In plants, roots are essential for water and nutrient acquisition. MicroRNAs (miRNAs) regulate their target mRNAs by transcript cleavage and/or inhibition of protein translation and are known as major post-transcriptional regulators of various developmental pathways and stress responses. In Arabidopsis thaliana, four isoforms of miR169 are encoded by 14 different genes and target diverse mRNAs, encoding subunits A of the NF-Y transcription factor complex. These miRNA isoforms and their targets have previously been linked to nutrient signalling in plants.
By using mimicry constructs against different isoforms of miR169 and miR-resistant versions of NF-YA genes we analysed the role of specific miR169 isoforms in root growth and branching.
We identified a regulatory node involving the particular miR169defg isoform and NF-YA2 and NF-YA10 genes that acts in the control of primary root growth. The specific expression of MIM169defg constructs altered specific cell type numbers and dimensions in the root meristem. Preventing miR169defg-regulation of NF-YA2 indirectly affected laterial root initiation. We also showed that the miR169defg isoform affects NF-YA2 transcripts both at mRNA stability and translation levels.
We propose that a specific miR169 isoform and the NF-YA2 target control root architecture in Arabidopsis.
MicroRNAs (miR) are small noncoding RNAs that regulate gene expression post-transcriptionally and therefore control development, stress responses and hormone signalling or metabolism in both animals and plants (Bartel, 2009; Voinnet, 2009). In Arabidopsis, miRNAs are mainly processed by Dicer-like 1 (DCL1) in 21 nt double-strand RNAs from long stem-loop precursor RNAs. Generally, the mature miR is incorporated into a RNA-induced silencing complex (RISC) which contains a specific pool of ARGONAUTE 1 (AGO 1) proteins (Schott et al., 2012). These loaded RISCs are subsequently able to guide post-transcriptional gene silencing (PTGS) of complementary or partially complementary mRNA by cleavage and/or inhibition of translation. In plants, miR-dependent translation inhibition could be separated from transcript cleavage activity (Brodersen & Voinnet, 2009).
Amongst plant miRNA genes, 22 families are conserved between monocots and dicots. These conserved miRs are usually produced by several loci and highly expressed (Rajagopalan et al., 2006), whereas recent or young genes are in single or low copy-number, species specific and expressed at low levels (Axtell, 2008). MiR169, the appearance of which in evolutionary time coincided with that of Angiosperms, has been identified in more than 40 species (Sunkar & Jagadeeswaran, 2008), where it often makes up the largest miR family. The Arabidopsis miR169 family contains 14 genes. However, only four mature miR isoforms (a, bc, defg and hijklmn), differing by one or two nucleotides at the 5′ or 3′ ends (Supporting Information Fig. S1), are produced (Li et al., 2010). The miR169 isoforms present distinct expression patterns during development (Gonzalez-Ibeas et al., 2011), in response to biotic (Singh et al., 2012) or abiotic stresses (Licausi et al., 2011; Zhao et al., 2011), suggesting a functional specialization.
