Further Disruption of the TAS3 Pathway via the Addition of the AGO7 Mutation to the DRB1, DRB2 or DRB4 Mutations Severely Impairs the Reproductive Competence of Arabidopsis thaliana

Abstract

The previous assignment of functional roles for AGO7, and the DOUBLE-STRANDED RNA BINDING (DRB) proteins, DRB1, DRB2, and DRB4, in either microRNA (miRNA) or trans-acting small-interfering RNA (tasiRNA) production allowed for use of the loss-of-function mutant lines, drb1, drb2, drb4, and ago7, to further functionally characterize the TAS3 pathway in Arabidopsis thaliana (Arabidopsis). Towards achieving this goal, we also describe the developmental and molecular phenotypes expressed by three newly generated Arabidopsis lines, the drb1ago7, drb2ago7, and drb4ago7 double mutants. We show that the previously reported developmental abnormalities displayed by the drb1, drb2, drb4, and ago7 single mutants, are further exacerbated in the drb1ago7, drb2ago7, and drb4ago7 double mutants, with rosette area, silique length, and seed set all impaired to a greater degree in the double mutants. Molecular assessment of the TAS3 pathway in the floral tissues of the seven analyzed mutants revealed that DRB1 is the sole DRB required for miR390 sRNA production. However, DRB2 and DRB4 appear to play secondary roles at this stage of the TAS3 pathway to ensure that miR390 sRNA levels are tightly maintained. We further show that the expression of the TAS3-derived tasiARF target genes, AUXIN RESPONSE FACTOR2 (ARF2), ARF3, and ARF4, was altered in drb1ago7, drb2ago7, and drb4ago7 flowers. Altered ARF2, ARF3, and ARF4 expression was in turn demonstrated to lead to changes in the level of expression of KAN1, KAN3, and KAN4, three KANADI transcription factor genes known to be transcriptionally regulated by ARF2, ARF3, and ARF4. Taken together, the demonstrated relationship between altered ARF and KAN gene expression in drb1ago7, drb2ago7 and drb4ago7 flowers, could, in part, explain the more severe developmental defects displayed by the double mutants, compared to milder impact that loss of only a single piece of TAS3 pathway protein machinery was demonstrated to have on drb1, drb2, drb4 and ago7 reproductive development.

1. Introduction

In Arabidopsis thaliana (L.) Heynh. (Arabidopsis), the complex and hierarchical regulatory networks that tightly control gene expression to underpin plant growth and development are still being elucidated. A central regulator of plant gene expression has however been identified in the form of small RNAs (sRNAs), with plant sRNAs being further divided into two primary species, the microRNAs (miRNAs) and the small-interfering RNAs (siRNAs) [1,2,3,4]. miRNAs are short (approximately 21 to 22 nucleotides (nt) in length), non-protein-coding RNAs that posttranscriptionally regulate the expression of a small number of closely-related target genes via directed binding of the miRNA-loaded, RNA-induced silencing complex (miRISC) to their target transcripts at regions of near perfect sequence complementarity [2,5]. miRISC contains the ARGONAUTE1 (AGO1) endonuclease at its catalytic core, and, once bound to the target transcript, miRISC initiates either a mRNA cleavage or translation repression mode of RNA silencing to regulate the expression of the targeted gene [1,5,6,7].

In plants, the siRNA species can be further divided into three main subclasses, including the trans-acting siRNA (tasiRNA), natural antisense transcript siRNA (natsiRNA), and repeat-associated siRNA (rasiRNA) subclasses [8,9,10,11,12]. The tasiRNA subclass of siRNA has been demonstrated to play an integral role in plant growth and organ development via regulation of key hormone pathways [13,14,15,16]. In Arabidopsis, four TAS gene families have been identified, specifically the TAS1 to TAS4 gene families, from which tasiRNA sRNAs are liberated via distinct production pathways [12,17,18,19]. The production of TAS1, TAS2, and TAS4 derived tasiRNAs is reliant on the association of a 22-nt triggering miRNA, with miR173 triggering TAS1 and TAS2 tasiRNA formation, and miR828 initiating the production of tasiRNAs from the TAS4 precursor [17,20,21]. More specifically, miR173 and miR828 are loaded into AGO1 post their production to direct the binding of miRISC at a single target site harbored in the 5′ region of the TAS1, TAS2, or TAS4 precursor transcripts [17,20]. AGO1 catalyzed cleavage of the TAS transcript triggers double-stranded RNA (dsRNA) synthesis downstream from the point of cleavage, a process mediated by SUPPRESSOR OF GENE SILENCING3 (SGS3) and RNA-DEPENDENT RNA POLYMERASE6 (RDR6) [17,22]. The resulting TAS1, TAS2, or TAS4-specific dsRNA substrate is then processed into a series of perfectly phased tasiRNA/tasiRNA* duplexes by the dsRNA BINDING4/DICER-LIKE4 (DRB4/DCL4) functional partnership [23,24], and post duplex strand separation, select tasiRNA sRNAs are loaded into AGO1-catalyzed RISC to direct target gene expression repression in trans.

The TAS3 pathway is unique among the Arabidopsis TAS pathways in that it contains features not observed in the TAS1, TAS2 or TAS4 pathways. Namely, the TAS3 precursor transcript, targeted by miR390, harbors two miR390 binding sites, termed a ‘two-hit trigger’, with both target sites necessary for efficient production of TAS3-derived tasiRNAs. The two-hit trigger mechanism for TAS3 tasiRNA production suggests more tightly controlled regulation of both the production of the tasiRNA sRNA itself, and of the regulation of the expression of the target gene transcripts of the TAS3 pathway [25]. In addition, initiation of TAS3 tasiRNA production requires an alternate AGO effector protein to AGO1, specifically AGO7 [14,22,25]. Once loaded with the miR390 sRNA (a 21-nt miRNA), AGO7 directs cleavage of the TAS3 transcript at only the 3′ miR390 target site [14,22,25]. The miR390/AGO7 complex then binds the TAS3 transcript at the 5′ binding site, a target site that harbors a single mismatched miR390/TAS3 base-pairing, and two G:U wobble pairings, between nucleotides 9 and 11 of the miR390 sRNA [22,25]. The resulting ‘bulge’ of the miR390/TAS3 hybrid formed is proposed to inhibit miR390-directed AGO7-catalyzed cleavage at the 5′ target site, as well as aiding in transcript stabilization for efficient SGS3/RDR6 recruitment, and subsequent TAS3-specific dsRNA synthesis upstream of the 3′ cleavage site [18,20,25,26,27,28]. As for the dsRNA substrates synthesized by the SGS3/RDR6 functional partnership from the TAS1, TAS2, and TAS4 precursor templates, the TAS3-specific dsRNA is then processed by the DRB4/DCL4 functional partnership to generate a series of precisely phased, 21-nt long, tasiRNA/tasiRNA* duplexes [16,23,24]. Post duplex strand separation, a single TAS3-derived tasiRNA, termed tasiARF, is loaded into AGO1-catalyzed RISC to direct posttranscriptional regulation of the expression of three AUXIN RESPONSE FACTOR (ARF) transcription factors, specifically the expression regulation of the ARF2, ARF3, and ARF4 transcripts [13,14,16,23,24,29].

The precise level and exact spatial distribution of the activity of each ARF transcription factor is essential to ensure normal Arabidopsis growth and development, with each ARF reprogramming the transcriptional response of large gene cohorts to auxin [30,31,32]. Therefore, the distribution and activity of each ARF is tightly controlled at both the transcriptional and posttranscriptional level [30,33]. For the three ARF transcription factors under tasiARF-directed expression regulation, ARF2 is required to control leaf senescence via its repression of auxin signaling in mature leaves, and has also been implicated in mediating the cross-talk between the auxin and abscisic acid (ABA) signaling pathways [34,35]. ARF3 (also termed ETTIN) and ARF4, however, are required to mediate auxin-directed determination of abaxial/adaxial polarity in rosette leaves, as well as ensuring the correct architectural formation of flowers, in addition to being required for floral organogenesis, with mutations to either the ARF3 or ARF4 locus resulting in vegetative and/or reproductive tissue organization defects [13,14,15]. Furthermore, aberrant ARF2, ARF3, and ARF4 gene expression has been demonstrated to result in morphological defects to rosette development and gynoecium architecture; for the gynoecium, the primarily alteration is to the length of the pistil, a phenotypic change that subsequently impacts fertility [15,36,37,38]. In wild-type Arabidopsis, tasiARF-directed AGO1 cleavage of the ARF3 and ARF4 transcripts regulates the localization, and therefore the activity of these two ARFs, to direct normal tissue patterning [39,40]. Members of the KANADI (KAN) clade of the GARP (Golden2, ARR-B, Psr1) transcription factor family have been similarly reported to be involved in the determination of floral organ symmetry, abaxial/adaxial leaf polarity, and vascular bundle patterning via their interaction with auxin and with ARF transcription factors. As such, the KANs have been identified as candidate regulators of reproductive organ development in Arabidopsis [41,42,43,44].

