AGO10 promotes floral determinacy

To identify genes with functions similar to those of AG in promoting stamen/carpel identities and terminating floral stem cells, we performed ethyl methanesulfonate (EMS) mutagenesis to screen for enhancers of the hua1-1 hua2-1 double mutant, which will be hereafter referred to as hua1 hua2. hua1 hua2 flowers are largely normal except that the gynoecia are shorter and more enlarged at the apical end than wild type (Figure 1A). The floral phenotypes of hua1 hua2 are sensitive to the dosage of functional AG. Combination of ag-1/+ or ag-4 with hua1 hua2 drastically enhances the homeotic and floral determinacy defects of hua1 hua2 [31]. Accordingly, a genetic screen in the hua1 hua2 background effectively identified genes that promote reproductive organ identity specification or floral determinacy [28], [32]–[35]. Among the new mutations we isolated in this screen were hua enhancer6 (hen6) and hua enhancer7 (hen7), recessive alleles in two genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. hen6 (ago10-12) and hen7 (ag-10) mutations enhance hua1 hua2.

(A) A hua1 hua2 flower with a gynoecium that was slightly enlarged at the top. (B) A hua1 hua2 hen6 (or hua1 hua2 ago10-12) flower with petaloid stamens and a bulged gynoecium containing ectopic floral organs inside. (C) A hua1 hua2 hen7 (or hua1 hua2 ag-10) flower that resembled that of an ag null mutant in that the third whorl stamens were transformed into petals and that the sepal-petal-petal pattern was reiterated. (D) Siliques of hua1 hua2. (E) Siliques of hua1 hua2 ago10-12. (F) A representative silique of hua1 ag-10 (left), hua1 ag-10 plants carrying the pAG::AG-GFP transgene (middle), and Ler (right). (G, H) Longitudinal sections of hua1 hua2 and hua1 hua2 ago10-12 stages 10-11 flowers. An ectopic floral meristem inside the fourth whorl carpels was present in hua1 hua2 ago10-12 (H) but not hua1 hua2 (G). ca, carpel; st, stamen. (I) A hua1 hua2 ago10-12 flower carrying AGO10 genomic DNA, which rescued the floral determinacy and homeotic defects of hua1 hua2 ago10-12. Some outer whorl organs were removed to expose the gynoecia in (A), (B), and (I). Scale bars, 50 µm in (G), (H) and 1 mm in all other panels.

https://doi.org/10.1371/journal.pgen.1001358.g001

hen6 enhanced the floral phenotypes of hua1 hua2 in terms of both reproductive organ identities and floral determinacy. While hua1 hua2 flowers have stamens in the third whorl, petaloid stemens were observed in hua1 hua2 hen6 flowers (Figure 1A,1B; Figure S3H), indicating a defect in stamen identity specification. In addition, hua1 hua2 hen6 gynoecia were much shorter than those of hua1 hua2 and appeared bulged or heart-shaped (Figure 1B,1D,1E). Extra floral organs internal to the fourth whorl carpels were present in hua1 hua2 hen6 but not in hua1 hua2 gynoecia; these extra organs were generated from an indeterminate floral meristem as observed in longitudinal sections of flowers (Figure 1G,1H; Table S1). hen6 was mapped to the BAC MQD19 on chromosome 5, from which AGO10 was selected as a candidate gene for sequencing. A C-to-T transition leading to the substitution of leucine (L) 674, which is highly conserved among all Arabidopsis AGO proteins, by phenylalanine (F) was found in AGO10 (Figure S1A,S1B). The homeotic and floral determinacy defects of hua1 hua2 hen6 were rescued by AGO10 genomic DNA and the fully rescued plants appeared morphologically identical to hua1 hua2 (Figure 1I). Therefore, hen6 was renamed ago10-12. While many previously isolated ago10 alleles show SAM defects with incomplete penetrance, the ago10-12 single mutant had no obvious SAM defects, suggesting that the L674F mutation does not completely compromise AGO10 function.

hen7 strongly enhanced hua1 hua2 such that hua1 hua2 hen7 flowers resembled those of ag null mutants. Stamens were transformed to petals and the floral meristems failed to terminate (Figure 1A,1C). Rough mapping showed that hen7 was linked to AG, and sequencing of AG revealed a G-to-A mutation resulting in an glutamate (E)-to-lysine (K) substitution at the 144th amino acid (Figure S1A). To show that AG could rescue the hen7 mutant phenotypes, we utilized the fertile hua1 hen7 double mutant. hua1 hen7 flowers had no organ identity defects, but the gynoecia were bulged with secondary floral organs inside the primary carpels (Figure 1F; Table S1). The floral determinacy defects of hua1 hen7 flowers could be rescued by a pAG::AG-GFP construct (Figure 1F). Therefore, hen7 is an ag allele, which we named ag-10. The ag-10 single mutant flowers had a normal complement of floral organs as in wild type (Figure 2A); as such, it is the weakest ag allele known to date. Most siliques on an ag-10 plant, however, can be distinguished from wild type by their curvature (Figure 2D). A small proportion of siliques were slightly bulged with additional organs inside (Figure 2D; Table S1). This contrasted with hua1 ag-10 plants, in which all siliques were bulged with internal floral organs inside (Table S1). Therefore, ag-10 was mildly defective in floral determinacy and this defect was drastically enhanced by hua1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. AGO10 and AGO1 promote floral determinacy.