In plants, the main targets of miR169 are genes that encode the subunit A of Nuclear Factor Y (Rhoades et al., 2002). This transcription factor (TF), also named heme-activated protein (HAP) or CCAAT-box binding factor (CBF), is a heterotrimeric TF composed of NF-YA (HAP2), NF-YB (HAP3/CBF-A) and NF-YC (HAP5/CBF-C) subunits. NF-Y TF recognizes the CCAAT motif frequently observed in eukaryotic promoters at position −60/−100 from the transcriptional start site (Mantovani, 1999). NF-YB and NF-YC subunits contain a histone fold domain very similar to H2A and H2B core histones (Baxevanis et al., 1995; Zemzoumi et al., 1999) and these two subunits must form a heterodimer for stable interaction with NF-YA. The NF-YA subunit confers the sequence specificity to the TF complex by an unknown mechanism (Mantovani et al., 1994; Dolfini et al., 2012). In animals, its fixation on DNA could affect specific post-translational modifications of histones and thus modify the chromatin status (Donati et al., 2008; Gatta & Mantovani, 2011). The NF-Y TFs have mainly been considered as activators of transcription; however, recent evidence suggests their involvement in gene repression (Ceribelli et al., 2008; Leyva-González et al., 2012). In contrast to animals where NF-Y are generally single genes, in all plant species examined so far, the three NF-Y subunits are encoded by multigenic families (Edwards et al., 1998). In Arabidopsis there are 10, 13 and 13 genes for the A, B and C subunits, respectively, and Cao et al. (2011) found 36 NF-Y genes in Brachypodium distachyon. The NF-Y genes present differential expression patterns during development (Gusmaroli et al., 2001, 2002; Siefers et al., 2009; Cao et al., 2011), or in response to environmental conditions (Pant et al., 2009; Des Marais et al., 2012), suggesting that, in different organs or under certain stimuli, only some combinations of subunits can be assembled to form the trimeric functional NF-Y factor. In plants, NF-Y TFs have been linked to development (Lotan et al., 1998; Combier et al., 2006; Wenkel et al., 2006), signalization (Warpeha et al., 2007) and responses to stresses (Nelson et al., 2007; Li et al., 2008; Liu & Howell, 2010; Zhao et al., 2011). Overexpression of one subunit may be sufficient to affect NF-Y function when the other NF-Y partners are also present in the same tissues, as several phenotypes were observed after individual NF-Y subunit overexpression. However, it could not be ruled out that the results observed were consequences of interactions with member(s) that normally do not interact with this specific subunit. Recently, two reviews have been published concerning plant NF-Y complexes (Petroni et al., 2012; Laloum et al., 2013).
Both miR169 and their NF-YA target genes present complex spatio-temporal expression patterns in response to abiotic stresses. Arabidopsis seedlings under cold stress show an over-accumulation of miR169 correlated with a reduction of some NF-YA target transcripts (Zhou et al., 2008; Lee et al., 2010). MiR169 may also play a role in long-distance signalling, as its abundance in phloem sap decreases during nitrogen (N) and phosphorus (P) limitation (Pant et al., 2009; Buhtz et al., 2010) and several miR169 NF-YA targets are induced (Pant et al., 2009; Zhao et al., 2011). Moreover, overexpression of miR169a resulted in both a reduction of N accumulation and an increased sensitivity to N starvation (Zhao et al., 2011).
Recent data clearly showed that small noncoding RNAs, including miRs, involved in response to environmental stimuli, play a role in root growth or lateral root (LR) organogenesis and therefore on root architecture (Khan et al., 2011). The root apical meristem (RAM) is established during embryogenesis and is composed of a quiescent center (QC) surrounded by initial cells. These stem cells subsequently supply different root cell files through divisions and differentiation. They also produce the columella and lateral root cap (Petricka et al., 2012). The RAM size determines primary root growth and mutants affected in the division/differentiation of the initials produce short roots. In Arabidopsis, at a certain distance from the root apex, LR are initiated from specific pericycle cells in front of the xylem poles (Dubrovsky et al., 2000, 2001). These pericycle cells, blocked in a specific phase of the cell cycle (Himanen et al., 2002), are primed in the elongation zone (EZ) by oscillations of auxin amounts in adjacent xylem cells (De Smet et al., 2007). Auxin is a major determinant of LR initiation. However, recently, it has also been proposed that oscillating gene expression patterns are involved in priming of pericycle cells and thereby specify the prebranch sites (Moreno-Risueno et al., 2010). In these sites, LR initiation starts with the asymmetric division of the two primed pericycle cells (Fukaki et al., 2002; Vanneste et al., 2005). Subsequently, both periclinal and anticlinal divisions allow the formation of the LR primordium (LRP). This primordium will emerge and the apical meristem is activated to allow growth of the novel LR. In Medicago truncatula, we have recently shown that root architecture is affected by miR396 activities. Indeed, miR396 inactivation by mimicry construct, which induced an over-expression of the miRNA targets, resulted in a clear increase of root biomass (Bazin et al., 2013).
