Background

Seeds are intricate units essential to establish the next sporophytic generation of plants. In Angiosperms, the development of a seed starts with a key process called double fertilisation, where one sperm cell fuses with the egg cell, originating the diploid zygote, while the other fuses with the central cell, giving rise to the triploid endosperm [1]. In the next stages of seed growth and maturation, an influx of nutrients, minerals, sugars and signals from the maternal plant through the vasculature of the funiculus is required [2]. Finally, when the seed is mature, the formation of a seed abscission zone (SAZ) occurs, composed of a separation layer with thin cell walls that will degenerate at the end of the process and an adjacent layer with lignified cells (lignified layer), which produces the tension necessary for the separation of the tissues [3]. This process allows seeds to detach from the plant.

The funiculus is a specialised sporophytic structure that anchors the ovule to the placenta within the gynoecium of flowering plants, being the contact route for communication between them [4]. It arises during ovule primordium and remains intact throughout ovule development until seed dehiscence occurs. Remarkably, there is a lack of information regarding Arabidopsis thaliana case-study compared to other plants such as Brassica napus [5,6,7] or Phaseolus vulgaris [8], in which topics like anatomy or genetic regulation of funiculus development and underlying biological function have been extensively studied. So far, only one studied focused on clarifying these topics using A. thaliana. Khan and colleagues work showed using light microscopy that the funiculus is composed of an outer epidermis, a parenchymatous cortex, and an internal vascular core with xylem and phloem [9].

A comparative transcriptomic study between the zygotic regions and subregions of the seed and the funiculus in A. thaliana revealed the accumulation of several exclusive mRNAs in this stalk structure, presenting the funiculus as a transcriptionally distinct region from the other seed tissues [9]. Of the 20 genes identified as funiculus-specific, some were associated with transport and others with carbon metabolism, but over 30% did not have an annotated function, thus, showing that regulatory networks in the funiculus require further investigation.

SEEDSTICK (STK) is one of the few genes identified to be involved in funicular patterning and growth [10, 11]. It belongs to the MADS-box transcription factor (TF) family and is considered a master regulator of ovule development. Observations using a reporter line with the entire STK genomic region fused to a green fluorescent protein revealed high expression of this gene specifically in the ovules’ sporophytic tissues, in which the funiculus is included, from the beginning of ovule development until after fertilisation, when seed is developing [12]. This expression pattern, together with many genetic-based studies performed over the years, supports the pivotal role of STK in orchestrating various processes during plant reproduction. STK redundantly controls ovule identity together with SHATTERPROOF1 (SHP1) and SHP2 [11] and participates in transmitting tract development by interacting with NO TRANSMITTING TRACT (NTT) [13] and CESTA (CES) [14]. STK also determines fruit size by regulating cytokinin homeostasis and controlling FRUITFULL expression [15]. Furthermore, STK is involved in seed development and size, affecting processes such as flavonoids biosynthesis [12], cell wall organisation and remodelling properties [16,17,18,19] and cell cycle [20]. Observations at floral stage 17 [21] showed that stk mutants develop thicker and longer funiculi compared to wild-type (WT) and present defects in the SAZ, resulting in a decrease in the ability to dehiscence [11]. Balanzà and colleagues [22] proposed a model with the STK network controlling the formation of the SAZ. According to this study, STK interacts with SEUSS co-repressor, forming a complex that will repress HECATE 3, enabling the correct deposition of lignin in the vasculature of funiculus cells. Nonetheless, only TFs have been reported as regulators for the establishment of the lignification pattern. However, in similar abscission processes occurring in the floral organ abscission zone (reviewed by Niederhuth and colleagues [23]) or silique dehiscence zone (reviewed by Yu and colleagues [24]), several intervenients besides TFs have been reported: hormones (like ethylene, auxin, jasmonic acid), small peptides and their receptors (INFLORESCENCE DEFICIENT IN ABSCISSION is perceived by HAESA or HAESA-LIKE 1), as well as cell wall modifying enzymes (cellulases, hemicellulases and pectinases).

Despite the significant achievements on STK role during seed shattering, many questions remain unanswered: Which downstream targets participate in seed abscission? Is STK controlling other processes in the funiculus? Is seed development being affected by them? Only one transcriptome on stk single mutant is currently available [12], however those data derive from inflorescences and fruits until five days after pollination, which do not include floral stage 17 [21]. Moreover, considering the different tissues contained in that transcriptome, it would be difficult to study, at the molecular level, processes happening in a specific structure such as the funiculus.

In this study, we explore the transcriptomic data from stk and WT funiculi at floral stage 17, in order to investigate at the molecular level, the processes affected by STK absence in the funiculus. We provide information about enriched biological processes and molecular functions underpin the mutant funiculus phenotype, and we validate with qPCR and/or promoter reporter lines the expression of candidate genes, encoding proteins such as pectinases, laccases, or sugar transporters. Overall, we believe that by leveraging our results, novel players in the STK-mediated network in the funiculus can be found, and, importantly, whose functions may be affecting seed development and shattering.

Materials and methods

Plant material and growth conditions

Wild-type (WT) Arabidopsis thaliana (L.) Heynh. Columbia-0 seeds and the stk-2 mutant line [11], referred to as stk, were sown directly on soil (COMPO SANA®, Germany) and kept for 48 h at 4 °C in the dark to induce stratification. Afterwards, all plants were grown under long-day conditions (16 h light at 22 °C and 8 h darkness at 18 °C) with 50–60% relative humidity and light intensity at 180 µmol m− 2 s− 1.