(A) An ag-10 flower; most ag-10 flowers were similar to wild-type ones. (B) An ag-10 ago10-13 flower; the gynoecia of ag-10 ago10-13 flowers were bulged and contained ectopic floral organs internally. (C) A flower of ag-10 ago10-13 pAGO10::AGO10-FLAG showing that AGO10 rescued the floral determinacy defects of ag-10 ago10-13. (D) Representative siliques from ag-10 plants. Most were similar to wild-type siliques, while some were constricted on one side but still long and thin. Bulged siliques with internal floral organs (represented by the one on the left) were observed at a low frequency (usually 1-2 per plant). (E) Siliques from ag-10 ago10-13 plants. 100% of the siliques on an ag-10 ago10-13 plant were bulged. Occasionally, the primary carpels that make up a silique were not properly fused, revealing the internal organs. (F) An ago10-13 ago1-11/+ gynoecium with internal floral organs (arrow) bursting out of the gynoecium. (G, H) Longitudinal sections of stage 12 ag-10 and ag-10 ago10-13 flowers. The floral meristem was not visible in most ag-10 flowers (G), whereas it was still active in ag-10 ago10-13 flowers to result in extra whorls of organs inside the fourth whorl carpels (H). ca, carpel; st, stamen. (I) A scanning electron micrograph of an ago1-11 ago10-13/+ flower. The severe floral determinacy defects were reflected by the presence of numerous floral organs inside unfused primary carpels (indicated by arrows). (J–L) ago1-11 enhanced the floral determinacy defects of hua1 hua2 and ag-10. (J) An ago1-11 flower. (K) A hua1 hua2 ago1-11 flower with bulged gynoecia and petaloid stamens (arrows). (L) A representative ag-10 ago1-11 flower with a severely bulged gynoecium. In (A), (B), (C), and (J), some outer whorl floral organs were removed to expose the gynoecia. Scale bars, 50 µm in (G) and (H), 500 µm in (I), and 1 mm in all other panels.

https://doi.org/10.1371/journal.pgen.1001358.g002

We performed another EMS screen in the hua1 ag-10 background to identify mutants with enhanced floral determinacy defects. One mutant had a flowers-within-flower phenotype reminiscent of ag null mutants. After removal of the hua1 mutation by crossing, the double mutant of ag-10 and the enhancer mutation still exhibited strong floral determinacy defects; all siliques on the double mutant were severely bulged and ectopic floral organs were present inside the primary carpels (Figure 2B,2E; Table S1). Flowers of the double mutant did not exhibit any homeotic transformation and were fertile. We noticed that seedlings of the double mutant exhibited SAM defects at a low penetrance similar to ago10 alleles. This, together with the mapping of the mutation to the top of chromosome 5, led us to suspect that the enhancer mutation was in AGO10. Sequencing AGO10 from the double mutant identified a G-to-A mutation that introduced a premature stop codon close to the end of the second exon of AGO10 (Figure S1A). Introduction of a pAGO10::AGO10-FLAG construct into the double mutant fully rescued the floral determinacy defect (Figure 2C). Therefore, we named this mutation ago10-13. From longitudinal sections, it was clear that the floral meristems persisted well beyond stage 6 in all ag-10 ago10-13 flowers but not in most ag-10 flowers (Figure 2G,2H), suggesting that AGO10 is required for floral stem cell termination. Like other ago10 alleles, the gynoecia and siliques of the ago10-13 single mutant were shorter and wider than wild type (Figure S3A,S3B).

In an EMS screen in the ag-10 background, we identified yet another ago10 allele, ago10-14, which harbored a G-to-A mutation that changes the 731th amino acid from aspartic acid (D) to asparagines (N) (Figure S1A). Flowers of ag-10 ago10-14 resembled those of ag-10 ago10-13 in that the gynoecia were bulged (data not shown). The following genetic and molecular studies were performed mainly with ago10-12 and ago10-13 alleles.

Given that ago10-12 enhanced hua1 hua2 and both ago10-13 and ago10-14 enhanced ag-10 in terms of floral determinacy, we conclude that AGO10 promotes floral stem cell termination. Consistent with such a role, the YFP-ZLL fusion protein from the functional pZLL::YFP-ZLL transgene [36] was found in the center of stages 5-6 floral meristems by immunolocalization (Figure 3A). By stage 7, YFP-ZLL was found on the adaxial side of carpel primordia (Figure 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Expression patterns of AGO10, AG, and WUS in floral meristems.

(A, B) Immunolocalization of AGO10 protein using anti-GFP antibodies in zll-1 pZLL::YFP-ZLL flowers. The grey signals represent AGO10. Before floral stage 6, AGO10 was predominantly localized at the floral meristem and in provascular tissue (A); AGO10 was adaxialized in carpels beyond stage 6 (B). (C, D) The ago10-13 mutation did not alter the spatial patterns of AG expression in developing flowers. AG mRNA was similarly present in the inner two whorls of an ag-10 (C) and an ag-10 ago10-13 (D) flower at stages 6–7. (E–H) ago10 mutations led to prolonged WUS expression. In hua1 hua2, the latest stage when WUS mRNA could be detected was stage 6 (E). WUS expression persisted to stages 8-9 in hua1 hua2 ago10-12 flowers (arrow in F). WUS expression was diminished by stage 7 in ag-10 (G). In ag-10 ago10-13, WUS expression could be observed in stages 8–9 flowers (arrow in H). Scale bars, 50 µm. ca, carpel; se, sepal; st, stamen.

https://doi.org/10.1371/journal.pgen.1001358.g003

AGO10 terminates floral stem cells through repression of WUS expression

To investigate how AGO10 regulates floral stem cells, we examined the genetic interactions between ago10 alleles and mutations in key stem cell regulators. In wus-1, the floral meristems are terminated prematurely, resulting in incomplete flowers with four sepals, four petals, and typically a single stamen [2] (Figure S3E). hua1 hua2 ago10-12 wus-1 flowers had identical phenotypes to those of wus-1 in terms of floral determinacy (Figure 4A), indicating that wus-1 was epistatic to hua1 hua2 ago10-12. Unlike wus-1 (Figure S3E), the single stamen in hua1 hua2 ago10-12 wus-1 was petaloid as in hua1 hua2 ago10-12 (Figure 4A), indicating that the homeotic transformation of stamens into petaloid stamens in hua1 hua2 ago10-12 was independent of WUS. Similarly, wus-1 was epistatic to ag-10 ago10-13 in terms of floral determinacy (data not shown). Therefore, AGO10 acts through WUS to regulate floral meristem activity. This conclusion was also supported by the prolonged expression of WUS in floral meristems caused by ago10 mutations. Using in situ hybridization, we found that WUS expression in hua1 hua2 ago10-12 flowers persisted well beyond floral stage 6, when WUS expression was terminated in hua1 hua2 flowers (Figure 3E,3F). Similarly, WUS expression was detected in very old ag-10 ago10-13 but not ag-10 flowers (Figure 3G,3H).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Genetic interactions between AGO10 and meristem regulators.