In this work, we have analysed the role of specific miR169 isoforms in root growth and branching. By blocking miR169 action with different mimicry constructs, we showed that a regulatory node involving the miR169defg isoform and its NF-YA2 and NF-YA10 targets acts to control RAM length and LR density. Furthermore, perturbing the regulation of NF-YA2 by miR169defg indirectly affects LR initiation, suggesting a role for this miR169 isoform in LR organogenesis and root architecture.
Materials and Methods
Plant material and growth conditions
All lines used are in the Arabidopsis thaliana Columbia (Col0) ecotype. Mimicry lines MIM169abc and defg have been described previously (Todesco et al., 2010). Seedlings were grown under long photoperiod (16 h : 8 h, light : dark) unless stated otherwise in half strength Murashige & Skoog (1962) medium (Duchefa) supplemented with 1% (w/v) sucrose.
For pNF-YA2:GUS, pNF-YA10:GUS, c. 2000-pb region upstream of the start codon was amplified from genomic DNA (all primers sequences used for cloning are in Table S1), to generate Gateway pENTR-D vectors which were recombined in the Gateway binary vector pKGWFS7 (Karimi et al., 2007). The same procedure was used for the pmiR169d:GUS and pmiR169e:GUS reporter lines; a 1800-pb region upstream of the miRNA stem loop was amplified from genomic DNA. For the p35S:miR169a and d lines, 300–400-bp fragments surrounding the miRNA sequence were amplified from genomic DNA, cloned into pENTR-D and transferred in Gateway binary vectors pH7WG2 or pK7WG2D, respectively.
For the transient experiments, the 3′UTR of NF-YA2 or NF-YA1 was fused to the GFP coding region and cloned into the pENTR-D. p35S:GFP-NF-YA2wt and the p35S:GFP-NF-YA1wt were generated by recombination in a homemade Gateway-compatible vector (Marin et al., 2010) or the binary vector pH7WG2, respectively. The miR resistant versions p35S:GFP-NF-YA1mut and p35S:GFP-NF-YA2mut were obtained by deleting the miR recognition site as shown. To overexpress miR169a the p35S:miR169a was used. For p35S:miR169d and p35S:miR169b a 200–400-bp fragment surrounding the miRNA sequences was amplified from genomic DNA, cloned into pENTR-D and recombined in the binary vector pH7WG2.
Lines expressing the GFP under the NF-YA2 promoter were transformed by plasmids that resulted from a double recombination of pDONR221 containing GFP-NF-YA2 (wild-type (WT) or mutated) and the NF-YA2 promoter into the binary vector pK7m24GW,3.
For NF-YA2 overexpressing lines, the HA tag was fused in frame to the NF-YA2 cDNA including the 3′UTR. A double recombination of this construct and the NF-YA2 promoter were cloned into the vector pK7m24GW,3. The miR resistant version was generated as previously described for GFP constructs.
Col0 transformation was performed by the floral dip method (Clough & Bent, 1998) and independent stable transgenic lines were selected.
Histochemical and histological analyses
The GUS assays were performed as described (Beeckman & Engler, 1994). To analyse expression in the whole organ, roots were cleared in chloral hydrate solution. Transversal and longitudinal sections were obtained after GUS staining and technovit 7100 resin embedding, using the protocol described by Vanneste et al. (2005) and De Smet et al. (2004).
Imaging and image analysis
For epifluorescence imaging, roots (mounted in 50% glycerol) were directly photographed on a widefield microscope (DMI6000B; Leica Microsystems, Wetzlar, Germany) with a ×20 objective. GFP images were made every 0.59 μm to scan the entire root. Deconvolution was done with Huygens software (Scientific Volume Imaging, Hilversum, the Netherlands).
To visualize GUS activity in whole roots we used a microscope equipped with Nomarski optics (DMI6000B; Leica Microsystems).