Sample collection for RNA-sequencing

A procedure for funiculi collection from dissected siliques was developed to address the question of which genes were differentially expressed in the funiculus in the absence of STK. Under a SZ61 stereo microscope (Olympus), WT and stk siliques from stage 17 (st 17) flowers [21] were first dissected, seeds were removed, and only the septum with the attached funiculi was placed into a 50 µL droplet of in vitro ovule culture medium [25] in a plastic dish (Φ 3.5 cm) (Supplementary Fig. 1A–C). Each funiculus was separated using a surgical knife and transferred with a tungsten needle to another plastic dish containing 10 µL of medium (Supplementary Fig. 1D, E). Afterwards, each funiculus was transferred to a 1.5 mL tube containing 40 µL of RNAlater™ Stabilization Solution (Ambion).

RNA-sequencing library preparation

mRNA was extracted from two biological replicates of 100 funiculi from st 17 of both WT and stk, using the Dynabeads™ mRNA DIRECT™ Micro Kit (Invitrogen), according to the manufacturer’s instructions. Sequencing libraries were prepared following the manufacturer’s instructions in the Low Sample protocol, using the TruSeq® RNA Sample Preparation v2 (Illumina Inc.) and sequenced with a NextSeq 500 sequencer (Illumina). Sequencing was performed as reported in Kadokura and colleagues work [26].

RNA-sequencing data analysis

Raw reads were cleaned and trimmed using Trimmomatic v.0.38 [27]. The resulting reads were aligned against the Arabidopsis genome (Araport11) using STAR v.2.7.0 software with default parameters [28]. The total number of reads as well as the uniquely mapped reads to the A. thaliana genome are reported in Supplementary Table 1. Afterwards, featureCounts v.1.6.3 [29] was used to generate the count matrix and to calculate gene expression values as raw read counts. Count read values were analysed using the DESeq2 package v.1.38.3 [30] from R software v1.30.1. A pre-filtering step was applied to remove lowly expressed genes. Specifically, genes with a total count of one or less across all samples were excluded from further analysis. Regularized logarithm (rlog) transformation of the count matrix was used for both principal component plot (PCA) analysis and heatmap clustering. PCA was generated using the plotPCA of DESeq2 package and data were visualised with ggplot2 package v.3.4.1 [31] in R. Heatmap showing clusters of samples was obtained with pheatmap package v.1.0.12 in R and using the top 1000 genes selected based on the mean of the rlog normalised counts. The total number of genes expressed in stk, WT or both samples was plotted using a Venn diagram, and plotMA function together with ggplot2 were used to show the log2FoldChange (log2FC) of all genes. To identify differentially expressed genes (DEGs), the raw count data were analysed using the DESeq function. A cut-off of p-value, p-adjust value and log2FC was applied to select differentially expressed genes: p-value and p-adjust value < 0.05 and a −1 < log2FC > 1.

Gene set enrichment analysis

To determine enriched gene ontology (GO) terms in the stk funiculus transcriptome compared to the WT, genes were ranked as: −Log10(p-value) multiplied by the sign of the Log2FC [32]. A Parametric Analysis of Gene Set Enrichment [33] with Hochberg multi-test adjustment method was performed using the ranked gene list on agriGO v.2.0 ([34]; http://systemsbiology.cau.edu.cn/agriGOv2/, accessed in July 2020), from which biological process and molecular function GO terms with false discovery rate (FDR) < 0.05 were obtained and analysed.

Enriched networks among DEGs

To uncover functional links between genes, a list containing up- and down-regulated DEGs was upload to GeneMANIA Cytoscape plugin v3.10.0 [35] software.GeneMANIA constructed individual interaction networks for each data source, including gene co-expression, physical interactions, shared protein domains, and GO biological process annotations. Each network was assigned a weight based on its relevance to the query genes, optimising these weights through a linear regression model to best fit the known associations of the input genes. To enhance the robustness of our analysis, only GO biological processes with a FDR < 0.05 were considered. The individual networks were then combined into a single composite network using the optimised weights. Label propagation algorithms were applied to this composite network to infer connections between genes, spreading functional labels from known genes to predict the functions of unknown genes.

Sample collection, RNA isolation and cDNA synthesis for validation by qPCR

A total of six samples for RNA extraction were collected from funiculi at st 17, representing two genotypes (WT and stk), from three biological samples. One hundred funiculi from one plant were used for each biological replicate. Samples were immediately frozen in liquid nitrogen and stored at −80 °C for follow-up experiments. Total RNA was extracted using the RNeasy Plant Mini Kit (QIAGEN) according to the manufacturer’s protocol, with a minor change: the recommended quantities for reagents were reduced by half. RNA concentration was measured using a spectrophotometer (DS-11 Series Spectrophotometer/Fluorometer). Following the manufacturer’s instructions, 21 ng of total RNA from funiculi at st 17 was used for cDNA synthesis using the Maxima H Minus First Strand cDNA Synthesis Kit (Thermo Fisher). The cDNA products were diluted to 0.21 ng/µL with nuclease-free water prior to qPCR.