(A) A hua1 hua2 ago10-12 wus-1 flower. wus-1 was completely epistatic to hua1 hua2 ago10-12 in floral determinacy such that the quadruple mutant flower had an incomplete set of floral organs as in a wus-1 flower. (B) An ag-1 flower. (C) An ag-1 pnh-1 flower that had apparently more floral organs than an ag-1 flower and showed signs of fasciation. (D,E) ap2-2 partially suppressed the floral determinacy defects of hua1 hua2 ago10-12 and ag-10 ago10-13. (D) A representative of the non-bulged gynoecia of some hua1 hua2 ago10-12 ap2-2 flowers. The petaloid stamens in hua1 hua2 ago10-12 were rescued back to stamens (arrow) by the ap2-2 mutation. (E) A representative of ag-10 ago10-13 ap2-2 gynoecia, which are longer than ag-10 ago10-13 gynoecia (Figure 2E). (F–H) phb-1d and phv-5d each enhanced the determinacy defects of ag-10. (F) ag-10 phb-1d/+ siliques. Ectopic floral organs, indicated by the arrow, were found in the siliques of most ag-10 phb-1d/+ flowers. (G) ag-10 phv-5d siliques were bulged or contained ectopic floral organs, as indicated by the arrow. (H) A longitudinal section of an ag-10 phb-1d/+ gynoecium. An ectopic floral meristem (indicated by the arrow) could be observed inside the carpels. (I–L) ag-10 plants harboring transgenic amiR165/166 exhibited loss of floral determinacy. (I–J) Weaker determinacy defects of some ag-10 amiR165/166 plants. The gynoecia were bulged as in ag-10 ago10-13. (K–L) Stronger determinacy defects of some ag-10 amiR165/166 plants. The gynoecia were replaced by an internal flower. Scale bars, 50 µm in (H), and 1 mm in all other panels. ca, capel; se, sepal.

https://doi.org/10.1371/journal.pgen.1001358.g004

CLV3 controls the spatial domain of WUS expression and the size of the stem cell domain [37]. clv3-1 flowers contain extra floral organs of all types, particularly stamens and carpels, due to enlarged floral meristems [38]. The combination of hua1 hua2 ago10-12 and clv3-1 had a synergistic effect on floral determinacy. The gynoecia of the quadruple mutant generated a massive amount of stigmatic tissue bursting out of the primary carpels (Figure S3F,S3G). Similar phenotypes were observed for ag-10 ago10-13 clv3-1 flowers (data not shown), implying that AGO10 and CLV3 act largely independently.

Since AG is crucial in the termination of floral stem cells, we evaluated the relationship between AG and AGO10 through extensive genetic interaction studies. We first introduced the null ag-1 allele into hua1 hua2 ago10-12. The quadruple mutant flowers were initially similar to ag-1 flowers, but the meristems continued to develop and appeared fasciated in old flowers (data not shown). Given that ago10-12 was a weak allele, we next combined the stronger ago10 alleles pnh-1 [25] or ago10-13 with ag-1. Most flowers of the ag-1 pnh-1 and ag-1 ago1-13 double mutants appeared more fasciated than ag-1 flowers (Figure 4B,4C; data not shown). The mild enhancement implies that AGO10 acts through the AG pathway to repress WUS expression but that AGO10 also has an AG-independent function in floral stem cell regulation. We conducted in situ hybridization to examine the patterns of AG expression in ag-10 and ag-10 ago10-13 flowers. AG transcripts were detected in the inner two whorls of ag-10 and ag-10 ago10-13 floral meristems (Figure 3C,3D). Therefore, ago10-13 did not affect the spatial domain of AG expression or lead to obvious changes in AG mRNA levels in floral meristems.

AGO1 is also required for floral stem cell termination

Since AGO1 is the most closely related to AGO10 among the ten Arabidopsis AGO genes, we sought to determine whether AGO1 is also required for floral stem cell termination. We introduced the partial loss-of-function ago1-11 mutation [39] into hua1 hua2 and ag-10. The infertile ago1-11 flowers develop all four types of floral organs, although the organs are abnormal in morphology (Figure 2J; Figure S3C). In hua1 hua2 ago1-11 flowers, either the gynoecia were severely bulged or the carpels were completely unfused with ectopic flowers developing internal to the primary carpels (Figure 2K). Similarly, indeterminate phenotypes were observed in the ag-10 ago1-11 double mutant (Figure 2L). Therefore, AGO1, like AGO10, is required for floral stem cell termination.

Genetic studies conducted by others as well as by ourselves show that AGO1 and AGO10 have overlapping functions in terminating floral meristems. First, it was previously shown that combining the strong ago1-7 and pnh-2 (an ago10 allele) alleles resulted in embryo lethality [23]. But ago1-7/+ pnh-2/pnh-2 and ago1-7/ago1-7 pnh-2/+ flowers showed an increase in the number of floral organs as well as ectopic growth of tissues inside the primary carpels (i.e., loss of floral determinacy). Second, we crossed the weak ago1-11 allele with the strong ago10-13 allele but also failed to obtain viable double mutants. The ago1-11 ago10-13/+ plants were smaller than ago1-11 in stature. Some inflorescences harbored flowers with filamentous floral organs and obviously enlarged floral meristems (Figure S3D). When carpels were formed, they were unfused and additional floral organs were observed inside the primary carpels (Figure 2I). In ago10-13 ago1-11/+ plants, the gynoecia were bulged and additional floral organs were found in some of the gynoecia upon dissection (Figure 2F). In summary, both ago1 ago10/+ or ago1/+ ago10 plants were more severe in floral determinacy defects than the corresponding single mutants. The dosage effects of ago1 and ago10 alleles indicate that the two genes have overlapping functions in floral stem cell termination.