Given the importance of ARF-mediated responses to auxin throughout Arabidopsis development, the involvement of members of the DRB protein family in controlling the TAS3 pathway requires further characterization. In Arabidopsis, DRB family members, DRB1, DRB2, and DRB4, have all been demonstrated to be required for miRNA and/or tasiRNA production [13,45,46,47,48]. However, post sRNA production, little is known of the roles that these three DRBs play in directing the sRNA to a specific AGO protein in order to direct the silencing fate of the target genes of the AGO-loaded sRNA. The TAS3 pathway therefore provides a unique opportunity to build on our current knowledge base of the requirement of DRB1, DRB2, or DRB4 for the production of the miR390 sRNA as well as the mode of RNA silencing directed by the miR390/AGO7 complex for tasiARF production, and, subsequently, the tasiARF/AGO1 complex to regulate ARF2, ARF3, and ARF4 gene expression.

Here we show that the developmental abnormalities previously reported for the Arabidopsis plant lines defective in the activity of DRB1, DRB2 or DRB4 (the drb1, drb2 and drb4 single knockout mutants, respectively), are further exacerbated by the addition of the ago7 loss-of-function mutation to these genetic backgrounds. More specifically, rosette area, silique length, and seed set are all reduced to a greater degree in the drb1ago7, drb2ago7 and drb4ago7 double mutants than observed in either the drb1, drb2, drb4 or ago7 single mutant lines, a finding that clearly revealed the additive effect of combining the drb1, drb2 and drb4 single mutations with the ago7 mutation. At the molecular level, reverse transcriptase quantitative polymerase chain reaction (RT-qPCR) assessment of miR390 precursor transcript abundance revealed that, in Arabidopsis floral tissues, DRB1 is the sole DRB required for miR390 sRNA production. However, DRB2 and DRB4 appeared to play secondary roles at this stage of the TAS3 pathway to ensure that miR390 levels are tightly maintained. RT-qPCR was further applied to additionally reveal altered ARF2, ARF3 and ARF4 expression in the floral tissues of drb1ago7, drb2ago7, and drb4ago7 plants. Furthermore, altered ARF gene expression in double mutant flowers was subsequently determined to result in reduced KAN1 and KAN3 expression, and elevated KAN4 expression. Taken together, the demonstrated relationship between altered ARF2, ARF3, and ARF4 expression, and that of the KAN1, KAN3, and KAN4 transcripts, could, in part, explain the more severe fertility defects displayed by the drb1ago7, drb2ago7 and drb4ago7 double mutants, compared to the milder negative impact on reproductive development associated with the loss of activity of only a single piece of TAS3 pathway protein machinery in the drb1, drb2, drb4, and ago7 single mutant backgrounds.

2. Materials and Methods

2.1. Plant Material

The four T-DNA insertion knockout mutant lines used in this study, including the drb1 (drb1-1; SALK_064863), drb2 (drb2-1; GABI_348A09), drb4 (drb4-1; SALK_000736), and the ago7 (ago7-1; SALK_037458) mutant lines have been described previously [13,49]. The double knockout mutant lines, drb1ago7, drb2ago7, and drb4ago7, were generated for use in this study via a standard genetic crossing approach. Homozygosity of each of the two mutant alleles harbored by the drb1ago7, drb2ago7, and drb4ago7 double mutants was confirmed via a standard PCR-based genotyping approach prior to any experimental analyses being performed on these newly generated Arabidopsis lines. The seeds of wild-type Arabidopsis (ecotype: Columbia-0 (Col-0)), and the seven mutant lines were surface sterilized using chlorine gas. Post sterilization, seeds were plated onto standard Arabidopsis plant growth media (half strength Murashige and Skoog (MS) salts), and post plating, seeds were stratified in the dark at 4 °C for 48 h (h). Post stratification, plates were transferred to a temperature-controlled growth cabinet (A1000 Growth Chamber, Conviron^®, Melbourne, VIC, Australia) and cultivated for a 14-day period under a 16/8 h light/dark cycle, and a day/night temperature of 22 °C /18 °C. Following this cultivation period, equal numbers of seedlings from each genetic background were transferred to 50 × 50 mm pots that contained a standard Arabidopsis soil mixture (Seeds & Cuttings Mix, Debco, Sydney, NSW, Australia) and were cultivated for an additional 14-day period, under standard growth conditions (16/8 h light/dark cycle; 22 °C /18 °C day/night temperature). At the completion of the 14-day growth period on soil, rosette assessments were completed on 28-day-old plants, while floral architecture (including molecular assessments) and silique assessments were completed on 42-day-old plants.

2.2. Phenotypic and Physiological Assessments

The rosette area of 28-day-old seedlings was determined for each mutant background to establish the effect of loss of DRB1, DRB2, DRB4 and/or AGO7 function on vegetative development. This quantitative assessment was completed using ImageJ software (Version 1.52q, National Institutes of Health, Bethesda, MD, United States) to trace the rosette leaf blade margins and the petioles of each rosette leaf to determine the total area of the rosette. At the time that this assessment was performed, Col-0 plants and the 7 knockout mutant lines had developed 9 or 10 true leaves in addition to the cotyledons, and all leaves were used in the rosette area calculations. Furthermore, 4 biological replicates of 6 plants per replicate were used to determine the rosette area of each plant line. The length and seed set of siliques of 42-day-old plants was assessed by firstly measuring the length of each silique before gently breaking open each silique sampled from along the entire length of the primary inflorescence stem and manually counting the number of seeds that each silique contained. As stated for the determination of the rosette area of each plant line, 4 biological replicates of 6 plants per replicate were used in the silique length and seed number assessments.

2.3. Total RNA Extraction and Molecular Analysis

For each molecular assessment reported here, total RNA was extracted from 4 biological replicates of 42-day-old Col-0, drb1, drb2, drb4, ago7, ago7drb1, ago7drb2, and ago7drb4 plants using TRIzol^TM Reagent according to the manufacturer’s instructions (Invitrogen^TM, Melbourne, VIC, Australia). Furthermore, each biological replicate contained the terminal floral buds sampled from 8 individual plants. The quality of the extracted RNA was visually assessed via a standard electrophoresis approach on a 1.2% (w/v) ethidium bromide stained agarose gel, and the quantity of each RNA extract determined using a NanoDrop spectrophotometer (NanoDrop^® ND-1000, Thermo Scientific, Sydney, NSW, Australia). Poor quality or low yielding samples were discarded and the extraction procedure repeated until 4 high quality and highly concentrated RNA preparations were obtained for the each of the eight plant lines under assessment.

To synthesize the miR390- or tasiARF-specific complementary DNA (cDNA), 200 nanograms (ng) of total RNA was treated with 0.2 units (U) of DNase I according to the manufacturer’s instructions (New England Biolabs, Brisbane, QLD, Australia). The DNase I-treated total RNA was then directly used as the template for the synthesis of sRNA-specific cDNAs with 1.0 U of ProtoScript^® II Reverse Transcriptase (New England Biolabs, Brisbane, QLD, Australia), and the cycling conditions of 1 cycle of 16 °C for 30 min; 60 cycles of 30 °C for 30 s, 42 °C for 30 s, and 50 °C for 2 s; and 1 cycle of 85 °C for 5 min.

A global high molecular weight cDNA library preparation to quantify gene expression was constructed via the initial treatment of 5.0 micrograms (µg) of total RNA with 5.0 U of DNase I according to the manufacturer’s protocol (New England Biolabs, Brisbane, QLD, Australia). The DNase I-treated total RNA was subsequently purified using an RNeasy Mini Kit according to the manufacturer’s protocol (Qiagen, Melbourne, VIC, Australia). One microgram of this preparation was used as template for cDNA synthesis along with 1.0 U of ProtoScript^® II Reverse Transcriptase and 2.5 millimolar (mM) of oligo dT₍₁₈₎, according to the manufacturer’s instructions (New England Biolabs, Brisbane, QLD, Australia).

