUCT7SG by Gibson cloning using NEBuilder HiFi DNA Assembly Master Mix (NEB, Ipswich, MA, USA). The Gibson cloning products were digested by PmeI before being transformed into Escherichia coli. Clones from independent colonies were then sequenced to identify the correct constructs. To prepare the sgRNAs in vitro, the pUCT7SG-sgRNA plasmids targeting the 5′-region of SK11 were equally mixed, after which T7-sgRNA cassettes were PCR amplified with M13F/M13R and then digested by HindIII to produce the 3′-ends. The purified linearized fragments were used as templates for in vitro transcription by T7 RNA polymerase. sgRNAs^ALL-Bio were generated by Biotin RNA Labeling Mix (Roche, Basel, Switzerland), while sgRNAs^Bio were generated by T7 RiboMAX Express Large-Scale RNA Production System (Promega, Madison, WI, USA) followed by 3′-end biotin labeling using Pierce RNA 3′ End Biotinylation Kit (Thermo Fisher Scientific, Waltham, MA, USA). The labeled sgRNAs were purified by chloroform:isoamyl alcohol (24:1) followed by ethanol precipitation (Supplemental Figure S2A). The labeling efficiency was then monitored by dot blot analysis (Supplemental Figure S2, B and C). The negative control (sgRNAs targeting the 3′-region of SK11) was prepared following the same procedure (Supplemental Figure S1C).

Dot blot analysis

A serial dilution of sgRNA probes and positive control were dropped onto the nylon membrane, which was UV-cross-linked twice using the Auto-crosslink mode in a Stratalinker 2400 UV Crosslinker. The membrane was then blocked in blocking buffer (1× phosphate-buffered saline [PBS] pH 7.4, 1% [w/v] sodium dodecyl sulfate [SDS]) for 30 min, and incubated with StrepTactin-HRP (1:5,000; Bio-Rad, Hercules, CA, USA; Cat: #1610381) diluted in blocking buffer for 30 min. After washing the membrane once in 0.1× blocking buffer and twice in 1× PBST buffer (1× PBS pH 7.4 with 0.05% [v/v] Tween-20), ECL substrate was added for chemiluminescent detection.

Cas9-ChAR

Siliques (about 5 g) at 4 DAP were harvested, frozen in liquid nitrogen, and ground to fine powder before being resuspended in 5 mL of crosslinking buffer (10-mM HEPES pH 7.5, 400-mM sucrose, 5-mM MgCl₂, 1-mM ethylenediaminetetraacetic acid [EDTA] pH 8.0, 1-mM phenylmethanesulfonyl fluoride [PMSF], 1× protease inhibitor cocktail, 1% [w/v] formaldehyde). The slurry was kept on ice and subjected to vacuum infiltration for 30 min. After adding glycine to a final concentration of 125 mM, the slurry was vacuum-infiltrated for an additional 5 min to stop the crosslinking reaction. The slurry was then diluted four-fold using nuclear fractionation buffer (15-mM PIPES pH 6.8, 250-mM sucrose, 5-mM MgCl₂, 60-mM KCl, 125-mM NaCl, 1-mM CaCl₂, 0.9% [v/v] Triton X-100, 1-mM PMSF, 1× protease inhibitor cocktail) before being passed through a 100-μm filter. The filtrate was centrifuged at 845 rcf for 10 min at 4°C to pellet the nuclei. The pellet was washed several times using nuclear fractionation buffer until the pellet was no longer green. The isolated nuclear pellet was then resuspended in 2-mL nuclear lysis buffer (50-mM HEPES pH 7.5, 150-mM NaCl, 1-mM EDTA, 1% [w/v] SDS, 1-mM PMSF) and sonicated to produce DNA fragments between 200 and 500 bp. After sonication, the sample was centrifuged at 13,200 rcf for 5 min at 4°C and the soluble chromatin fraction was harvested. sgRNA^ALL-Bio, RNase inhibitor (SUPERase In RNase Inhibitor; Thermo Fisher Scientific), and streptavidin beads (Dynabeads MyOne Streptavidin C1; Invitrogen, Waltham, M, USA) were added to the chromatin fraction and incubated at 4°C for 2 h. After beads were removed, the cleared chromatin was mixed with sgRNA^Bio, RNase inhibitor, dCas9 (Cas9 Null Mutant Protein; Applied Biological Materials), streptavidin beads, and 0.1 volume of 10×dCas9 buffer (200-mM HEPES pH 6.5, 50-mM MgCl₂, 1-M NaCl, 1-mM EDTA), and incubated overnight at 4°C. The beads were subsequently washed 6–8 times using washing buffer (50-mM Tris–HCl pH 7.4, 150-mM NaCl, 1-mM EDTA pH 8.0, 5% [v/v] glycerol, 0.5% [v/v] Triton X-100, 1-mM PMSF) and incubated in loading buffer at 95°C for 10 min to release all bound proteins in the eluate, which was subjected to LC–MS/MS analysis (Supplemental Figure S1A). Peptides with high confidence (>95%; unused score >1.33) and undetectable in the negative control group were considered as candidates (Supplemental Figure S1C).