AGO10 acts partly through miR172 to promote floral determinacy

miR172 negatively regulates AP2 mainly through translational repression [10], [11] to result in proper floral patterning including the termination of floral stem cells [12] and stamen identity specification [32]. In the hua1 hua2 background in which AG expression is reduced, loss of function in HEN1, a player in miRNA biogenesis [40], results in stamen-to-petal transformation. The homeotic transformation is mainly due to de-repressed AP2 expression since the ap2-2 mutation restores stamen identity to hua1 hua2 hen1-1 flowers [32].

To determine whether AGO10 acts through miR172 in floral patterning, we examined the genetic interactions between ago10 and ap2 mutations by generating hua1 hua2 ago10-12 ap2-2 and ag-10 ago10-13 ap2-2. First, ap2-2 suppressed the homeotic defects of hua1 hua2 ago10-12 in that the third whorl petaloid stamens in the triple mutant were restored to stamens in the quadruple mutant (Figure 4D). While the third whorl organs in hua1 hua2 ago10-12 had patches of cone-shaped cells reminiscent of petal epidermal cells (Figure S3I, S3J), the third whorl organs in hua1 hua2 ago10-12 ap2-2 had epidermal cells with the shapes of interlocking puzzle pieces resembling anther epidermal cells (Figure S3K,S3L). This indicates that the homeotic transformation in hua1 hua2 ago10-12 flowers is due to de-repressed AP2 and that AGO10 mediates the function of miR172 in repressing AP2. Second, ap2-2 partially suppressed the floral determinacy defects of hua1 hua2 ago10-12 and ag-10 ago10-13 flowers. ap2-2 did not affect the shapes of hua1 hua2 or ag-10 gynoecia ([31]; Figure S3M), but caused hua1 hua2 ago10-12 and ag-10 ago10-13 gynoecia to be less bulged (compare Figure 4D to Figure 1B and Figure 4E to Figure 2E), suggesting that the determinacy defects were not as severe as in hua1 hua2 ago10-12 or ag-10 ago10-13. To quantify this effect, we dissected gynoecia to observe the presence/absence of ectopic organs inside. Compared with hua1 hua2 ago10-12 and ag-10 ago10-13 plants that had ectopic floral organs in 91% and 100% of the siliques, respectively, ap2-2 reduced the percentage of siliques with internal organs to 41% and 86% in hua1 hua2 ago10-12 and ag-10 ago10-13, respectively (Table S1). Although the suppression effect for the latter genotype was not large, we found that ap2-2 completely suppressed the presence of an elongated gynophore in ag-10 ago10-13 (Figure S3N-S3P). An elongated gynophore probably reflects a partial conversion of the floral meristem to an inflorescence meristem, hence, prolonged stem cell activity. Therefore, the floral stem cell defects of ago10 mutants were partially rescued by ap2-2, indicating that AP2 is likely de-repressed in ago10 mutants and that the de-repression is partially responsible for the floral stem cell defects. This conclusion is consistent with the similar genetic behavior of AP2m3 and ago10 in that both enhanced ag-1 ([12]; Figure 4B,4C).

Despite the genetic evidence supporting de-repressed AP2 expression in ago10 mutants, we were unable to consistently detect an increase in AP2 protein levels in ag-10 ago10-13 vs. ag-10 or ago10-13 vs. wild type by western blotting (data not shown). This is likely attributable to insufficient sensitivity of the assay in the detection of small differences. Since the levels of miR172 were not affected by ago10-13 (Figure S2A), it is likely that miR172-mediated repression of AP2 requires AGO10.

AGO10 is associated with miRNAs in vivo

Although our genetic evidence indicates that AGO10 mediates the activities of miR172 and previous genetic studies implicate AGO10 in mediating the translational repression of target mRNAs by multiple miRNAs [26], molecular evidence of AGO10 acting directly with miRNAs has been lacking. We sought to determine whether AGO10 was associated with miRNAs in vivo. We immunoprecipitated (IP) YFP-ZLL from inflorescences of zll-1 pZLL::YFP-ZLL plants [36] with anti-GFP antibodies. pAG::AG-GFP and Ler (wild type) were used as a positive and a negative control, respectively, for the IP. Proteins of the expected sizes from the pZLL::YFP-ZLL and pAG::AG-GFP IP samples were detected by western blotting (Figure 5A). We also included pAGO1::FLAG-AGO1 [20] and Col (a control for pAGO1::FLAG-AGO1) for comparison. We extracted RNAs from the IP samples and performed small RNA northern blotting to detect selected miRNAs and siRNAs. AGO10 bound miRNAs such as miR165/166, miR172, miR173, miR319 and miR168 (Figure 5B), but not miR390 (Figure S4), which specifically associates with AGO7 [18]. Northern blotting failed to detect signals for 24 nt endogenous siRNAs from several loci from the IP samples (data not shown). The AGO10-associated miRNAs are also associated with AGO1 in vivo (Figure 5B; [17], [20]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. AGO10 is associated with miRNAs in vivo and is catalytically active as a “slicer” in vitro.

(A) Immunoprecipitation (IP) followed by western blotting using anti-GFP antibodies from three genotypes as indicated. The immunoprecipitated YFP-ZLL fusion protein could be detected as a band of the expected size (∼170 kD). AG-GFP (∼60 kD) and Ler served as a positive and a negative control for the IP, respectively. (B) Detection of miRNAs by northern blotting from FLAG-AGO1 and YFP-ZLL IP. Col and AG-GFP served as negative controls for FLAG-AGO1 and YFP-ZLL, respectively. (C) An in vitro “slicer” assay. Both FLAG-AGO1 and YFP-ZLL immune complexes cleaved the PHB wild-type probe, but not the miR165/166-resistant probe. The full-length PHB and PHBm probes were 237 nt. The 5′ and 3′ cleavage fragments were 112 nt and 125 nt, respectively, and were not separated in the gel (represented by the band of approximately 120 nt). (D) Northern blot for miR165/166 in the genotypes as indicated. U6 served as a loading control. The numbers below the gel image indicate the relative abundance of miR165/166. (E) Expression of the HD-Zip genes PHB, PHV and CNA in wild type (Ler), ago10-13, and ago1-11 inflorescences as examined by realtime RT-PCR. (F) Realtime RT-PCR to determine the levels of PHB, PHV, and CNA mRNAs in ag-10 and ag-10 ago10-13 inflorescences.