All generated, single-stranded cDNAs were next diluted to a working concentration of 50 ng/µL in RNase-free H₂O prior to their use as a template for the quantification of the abundance of either the miR390 or tasiARF sRNA, or the expression of either the miR390 or tasiRNA precursor transcripts, or the ARF or KAN mRNAs. In addition, all RT-qPCRs used the same cycling conditions of 1 cycle of 95 °C for 10 min, followed by 45 cycles of 95 °C for 10 s and 60 °C for 15 s. The GoTaq^® qPCR Master Mix (Promega, Sydney, NSW, Australia) was used as the fluorescent reagent for all performed RT-qPCR experiments. Small RNA abundance and gene expression was quantified using the 2^−ΔΔCT method with the small nucleolar RNA, snoR101, and UBIQUITIN10 (UBI10; AT4G05320) used as the respective internal controls to normalize the relative abundance of each assessed RNA molecule. The sequence of each DNA oligonucleotide used in this study for either the synthesis of a sRNA-specific cDNA, or to quantify either sRNA abundance or gene expression is provided in Supplemental Table S1.

2.4. Statistical Analysis

A standard Student’s t-test was used to determine the degree of significance of phenotypic (Figure 1B) or molecular (Figures 3–5) variance for each of the seven mutant lines analyzed in this study compared to wild-type Arabidopsis.

3. Results

3.1. Arabidopsis Plant Lines Defective in DRB1, DRB2, DRB4 and AGO7 Activity Express Phenotypic Defects during Vegetative and Reproductive Development

To investigate functional interactions between the DRB family members, DRB1, DRB2, and DRB4, and the TAS3 pathway associated AGO7, and to further study the effect that the loss of their function has on Arabidopsis vegetative and reproductive development, the phenotypes displayed by the drb1, drb2, drb4, and ago7 single mutants were initially assessed for comparison to wild-type Arabidopsis (Figure 1). In addition, and in an attempt to assign further functional roles to DRB1, DRB2, DRB4, and AGO7 in the TAS3 pathway, the developmental phenotypes expressed by three previously undescribed Arabidopsis mutant lines, the drb1ago7, drb2ago7, and drb4ago7 double mutants, were also included in this assessment.

As previously described, the severe developmental phenotype of the drb1 rosette is characterized by an overall reduction in size [48,49,50,51]. The smaller sized rosette of drb1 plants is due to this mutant producing smaller sized leaves that have upwardly curled (hyponastic) margins, and that have formed on petioles of reduced length (Figure 1A). Comparatively, the vegetative phenotypes displayed by the drb2 and drb4 single mutants are mild [13,24,45,49]. Specifically, the rosette leaves of drb2 plants are of an equivalent size to those of Col-0 plants but are ovate in shape, with mild margin serration, and which have formed on extended petioles (Figure 1A). Similarly, although the rosette leaves of drb4 plants have a lanceolate shape, and which have formed on petioles of a longer length, drb4 rosette leaves are of an equivalent size to those of Col-0 plants. Figure 1A also shows that the vegetative tissues of the drb4 single mutant display a pale, light green to yellowish coloration compared to the darker green coloration of the vegetative tissues of wild-type Arabidopsis.

The vegetative phenotype displayed by the ago7 single mutant is also mild in comparison to the severe developmental phenotype expressed by the drb1 single mutant. Like the rosette leaves of drb4 plants, ago7 rosette leaves are lanceolate in shape and are equivalent in size to Col-0 rosette leaves, if not becoming slightly larger in size upon maturity. In addition, ago7 vegetative tissue has a pale green, brownish-to-yellowish coloration. Furthermore, and as reported for drb2 plants, the margins of ago7 rosette leaves develop a mild degree of serration. However, unlike the drb2 and drb4 single mutants, ago7 rosette leaf petioles are not increased in length (Figure 1A). The expression of similar phenotypic characteristics by the drb2, drb4, and ago7 mutant lines is not a surprise observation considering that DRB2, DRB4, and AGO7 have all been demonstrated to play a role in regulating tasiARF sRNA production from the TAS3 precursor previously [11,13,47].

In order to further establish the degree of involvement of DRB1, DRB2, DRB4, and AGO7 in the TAS3 pathway, the drb1ago7, drb2ago7, and drb4ago7 double mutants were next analyzed, with these three double mutants generated via a standard genetic crossing approach. Compared to their single mutant counterparts, the additive effect of different combinations of individual drb mutations together with the ago7 mutation, is readily apparent (Figure 1A). Of the seven mutant lines assessed in this study, the drb1ago7 double mutant displayed the most severe vegetative phenotype. The drb1ago7 double mutant expressed all of the hallmarks of loss of DRB1 function during vegetative development, including smaller sized and hyponastic rosette leaves on petioles of shorter length, but to a greater degree than expressed by Arabidopsis plants where only DRB1 activity is defective (Figure 1A). More specifically, compared to a rosette area of 65.7 mm² for Col-0 plants, the rosette area of the drb1 mutant was reduced by 23.4%, with a further 17.4% reduction determined for the drb1ago7 double mutant; an overall 40.8% reduction in rosette area compared to Col-0 plants (Figure 1B).

Figure 1 also shows that the addition of the ago7 mutation to the drb2 background severely impaired the vegetative development of drb2ago7 plants with the double mutant developing a rosette that was reduced in area by 25.9% compared to Col-0 rosettes (Figure 1B), and a 31.1% and 23.2% reduction in total area compared to drb2 and ago7 rosettes, respectively. In addition to being reduced in size, the rosette leaves of drb2ago7 plants adopted the lanceolate shape of ago7 leaves, and, furthermore, the margins of these leaves displayed a slightly higher degree of serration than the margins of rosette leaves of either drb2 or ago7 plants (Figure 1A). Curiously, of the three double mutant lines generated in this study, addition of the ago7 mutation to the drb4 background ‘visually’ appeared to effect vegetative development of the resulting double mutant to the lowest degree. However, quantification of the rosette area of the drb4ago7 double mutant revealed a 24.7% reduction in area compared to Col-0 plants, a similar degree of reduction to rosette area as determined for the drb2ago7 double mutant. This analysis further revealed that the rosette area of the drb4ago7 double mutant was reduced by 15.2% and 21.9% compared to the rosette areas of 58.4 and 63.4 mm² for the drb4 and ago7 single mutants, respectively (Figure 1B). Figure 1A additionally shows that although drb4ago7 rosette leaves maintained the lanceolate shape of drb4 and ago7 rosette leaves, the area of the blade of drb4ago7 rosette leaves was reduced; however, the length of rosette leaf petioles was increased. The degree of drb4ago7 rosette leaf margin serration also remained similar to that of the ago7 single mutant.

As observed for the vegetative development of the four single mutants analyzed in this study, reproductive development was negatively impacted to the greatest degree by loss of DRB1 function. Specifically, compared to Col-0 flowers, drb1 flowers exhibited an elongated shape due to the narrowing of both the sepals and petals (Figure 1C). In addition, the stigma of the pistil of drb1 flowers extended past the stamens; the anther filaments of drb1 stamens were shorter in length and the pollen sacs greatly reduced in size. Figure 1D clearly shows that together, these defects in drb1 flower morphology, in turn negatively impacted drb1 fertility, as evidenced via the formation of siliques that were reduced in their length (to different degrees) along the length of the drb1 primary inflorescence. Although drb2 flowers were equivalent in size to those of Col-0 plants, drb2 sepal shape was altered. Specifically, compared to Col-0 sepals, drb2 sepals had a broader width and a more rounded distal tip (Figure 1C). In addition, due to their increased width, drb2 sepals overlapped one another. However, the overlapping sepals of drb2 flowers housed petals of an equivalent size and shape to Col-0 petals. Figure 1C further shows that the pistil of drb2 flowers was elongated, extending the stigma past the pollen sacs of morphologically normal stamens. Stigma exsertion in drb2 flowers had a very mild impact on drb2 fertility, with usually only the first one to three siliques that formed on the proximal end of the primary inflorescence failing to expand to the normal length of the corresponding siliques that formed on the primary inflorescence of wild-type plants (Figure 1D).