Measurements of primary root length and lateral root number were made on scanned images of Petri-dishes containing the seedlings grown on the surface of the agar medium using ImageJ (http://rsb.info.nih.gov/ij/). The number of LRP, in the lateral root formation zone as defined by Dubrovsky & Forde (2012), was counted by using the Nomarski microscope in roots cleared by choral hydrate. The classification of LRP developmental stages was performed according to Malamy & Benfey (1997).
Measurements of meristem size and cell number
The cell wall of the roots was stained according to Truernit et al. (2008). A Leica SP8 confocal laser scanning microscope (Leica Microsystems) was used to image root tips. Serial optical sections were reconstituted into 3D image stacks and measurements were made using ImageJ. Cell length and meristem length are averages of two opposite cell files; cell width and thickness are averages of all visible cells in a cross section.
RNA extraction, RNA blot assays, quantitative RT-PCR
Total RNA was extracted as described (Mallory et al., 2001) or by using the Trizol procedure as described by the manufacturer (Life Technologies SAS, Saint Aubin, France). For RNA gel blot analysis, 10 μg of RNA was separated by denaturing (7 M Urea) 15% polyacrylamide gel electrophoresis, blotted to a nylon membrane (Hybond NX; GE Healthcare, Vélizy-Villacoublay, France), and EDC cross-linked as described (Pall et al., 2007). miRNA probes were prepared by end-labelling antisense oligonucleotides with 32P using T4 polynucleotide kinase (Thermo Fischer Scientific Bioscience, Villebon-sur-Yvette, France). RNA gel blots were hybridized (Mallory et al., 2001) with the miR169 probe together with U6 probe. Nonsaturated signals were scanned on a Molecular Dynamics Storm 840, and analysed with ImageJ.
For quantitative RT-PCR analysis (all qPCR primers sequences used can be found in Table S2), total RNA (5 μg) was subjected to DNase I treatment (Thermo Fisher Scientific) and reverse transcribed by using oligodT with Revertaid (Thermo Fisher Scientific). mRNA concentrations were assayed in two to four independent biological replicates and, for each gene, two technical replicates were done. Real time RT-PCR was performed on a Roche LightCycler480 using Roche reagents (http://www.roche.com). For the Arabidopsis experiment the At1g13320 and At4g26410 (Czechowski et al., 2005) genes were used as housekeeping genes (HKG) for the normalization of the expression of the target gene (TG). The use of geNORM software (Vandesompele et al., 2002), established that these genes were the most stable whatever the organs, genotypes or hormone treatments. Relative expression levels were calculated following this formula ΔCycle Threshold (Ct) = CtTG – mean (CtHKG) and transformed in relative value expression (REV), where REV = 2−ΔCt. The ΔΔCt method was used to analyse the relative change in gene expression between two conditions, where ΔΔCt = ΔCtmutant/NAA − ΔCtWT/NPA, and transformed in relative value expression, where REV = 2−ΔΔCt (Livak & Schmittgen, 2001). For the transient experiments in Nicotiana benthamiana, UBI3 (AY912494) and EF1α (D63396) were used as HKG and ΔCt and ΔΔCt were used as already described.
The miR169 and miR171bc profiles were obtained by stem–loop RT-PCR as previously described by Varkonyi-Gasic et al. (2007) with reverse transcription primers designed according to Chen et al. (2005; Table S2) and a pair of qPCR primers (Table S2; Varkonyi-Gasic et al., 2007). qPCR analysis was performed as already described.
Agroinfiltration and transient experiment
Agroinfiltration was performed in 3-wk-old N. benthamiana leaves. p35S:miR169a, p35S:miR169b, p35S:miR169d, p35S:miR390b, p35S:GFP-NF-YA1wt, p35S:GFP-NF-YA1mut, p35S:GFP-NF-YA2wt, 35S:GFP-NF-YA2mut and p35S:mcherry were transformed into Agrobacterium tumefaciens. For co-infiltrations, individual overnight cultures were harvested and suspended at OD 1 in 10 mM MgCl2, 10 mM MES, pH 5.8 and 150 μM acetosyringone. After 3 h of incubation at room temperature, various combinations were mixed at a 1 : 1 : 1 ratio; these mixes were used to infiltrate leaves. These leaves were harvested 3 d after the infiltration. Half of them were used immediately for protein extraction and the other half was frozen and used for RNA extraction as described above. The p35S:mcherry was always used as an infiltration control and p35S:miR390b (Marin et al., 2010) as a control of protein and RNA stability in absence of the targeting miR.