Primer design and qPCR analysis for validation of RNA-sequencing

Primers were designed using Primer3 v.4.1.0 [36,37,38], following the parameters of Ferreira and colleagues [39]. For MLP28, Litholdo and colleagues [40] primers were used, and for ACTIN 7 (ACT7), YELLOW-LEAF-SPECIFIC GENE 8 (YLS8) and HISTONE 3.3 (HIS3.3), Ferreira and colleagues [39] primers were used. Supplementary Table 2 presents all the primers used for qPCR. Primer specificity was confirmed by conventional PCR and electrophoresis on 1% (w/v) agarose gel. qPCR was performed according to Ferreira and colleagues [39]. Briefly, reactions were prepared in a 10 µL final volume containing 5 µL of 2x SsoAdvanced™ Universal SYBR® Green Supermix (Bio-Rad), 0.125 µL of each specific primer pair at 250 nM, 0.75 µL of nuclease-free water, and 4 µL of diluted cDNA template. The reactions were performed in 96-well plates and run on a CFX96 Real-Time System (Bio-Rad) under the following cycling conditions: initial denaturation at 95 °C for 30 s, followed by 40 cycles at 95 °C for 15 s and 60 °C for 30 s, and an additional data acquisition step of 15 s at the optimal acquisition temperature. All reactions were run in three technical replicates and all assays included non-template controls (NTCs). Using three points of a 5-fold dilution series (1:5, 1:25 and 1:125) from a pool of funiculi cDNA (containing both WT and stk cDNA), a standard curve was generated to estimate the PCR efficiency of each primer pair using CFX Maestro software v.2.0 (Bio-Rad). The slope and coefficient of determination (R2) were obtained from the linear regression line and the amplification efficiency (E) was calculated according to the formula E = (10− 1/slope − 1) x 100%. The R2 value should be higher than 0.980 and the E value should be between 90% and 110%. After completion of the amplification reaction, melt curves were generated by increasing the temperature from 65 °C to 95 °C, with fluorescence readings acquired at 0.5 °C increments. From the melt curve, the optimal temperature for data acquisition (3 °C below the melting temperature of the specific PCR product) was determined and the specificity of the primers was confirmed (Supplementary Table 3). The sample maximisation method was chosen as the run layout strategy, in which all samples for each defined set were analysed in the same run, thus, different genes were analysed in distinct runs [41]. The quantitative cycle, baseline correction and threshold setting were automatically calculated using CFX Maestro software v.2.0 (Bio-Rad). The qPCR products were verified using 1% (w/v) agarose gels. The expression levels of the target genes were quantified in all samples using three reference genes (ACT7, YLS8 and HIS3.3; [39]), whose expression stability between stk and WT funiculi at st 17 was verified using the RNA-sequencing (RNA-seq) data (Supplementary Table 4). Relative gene expression was calculated using the 2−ΔΔCt method [42]. Data were statistically analysed using the CFX Maestro software v.2.0 (Bio-Rad), which compared the relative expression differences using a Student’s t-test. Statistical significance was set to a p-value threshold of 0.05. The Pearson correlation coefficient (r) was calculated to determine statistical correlation between the RNA-seq and qPCR data.

Construct generation and plant transformation

Genomic regions corresponding to the promoters of three genes (ADPG1, MLP28 and SCPL41) were amplified using DreamTaq DNA polymerase (Thermo Fisher) for ADPG1 and SCPL41, and Phusion DNA polymerase (Thermo Fisher) for MLP28, using the primer pairs described in Supplementary Table 2. The PCR products were cloned into pENTR™/D-TOPO (Invitrogen) and subcloned into the binary vector pBGWFS7 containing the β-glucuronidase (GUS) gene [43]. All constructs were confirmed by DNA sequencing. Expression vectors were delivered into Agrobacterium tumefaciens GV3101 (pMP90RK) and were used to transform A. thaliana (WT and stk genotype) by the floral dip method [44]. Transformant lines were obtained using glufosinate ammonium (BASTA®) as a selection agent.

Detection of GUS activity

For histochemical analysis, siliques at st 17 from T3 generation plants were fixed in acetone 90% (v/v) (Thermo Fisher) at -20 ºC overnight (ON). Samples were then washed twice with sodium buffer [0.2 M sodium dihydrogen phosphate (NaH2PO4; Panreac), 0.2 M sodium hydrogen phosphate (Na2HPO4; Sigma Aldrich)] for 10 min and incubated in a X-Gluc solution [50 mM Na2HPO4, 50 mM NaH2PO4, 0.2% (v/v) Triton X-100 (Sigma Aldrich), 2 mM potassium hexacyanoferrate(II) trihydrate (C6FeK3N6 + 2 · 3H2O; Sigma Aldrich), 2 mM potassium hexacyanoferrate(III) (C6FeK3N6 + 3; Sigma Aldrich), 1 mg/mL X-Gluc (5-bromo-4-chloro-3-indolyl β-D-glucuronic cyclohexylammonium salt; Biosynth)] ON at 37 ºC. After chemical GUS detection, samples were rinsed with a 90% (v/v) ethanol, followed by 70% (v/v) ethanol for 10 min and incubated in a clearing solution [160 g of chloral hydrate (Sigma Aldrich), 100 mL of deionised water, and 50 mL of glycerol (Sigma Aldrich)], where they were kept at 4 °C ON. The following day, siliques were dissected under a stereomicroscope (model C-DSD230; Nikon) using hypodermic needles (0.4 × 20 mm; Braun). The funiculi attached to the septum were maintained in a drop of clearing solution and covered with a cover slip. Samples were observed under an upright Axio Imager AZ microscope (Zeiss) equipped with differential interference contrast optics. Images were captured with an Axiocam MRc3 camera (Zeiss) and processed using Zen 2011 Software (blue edition; Zeiss).