While drb4 flowers were of an equivalent size to Col-0 flowers, drb4 sepals had a more upright orientation than the sepals of Col-0 flowers, resulting in the flowers of the drb4 mutant forming a more compact overall morphology (Figure 1C). The upright orientation of drb4 sepals resulted in the petals of this mutant background to also adopt a more vertical orientation than the petals of Col-0 flowers; Col-0 petals are of an equivalent size to drb4 petals but bend outwards and lay horizontally at their distal tip (Figure 1C). The petals of drb4 flowers also adopted an overall obcordate shape characterized by the formation of a single indentation in the middle of their margin to form a heart shape at their distil tip. In comparison, Col-0 petals display uniform curvature of the margin of the distil tip. Interestingly, the drb4 pistil was equivalent in length to the pistil of Col-0 flowers; however, the anther filaments of drb4 stamens were longer than those of Col-0 flowers that resulted in the pollen sacs of drb4 stamens extending slightly past the stigma of the drb4 pistil (Figure 1C). However, this mild alteration to stamen development did not appear to negatively impact drb4 fertility to any great degree, with the siliques that formed along the length of the drb4 primary inflorescence being of a near equivalent size and shape to Col-0 siliques (Figure 1D). Phenotypically, the only consequence of the loss of AGO7 activity on Arabidopsis reproductive development is stigma exsertion, with the sepals, petals, and stamens of ago7 flowers all of normal morphological appearance (Figure 1C). When compared to the drb2 single mutant, the degree of stigma exsertion was slightly more pronounced in ago7 flowers. Accordingly, fertility was impacted to a greater degree in the ago7 single mutant than in drb2 plants. Specifically, the development of the first five to six siliques that formed from the proximal region of the ago7 primary inflorescence were negatively impacted by the loss of AGO7 activity, as evidenced by their failure to expand (Figure 1D).

As reported for vegetative development, addition of the ago7 mutation to the drb1, drb2, and drb4 backgrounds, had a further negative impact on the reproductive development of the three double mutants analyzed in this study (Figure 1). The flowers of the drb1ago7 double mutant were reduced in their overall size; however, the degree of reduction was similar in the double mutant, compared to the drb1 single mutant. Interestingly, the sepals of drb1ago7 flowers were wider than drb1 sepals and had adopted an overall shape that was more similar to the shape of sepals of either ago7 or Col-0 plants (Figure 1C). It was therefore curious to observe that the petals of drb1ago7 flowers retained the shape of the petals of drb1 flowers. Considering that both drb1 and ago7 display stigma exsertion, it was unsurprising to observe that drb1ago7 flowers also expressed this phenotype (Figure 1C). However, the degree of stigma exsertion in drb1ago7 flowers did not appear to be any more pronounced in the double mutant than in either of the single mutant backgrounds. The additive effect of combining the drb1 and ago7 mutations together to generate the drb1ago7 double mutant was, however, readily evident via the formation of siliques that were further reduced in length, compared to drb1 siliques, combined with an increased frequency of siliques along the length of the drb1ago7 primary inflorescence that failed to expand at all, post their initial formation (Figure 1D); a phenotype that strongly suggested that fertility was completely dysfunctional in the drb1ago7 flowers from which these unexpanded siliques had formed.

The drb2ago7 double mutant developed flowers that were morphologically similar to the flowers of the drb2 single mutant. The sepals, petals, and stamens of drb2ago7 flowers all adopted the phenotypic characteristics of the flowers of the drb2 single mutant, a finding that indicated that the drb2 loss-of-function phenotype had a greater degree of penetrance in Arabidopsis reproductive tissues than the ago7 loss-of-function mutation (Figure 1C). However, although drb2ago7 sepals retained the size and shape of drb2 sepals, the sepals of drb2ago7 flowers were not overlapping, a clear phenotypic distinction between drb2ago7 and drb2 flowers. A further difference in the development of drb2, ago7, and drb2ago7 flowers was that the drb2ago7 stigma not only displayed exsertion but was increased in its width, as was the width of the style of the drb2ago7 pistil (Figure 1C). This phenotypic distinction of the double mutant to either the drb2 or ago7 single mutant, could, in part, account for the readily apparent additive effect on Arabidopsis reproductive development that resulted from combining these two mutations together, with many more drb2ago7 siliques being reduced in length, and/or failing to expand, compared to the siliques that formed on the primary inflorescence of either the drb2 or ago7 single mutant (Figure 1D). As noted for the drb1ago7 double mutant, the failure of these drb2ago7 siliques to expand post their formation strongly suggested that the reproductive competence of the drb2ago7 flowers from which these siliques had formed was completely compromised.

The sepals and petals of drb4ago7 flowers were of an equivalent size and shape to those of Col-0 flowers, however, as observed for drb4 flowers, drb4ago7 sepals had a more upright orientation, which in turn orientated the petals of drb4ago7 flowers vertically (Figure 1C). As reported for drb4 petals, the margin of the distal tip of drb4ago7 petals formed a distinct single indentation, a phenotypic alteration that gave drb4ago7 petals an obcordate shape. Interestingly, the pistil and stamens of the drb4ago7 double mutant extended to the same degree, despite the stamens of the drb4 mutant extending past the stigma in drb4 flowers and the ago7 single mutant displaying stigma exsertion (Figure 1C). However, the fertility of the drb4ago7 double mutant was clearly compromised to a greater degree than the fertility of either the drb4 or ago7 single mutants. Specifically, almost all of the siliques that formed on the primary inflorescence of the drb4ago7 double mutant were reduced in length, although the degree of reduction in silique length was dependent on the position of the silique on the drb4ago7 primary inflorescence (Figure 1D). In addition, a small number (1 or 2) of drb4ago7 siliques failed to expand, and these siliques were always observed to form from the early flowers that initially formed on the proximal region of the drb4ago7 primary inflorescence (Figure 1D).

3.2. The Reproductive Competence of the Double Mutant Plant Lines, drb1ago7, drb2ago7, and drb4ago7, Was Severely Compromised

The readily observable alterations to drb1ago7, drb2ago7, and drb4ago7 flower and silique morphology inferred that the most significant negative impact resulting from the addition of the ago7 mutation to the drb1, drb2, and drb4 backgrounds, was on reproductive development (Figure 1). Therefore, silique length and the number of seeds that each silique contained was quantified. Figure 2A shows that, upon maturity, 60% of Col-0 siliques were either 12 (26%) or 13 mm (34%) in length with the majority, 86%, of these consistently sized siliques harboring more than 40 seeds. Compared to Col-0 plants, the fertility of the drb1 mutant is clearly compromised with 80% of mature siliques measuring either 5 (20%), 6 (30%), or 7 mm (30%) in length. Furthermore, the short siliques that formed on the primary inflorescence of the drb1 mutant were determined to house a greatly reduced number of seeds, specifically; 67% of siliques contained 11–20 seeds, and 23% of drb1 siliques only harbored 1 to 10 seeds (Figure 2B). The severe retardation of drb1 reproductive development has previously been attributed to the impairment of miRNA production in the absence of DRB1 function [52,53].

Figure 2C shows the mild impact that loss of DRB2 function had on Arabidopsis reproductive development with the majority of drb2 siliques measuring 10 (17%), 11 (29%), or 12 mm (25%) in length. Although the siliques that formed on the primary inflorescence of the drb2 mutant were slightly reduced in length, the majority of these siliques (80%) were determined to harbor more than 40 seeds. However, a distinct population of siliques that were not observed in Col-0 plants was evident for the drb2 mutant (Figure 2C). Namely, 4% of drb2 siliques measured 6 or 7 mm in length and housed 11 to 20 seeds, a silique length and silique seed number not detected in wild-type Arabidopsis (Figure 2A). Loss of DRB4 function was determined to have a similar mild degree of negative impact on the reproductive competence of Arabidopsis as evidenced by the majority of drb4 siliques (73%) measuring 10 (30%) or 11 mm (43%) in length (Figure 2D). As reported for the drb2 mutant, although silique length was reduced in the drb4 mutant, the majority of siliques (79%) were determined to house more than 40 seeds (Figure 2D). A further similarity between the loss of DRB2 and DRB4 function was noted for the drb4 background; that is, the development of a silique population distinct to that of Col-0 plants. More specifically, 1% and 8% of siliques elongated to 7 and 9 mm in length, respectively, and these siliques were determined to contain 11 to 20 seeds (Figure 2D).

Silique length varied widely in the ago7 mutant background, ranging in size from 6 (4%) to 15 mm (1%). However, as determined for Col-0 plants, the majority of ago7 siliques were either 12 (20%) or 13 mm (22%) in length (Figure 2E). The variable, yet overall reduced length of ago7 siliques was further supported by the finding that only 12% of ago7 siliques harbored more than 40 seeds, with the majority of ago7 siliques (65%) determined to contain 31–40 seeds. This finding is in direct contrast to the Col-0 finding that 86% of siliques harbored more than 40 seeds and that only 10% of Col-0 siliques harbored 31–40 seeds (Figure 2E). When taken together, silique length and seed number per silique readily indicated that, although the reproductive competence of the ago7 mutant was only mildly impacted, the degree of this impact remained uniform throughout reproductive development of the ago7 single mutant. The additive negative effect on the reproductive competence of Arabidopsis resulting from combining the drb1 and ago7 mutant backgrounds was most readily apparent via the comparison of silique length and seed number per silique between the drb1 single and drb1ago7 double mutants. That is, 20% and 30% of drb1 siliques were 5 and 6 mm in length, respectively; however, for the drb1ago7 double mutant, 30% and 37% of siliques were 5 and 6 mm in length (Figure 2B,F). Correspondingly, 23% and 67% of drb1 siliques contained 1–10 and 11–20 seeds, respectively (Figure 2B). In the double mutant, however, 36% of siliques were determined to house 1–10 seeds and 61% of drb1ago7 siliques contained 11–20 seeds (Figure 2F). The 13% increase in the number of siliques that harbored 1–10 seeds for the drb1ago7 line, compared to those of the drb1 single mutant, was clearly due to only 3% of drb1ago7 siliques containing 21–30 seeds, compared to 10% of drb1 siliques housing the same seed number (Figure 2B,F).