https://doi.org/10.1371/journal.pgen.1001358.g005

AGO10 has “slicer” activity

Because AGO10 contains the DDH catalytic residues that are critical for the “slicer” activity of AGO proteins, we decided to determine whether AGO10 is catalytically active in vitro. We performed a “slicer” activity assay on the miR165/166 target PHABULOSA (PHB). A portion of the PHB transcript containing the miR165/166 binding site was generated through in vitro transcription and incubated with IP samples from pAGO1::FLAG-AGO1 and pZLL::YFP-ZLL plants as well as the corresponding negative controls Col and pAG::AG-GFP, respectively. A miR165/166-resistant PHB transcript, PHBm, was also included as a negative control. The PHB transcript was cleaved into two fragments of expected sizes by both FLAG-AGO1 and YFP-ZLL but not by the corresponding control IPs (Figure 5C). As expected, the PHBm transcript failed to be cleaved by either FLAG-AGO1 or YFP-ZLL (Figure 5C). This shows that AGO10, like AGO1, can cause the cleavage of miRNA-targeted mRNAs.

Given that AGO10 has “slicer” activity, we next examined the levels of small RNA-targeted mRNAs in ago10 mutants. The levels of most examined mRNAs were not obviously different between ago10-13 and wild type (Figure S2B), consistent with findings from a previous study [26]. A small elevation in the levels of CUC1 mRNA, which is targeted by miR164, was consistently detected in ago10-13 (Figure S2B). Interestingly, a small reduction in the levels of miR164 in ago10 mutants was consistently observed (Figure S2A). The miR164-resistant CUC1 transgene (pCUC1::CUC1m-GFP; [41]) did not enhance the determinacy defects of ag-10 (data not shown), suggesting the ag-10 ago10-13 floral determinacy defects were unlikely attributable to de-repressed CUC1 expression. There are several explanations for the lack of differences in miRNA target mRNA levels between ago10 and wild type. First, changes are masked by the target mRNAs outside of the highly restricted expression domains of AGO10. Second, AGO1 is sufficient to regulate most miRNA-targeted mRNAs at the transcript level, and AGO10 acts primarily through translational inhibition in vivo.

miR165/166 and its targets, the HD-Zip genes, are crucial players in floral determinacy

The fact that the ap2-2 mutation failed to completely suppress hua1 hua2 ago10-12 or ag-10 ago10-13 suggests that AGO10 regulates floral stem cell termination also through other miRNAs. Because AGO10 associates with miR165/166 and the HD-Zip genes PHB, PHAVOLUTA (PHV), and REVOLUTA (REV) targeted by this miRNA strongly influence the formation of the SAM [29], we tested whether proper regulation of PHB by miR165/166 was crucial for floral stem cell termination. We crossed the miR165/166-resistant form of PHB, phb-1d [42], with ag-10 ago10-13. Both ag-10 phb-1d/+ and ag-10 ago10-13/+ phb-1d/+, but not phb-1d/+ flowers had bulged gynoecia with ectopic floral organs inside (Figure 4F; Figure S3Q,S3R). Sections of ag-10 phb-1d/+ flowers revealed indeterminate floral meristems inside the primary carpels (Figure 4H). In addition, in the genetic screen in the ag-10 background, we isolated a semi-dominant phv allele, phv-5d, as an ag-10 enhancer. The ag-10 phv-5d double mutant exhibited bulged siliques throughout the plant (Figure 4G) whereas most but not all siliques in ag-10 phv-5d/+ plants were bulged (Figure S3S). phv-5d contained a G-to-A mutation in the miR165/166 binding site (Figure S1A,S1C) and this lesion was identical to those of the previously characterized phv-1d, -2d, -3d, and -4d alleles [29]. Although the mutation resulted in a glycine-to-aspartic acid substitution, extensive studies of an identical mutation in PHB (phb-3d) showed that it was the disruption of miR165/166 targeting rather than the amino acid change that caused the developmental defects in phb-3d [30]. Collectively, these results demonstrate that miR165/166-mediated regulation of HD-Zip genes is necessary for floral stem cell termination.

Having shown that the repression of PHB and PHV by miR165/166 is important for floral stem cell termination, we sought to determine whether ago1 or ago10 mutations resulted in defects in miR165/166-mediated repression of HD-Zip genes. We first examined the accumulation of miR165/166 in ago1-11, ago1-11 ago10-13/+, ago10-13, and ago1-11/+ ago10-13 inflorescences. Intriguingly, miR165/166 levels were reduced in ago1-11 but increased in ago10-13 (Figure 5D). The increase in miR165/166 levels was consistently observed in all ago10 alleles tested (ago10-12, ago10-13, and pnh-1; Figure S2A and data not shown), and was also previously observed in ago10 seedlings [43].

Next, we examined the levels of HD-Zip mRNAs in Ler, ago10-13, ago1-11, ag-10, and ag-10 ago10-13 flowers by realtime RT-PCR. The levels of PHB, PHV and CORONA (CNA) mRNAs were reduced in ago10 genotypes as compared to controls (Figure 5E,5F), which correlated with the increased abundance of miR165/166. This indicated that AGO1 was sufficient to control the mRNA levels of the HD-Zip genes in the absence of AGO10. In ago1-11, in which miR165/166 levels were low, PHB and PHV mRNA levels were slightly higher than wild type (Figure 5E), consistent with AGO1 being the major slicer of PHB and PHV in vivo. Therefore, ago1 and ago10 mutations have opposite effects on miR165/166 accumulation and expression of the HD-Zip genes.