Protein extraction was performed according to Elliott et al. (2006). Protein concentration was quantified by Bradford assay. Fluorescence was performed, in duplicate, on 10–20 or 30 μg of total proteins for the transient or stable experiments, respectively, by using the Infinite 200 fluorometer (TECAN Group Ltd, Männedorf, Switzerland). The excitation and emission parameters were 475 nm and 510 nm for GFP and 565 nm and 610 nm for mcherry. For transient experiments, the GFP fluorescence was normalized by the mcherry fluorescence. For the stable experiments, the GFP fluorescence was calculated for 1 μg of proteins.
Statistical analysis and comparison of the agroinfiltration
As the level of expression in the transient experiments could be highly different between each infiltrated leaf, we compared the results obtained from mRNA concentration and fluorescence intensity in leaves transformed with the mutated form of the NF-YA 3′UTR against those obtained with the WT form of the NF-YA 3′UTR. All pair ratios were calculated. For mRNA comparison, the formula used is ratiomRNA = 2−(ΔCtGFP − ΔCtmcherry)mut/2−(ΔCtGFP − ΔCtmcherry)WT, where ΔCtGFP = CtGFP − CtHKG and ΔCtmcherry = Ctmcherry − CtHKG. For the GFP fluorescence, the formula is ratiofluorescence = (GFP fluorescence/mcherry fluorescence)mut/(GFP fluorescence/mcherry fluorescence)WT. Statistical analysis were performed on R 2.14.2 (R Development Core Team, 2012).
Root architecture is affected in MIM169defg plants
In order to study the potential role of miR169 in root architecture, we analysed mimicry plants generated by Todesco et al. (2010) for their root phenotype. These transgenic plants expressed specific constructs that sequester either both miR169abc and miR169hijklmn isoforms or only the miR169defg isoform, driven by the cauliflower mosaic virus 35S promoter. These lines named MIM169abc or MIM169defg by Todesco et al. (2010) were reported to present smaller rosettes. In contrast to MIM169abc (Fig. S2), MIM169defg plants exhibited shorter primary roots when compared with wild-type (WT) plants (Fig. 1a,b). This root phenotype was linked with slower root growth kinetics after seedling germination (Fig. S3). As PR were shorter and the LR number remained unchanged, LR density was enhanced in the MIM169defg lines (Fig. 1b,c). MIM169defg LR length was also smaller (Fig. 1d). For PR, smaller root length correlated with a shorter meristem length (Fig. 1e,f) at 5 but also at 7 or 12 d-after-germination (DAG). Interestingly, as shown in Fig. 1(f–g), global root thickness was also reduced in these plants. Indeed, the MIM169defg root diameter was smaller than in WT (Fig. S4). A reduction in the number of endodermal cells was observed (n =10) (Fig. 2a). Moreover, as shown in Fig. 2(b), significant differences in the thickness of nearly all cell types were detected in 7-DAG roots (n =15). Interestingly, only the width of the two cells which arise from the division of a same initial cell was significantly affected (n =15, Fig. 2c). Therefore, expression of MIM169defg constructs (but not MIM169abc) altered specific cell type numbers and dimensions, root meristem size and, consequently, primary root growth.