The added negative impact on the reproductive competence of the drb2ago7 double mutant (Figure 2G) was again readily evidenced via comparison of the silique length and seed number metrics determined for the drb2 single mutant with those determined for the double mutant (Figure 2C). More specifically, 35% of drb2ago7 siliques were 4 to 9 mm in length compared to only 11% of drb2 siliques that elongated to a length of 6 to 9 mm (Figure 2C,G). It is important to note here that siliques of 4 or 5 mm in length that formed on the primary inflorescence of the double mutant were not observed for the drb2 mutant, nor did they form on the primary inflorescence of the ago7 mutant (Figure 2C,E,G). In addition, 44% of drb2ago7 siliques were determined to house between 1 and 30 seeds (Figure 2G) compared to only 20% of drb2 siliques housing a corresponding seed number (Figure 2C). It is again important to note here that, while 3% of drb2ago7 siliques were determined to harbor 1 to 10 seeds (Figure 2G), no siliques harboring this low seed number were observed for the drb2 mutant background (Figure 2C). Compared to the drb4 single mutant, the drb4ago7 double mutant also exhibited a considerable reduction in fertility (Figure 2D,H). Namely, 31% of drb4ago7 siliques were 5 to 9 mm in length (Figure 2H) compared to only 9% of drb4 siliques that elongated to either 7 (1%) or 9 (8%) mm in length (Figure 2D). Reduced silique length in this double mutant background was again demonstrated to result in a reduction to the number of seeds housed per silique, with 31% of drb4ago7 siliques determined to harbor between 1 and 30 seeds (Figure 2H). In comparison, only 6% of drb4 siliques contained a corresponding number of seeds (Figure 2D). This comparative analysis additionally revealed that, while 79% of drb4 siliques were determined to house over 40 seeds, only 7% of the siliques of the drb4ago7 double mutant contained more than 40 seeds (Figure 2D,H). Taken together, the Figure 2 analyses clearly revealed the additive negative impact that combining the ago7 mutation to the drb1, drb2, and drb4 mutant backgrounds had on Arabidopsis reproductive development.

3.3. DRB1 Is Required for miR390 Production in Arabidopsis Floral Tissues

The previous demonstration that DRB1, DRB2, DRB4 and AGO7 each play functional roles in either miRNA or tasiRNA production and/or action [11,13,22,25,45,46,47], led us to next molecularly profile the TAS3 pathway to attempt to associate TAS3 pathway dysfunction with the degree of severity of the reproductive phenotypes expressed by the seven mutant lines assessed in this study. In Arabidopsis, the ‘triggering’ miRNA of the TAS3 pathway, miR390, is processed from two distinct precursor transcripts, PRI-MIR390A and PRI-MIR390B [22]. Therefore, the expression of these two precursors was assessed via a RT-qPCR approach to determine the requirement of DRB1, DRB2, DRB4, and/or AGO7 for the; (1) processing of the miR390 sRNA from the PRI-MIR390A and PRI-MIR390B precursors, or; (2) regulation of miR390 abundance post the liberation of this sRNA from its precursors.

In the floral tissues of the assessed mutants, RT-qPCR revealed that the abundance of the miR390 precursor transcripts, PRI-MIR390A and PRI-MIR390B, remained at levels approximate to those of wild-type flowers in drb2, drb4, ago7, drb2ago7, and drb4ago7 flowers (Figure 3A,B). This analysis also clearly showed that PRI-MIR390A and PRI-MIR390B abundance was significantly elevated by 17.46- and 41.02-fold in drb1 flowers, and by 11.45- and 37.16-fold in drb1ago7 flowers, respectively (Figure 3A,B). Taken together, the RT-qPCR data presented in Figure 3A,B, clearly indicated that, in Arabidopsis flowers, DRB1 is the sole DRB protein required for miR390 precursor transcript processing. The requirement of DRB1 for efficient precursor transcript processing in Arabidopsis reproductive tissues is further enforced by the 73% reduction to miR390 abundance detected in drb1 flowers (Figure 3C). RT-qPCR further revealed that the miR390 sRNA remained at its wild-type levels in drb2 flowers but was reduced in abundance by 25% in drb4 flowers. Reduced miR390 abundance in drb4 flowers, in the absence of change in the expression level of either the PRI-MIR390A or PRI-MIR390B precursor (Figure 3A,B), suggests that, although DRB4 is not required for the production of the miR390 sRNA, DRB4 could potentially mediate a secondary regulatory role in the TAS3 pathway to ensure that miR390 sRNA abundance is correctly maintained. In ago7 flowers, miR390 abundance was demonstrated by RT-qPCR to be elevated by 1.47-fold (Figure 3C). Elevated miR390 sRNA abundance in this mutant background was not a surprise finding as more ‘free’ or ‘unbound’ miR390 would be expected to be present in the cell if not bound by AGO7 post its DRB1 (and DCL1)-mediated production.

A mild 10% reduction in miR390 abundance was detected in drb1ago7 flowers (Figure 3C). A more dramatic change to miR390 abundance could have been expected for the drb1ago7 double mutant considering that PRI-MIR390A and PRI-MIR390B abundance was elevated by 11.45- and 37.16-fold, respectively (Figure 3A,B). However, considering that miR390 abundance was reduced by 73% in drb1 flowers, yet elevated by almost 50% in ago7 flowers, a mild 10% reduction to miR390 sRNA abundance in drb1ago7 flowers was considered appropriate. Surprisingly, in the absence of altered PRI-MIR390A (Figure 3A) or PRI-MIR390B (Figure 3B) expression, miR390 abundance was revealed to be elevated by 2.38- and 3.48-fold in drb2ago7 and drb4ago7 flowers, respectively (Figure 3C). This curious result could potentially indicate that, although DRB2 and DRB4 are not required for the processing of the miR390 sRNA from its precursor transcripts, both DRB proteins mediate secondary roles in the TAS3 pathway to regulate the abundance of the miR390 sRNA in Arabidopsis flowers post its production to ensure tight control of its levels. The previous demonstration of antagonism between DRB1, DRB2, and DRB4 function [45,47] adds further weight to this suggestion; that is, the elevated miR390 abundance in drb2ago7 and drb4ago7 flowers was most likely the result of (1) more efficient miR390 production in the absence of DRB2- or DRB4-mediated antagonism of DRB1 function, and (2) a higher level of unbound miR390 being present in the cells of this tissue in the absence of the miR390 sRNA being loaded by AGO7.

3.4. tasiARF Target Gene Expression Is Altered in the Absence of DRB1, DRB2, DRB4, and AGO7 Activity

The quantification of miR390 abundance by RT-qPCR revealed that the accumulation of this miRNA remained at wild-type levels in drb2 and drb1ago7 flowers, was mildly altered in drb4 and ago7 flowers, and was significantly altered in the flowers of the drb1, drb2ago7, and drb4ago7 mutants (Figure 3C). The RT-qPCR approach was therefore next applied to determine if there was any correlation between the level of the triggering miRNA, miR390, and the expression of its target, the TAS3 precursor transcript. Considering that the miR390 sRNA directs AGO7-catalyzed cleavage of the TAS3 precursor at the 3′ target site of this transcript [14,22,25], a reduction in the abundance of the triggering miRNA would be expected to lead to target transcript over-accumulation. RT-qPCR analysis however revealed little correlation between the abundance of the TAS3 transcript and the level of the miR390 sRNA in the floral tissues of each assessed mutant. Taking the three plant lines where miR390 levels were determined to be significantly altered by −3.70- (drb1), 2.38- (drb2ago7), and 3.48-fold (drb4ago7), respectively, as an example (Figure 3C), the abundance of the TAS3 transcript was only mildly altered by 1.40-, 1.08- and −1.08-fold in drb1, drb2ago7, and drb4ago7 flowers (Figure 4A). A similar lack of correlation between miR390 levels and TAS3 abundance was also documented for drb2, drb4, ago7, and drb1ago7 flowers (Figure 3C, Figure 4A). Failure to establish a correlative link between the levels of the miR390 sRNA and the TAS3 target transcript in the floral tissues of the plant lines analyzed in this study is potentially explained by the significant difference in the abundance of these two RNA molecules. RT-qPCR indicated that, in Arabidopsis floral tissues, the miR390 sRNA is 100 times more abundant than the TAS3 target transcript (data not shown). Therefore, even in drb1 flowers where miR390 levels were determined to be reduced by the greatest degree (−3.70-fold), the miR390 sRNA would still be in considerable excess compared to the abundance of its targeted transcript, the TAS3 precursor. This largely biased relationship between the level of the targeting miRNA, and the abundance of the transcript targeted by this miRNA, provides a possible explanation for our inability to establish a correlation between miR390 levels and TAS3 abundance in the floral tissues of any of the mutant lines analyzed (Figure 3C, Figure 4A).