Immunostaining

The immunostaining assay was extensively modified according to a previous protocol (Escobar-Guzman et al., 2015). The pistils (before fertilization) or developing siliques were dissected and fixed in fixative buffer (1× PBS buffer with 4% [w/v] paraformaldehyde, 10% [v/v] DMSO, and 0.1% [v/v] Triton X-100) for 1 h under vacuum. The fixation process was then stopped by washing with 1× TBS buffer twice. The samples were transferred to a clean slide and embedded in a 200-μL slice of 10% (w/v) polyacrylamide. The solidified samples on slides were digested with an enzyme solution (4% [w/v] cellulase RS [Yakult, Cat: C8003], 1% [w/v] driselase [Sigma, St Louis, MO, USA; Cat: D9515], 2% [w/v] pectinase [Yakult, Cat: M8002] in 0.45-M mannitol) and incubated at 37°C for 60 min. After washing the slides with 1× PBS for 5 min, the samples on slides were blocked in blocking solution (1× PBS with 1% [w/v] BSA and 0.1% [v/v] Triton X-100) for 1 h at room temperature. The samples on slides were subsequently incubated with anti-FLAG antibody (Sigma, Cat: F3165; 1:100 dilution) in blocking solution overnight at 4°C. After washing the slides with blocking solution at 4°C with gentle shaking for 5 h, the samples on slides were incubated with goat anti-mouse IgG (H + L) coupled with Alexa Fluor 488 (Invitrogen, Cat: A-11001; 1:200 dilution) in blocking solution overnight at 4°C. The slides were then washed twice with 1× PBS plus 0.1% (v/v) Triton X-100 at 4°C and mounted for observation by confocal microscopy.

Ruthenium red staining and DMACA staining

Visualization of seed reserves deposited at the seed coat was performed by ruthenium red staining and DMACA staining as previously described (Li et al., 2018) with modifications. For ruthenium red staining, dry seeds were incubated in water at room temperature for 1 h. The rehydrated seeds were submerged in 0.01% (w/v) ruthenium red solution at room temperature for 1 h, and then the stained seeds were mounted in water for observation under a microscope (Axio Imager 2, Carl Zeiss). For DMACA staining, dry seeds were stained in the DMACA reagent (2% [w/v] DMACA in 3-M HCl, 50% [w/v] methanol) at room temperature under dark conditions overnight, and subsequently washed 3 times with 70% (v/v) ethanol. Stained seeds were observed using a Nikon SMZ1500 stereomicroscope (Nikon).

Measurement of fatty acids

Fatty acids of mature seeds were quantified by GC following the reported procedure with minor modifications (Li et al., 2018). Two micrograms of intact seeds were used for acid-catalyzed transmethylation in 500-μL reaction solution (5% [v/v] H₂SO₄ in methanol with 10-μg heptadecanoic acid) at 90°C for 3 h. The fatty acid methyl esters were extracted with hexane and analyzed on a GC-2010 GC–MS system (Shimadzu, Kyoto, Japan) with an HP-88 fused silica column (Agilent, St Clara, CA, USA).