Image processing

All images were processed for publication using Fiji [45] and Adobe Photoshop CC 2019.

Results

RNA-sequencing analysis of stk funiculi at St 17

To investigate if the transcriptional profiles of stk and WT funiculi were similar, a principal component analysis (PCA) and hierarchical clustering were conducted (Fig. 1A). PCA showed that 90% of the total variance could be explained by the first two components (53% by PC1 and 37% by PC2). A cluster of stk replicates could be identified, but WT replicates showed a slight discrepancy between each other, with one of the replicates being closer to the stk replicates. However, the expression analysis of the top 1000 genes clearly showed two distinct clusters, one for stk replicates and the other for WT replicates (Fig. 1B). The heatmap also demonstrated that the expression of these genes between stk and WT samples was similar, suggesting a low number of differentially expressed genes (DEGs) between the two genotype samples. Overall, the results showed clusters between replicates, even though the expression profiles of stk and WT samples were not completely distinguishable.

Fig. 1
figure 1

Assessment of samples quality. (A) Principal component analysis (PCA) conducted on rlog normalised gene expression values of stk and WT samples. X − and Y − axes show PC1 and PC2, respectively, with the amount of variance contained in each component being 53% and 37%, respectively. Each point in the plot represents a biological replicate of 100 funiculi, with a total of four biological replicates in the plot. Symbols of the same colours are replicates of the same experimental group, where orange represents stk funiculi, and blue represents WT funiculi. (B) Heat map of the top 1000 genes selected based on the mean of the rlog normalised counts. Individual samples are shown in columns and genes in rows. The upper axis shows the clusters of samples, and the left vertical axis shows the clusters of genes. The colour scale represents the relative expression of genes: purple indicates low relative expression levels and yellow indicates high relative expression levels

Full size image

To obtain an overview of the funiculi transcriptome, the total number of expressed genes in stk and WT samples was examined. Both datasets presented a similar number of genes (21841 for stk and 21723 for WT), from which 1293 genes (5.9%) were exclusively expressed in stk funiculi, and 1175 genes (5.4%) were only present in WT funiculi. The commonly expressed genes expressed in both datasets numbered 20,548 (Fig. 2A). Differential expression analysis revealed a total of 169 DEGs (Supplementary Table 5), with p-value and p-adjust value < 0.05 and a −1 < log2FoldChange (log2FC) > 1, from which 119 genes were upregulated and 50 were downregulated, as seen in the MA plot (Fig. 2B).

Fig. 2
figure 2

Global analysis of stk vs. WT transcriptome from funiculi at st 17. (A) Venn diagram showing the number of genes expressed in each genotype. The number of genes expressed only on stk funiculi or WT funiculi are shown, as well as the number of overlapping genes between transcriptomes. WT is represented by a blue circle and stk by an orange circle. (B) MA plot demonstrates the relationship between the log10 average normalised expression on the x − axis and the significance of the differential expression test expressed as log2FoldChange (log2FC) on the y − axis for each gene in the transcriptome. It illustrates the number of DEGs in the stk funiculi. Gray dots represent genes that are not significantly differentially expressed, while pink and purple dots represent genes that are significantly up- and downregulated, respectively, with p-adjust value < 0.05

Full size image

Parametric analysis of gene set enrichment

To better understand the processes being affected in stk funiculus compared to WT, a parametric analysis of gene set enrichment (PAGE) was performed. PAGE is a statistically sensitive method that considers gene expression levels to rank an annotated gene cluster. Genes were ranked following the formula: -Log10(p-value) multiplied by the sign of the Log2FC. Each gene ontology (GO) term has an associated z-score value, which provides information about the significance of the enriched term among the data. Moreover, the z-score can assume a positive or negative value, depending on whether the term includes more up- or downregulated genes, respectively. For a full list of biological and molecular function GO terms, please see Supplementary Table 6. Significant enrichment of 53 and seven GO terms for biological function and molecular function, respectively, was observed among the gene datasets. The most significantly upregulated enriched GO category was cellular cell wall organization or biogenesis (GO:0070882) (Fig. 3A), with more than 100 genes associated. Terms related to dehiscence were active in stk transcriptome: dehiscence (GO:0009900), lignin biosynthetic process (GO:0009809) and phenylpropanoid metabolic process (GO:0009698). In addition, many cell wall related terms were found such as hemicellulose metabolic process (GO:0010410), xylan biosynthetic process (GO:0045491) and glucuronoxylan biosynthetic process (GO:0010417). Apart from these, regulation of DNA binding (GO:0051101) was a term that stood out. Concerning the molecular function, only two terms were upregulated, oxidoreductase (GO:0016722) and acyl-CoA hydrolase activity (GO:0047617), with the first one presenting a higher fold enrichment (Fig. 3B).

Fig. 3
figure 3

Parametric GO gene set enrichment analysis of activated pathways in stk vs. WT transcriptome from funiculi at st 17. GO enrichment was evaluated at two different levels: (A) Biological Processes and (B) Molecular Function. Bubble plot diagrams show the significance [-Log10(false discovery rate; FDR)] on the x-axis, the fold enrichment (z-score) in a gradient of colour from purple (underrepresented) to yellow (overrepresented), and the number of genes in each pathway

Full size image

PAGE analysis identified the GO term glycoside catabolic process (GO:0016139) as the most inhibited (Fig. 4A). The absence of STK in the funiculus affected the sucrose metabolism (GO:0005985), starch metabolism (GO:0005982) and the immune effector process (GO:0002252). Furthermore, the term cell cycle (GO:0007049) appeared in the analysis, containing more than 400 genes, as well as other related terms like posttranscriptional regulation of gene expression (GO:0010608) and gene silencing (GO:0016458). The glutathione metabolic process and gynoecium development were also downregulated in the stk funiculi transcriptomic profile. The enriched molecular functions for the downregulated genes revealed terms exclusively related to enzymatic activity: transferase activity, sucrose alpha-glucosidase and beta-glucosidase activity (Fig. 4B).