Although the RT-qPCR approach failed to establish correlation between miR390 sRNA levels and TAS3 transcript abundance, RT-qPCR did however reveal that the tasiARF sRNA was reduced in abundance by 8 (drb1 flowers) to 43% (drb4ago7 flowers) in the floral tissues of the seven mutant backgrounds analyzed (Figure 4B). This finding readily indicated that tasiARF sRNA production is disrupted to different degrees in the absence of the activity of the TAS3 pathway machinery proteins, DRB1, DRB2, DRB4, and AGO7. Reduced tasiARF abundance in drb1, drb2, drb4, ago7, drb1ago7, drb2ago7, and drb4ago7 flowers (Figure 4B) led us to next apply RT-qPCR to assess the level of expression of ARF2, ARF3 and ARF4, the three target transcripts of the tasiARF sRNA. The expression of ARF2 remained unchanged in drb1 flowers (Figure 4C). This was not an unexpected result considering that tasiARF levels were only reduced by 8% in drb1 flowers. In drb2 flowers where tasiARF levels were determined to be reduced by 25% (Figure 4B), RT-qPCR revealed ARF2 expression to also be reduced (Figure 4C), albeit to a lesser degree (14%). Considering that tasiRNA sRNAs have been reported to regulate the expression of their target genes in trans, via an AGO1-catalyzed mRNA cleavage mode of RNA silencing [48], reduced ARF2 expression in drb2 flowers where the levels of the tasiARF sRNA were also determined to be reduced, provided an unexpected result. However, a similar relationship between reduced tasiARF abundance (Figure 4B) and decreased ARF2 expression was also determined for drb4, ago7, drb1ago7, drb2ago7, and drb4ago7 flowers (Figure 4C). Figure 4C also clearly shows that ARF2 expression was reduced to the greatest degree in the three double mutants, with ARF2 transcript abundance decreased by 59%, 44% and 54% in drb1ago7, drb2ago7, and drb4ago7 flowers, respectively (Figure 4C).

The expression of the tasiARF target gene, ARF3, was determined to remain at approximate wild-type levels in the drb1, drb2 and drb4 single mutants (Figure 4D). However, ARF3 expression was elevated by 2.20-, 2.53-, 3.11-, and 1.69-fold in ago7, drb1ago7, drb2ago7, and drb4ago7 flowers, respectively; the four Arabidopsis lines analyzed in this study that harbored the ago7 mutation (Figure 4D). A more variable expression profile was constructed for the ARF4 target transcript across the floral tissues of the analyzed mutants. That is, ARF4 expression was reduced by 30% in drb4 flowers, mildly elevated in drb1 (1.47-fold), drb2 (1.35-fold), ago7 (1.84-fold), and drb4ago7 (1.51-fold) flowers, and significantly elevated by 3.88- and 2.55-fold in drb1ago7 and drb2ago7 flowers, respectively (Figure 4E). Considering the seven mutant lines analyzed, ARF4 expression was only determined to be reduced by 1.43-fold in drb4 flowers, and it was not a surprise observation that ARF4 expression was only mildly elevated by 1.51-fold in drb4ago7 flowers, compared to the elevated level of ARF4 upregulation documented for the floral tissues of the drb1ago7 and drb2ago7 double mutants (Figure 4E). When taken together, the Figure 4 expression analyses revealed a consistent trend for the three assessed double mutants, that is; in drb1ago7, drb2ago7, and drb4ago7 floral tissues, reduced tasiARF abundance led to repressed ARF2 expression and the enhanced expression of ARF3 and ARF4.

3.5. The Expression of the KANADI Transcription Factors, KAN1, KAN3, and KAN4 Is Altered in Arabidopsis Flowers Where tasiARF Target Gene Expression Is Changed

Previous research has demonstrated the involvement of members of the KAN clade of the GARP transcription factor family in the determination of floral organ symmetry via their interaction with auxin and with ARF transcription factors [41,42,43,44]. More specifically, ARF3 and ARF4 have been shown to repress the transcriptional activity of KAN transcription factor genes in Arabidopsis floral tissues [54,55]. Therefore, RT-qPCR was next applied to analyze the expression of the four KAN gene loci previously demonstrated to be expressed in Arabidopsis flowers [43,44]. Figure 5A readily shows that compared to Col-0 flowers, KAN1 expression was repressed in the floral tissues of each mutant background. Namely, RT-qPCR revealed KAN1 transcript abundance to be moderately reduced by 36% and 33% in drb1 and drb2 flowers respectively, and significantly decreased in abundance by 63%, 59%, 67%, 64%, and 58% in drb4, ago7, drb1ago7, drb2ago7, and drb4ago7 flowers, respectively (Figure 5A). It is important to note here that, although KAN1 expression was reduced in the flowers of each of the mutant backgrounds analyzed, the greatest degree of KAN1 expression repression was detected in the two mutant lines, the drb1ago7 and drb2ago7 double mutants, where ARF3 and ARF4 expression was determined to be most highly upregulated (Figure 4D,E). Unlike the uniform repressive influence on KAN1 expression that the loss of DRB1, DRB2, DRB4, and/or AGO7 function was observed to have (Figure 5A), the abundance of the KAN2 transcript was, in comparison, only mildly altered in the floral tissues of each mutant (Figure 5B). Specifically, KAN2 expression remained unchanged in ago7 flowers, was elevated by 18% in drb2 flowers, and was reduced by 10%, 10%, 32%, 21%, and 28% in drb1, drb4, drb1ago7, drb2ago7, and drb4ago7 flowers, respectively (Figure 5B). The further 22% and 18% reduction to KAN2 expression in drb1ago7 and drb4ago7 flowers, in addition to the documented 10% reduction to KAN2 expression in drb1 and drb4 flowers, respectively, formed an interesting finding. Namely, RT-qPCR revealed that ARF2 expression was most highly repressed in the same two double mutant backgrounds (Figure 4C). Taken together, this result tentatively identifies ARF2 as a putative positive regulator of KAN2 expression, however, further research is required to confirm whether the ARF2 transcription factor is a regulator of KAN2 gene expression in Arabidopsis flowers.

Compared to Col-0 flowers, KAN3 expression was only mildly altered in drb1 (elevated by 11%) and drb2 flowers (reduced by 14%; Figure 5C). In drb4, ago7, drb1ago7, drb2ago7, and drb4ago7 flowers, however, KAN3 expression was highly reduced by 44%, 50%, 43%, 57%, and 65%, respectively (Figure 5C). Comparison of KAN3 expression in the flowers of the three drb single mutant lines analyzed in this study, to its expression in the flowers of the three double mutant backgrounds assessed in parallel, provides a strong demonstration of the additive negative impact that combining the ago7 mutation with the drb1, drb2, and drb4 mutations had on the molecular phenotype expressed by each double mutant. That is, addition of the ago7 mutation to the drb1 mutant background resulted in KAN3 expression being repressed by 54% in drb1ago7 flowers, compared to its mildly elevated expression in drb1 flowers (Figure 5C). Similarly, the addition of the ago7 mutation to the drb2 and drb4 mutant backgrounds was demonstrated to cause a further 43% and 21% reduction to the abundance of the KAN3 transcript in drb2ago7 and drb4ago7 flowers, compared to the already reduced level of KAN3 expression detected in drb2 and drb4 flowers, respectively (Figure 5C). Unlike the overall general trend of reduced KAN1, KAN2, and KAN3 expression in single and double mutant flowers, RT-qPCR revealed that KAN4 expression was elevated in each assessed mutant (Figure 5D). More specifically, KAN4 expression was determined to be upregulated by between 1.71-fold (drb2ago7) and 2.19-fold (ago7) in the floral tissues of the seven assessed mutants. Curiously, although KAN4 expression was elevated by approximately 2.1-fold in drb2 and drb4 flowers, and by 2.19-fold in ago7 flowers, RT-qPCR revealed a lower degree of upregulated expression in drb2ago7 (1.71-fold) and drb4ago7 (1.78-fold) flowers, and not a greater degree of upregulated expression as was expected from addition of the ago7 mutation to either the drb2 or drb4 mutation (Figure 5D). Taken together, the RT-qPCR data presented in Figure 5 documents readily apparent trends in expression in single and double mutant floral tissues for three out of the four KAN transcription factor loci assessed, namely, reduced KAN1 and KAN3 expression, and upregulated KAN4 expression.