NF-YA2 and NF-YA10 levels are strongly altered in MIM169defg lines
We analysed the levels of NF-YA transcripts in MIM lines. All NF-YA transcripts except NF-YA4 and NF-YA7 contained putative miR169 binding sites in their 3′UTR, even NF-YA6 which was not previously reported as a putative miR169 target (Fig. S1c). In contrast to NF-YA4 and NF-YA7, all miR169 NF-YA targets were faintly expressed in the tissues tested (Fig. S5a), suggesting that miR169-mediated regulation acts in the global regulation of this gene family. The detected NF-YA2, NF-YA10 and miR169 expression patterns were somehow distinct from those of Zhao et al. (2011) for roots and shoots, likely due to changes in growth conditions or seedling ages.
Siefers et al. (2009) reported that NF-YA genes can be separated into three different clades group 1 (NF-YA3, NF-YA5, NF-YA8 and NF-YA6), group 2 (NF-YA1 and NF-YA9) and group 3 (NF-YA2 and NF-YA10). Bioinformatics analyses (http://mpss.udel.edu/at_sRNA/), indicated that NF-YA transcripts might potentially be recognized by different miR169 isoforms (Fig. S1b). Indeed, group 1 is targeted by three isoforms (miR169a-bc-hijklmn), group 2 by all four isoforms (miR169a-bc-defg-hijklmn) and group 3 by three isoforms (miR169bc-defg-hijklmn) (Fig. S1b). We thus explored the degree of specificity of the MIM constructs on the action of miR169 isoforms. Expression of the 10 NF-YA genes in 5-d-old roots of MIM169abc and MIM169defg seedlings showed an over-expression of all the miR169 targeted transcripts in both types of MIM roots when compared with WT. However, major differences between the MIM lines were detected for NF-YA2 and NF-YA10 which are induced more than 20-fold in MIM169defg plants (Fig. 3a).
By using stem-loop RT-qPCR with primers similar to those used by Liang et al. (2012), we were able to observe a significant and homogeneous decrease of all miR169 isoforms in roots of MIM169defg (Fig. S6b). Although a similar variation was observed for MIM169abc roots; we always detected higher amounts of the miR169defg isoform in these roots (Fig. S6a), in agreement with Todesco et al. (2010). The differences observed for miR169defg between MIM lines and the highest over-accumulation of NF-YA2 and NF-YA10 transcripts only in MIM169defg lines, suggest that the miR169defg isoform is responsible for this regulation. The over-expression of corresponding proteins (eventually together with other specific NF-YA proteins) could explain the root apex phenotype which is not observed in MIM169abc plants although we cannot exclude an effect of other miR169 targets. Indeed, analysis of transcript degradation products in degradome analyses of Arabidopsis inflorescences (German et al., 2008; Meng et al., 2012) suggest the potential existence of non-NF-YA targets for miR169. Finally, as our results were identical for the four isoforms in MIM169defg lines, an alternative possibility is that, in our RT-qPCR conditions, the primers used were not able to discriminate strictly between the miR169 isoforms, which differ only by two nucleotides at the 3′ or 5′ end of mature miRs. Hence, the NF-YA2 and NF-YA10 transcription factors are specifically overexpressed in MIM169defg lines.
Different NF-YA genes are primarily regulated by specific miR169 isoforms
In order to test whether particular miR169 isoforms could target specific NF-YA transcripts, we performed stable transformations of Arabidopsis plants. For that, we overexpressed two miRNA precursors that produce two different isoforms, miR169a or miR169d (Fig. S7a,b). In 5-d-old roots of three independent 35S:miR169a OE plants, qRT-PCR experiments detected a slight reduction of transcript abundance for all miR169 targeted NF-YA transcripts but NF-YA2, (Fig. 3b), which was not predicted as a target of this isoform. By contrast, for NF-YA10, another gene not predicted as miR169a target, we detected the lowest transcript accumulation (Fig. 3b). Hence, as previously proposed by Zhao et al. (2011) NF-YA10 transcripts are also cleaved by miR169a isoform. By the same type of experiment, reductions of NF-YA2 and NF-YA10 transcript abundances were detected in roots of the miR169d OE lines, whereas slight differences were observed for the other NF-YA transcripts (Fig. 3c). However, no root growth or lateral root phenotype has been observed for these miR169 OE lines (Fig. S8). Thus, a 50% reduction in NF-YA2 and NF-YA10 transcript accumulation seems insufficient to affect root growth. In this respect, in MIM169defg an increase of at least × 4 or 5 (log2 fold changes) of transcript accumulation has been detected for these two genes. Moreover, we cannot exclude the idea that the amounts of other NF-YA proteins (and/or other potential targets) could be involved in the phenotype. Taken together, opposite regulation of NF-YA2 and NF-YA10 in MIM169defg and miR169d OE lines confirmed the specificity of targeting of these two transcripts by this isoform. In roots or seedlings of miR169 OE lines, a partial reduction of NF-YA transcripts was observed (our work and Li et al., 2008; Zhao et al., 2011). This suggests that mRNA cleavage is only part of miR169 action on its targets.