ago10 mutants were previously found to have increased levels of miR165/166 in seedlings and the increase in miR165/166 and the resulting decrease in HD-Zip expression were responsible for the SAM and leaf polarity defects of ago10 mutants [43]. Therefore, it is possible that reduced HD-Zip expression caused by elevated levels of miR165/166 also underlies the floral determinacy defects of ago10 mutants. However, it is also possible that AGO10 regulates the HD-Zip genes through translational repression since it is known to do so on other miRNA target genes [26]. If this were the case, the levels of HD-Zip proteins would be increased in ago10-13 despite the reduction in the mRNA levels. The lack of antibodies against the HD-Zip proteins prevented us from directly determining the levels of HD-Zip proteins in ago10 mutants. We used the mRNA levels of LITTLE ZIPPER3 (ZPR3), which is a direct target of PHB [44], as a proxy for PHB protein levels. It was shown that ZPR3 mRNA levels are reduced in HD-Zip loss-of-function mutants and increased in miRNA165/166-resistant HD-Zip mutants [44]. Realtime RT-PCR showed that ZPR3 levels were severely reduced in ago10-13 (Figure S5), suggesting that HD-Zip protein levels were low in ago10-13. Next, we investigated whether reduced HD-Zip gene expression could result in loss of floral determinacy. ag-10 plants were transformed with an artificial miR165/166 construct in which pre-miR165 and pre-miR166 were placed in tandem and driven by the 35S promoter (Figure S6). Among 64 primary transformants, 16 exhibited premature termination of the SAM, a phenotype reminiscent of ago10 mutants (Figure S3T). A few of them couldn't survive to generate true leaves. 12 plants exhibited bulged gynoecia similar to ag-10 ago10 and 4 plants had more severe floral determinacy defects in that the gynoecia were replaced by an internal flower (Figure 4I–4L). This demonstrated that down-regulation of HD-Zip gene expression by over-expression of miR165/166 compromises floral determinacy.

AGO10 and AGO1 have similar small RNA–binding specificities but can act differently on miRNAs in vivo

Among the ten argonaute proteins in Arabidopsis, AGO10 is the most similar in sequence to AGO1, the major miRNA effector. Genetic evidence indicates that AGO10 and AGO1 have redundant, overlapping, or even antagonistic functions in development and in RNA silencing [23], [27]. For several tested miRNA target genes, protein but not mRNA levels are elevated in ago10 mutants, suggesting that AGO10 is necessary for miRNA-mediated translational repression of targets [26]. However, direct molecular evidence supporting AGO10 as a miRNA effector has been lacking, perhaps owing to the spatially highly restricted expression of AGO10, which makes it challenging to assay small RNA binding and cleavage activities of AGO10. In this study, we provide direct evidence that AGO10 associates with miRNAs in vivo and possesses miRNA-directed cleavage activity in vitro. So far, three other Arabidopsis AGO proteins, AGO1, AGO4 and AGO7, have been shown to possess “slicer” activity [18], [20], [45]. AGO1 preferentially associates with most miRNAs, ta-siRNAs and transgene siRNAs and mediates target mRNA cleavage and translational inhibition [17], [20], [26]. AGO7 specifically binds and mediates the function of miR390 in ta-siRNA production through both cleavage and non-cleavage functions [18]. AGO4 binds 24 nt endogenous siRNAs and causes transcriptional silencing of target loci in either a slicer-dependent or slicer-independent manner [45]. From our limited examination of small RNAs bound by AGO10, it appears that AGO10 has similar small RNA binding specificities as AGO1 in that it binds all examined miRNAs except for miR390 and it does not bind 24 nt endogenous siRNAs.

Despite the similar small RNA binding specificities of the two proteins, mutations in the two genes can have different effects on miRNA accumulation and the expression of miRNA target genes. The abundance of many miRNAs is reduced in severe ago1 mutants [46] but not affected in ago10 null alleles ([26] and this study). While many miRNA target mRNAs are increased in abundance in ago1 mutants [21], most assayed miRNA targets are not significantly affected in ago10 mutants ([26] and this study). miR165/166 is an extreme example of the different effects of ago1 and ago10 mutations. While miR165/166 levels are reduced in ago1-11, they are increased in ago10 mutants. The targets of miR165/166 are also affected in opposite directions in the two mutants – they are de-repressed in ago1-11 and further repressed in ago10. This shows that AGO1 is the major slicer acting upon the HD-Zip mRNAs in vivo, although AGO10 has slicer activity in vitro. While the two argonaute proteins act differently on miR165/166, they both mediate the activities of miR172. Both ago1-11 and ago10-12 result in petaloid stamens in the hua1 hua2 background, suggesting that both ago1 and ago10 mutations lead to de-repression of AP2. Therefore, AGO1 and AGO10 mediate the activities of some miRNAs but also have different effects on others.

Floral stem cell termination versus differentiation

Floral stem cell termination is tightly coupled to the formation of carpel primordia in flower development. AG, which specifies carpel identities, acts to terminate the floral stem cells. It may thus be presumed that the termination of floral stem cells is simply via differentiation of these cells into carpel cells. However, the fact that ag-10 ago10-13 and hua1 hua2 ago10-12 flowers have 4th whorl carpels but are indeterminate indicates that carpel identity specification and floral stem cell termination, although both directed by AG, are two separable processes. It is expected that factors acting downstream of AG in the two processes must be distinct. KNUCKLES, for example, is a target of AG that is only required for floral determinacy specification [47].

HD-Zip genes, targets of miR165/166, are crucial for the temporal program of floral stem cells

Our previous studies show that both miR172 and its target AP2 regulate the temporal program of floral stem cells. Plants expressing miR172-resistant AP2 cDNA have prolonged WUS expression and indeterminate floral meristems that produce numerous stamens [12]. In this study, we show that the HD-Zip genes, which are targets of miR165/166 and which are previously known to specify adaxial identity in lateral organs, are also crucial factors in floral stem cell regulation. Over expression of miR165/166 in the ag-10 background results in indeterminate flowers. Consistent with this, the phb phv cna triple mutant was observed to have occasional flowers with enlarged gynoecia containing ectopic carpels inside [48]. On the other hand, ag-10 phb-1d/+ and ag-10 phv-5d gynoecia are also indeterminate. Given the similar phenotypes of de-repressed PHB or PHV and loss of function in PHB, PHV and CNA in terms of floral determinacy, it is likely that either too much or too little HD-Zip activity is detrimental to the precise regulation of floral stem cells. Given that the HD-Zip genes promote adaxial identities of lateral organs, our findings raise the intriguing possibility that floral stem cell termination depends on correct polarity specification of the fourth whorl organs. This hypothesis will be tested in the future.