4. Discussion

Previous research has established the requirement of DRB1, DRB2, DRB4, and AGO7 for miRNA and/or tasiRNA production in Arabidopsis [11,13,22,25,45,46,47]. Therefore, this study aimed to shed additional light on the functional roles played by these three DRB proteins and by the AGO7 effector protein in the TAS3 pathway, specifically focusing on the; (1) requirement of DRB1 and/or DRB2 for miR390 production, the triggering miRNA of the TAS3 pathway; (2) requirement of DRB2, DRB4 and AGO7 for the production of the tasiARF sRNA from the TAS3 precursor, and; (3) mode of RNA silencing directed by the tasiARF sRNA to regulate the expression of the three target genes of the TAS3 pathway, ARF2, ARF3, and ARF4. Towards this goal, we generated three previously undescribed double mutant lines in an attempt to determine the added degree of TAS3 pathway dysfunction resulting from the loss of function of two pieces of pathway protein machinery versus the loss of activity of a single piece of pathway protein machinery.

The additive effect of combining the ago7 mutation with the drb1, drb2, or drb4 mutant backgrounds on Arabidopsis vegetative and reproductive development was readily apparent. During vegetative development, evidence of the negative impact of combining these different mutant backgrounds together was provided via comparison of the rosette area of drb1ago7, drb2ago7, and drb4ago7 plants to that of Col-0 plants (Figure 1A). Namely, the rosette area of the drb1ago7, drb2ago7, and drb4ago7 double mutants was reduced by 40.9%, 25.9%, and 24.7%, respectively, compared to the rosette area of wild-type Arabidopsis (Figure 1B). The lower degree of retardation to the vegetative development of the drb2ago7 and drb4ago7 double mutants compared to that of the drb1ago7 double mutant, could in part be accounted for by the different level of involvement of DRB1, DRB2, and DRB4 in miRNA production. That is, together with DCL1, DRB1 is required for the production of the majority of Arabidopsis miRNAs [46,48,50,56,57], whereas, comparatively, DRB2 and DRB4 have only been associated with the production of specific miRNA cohorts within the global population of Arabidopsis miRNAs [21,45,47,58,59].

Although vegetative development was clearly impaired in each assessed double mutant compared with either Col-0 plants, or to the respective single mutants, the impact to the reproductive phase of Arabidopsis development was much more striking. More specifically, the morphology of drb1ago7, drb2ago7, and drb4ago7 flowers was considerably altered compared to Col-0 flower morphology with each double mutant displaying a combination of architectural changes expressed by the flowers of the drb1, drb2, drb4, and ago7 single mutants (Figure 1C). Altered floral architecture was in turn demonstrated to have a severe negative impact on the reproductive competence of each double mutant via the formation of siliques on the primary inflorescence of the drb1ago7, drb2ago7, and drb4ago7 double mutants that were either greatly reduced in length upon maturity or that failed to elongate post their initial formation (Figure 1D). The significantly reduced fertility of the drb1ago7, drb2ago7, and drb4ago7 double mutants, compared to that of either Col-0 plants or the drb1, drb2, drb4, and ago7 single mutants, was further evidenced via the quantification of silique length and the number of seeds that each silique contained (Figure 2). Specifically, the three double mutants produced a population of siliques that were reduced in their fully expanded length, namely siliques 4 to 7 mm; a silique length that was not observed for Col-0 plants (Figure 2A,F–H). In addition, this silique population was determined to house between 1 to 10, or 11 to 20 seeds, silique seed numbers that were also not observed in wild-type Arabidopsis. Furthermore, although siliques of this size, and that harbored an equivalent number of seeds were observed in the drb1, drb2, drb4, and ago7 single mutant backgrounds, both of these metrics used to assess reproductive competence in this study were substantially increased in the drb1ago7, drb2ago7, and drb4ago7 double mutants (Figure 2).

The architectural alteration to floral tissue morphology displayed by drb1, drb2, ago7, drb1ago7, and drb2ago7 pistils, stigma exsertion, has been previously associated with aberrant expression of the three target genes of the TAS3 pathway, ARF2, ARF3, and ARF4 [15,18,36,37,38,60]. Considering that DRB1, DRB2, DRB4, and AGO7 have each been assigned functional roles in miRNA and/or tasiRNA production, the TAS3 pathway was next molecularly profiled via an RT-qPCR-based approach (Figure 3 and Figure 4) in an attempt to correlate TAS3 pathway disruption with the reproductive tissue phenotypes expressed by the mutant plant lines assessed in this study. Initial molecular profiling clearly revealed that, in Arabidopsis floral tissues, DRB1 is the sole DRB protein required to assist DCL1-catalyzed processing of the miR390 precursor transcripts, PRI-MIR390A and PRI-MIR390B. More specifically, the sole requirement of DRB1 for miR390 production in Arabidopsis floral tissues was evidenced by the 17.46- and 41.02-fold elevation in PRI-MIR390A (Figure 3A) and PRI-MIR390B (Figure 3B) transcript abundance in parallel with the 73% reduction to the level of the miR390 sRNA in drb1 flowers (Figure 3C). This was not a surprise finding considering that DRB1 has been extensively established to be required by DCL1 for the accurate and efficient production of the majority of Arabidopsis miRNAs [46,48,50,56,57].

Although miR390 precursor transcript expression remained at wild-type equivalent levels in drb2, drb4, ago7, drb2ago7, and drb4ago7 flowers, miR390 abundance was reduced by 30% in drb4 flowers and significantly elevated by 2.38- and 3.48-fold in drb2ago7 and drb4ago7 flowers (Figure 3C). In the root tissue of Arabidopsis plants molecularly modified to overexpress the tasiARF target genes, ARF2, ARF3 and ARF4, an auto-regulatory feedback loop for the TAS3 pathway has been previously identified whereby the ectopic expression of these three ARFs was demonstrated to promote the expression of the miR390 encoding loci, MIR390A and MIR390B [61]. Enhanced transcription from the MIR390A and MIR390B loci resulted in the elevated abundance of both the PRI-MIR390A and PRI-MIR390B precursor transcripts and the enhanced accumulation of the miR390 sRNA in these molecularly modified plant lines, presumably in an auxin-mediated attempt to return ARF2, ARF3 and ARF4 expression back to their wild-type levels [61]. However, although miR390 abundance was altered in drb4, ago7, drb2ago7, and drb4ago7 flowers, PRI-MIR390A and PRI-MIR390B expression remained largely unchanged (Figure 3). This finding indicated that the previously reported auxin-mediated auto-regulatory feedback loop for the TAS3 pathway was not the cause of the observed alterations to miR390 abundance in the flowers of these mutant lines. An alternate explanation for the observed alteration to miR390 abundance is offered by the demonstrated antagonism of DRB2 on DRB1 and/or DRB4 function [45,47,50]; an interplay between these three DRB proteins that potentially ensures that miRNA and siRNA abundance is tightly maintained throughout Arabidopsis development. This explanation is further supported by our recent molecular characterization of the miR399/PHOSPHATE2 (PHO2) expression module in non-stressed and phosphate stressed Arabidopsis [62]. That is, in Arabidopsis root and shoot tissues, DRB1 is the primary DRB family member required for miR399 production under both growth regimes; however, both DRB2 and DRB4 play secondary roles in this expression module to ensure that miR399 abundance, and, therefore, PHO2 expression, is tightly maintained. The miR390 precursor transcript and sRNA profiles presented in Figure 3 therefore similarly identifies the requirement of DRB2 and DRB4 to ensure that miR390 levels are strictly maintained in Arabidopsis floral tissues.