IVT R-loop gel shift assay

The IVT R-loop gel shift assays were modified from a previous study (Powell et al., 2013). The first introns of SK11 and SK12 were cloned into pGEM-T (Promega) after the SP6 promoter site and transcribed using SP6 RNA Polymerase (Roche). The products were purified by chloroform extraction and ethanol precipitation. Purified products were divided equally into either mock-treated samples or samples treated with RNase H (NEB). The products were purified again by chloroform extraction and ethanol precipitation before being separated by agarose gel electrophoresis.

ChIP assay

ChIP assay was performed following the reported protocol with minor modifications (Kaufmann et al., 2010; Li et al., 2018). The extracted chromatin was sonicated to produce DNA fragments of 200–500 bp and incubated with anti-FLAG beads (Sigma) or Protein G magnetic beads (Thermo Fisher Scientific) supplied with anti-CTD (ab26721; Abcam, Cambridge, UK), anti-CTD (S2P) (ab5095; Abcam), or anti-CTD (S5P) (ab5131; Abcam) antibodies for 2 h at 4°C. A genomic fragment of TUBULIN2 (TUB2) was amplified as a control. Fold-enrichment was calculated first by normalizing the amount of a target DNA fragment in the immunoprecipitated fraction against that in the input and then by normalizing the value of this fragment against that for TUB2 as a negative control. ChIP assays were repeated with three biological replicates. The primers used are listed in Supplemental Data Set 2.

DRIP assay

The DRIP assays were modified from the previous study (Li et al., 2020a). Tissues of interest were ground to fine powder in liquid nitrogen, and resuspended in genomic DNA extraction buffer (10-mM HEPES pH 7.5, 400-mM sucrose, 25-mM EDTA, 1-mM MgCl₂, 0.5% [w/v] SDS, 1-mM PMSF, RNase inhibitor [SUPERase·In; Invitrogen]). DNA was recovered by phenol/chloroform extraction and then fragmented into pieces around 500 bp in size by sonication on ice. After DNA concentrations of different samples were measured by Nanodrop (Thermo Scientific), an equal amount of DNA for each sample was subjected to immunoprecipitation with the anti-DNA–RNA hybrid S9.6 antibody (Kerafast). RNase H (NEB) was added for negative controls during immunoprecipitation. The immunoprecipitation and subsequent qPCR procedures were followed as described in the ChIP assay.

Nascent RNA isolation

Nascent RNA was isolated as previously described with minor modifications (Wu et al., 2016). Siliques harvested at 4 DAP were crosslinked as described in the Cas9-ChAR. The nuclei were isolated in Honda buffer (20-mM HEPES pH 7.4, 440-mM sucrose, 1.25% [w/v] Ficoll, 2.5% [w/v] Dextran T40, 10-mM MgCl₂, 5-mM DTT, 0.5% [v/v] Triton X-100, 1× protease inhibitor cocktail, 20-U/mL RNase inhibitor [SUPERase·In; Invitrogen], 50-ng/μL tRNA), and the nuclear pellet was resuspended in resuspension buffer (25-mM Tris–HCl pH 7.5, 100-mM NaCl, 0.5-mM EDTA, 1-mM DTT, 50% [v/v] glycerol, 20-U/mL RNase inhibitor). The slurry was then washed 3 times in urea wash buffer (25-mM Tris–HCl pH 7.5, 300-mM NaCl, 0.5-mM EDTA, 1-mM DTT, 1-M urea, 1% [v/v] Tween-20, 20-U/mL RNase inhibitor). The nuclei were spun down at 845 rcf for 10 min at 4°C, resuspended in Proteinase K digestion solution (25-mM Tris–HCl pH 8.0, 1-mM CaCl₂, 1% [v/v] Tween-20, 20-μg Proteinase K), and incubated at 55°C for 1 h. After digestion, the solution was purified twice with acidic phenol/chloroform (pH 4.3) and precipitated by cold ethanol. The nucleic acid was then dissolved in RNase-free water and digested with DNase I (Roche) for 1 h at 37°C. The solution was further purified with chloroform and was precipitated by cold ethanol. The resolved RNA was used as template for reverse transcription with gene-specific primers.