AGO10 and AGO1 promote floral stem cell termination via similar and different mechanisms

The relationship between AGO10 and AGO1 varies depending on the developmental processes. During embryogenesis, AGO10 and AGO1 share overlapping functions in maintaining STM expression [23]. In the weak zll-15 mutant background, increasing the dosage of AGO1 suppressed the defects in embryonic stem cells, while reducing AGO1 dosage enhanced the defects [27]. Our studies provide the molecular evidence supporting the overlapping biological roles of AGO1 and AGO10 by showing that they have similar small RNA binding specificities and that they both have “slicer” activity. A previous study also documented that AGO10 antagonizes AGO1 function. ago10 mutations partially suppressed the leaf margin serration phenotype of the hypomorphic ago1-27 allele [27]. Moreover, defects in transgene PTGS and miRNA-mediated gene silencing in the hypomorphic ago1 mutant were also restored by ago10 mutations [27]. The elevated AGO1 protein levels in the double mutants reveal that AGO10 is a negative regulator of AGO1 at the translational level. By showing that AGO10 binds miR168, which targets AGO1, we provide the molecular evidence for the direct regulation of AGO1 by AGO10 through miR168.

Our studies support overlapping roles for AGO10 and AGO1 in the temporal regulation of floral stem cells but also reveal that the two proteins do so with both similar and different molecular mechanisms (Figure 6). It is well established that AGO1 mediates the functions of miR172 and miR165/166 in different aspects of plant development. It is likely that AGO1 promotes floral stem cell termination by repressing AP2 and HD-Zip expression. Consistent with this, de-repression of AP2 or HD-Zip genes compromises floral stem cell termination. We show here that AGO10 also associates with these two miRNAs in vivo. Genetic evidence suggests that AGO10 mediates the functions of miR172 in both stamen identity specification and floral determinacy. Therefore, AGO10 reinforces the functions of AGO1 in the regulation of AP2. Perhaps AGO1 alone is not sufficient to mediate the functions of miR172 in floral meristems such that AGO10, which is specifically expressed in certain cells in the floral meristem, is necessary to act upon the same miRNA. Intriguingly, AGO1 and AGO10 exert opposite effects on miR165/166 accumulation and the expression of the HD-Zip genes. The increased miR165/166 levels and reduced expression of HD-Zip genes in ago10 mutants suggest that AGO10 promotes HD-Zip gene expression by reducing miR165/166 levels. How AGO10 does so is currently unknown. Despite the opposite effects on HD-Zip expression, AGO1 and AGO10 both promote floral determinacy. This can be reconciled by the fact that both reduced expression and de-repression of HD-Zip genes compromise floral determinacy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. A model of AGO1 and AGO10 in floral stem cell regulation.

miR172 promotes floral stem cell termination by repressing the expression of AP2, which promotes WUS expression. Both AGO1 and AGO10 are associated with miR172 in vivo and genetic evidence supports a role of AGO10 in mediating the activities of miR172. miR165/166-mediated repression of PHB and PHV is also necessary for floral stem cell termination. AGO1 and AGO10 exert opposite effects on miR165/166. While AGO1 mediates the activities of miR165/166, AGO10 represses the expression of miR165/166. The relationship between PHB/PHV and WUS is unknown. Interactions that are potentially indirect are represented by dotted lines.

https://doi.org/10.1371/journal.pgen.1001358.g006

SAM and floral meristems—differences in stem cell maintenance

It has not escaped our attention that loss of function of AGO10 results in opposite effects in stem cell regulation between the SAM and the floral meristems. While ago10 mutations lead to premature termination of stem cells in the SAM, they result in prolonged floral stem cell maintenance. A previous study found that over expression of miR165/166 recapitulates the SAM defects of ago10 mutants [49]. In this study, we show that over expression of miR165/166 recapitulates not only the SAM defects but also the floral stem cell defects of ago10 mutants. Therefore, over accumulation of miR165/166 and reduced expression of the HD-Zip genes probably underlie the failure to maintain the SAM and to terminate the floral stem cells. How reduced expression of the HD-Zip genes leads to opposite effects in the two types of meristems is currently unknown.

Plant strains and EMS mutagenesis

The plant strains used in this study are all in the Landsberg erecta (Ler) ecotype except for ago1-36 pAGO1::FLAG-AGO1 [20], which is in the Columbia (Col) background. All plants were grown at 23°C under continuous light except for those containing the phb-1d mutation, which were grown at 17°C. zll-1 pZLL::YFP-ZLL [36], clv3-1 [38], wus-1 [2], ap2-2 [50], phb-1d [42], hua1-1 [31], hua2-1 [31], and pnh-1 [25] were all previously characterized.

ag-10, ago10-12, ago10-13, ago10-14 and phv-5d were isolated in this study. An EMS mutagenesis screen was carried out in the hua1 hua2 background as described [32]. Single M2 families were screened and two independent mutants, hua1 hua2 ago10-12 and hua1 hua2 ag-10, with bulged gynoecia and ag null mutant-like phenotypes, respectively, were identified. The ag-10 mutation was recovered from fertile siblings. The two lines were backcrossed to hua1 hua2 three times to clean up the genetic backgrounds. They were also crossed to either hua1 or hua2 to generate hua1 ago10-12, hua2 ago10-12, hua1 ag-10 and hua2 ag-10 double mutants. The ago10-12 or ag-10 single mutant was obtained by crossing hua1 ago10-12 or hua1 ag-10 to wild type. The ago10-13 allele was isolated in a separate EMS mutagenesis screen in the hua1 ag-10 background as a mutation that enhanced the floral determinacy phenotypes of hua1 ago-10. The mutant was backcrossed to ag-10 three times to obtain the ag-10 ago10-13 double mutant, with which most subsequent analyses were conducted. The ago10-13 single mutant was obtained by crossing ag-10 ago10-13 to Ler. The ago10-14 and phv-5d alleles were isolated in an EMS screen in the ag-10 background as enhancers of its weak determinacy defects.