Considering that the molecular profiling of the individual transcripts of each step of the TAS3 pathway revealed that, in Arabidopsis floral tissues, the abundance of the miR390 sRNA is 100-fold higher than the level of its targeted transcript, TAS3, it was unsurprising to observe no tight correlation between the abundance of these two RNA molecules. For example, in drb1 flowers, where the level of the miR390 sRNA was reduced by 3.70-fold (Figure 3A), TAS3 abundance was only elevated by 1.40-fold (Figure 4A). This mild alteration in target transcript abundance in response to the significant change in the level of the targeting sRNA, indicated that in drb1 flowers the abundance of the miR390 sRNA did not decrease to a level sufficient to strongly influence the abundance of the TAS3 transcript. However, in spite of the failure to establish a correlative relationship between miR390 abundance and TAS3 expression, tasiARF abundance was determined to be reduced by 8% (drb1) to 43% (drb4ago7) in the floral tissues of all seven mutant lines analyzed in this study (Figure 4B), a finding that readily revealed TAS3 pathway dysfunction in each of these mutant lines defective in the activity of one, or multiple pieces of pathway protein machinery.

A change in level of the tasiARF sRNA in drb2, drb4, ago7, drb1ago7, drb2ago7, and drb4ago7 flowers was subsequently demonstrated to repress the expression of the tasiARF target gene, ARF2, by 14% (drb2) to 59% (drb1ago7) (Figure 4C). Elevated, and not reduced expression, would be expected for a miRNA target gene solely regulated via a mRNA cleavage-based mode of miRNA-directed RNA silencing in response to reduced miRNA abundance [5,7]. However, we have previously demonstrated that Arabidopsis miRNAs that require DRB2 for their production and/or action additionally regulate the expression of their target genes via a translational repression mode of RNA silencing [6,45,58]. In addition, animal miRNAs which predominantly regulate the expression of their target genes via a translational repression mode of RNA silencing, often scale in their abundance in accordance with the level of their target transcripts [63]. Therefore, repressed ARF2 expression in parallel with reduced tasiARF abundance putatively suggests that the tasiARF sRNA directs this alternate mode of RNA silencing to regulate ARF2/ARF2 abundance in Arabidopsis flowers. Furthermore, stigma exsertion is observed in the early developing flowers of the Arabidopsis arf2 knockout mutant, an architectural alteration that, in turn, greatly reduces the fertility of arf2 plants [64]. Here, we demonstrate that ARF2 expression is reduced by 14% and 47% in drb2 and ago7 flowers, respectively (Figure 4C), a gene expression alteration that could, in part, account for the stigma exsertion phenotype displayed by drb2 and ago7 flowers (Figure 1C) as well as the fertility defects of siliques that initially form from the proximal region of the primary inflorescence of drb2 and ago7 plants (Figure 1D).

In contrast to the reduced expression of the ARF2 target transcript in the floral tissues of six out of the seven mutant plant lines assessed (Figure 4C), ARF3 expression remained at approximate wild-type levels in drb1, drb2, and drb4 flowers and was elevated by 1.69- to 3.11-fold in ago7, drb1ago7, drb2ago7, and drb4ago7 flowers. The reproductive phase of development was most severely impacted in the drb1ago7 and drb2ago7 double mutants, the two plant lines where ARF3 expression was elevated to the greatest degree, a 2.53- and 3.11-fold increase in ARF3 transcript abundance, respectively (Figure 4D). Intriguingly, elongation of the pistil of early developing flowers and reduced fertility has been previously reported for Arabidopsis plants molecularly engineered to overexpress ARF3 [14], a reproductive phenotype shared with the arf2 mutant [64]. Reduced ARF2 expression and significantly elevated ARF3 transcript abundance in the floral tissues of ago7, drb1ago7 and drb2ago7 plants not only raises the possibility of negative cross-talk between ARF2 and ARF3 in Arabidopsis flowers, but adds further evidence that the fertility defects displayed by these three mutant lines is the result of the elongated pistil of ago7, drb1ago7, and drb2ago7 flowers due to TAS3 pathway dysfunction in this tissue. Like ARF3, ARF4 is required for the maintenance of cell identity [55]. Of particular interest is the report that the arf4 single mutant does not display any obvious developmental defects; however, the arf3arf4 double mutant expresses a range of vegetative and reproductive phenotypic abnormalities. This finding indicates redundant, or at least partially overlapping, functional roles for ARF3 and ARF4 in auxin-mediated developmental processes in Arabidopsis [55]. The almost identical expression profile constructed for ARF3 and ARF4 by RT-qPCR across ago7, drb1ago7, drb2ago7, and drb4ago7 flowers (Figure 4D,E), was highly supportive of the suggestion that ARF3 and ARF4 perform redundant and/or overlapping functional roles in auxin-mediated development [55], as well as to again indicate that TAS3 pathway dysfunction is the likely cause of the defects in flower development and fertility observed for the assessed mutant plant lines.

The three tasiARF targets, ARF2, ARF3, and ARF4, each function to repress the expression of their transcriptionally regulated genes in response to auxin [14,29]. Considering that ARF3 and ARF4 expression was determined to be elevated in ago7, drb1ago7, drb2ago7, and drb4ago7 flowers (Figure 4D,E), it was not surprising to subsequently observe repressed KAN1 and KAN3 expression (Figure 5A,C), two KAN transcription factor genes under ARF3/ARF4-mediated transcriptional regulation [54,55], in the floral tissues of the four plant lines where AGO7 activity was defective. KAN1 and KAN3 expression was also determined to be reduced in drb4 flowers by 63 and 44%, respectively (Figure 5A,C). However, reduced KAN1 and KAN3 expression in the drb4 mutant background was potentially due to factors independent of ARF2-, ARF3- and/or ARF4-directed transcriptional control, with the expression of all three tasiARF target genes determined to be reduced in drb4 flowers (Figure 4). In direct contrast to the reduced expression of KAN1 and KAN3 in drb4, ago7, drb1ago7, drb2ago7, and drb4ago7 flowers (Figure 5A,C), RT-qPCR revealed KAN4 expression to be elevated in the floral tissues of all seven mutant lines analyzed (Figure 5D). Considering that ARF2 expression was reduced in drb2, drb4, ago7, drb1ago7, drb2ago7, and drb4ago7 flowers (Figure 4C), and that ARF2 functions as a transcriptional repressor, this result putatively suggests that ARF2, and not ARF3 or ARF4, is required to direct auxin-mediated KAN4 expression responses in Arabidopsis floral tissues. However, further research is required before an ARF2/KAN4 regulatory circuit can be confidently established for Arabidopsis flowers. The KAN clade of the GARP transcription factor family has been demonstrated to function together with a group of Class III Homeodomain Leucine Zipper (HD-ZIP III) transcription factors to determine abaxial/adaxial polarity via defining expression boundaries for developmental genes in response to cellular signals, including responding to auxin [65,66,67,68]. In addition, the KAN transcription factors transcriptionally control the activity of many of the genes involved in the auxin biosynthesis pathway [69,70,71]. Therefore, the finding that we report here that altered ARF2, ARF3, and ARF4 expression influences the transcriptional activity of KAN1, KAN3, and KAN4, three KAN transcription factors that regulate auxin production, provides an excellent demonstration of the complex and interrelated regulatory networks that underpin plant development as well as revealing the detrimental phenotypic consequences of their dysfunction.

5. Conclusions

The comprehensive phenotypic assessment of the reproductive development of the drb1, drb2, drb4, and ago7 single mutants and of the drb1ago7, drb2ago7, and drb4ago7 double mutants, combined with the extensive molecular profiling of the TAS3 pathway in the floral tissues of these mutant lines revealed a number of new insights into this unique tasiRNA-directed gene expression regulation pathway. Namely, DRB1 is the sole DRB protein involved in the production of the TAS3 pathway triggering miRNA, miR390; however, both DRB2 and DRB4 appear to play secondary roles at this stage of the pathway to ensure that the level of the miR390 sRNA is correctly maintained. In addition, altered tasiARF abundance was demonstrated to have an opposing effect on target gene expression, with ARF2 expression reduced, and ARF3 and ARF4 transcript abundance elevated. Opposing changes in ARF2, ARF3 and ARF4 expression were putatively the result of differences in the mechanism of RNA silencing directed by the tasiARF sRNA; that is, ARF2 activity is controlled by a translational repression mode of RNA silencing, while the expression of the ARF3 and ARF4 target genes is regulated by the canonical mRNA cleavage mode of RNA silencing. In turn, altered ARF2, ARF3, and ARF4 expression directed a change in the transcriptional activity of the KAN transcription factors, KAN1, KAN3, and KAN4, transcription factors known to target auxin biosynthetic genes for expression regulation. Taken together, the results presented here demonstrate that disruption of the protein machinery at any stage of the TAS3 pathway has severe negative consequences on Arabidopsis development, with the impact of TAS3 pathway dysfunction being especially pronounced in Arabidopsis reproductive tissues.