Fig. 4
figure 4

Parametric GO gene set enrichment analysis of inhibited pathways in stk vs. WT transcriptome from funiculi at st 17. GO enrichment was evaluated at two different levels: (A) Biological Processes and (B) Molecular Function. Bubble plot diagrams show the significance [−Log10(false discovery rate; FDR)] on the x-axis, the fold enrichment (z-score) in a gradient of colour from purple (overrepresented) to yellow (underrepresented), and the number of genes in each pathway

Full size image

Expression of funiculus-specific genes in the transcriptome

The expression profiles of genes previously identified as funiculus-specific [9] were used to assess the reliability of the transcriptome. All 20 funiculus-specific genes were detected in the RNA-seq data (Table 1), with two of them (EXPANSIN-LIKE B1 and GRETCHEN HAGEN 3.3) being downregulated in the stk mutant. These genes represent an expansin-like, involved in cell wall extension, and an IAA-amido synthase, implicated in auxin responses, respectively.

Table 1 Expression of funiculus-specific genes [9] on stk vs. WT transcriptome from funiculi at St 17
Full size table

Enriched networks on DEGs list

To further investigate enriched networks among DEGs and how genes correlated with each other’s regarding physical interactions, co-expression and shared protein domains, the up- and downregulated DEGs were uploaded separately to the GeneMANIA program. All significant biological functions predicted by this tool with false discovery rate (FDR) < 0.05 were listed in Supplementary Table 7. Regarding the downregulated genes in the stk transcriptome, the program detected a network related to carbohydrate transmembrane transporter activity (Fig. 5A). SUGARS WILL EVENTUALLY BE EXPORTED TRANSPORTER 4 (SWEET4) and SWEET7 were the two DEGs identified as involved in this process, being co-expressed and physically interacting with other members of SWEET family. Another network that stood out was associated with floral development (Fig. 5A), containing four DEGs: CELL WALL INVERTASE 4 (CWINV4), ARABIDOPSIS DEHISCENCE ZONE POLYGALACTURONASE 1 (ADPG1), CYTOCHROME P450, FAMILY 90, SUBFAMILY D, POLYPEPTIDE 1 (CYP90D1) and ARGONAUTE 9 (AGO9). However, no co-expression was detected between these genes. These two selected enriched networks share not only a common gene, SWEET13, but are also linked by co-expression. For instance, CWINV4 is co-expressed with SWEET4, while ADPG1 is co-expressed with SWEET7. For the upregulated DEGs, the most enriched network with smallest FDR identified by the program was associated with plant type-cell wall biogenesis (Fig. 5B). Nine DEGs were identified such as NAC DOMAIN CONTAINING PROTEIN 43 (NAC043), NAC012, LACCASE 4 (LAC4) and KNOTTED-LIKE HOMEOBOX OF ARABIDOPSIS THALIANA 7 (KNAT7). In addition, other DEGs and genes that the program identified as involved with cell wall biogenesis can be visualised. For instance, WRKY12 is co-expressed with both NAC043 and NAC012, LAC4 is co-expressed with LAC2, which is also a DEG in the present transcriptome, and is co-localised with KNAT7.

Fig. 5
figure 5

Expression networks of stk vs. WT transcriptome from funiculi at st 17 derived from GeneMANIA. (A) Networks known from the downregulated differentially expressed genes (DEGs). (B) Networks identified from the upregulated DEGs. Genes added as relevant interactors by the GeneMANIA program are shown with a white dash in the middle of the circle, and the size is proportional to the number of interactions they have. The purple edges denote co-expression network interactions, the orange edges represent physical interactions between genes and the green edges indicate shared domain. Different colours in each circle show different functions, as indicated

Full size image

RNA-sequencing data validation through qPCR and promoter expression analysis

To validate the RNA-seq data, we selected a group of up- and downregulated genes from the transcriptome (Table 2), which we considered as potential candidates to be further studied. First, we analysed the GO terms and enriched networks, and we chose genes known to be involved in abscission and seed development. LAC2 and LAC4, members of laccase family, are involved in lignin biosynthesis [46, 47], SWEET4 and SWEET7, belonging to sweet family, are related to sugar transport [48, 49] and ADPG1, with pectinase activity, is implicated in abscission [50]. A comprehensive literature search was conducted on all deregulated genes from the transcriptome to identify other potential candidates that were not recognised on the enriched networks analysis. Two distinct gene families were subsequently selected for further investigation. MAJOR LATEX PROTEIN-like 28 (MLP28) and MLP168 are part of the major latex protein family, which is involved with hormone signalling [51], and SERINE CARBOXYPEPTIDASE-like 41 (SCPL41), an enzyme with membrane lipid metabolism functions [52]. All genes selected were expressed in pistil (Supplementary Fig. 2) and silique (Supplementary Fig. 3) according to the online database ePlant [53] and Arabidopsis eFP Browser [54], respectively.