miR169 isoforms could affect the translation of NF-YA2 gene
As miR169-recognition sites are in the 3′UTR of NF-YA genes, this miRNA may act at two levels: RNA stability and translation inhibition. To analyse the mode of action of miR169, we used transient N. benthamiana transformations with different p35S:GFP-NF-YA constructs. For that, we expressed, under the control of the 35S promoter, the GFP reporter gene fused to the 3′UTR of NF-YA1 (p35S:GFP-NF-YA1wt) or NF-YA2 (p35S:GFP-NF-YA2wt) genes. In addition, we constructed a miR-resistant version (p35S:GFP-NF-YA1mut) and (p35S:GFP-NF-YA2mut) in which the miR169 binding site was deleted from the NF-YA1 or NF-YA2 3′UTR (Fig. S9). These constructs were used for co-infiltrations with strains that overexpress miR169a, miR169bc, miR169defg isoforms or miR390b (as a negative control) and p35S:mcherry (as a transfection control). GFP protein and mRNA concentrations were measured 72 h post-infiltration and normalized by quantification of mCherry fluorescence or mRNA concentrations (see the 'Materials and Methods'). For all these samples, the quantification of miRNA accumulation was done by northern blots (Fig. S10). For each miRNA tested against NF-YA1 or NF-YA2 constructs, we analysed different independent experiments for mRNA or protein (fluorescence) ratio (p35S:GFP-NF-YA1mut/p35S:GFP-NF-YA1wt) and (p35S:GFP-NF-YA2mut/p35S:GFP-NF-YA2wt). This data presentation is well adapted when results show large dispersion in absolute values among experiments. Figure 4(a) left and right corresponds to the ratio of NF-YA1 mRNA and protein, respectively. In this system, as miR390b cannot target NF-YA1 3′UTR, co-infiltrations with miR390b had no effect on the ratio of GFP (mut)/GFP (wt) mRNA (Fig. 4a left) or fluorescence (Fig. 4a right). Our results for miR169a and miR169d were very similar to this control; thus, we concluded that these two miRNAs have no effect on NF-YA1 transcripts or protein concentrations, whereas, for miR169b, we observed a slight variation. Figure 4(b) left and right correspond to the ratio of NF-YA2 mRNA and protein, respectively. In this system, miR390b cannot target NF-YA2 3′UTR, thus co-infiltrations with miR390b had no effect on the ratio of GFP (mut)/GFP (wt) mRNA (Fig. 4b left) or fluorescence (Fig. 4b right). Again miR169a and miR169b gave similar results as the control, suggesting no effect on NF-YA2 expression. Interestingly, we observed robust differences for miR169d at both the NF-YA2 transcript and protein levels (Fig. 4b left and right). Therefore, the miR169defg isoform likely targets NF-YA2 transcripts both at the mRNA stability and translation levels. This double modality of action was also observed in roots of seedlings that stably overexpressed p35S:GFP-NF-YA2wt or p35S:GFP-NF-YA2mut (Fig. 4c). Indeed, in these transgenic plants, the impact of miR169 regulation on NF-YA2 was 2.5× and 5× at the transcript and protein levels, respectively. Taken together, these results show that the miR169defg isoform acts on NF-YA2 gene expression at both the transcript stability and translational levels in Arabidopsis roots as well as in N. benthamiana leaves.