Construction of amiR165/166

The amiR165/166 plasmid contains modified pre-miR165 and pre-miR166, two stem-loop structures placed in tandem and inserted between the 2X 35S promoter and the 35S terminator in a small-sized pOT2 vector (Figure S6). The amiR165 and amiR166 construction was done by two rounds of PCR cloning using a proofreading Taq polymerase and two pairs of long primers (∼60–100 nt) engineered to contain a SwaI and a PmeI site, respectively. The two pairs of primers (M0SL-at-miR165-SwaI-PR and M0SL-at-miR165-SwaI-PF or M4SL-at-miR166-PmeI-PR and M4SL-at-miR166-PmeI-PF) covered the entire artificial stem-loop sequences of miR165 or miR166 to minimize errors in the miR165 or 166 duplex regions during the PCR reactions. First, the primers M0SL-at-miR165-SwaI-PR and M0SL-at-miR165-SwaI-PF were used to amplify the pOT2 vector that contains the miR168 backbone. The PCR products were cleaved by SwaI and self-ligated to generate a plasmid that contains pre-miR165. Second, PCR was performed using this plasmid as the template with the primers M4SL-at-miR166-PmeI-PR and M4SL-at-miR166-PmeI-PF. The PCR products were cleaved by PmeI and self-ligated to generate a plasmid containing the tandem pre-miR165 and pre-miR166 sequences. Finally, this plasmid in pOT2 was amplified by a pair of primers (Origin-del-PacI-PF and Origin-del-PacI-PR) that contained PacI sites to delete the plasmid replication origin. The PCR products that contained the amiR165/166 and a chloramphenicol selection marker were introduced into a modified pFGC5941 binary vector through the unique PacI site. Recombinant binary plasmids were verified by DNA sequencing before being used for plant transformation.

RNA extraction and realtime RT-PCR

Total RNAs were extracted from inflorescences or seedlings using TRI Reagent (Molecular Research Center, Cat# TR118) as per manufacturer's instructions. Contaminating DNA was removed by DNase I (Promega, Cat# M610A; Roche, Cat# 04716728001) digestion. cDNA was synthesized from 2 µg total RNAs using reverse transcriptase (Fermentas, Cat# EP0441) and an oligo-dT primer. Quantitative PCR was performed in triplicates on a Bio-Rad IQ cycler apparatus with the iQ SYBR green supermix (Bio-Rad, Cat# 170-8882). The primers used are listed in Table S2.

Small RNA northern blotting

Northern blotting to detect small RNAs was performed as described [40], [51]. 5′-end-labeled (³²P) antisense DNA oligonucleotides were used to detect miRNAs from total RNAs. 5′-end-labeled (³²P) antisense LNA oligonucleotides were used to detect miRNAs from AGO1 or AGO10 immunoprecipitation samples. The probe sequences are listed in Table S2.

Immunoprecipitation

FLAG-AGO1 immunoprecipitation was performed as described [20]. ag-1 pAG::AG-GFP and zll-1 pZLL::YFP-ZLL inflorescences were ground in liquid nitrogen and homogenized in an equal volume of extraction buffer (20 mM Tris HCl pH 7.5, 300 mM NaCl, 5 mM MgCl₂, 5 mM DTT, 1% (v/v) protease inhibitor cocktail (Roche, Cat# 11244800; 1 tablet was dissolved in 1 ml nuclease-free water)). The extracts were centrifuged for 10 min at 16,000 g three times to pellet any debris. The lysates were then precleared with protein-A agarose (Thermo Scientific, Cat# 20333) for 1 hr. Anti-GFP antibodies (Clontech, Cat# 632460) were added to the samples at a 1∶200 dilution and the samples were incubated on a rotating shaker for 2 hrs. Protein-A agarose was added at a 1∶50 ratio and incubation was continued for another 2 hrs. The immune complexes were washed four times with 1.5 ml of extraction buffer supplied with 0.5% NP-40 and two times with RISC (RNA-induced silencing complex) buffer (40 mM Hepes pH 7.4, 100 mM KOAc, 5 mM MgOAc, 4 mM DTT). Finally, the immune complexes were resuspended into 100 µl RISC buffer.

In vitro transcription and RNA cleavage assay

The PHB and PHBm cDNAs were amplified by PCR with primers PHBprobeF and PHBprobeR (see Table S2 for sequence information) from the wild type PHB and the miR165/166-resistant PHB G202G [30] templates, respectively. The PCR products were cloned into pGEM-T Easy (Promega, Cat#A1360) and selected clones were sequenced to ensure the absence of unwanted mutations. The clones were linearized with SpeI (NEB, Cat# R0133L) and the linearized DNA was gel purified. In vitro transcription and RNA cleavage assays were carried out as described [20]. In vitro transcription was performed by incubating 800 ng DNA in a 25 µl reaction with T7 RNA polymerase (Promega, Cat# P2077) and α-P³²-UTP at 37°C for 1.5 hrs. Labeled PHB and PHBm probes were gel purified and dissolved in 50 µl nuclease-free water. 3 µl probes were used in a 25 µl cleavage reaction mix containing 20 µl AGO1 or AGO10 immune complexes in RISC buffer, 1 µl 25 mM ATP and 1 µl RNase inhibitor (Fermentas, Cat# EO0381). The reaction mix was incubated at 37°C for 1.5 hrs. The RNAs were resolved in an 8 M Urea/5% polyacrylamide gel and detected with a Typhoon Phosphorimager.

Protein extraction and western blotting

To extract proteins from plants, 25 mg inflorescence material was ground in liquid nitrogen and homogenized with 150 µl 1XSDS loading buffer supplied with 0.5% (v/v) 2-mercaptoethanol and 1% (v/v) protease inhibitor cocktail mix. The extracts were boiled for 5 min, followed by centrifugation at 16,000 g at 4°C for 5 min. 15 µl supernatant was resolved in a 10% SDS-PAGE gel. To assay the AGO1 or AGO10 immunoprecipitation samples, 8 µl immune complexes were boiled with an equal volume of 2XSDS loading buffer with 1% (v/v) 2-mercaptoethanol and 2% protease inhibitor cocktail mix. The proteins were resolved in a 7.5% SDS-PAGE gel and a monoclonal anti-GFP antibody (Abcam, Cat# ab3277) was used for immunodetection.