Table 2 List of differentially expressing genes candidates from the stk vs. WT funiculi transcriptome
Full size table

The relative expression levels of eight selected genes, MLP168, LAC2, LAC4, SWEET4, SCPL41, MLP28, ADPG1 and SWEET7 were quantified by qPCR in stk and WT funiculi at st 17 [21]. The transcript levels were normalised to three reference genes (ACT7, YLS8 and HIS3.3), previously tested according to Ferreira and colleagues [39] guidelines to infer about their appropriateness for funiculi samples, and are presented relative to WT funiculi transcript levels (Fig. 6). The analysis revealed that MLP168, LAC2 and LAC4 expression was higher on stk funiculi compared to WT (Fig. 6A). In contrast, SWEET4, SCPL41, MLP28, ADPG1 and SWEET7 were downregulated on stk samples (Fig. 6B). These results were consistent with the transcriptome dataset (Table 2), with a correlation coefficient of 0.809 (Supplementary Fig. 4), indicative of good correlation between RNA-seq and qPCR expression values for candidate genes.

Fig. 6
figure 6

Relative normalised expression levels of MLP168, LAC2, LAC4, SWEET4, SCPL41, MLP28, ADPG1 and SWEET7 in WT and stk funiculi at st 17 [21]. (A) and (B) Relative normalised expression of upregulated and downregulated genes, respectively, in stk funiculi according to the RNA-seq data. The transcript levels were normalised to ACT7, YLS8 and HIS3.3 reference genes [39]. The data correspond to the ratio of the expression compared to the WT. Error bars represent the standard error of the mean of three independent biological replicates, each with three technical replicates. Statistical analysis was performed using a Student’s t-test. Asterisks indicate significant differences from WT (*, p < 0.05)

Full size image

A promoter-driven GUS assay was also performed to further confirm the RNA-seq transcriptome. For instance, we analysed the promoter activity of MLP28 (pMLP28), SCPL41 (pSCPL41) and ADPG1 (pADPG1). The promoter regions of these genes were fused to the GUS reporter gene, and funiculi from both stk and WT transgenic plants were observed at st 17 of flower development [21], corresponding to the same type of sample used to obtain the transcriptome (Fig. 7). The selection of downregulated genes facilitated the observation of differences in the promoters’ expression between stk and WT funiculi. GUS activity driven by pMLP28 was observed in the entire WT funiculus and septum, whereas in the stk background, the detection of the promoter ceased in these two tissues (Fig. 7A and D). In the case of pSCPL41, GUS expression was strong in the WT funiculus, however, in the stk background, there was no expression of GUS in the funiculus (Fig. 7B and E). For pADPG1, the GUS signal was restricted to the SAZ established in the WT funiculus and no GUS staining was visible in the stk funiculus (Fig. 7C and F). Overall, these results confirmed the transcriptome data and qPCR, which showed a decrease in transcript levels of MLP28, SCPL41 and ADPG1 in the stk funiculus.

Fig. 7
figure 7

Histochemical localisation of GUS activity in WT (A–C) and stk (D–F) funiculi at st 17 [21], expressing pMLP28-GUS, pSCPL41-GUS or pADPG1-GUS constructs. (A and D) pMLP28 expression was detected in WT funiculus but not in the stk background. (B and E) pSCPL41 was strongly expressed in WT funiculus, while the GUS activity ceased in stk funiculus. (C and F) pADPG1 signal was restricted to the SAZ of WT funiculus, but it was not detected on the stk funiculus. Three independent transgenic lines were observed for each situation. Abbreviations: f – funiculus; ms – mature seed; saz – seed abscission zone; sp – septum. Scale bar = 50 μm

Full size image

Discussion

Seeds are essential for plants, as they harbour the new offspring, and for humans, as a major source of calories. The processes involved in seed development are complex and require an accurate coordination of several regulatory networks established in different reproductive tissues [2]. The funiculus connects the maternal plant to the ovule and lately to the seed, contributing to nutrient transport, mechanical support, and seed positioning. When one of these functions is altered, as observed in stk mutants [11], seed development is disturbed. This influences seed size and viability as well as seed dispersal, ultimately compromising the reproductive success of the plant. Deciphering the genetic networks controlled by STK in the funiculus is, therefore, of major importance in the frame of elucidating how seed development is being disrupted.

Here we analysed the funiculus transcriptional profile of stk and WT st 17 of flower development to infer the processes affected by STK absence and gene regulatory networks enriched among the DEGs. We show that cell wall biogenesis, lignification and sugar transport are enriched in the transcriptome and which DEGs are known to be involved in these processes. We then focus on selecting candidate genes for future studies, by validating their expression using qPCR and/or promoter reporter lines.

We started by performing a collection assay to obtain RNA-seq data from stk and WT funiculi of st 17 of flower development. A total of 169 genes (p-adjust value < 0.05) were found to be deregulated in the transcriptome (Supplementary Table 5), indicating that the transcriptome profiles of stk and WT were not completely distinct, as supported by PCA. Commonly, RNA-seq studies comparing different tissues or developing stages tend to show a more evident separation between different biological samples and, thus, a high number of DEGs [55,56,57]. Moreover, the transcriptomic data generated hereby were from low-input mRNA, which could explain the reduced coverage of low expressed transcripts and, consequently, the detection of low expressed genes as DEGs. Generally, the number of DEGs is slightly lower when compared to libraries obtained from higher amounts of input RNA [58]. It is worth noting that all genes classified as funiculus-specific by Khan and colleagues [9] were present in the transcriptome, providing a high level of confidence for further analysis.