Co-IP assay

Co-IP was carried out according to the previous study (Li et al., 2018), except that nuclear protein extracts were used in this study. Siliques at 4 DAP were ground in liquid nitrogen and lysed with nuclear fractionation buffer (20-mM Tris–HCl pH 7.0, 250-mM sucrose, 25% [v/v] glycerol, 20-mM KCl, 2-mM EDTA, 2.5-mM MgCl₂, 30-mM β-mercaptoethanol, protease inhibitor cocktail, and 0.7% [v/v] Triton X-100) for 15 min on ice. After the obtained slurry was filtered on a 100-μm cell strainer to remove tissue debris, the total filtrate was centrifuged at 1,000 g for 5 min at 4°C. The pellet was washed with resuspension buffer (20-mM Tris–HCl pH 7.0, 20-mM KCl, 2-mM EDTA, 2.5-mM MgCl₂, 30-mM β-mercaptoethanol, and protease inhibitor cocktail) 3–4 times, resuspended in extraction buffer (50-mM Tris–HCl pH 7.5, 150-mM NaCl, 1-mM EDTA, 5% [v/v] glycerol, 0.5% [v/v] Triton X-100, 1-mM PMSF, proteinase inhibitor cocktail, and 25-μM MG132), incubated with protein G magnetic beads (Thermo Fisher Scientific) supplied with anti-CTD (ab26721; Abcam), anti-CTD (S2P) (ab5095; Abcam), or anti-CTD (S5P) (ab5131; Abcam) antibodies at 4°C for 2 h, and then washed 4–6 times in extraction buffer. Immunoprecipitated proteins and nuclear protein extracts as the input were resolved by SDS–polyacrylamide gel electrophoresis and detected by various antibodies.

Expression analysis

Total RNA from various tissues was extracted using a FavorPrep Plant Total RNA Mini Kit (Favorgen) and reverse transcribed using the M-MLV Reverse Transcriptase kit (Promega). Quantitative PCR was performed on three biological replicates using a CFX384 Real-Time PCR Detection System with iQ SYBR Green Supermix (Bio-Rad). The expression of U-BOX was used as an internal control. The primers used for expression analysis are either listed in Supplemental Data Set 2 or according to the previous study (Li et al., 2018)

Statistical analysis

The detailed statistical analyses of the experiments are available in the figure legends and Supplemental Data Set 3. Statistical tests were conducted using Microsoft Excel and Origin.

Accession numbers

Sequence data used in this study can be found in the Arabidopsis Information Resource (https://www.arabidopsis.org) under the following accession numbers: SmD1a (At3g07590), SmD1b (At4g02840), SmD3b (At1g20580), SmE1 (At2g18740), SK11 (At5g26751), SK12 (At3g05840), and U-BOX (At5g15400).

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. Cas9-ChAR is designed for identifying trans-acting factors of SK11.

Supplemental Figure S2. Production of probes for Cas9-ChAR.

Supplemental Figure S3. Characterization of SmD1 mutants.

Supplemental Figure S4. Seed phenotypes of smd1a-380 and smd1b-601, and the complementation of smd1b-601.

Supplemental Figure S5. Seed phenotypes of smd3b-410 and sme1-2.

Supplemental Figure S6. Comparison of sequence similarity between SK11 and SK12.

Supplemental Figure S7. Characterization of the reporter lines containing a wild-type or mutated first intron in the 5′-UTR of SK11.

Supplemental Data Set 1. List of proteins identified from Cas9-ChAR.

Supplemental Data Set 2. Oligonucleotides used in this study.

Supplemental Data Set 3. Summary of statistical analysis.

 

C.L. and H.Y. conceived and designed the study. C.L. performed most of the experiments. B.C. conducted the GC–MS analysis of fatty acids. C.L. and H.Y. analyzed the data and wrote the paper. All authors read and approved the article.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the instructions for Authors (https://academic.oup.com/plcell) is: Hao Yu (dbsyuhao@nus.edu.sg).