The most actively enriched function in stk funiculi was related to cell wall biogenesis (GO:0070882), consistent with previous phenotypic observations [11, 22]. After fertilisation, the WT funiculus vasculature proliferates in a synchronised manner with the cortex and epidermis [9], with growth being more accentuated in stk mutants. Therefore, terms related to the production of cell wall components such as polysaccharides [xylan (GO:0045491), glucuronoxylan (GO:0010417) and hemicellulose (GO:0010410)] were found to be enriched in PAGE analysis. The proliferation of cells is characterised by activation of the transcriptional machinery and, in the case of stk, transcription needs to be intensified to allow the overgrowth of the funiculus. This may explain the DNA binding (GO:0051101) GO term enrichment. The stk mutant has defects in seed dehiscence [11] and the reason was later revealed by Balanzà and colleagues [22], who concluded that an irregular accumulation of lignin in stk SAZ caused this phenotype. This control of lignin production by STK was reflected in the enrichment of GO terms for dehiscence (GO:0009900), lignification (GO:0009809) and phenylpropanoid metabolism (GO:0009698), which is a lignin precursor [59]. The analysis of molecular functions revealed the induction of genes related to oxidoreductase activity (GO:0016722). In plants, enzymes such as catalases or peroxidases present this type of activity, and are responsible for regulating the cellular level of hydrogen peroxide, a reduced oxygen species [60]. In the present transcriptome, PEROXIDASE 64 (PRX64) was upregulated in stk, and many studies using loss-of-function mutants or reporter lines observation [61,62,63] have concluded that PRX64 is involved in lignification. The link between this peroxidase and lignification in SAZ has not yet been described, making it an interesting gene to be further analysed. The most inhibitory process in stk transcriptome was the glycoside catabolic process (GO:0016139). Glycosides are a wide range group of plant secondary metabolites with multifaceted roles in regulating growth, defense and reproduction in plants [64,65,66]. Glycoside hydrolases are a family of enzymes responsible for the removal of specific sugar moieties from glycosides. The mutant analysis of a specific enzyme, BRANCHING ENZYME 1, revealed an impairment in carbohydrate metabolism [67], which is another GO term revealed in this transcriptome. Sugars are essential for embryogenesis, and the fact that STK is controlling both glycosylation and sugar catabolism (GO:0005985) might related to the small seed phenotype that stk presents. Also related to stk seed size are the terms starch metabolism (GO:0005982) and glutathione metabolism (GO:0006749). Starch is a source of sugar production in the funiculus, that can possibly be transported to the endosperm to nourish the embryo [9]. On the other hand, glutathione is a cell redox that helps in preventing an accumulation of reactive oxygen species within the cell [68]. A previous study showed that glutathione is expressed and accumulates in the funiculus as well as it is crucial for proper embryo development and seed maturation [69]. Collectively, these results suggest that STK controls the metabolism and influx of molecules from the funiculus to the seed, which are important for the correct development of this unit.

A closer look at the DEGs list, using GeneMANIA, allowed the identification of different networks and which specific DEGs were part of them, as well as which genes were being co-expressed, shared protein domains or had a known physical interaction with the DEGs. Not all DEGs were associated with a biological process, which may be a result of an uncharacterised gene or the program itself, since GeneMANIA shows improved outcomes if the input gene list is functionally related. Nonetheless, this analysis contributed for the criteria of selecting genes known to be involved in abscission or seed development. One of the enriched networks for downregulated DEGs was associated with carbohydrate transport activity, a process that allows the delivery of sucrose with maternal origin to the seed coat via the funiculus phloem, important for seed development [70,71,72]. Two genes from the transcriptome were identified in this network, SWEET4 and SWEET7, both with glucose transport activity, indicating that STK promotes the sugar efflux required for seed filling (Eom et al., 2015), particularly in the form of glucose instead of sucrose. This finding, combined with other previously reported data [16, 17, 19, 20], may explain the small seed phenotype that stk exhibits [11]. The other network that stood out included terms associated with floral organ development. CWINV4 is an enzyme involved in sugar signalling [73] that hydrolyses sucrose to generate glucose and fructose. It was first found to be involved in nectar production in Arabidopsis [74] and, afterwards, a study reported that CWINV4 together with CWINV2 (mentioned in the paper as CWIN4 and CWIN2) are positive regulators of ovule initiation, through sugar signalling [75]. Remarkably, transcriptomic data obtained from the double mutant transgenic line pSTK-amiRNACWIN24 showed several members of the SWEET family as downregulated DEGs, including SWEET4 and SWEET7, as well as TFs, such as STK [75]. Therefore, it would be valuable to investigate whether STK involvement in sugar efflux is due to the control of CWINV, which would produce sucrose that could be transported to the seed via SWEET transporters. So far, we had confirmed by qPCR that SWEET4 and SWEET7 are downregulated in stk funiculi at st 17. Another gene, part of the floral development network, was ADPG1 that belongs to the polygalacturonase family, which is responsible for disassembling the pectin part of plant cell walls [76]. A genetic approach demonstrated that ADPG1 together with ADPG2 are important for silique dehiscence, whereas the combined action of ADPG1, ADPG2 and QUARTET 2 contributes to anther dehiscence [50]. Observations based on transcriptional β-glucuronidase activity were in agreement with the phenotype mentioned above: ADPG1 was mainly expressed in silique dehiscence zones and in anthers, immediately before anthesis, but also in SAZ. Using the reporter line generated in the present study, we simultaneously confirmed that ADPG1 promoter is specifically expressed in the SAZ and that GUS activity ceases in the stk background (in accordance with RNA-seq and qPCR results). Notably, the authors showed that ADPG1 expression in the SAZ is disrupted in the hecate 3 mutant, a TF controlled by STK and involved in SAZ formation [22]. Despite the absence of an “obvious defect” in the adpg1 mutants mentioned by the authors, they did not reveal the type of analysis that was performed. Hence, we believe that these findings point to the involvement of ADPG1 with STK during seed abscission. Among the upregulated DEGs, the most enriched network recognised by GeneMANIA was the plant cell wall biogenesis, in which a laccase family member was identified, LAC4. Laccases have been associated with lignin biosynthesis [77], functioning as activators of lignin production [47, 78]. Studies have shown that disruption of LAC4, along with other laccases, leads to a decrease in lignin content of several organs, such as stems, siliques and roots [47, 78]. Interestingly, only one study has reported the activity of a specific laccase, LAC2, as a negative regulator of lignin deposition in root vascular tissues under abiotic stress [46]. LAC2 was identified as being co-expressed with LAC4, and qPCR indicated that there was more gene expression in stk with respect to WT. Since stk SAZ has an overproduction of lignin [22], these findings suggest that LAC2 and LAC4 participate in the STK-network controlling lignin biosynthesis in the funiculus.

Three additional candidate genes were selected for validation: MLP168 and MLP28, members of the pollen allergen Bet v 1 family [79] and the pathogenesis-related protein class 10 as well. Their protein conformation exhibits a hydrophobic cavity capable of binding to ligands, such as cytokinins, brassinolides and secondary metabolites, which MLPs transport to other organs via the plant vasculature, triggering, in some cases, downstream transduction signals in those organs [51]. Extensive studies on MLPs functions, mostly under biotic and abiotic stresses, have been conducted in several plant species, like A. thaliana, Brassica rapa, and Vitis vinifera [51]. Regarding MLP168, it was reported an interaction of this protein with the ABA INSENSITIVE 5 protein in a yeast two-hybrid assay [80]. We validated by qPCR that MLP168 was upregulated in stk funiculi. As for MLP28, observations of an amiRNA-MLP28 line showed defects on vegetative growth with alterations in leaf morphology and shoot apex, dwarfness and eventual premature plant death [40]. The present transcriptomic data revealed that MLP28 was downregulated in stk funiculi, which was confirmed by qPCR and promoter-driven GUS activity assays. Due to MLPs nature of binding to plant hormones already implicated in several processes of reproduction [81], it seems likely their involvement with ligand transportation from the funiculus to the seed. The last gene selected was SCPL41, a serine carboxypeptidase-like protein containing a catalytic triad responsible for cleaving the C-terminal peptide bond in proteins or peptides [82]. The analysis of scpl41 mutants has revealed an increase in lipid content in seedlings [52]. Since stk funiculus cells undergo an augmentation in size that must be accompanied by overproduction of cell components, such as lipids, it is reasonable to think that SCPL41 control of lipid content in the funiculus may be regulated by STK. The qPCR results showed a subexpression of SCPL41 in stk funiculi from st 17 flowers and the promoter line showed expression of GUS in the WT funiculus, which was absent in the stk background.

Conclusion

Overall, the presented work provides the first RNA-seq data from stk and WT funiculi at st 17. The RNA-seq analysis complemented a previous phenotyping study of stk funiculi by showing GO terms and enriched networks related to cell wall biogenesis, sugar transport and abscission. The results demonstrated that STK is mainly a repressor of cell wall and lignin biosynthesis-related genes, but simultaneously an activator of glucose transporters and seed dehiscence genes acting at the separation layer of the SAZ (Fig. 8). The validation of the transcriptome reinforced the reliability of the data obtained and confirmed that genes never associated with funiculus functions such as MLPs or SCPLs are, in fact, deregulated in stk funiculus (Fig. 8). These results provide new insights into the regulatory molecular network controlled by STK in the funiculus and highlight the existence of novel downstream targets of STK, whose functions require further exploration in the context of funiculus and seed development processes.

Fig. 8
figure 8

A simplified model proposing the involvement of candidate genes from stk vs. WT transcriptome from funiculi at st 17 in the STK-mediated network established in the funiculus. (Blue box) STK activates the expression of ADPG1 at the separation layer of the SAZ. This activation leads to pectin degradation of the cell wall, which is essential for seed abscission. (Yellow box) STK inhibits the expression of LAC2 and LAC4 at the lignification layer of the SAZ. This repression prevents an overproduction of lignin, facilitating seed abscission. (Pink box) STK activates SWEET4 and SWEET7 expression. These transporters facilitate glucose movement from the funiculus to the seed, that is important for proper seed development. (Orange box) STK also regulates MLP28 and MLP168 expression. MLP28 is activated, while MLP168 is repressed. This fine-tuned regulation ensures correct transport of ligands, such as hormones, from the funiculus to the seed, further supporting seed development. (Purple box) STK activates SCPL41 expression, which repress the overproduction of lipids in funiculus cells. Green dashed arrow – activates gene expression; Pink dashed line with end bars – represses gene activation; black arrow – promotes action. Abbreviations: f – funiculus; ms – mature seed; saz – seed abscission zone; sp – septum

Full size image