NF-YA2 and NF-YA10 are expressed in vascular tissues
In order to explore the expression patterns of NF-YA2 and NF-YA10, the two miR169defg targets, we transformed WT plants with promoter:GUS-GFP constructs that contained the 2-kb genomic sequence upstream from their initiation codons (Fig. 5). The expression patterns in shoots were identical to those of Siefers et al. (2009), with a high level of expression in the cotyledon vasculature and hypocotyl for pNF-YA2:GUS and a faint expression at the apical zone for pNF-YA10:GUS (Fig. 5a). In roots, high staining of pNF-YA2:GUS was visible in vascular tissues (Fig. 5a,b,d) whereas a weaker signal was observed for pNF-YA10:GUS (Fig. 5a,c,e). For these two genes, expression in vascular tissues started at the EZ (Fig. 5b–e). In 5-d-old seedlings, a strong pNF-YA2:GUS signal was transitorily detected in the columella (Fig. 5a,b). Moreover, in 5-DAG roots, a clear expression of NF-YA10 was discerned in xylem cells at the transition zone (Fig. 5c), whereas the expression of NF-YA2 was muted in pericycle and very faint in endodermal cells (Fig. 5b). Transversal sections and longitudinal views revealed low staining in pericycle and vasculature (Fig. 5d,e). These patterns are in agreement with those of Siefers et al. (2009). GUS staining was also observed in vascular tissues of emerged LR (Fig. 5f,g) and at the LR tip for pNF-YA2:GUS.
Deregulation of NF-YA2 increases LR initiation
Our attempts to identify knockdown plants of NF-YA2 and NF-YA10 in SALK lines and/or to prepare RNAi lines using a constitutive promoter to downregulate either one or the two genes were unsuccessful. Pagnussat et al. (2005) previously showed that a T-DNA insertion in the NF-YA2 gene resulted in defect of fertility (mutant line une8) and our lines overexpressing NF-YA2 under the 35S constitutive promoter showed strong sterility and retarded growth (Fig. S11). As shown above, 35S:mir169a and 35S:mir169d did not induce very strong reductions of NF-YA transcripts in roots and hence cannot be considered as mutant KO lines (Fig. 3b,c).
In order to further characterize the function of NF-YA2, we generated transgenic plants which expressed NF-YA2 gene under its own promoter. Again two constructs, pNF-YA2:HA-NF-YA2wt having the intact 3′UTR region of the gene or pNF-YA2:HA-NF-YA2mut (with miR site deletion in the 3′UTR) were used. Expression of the intact NF-YA2 gene, sensitive to miR169 repression, had no effect on root architecture (Fig. S12). For pNF-YA2:HA-NF-YA2mut lines, the size of all aerial organs of adult plants (3 wk old) was always smaller in the glasshouse. By contrast, in vitro, during the first 10 d of culture, we never saw a difference between the rosette leaves size of pNF-YA2:HA-NF-YA2mut lines and WT. Nine-day-old plants grown in vitro produced more LRs than the WT (Fig. 6a,c) without affecting the PR size (Fig. 6a,b). Hence, the emerged lateral root density was higher in these plants (Fig. 6c). Statistic analysis of LR densities obtained for three independent lines strengthened the fact that the NF-YA2 gene plays an important role in LR development. We further analysed the effects of NF-YA2 OE on LR formation by counting the numbers of LR primordia from stages I to VII as described by Dubrovsky & Forde (2012). Increase in stages I to IV primordia could suggest a role of NF-YA2 on LR initiation (Fig. 6d). LR initiation requires auxin responses, mediated by several AUX/IAA modules (Goh et al., 2012). In order to discriminate between direct or indirect action on NF-Y complex in LR development, we used the lateral root induction system described by Himanen et al. (2002). Under these conditions, we were not able to observe any differences in the transcript levels of genes involved in LR initiation (IAA2, IAA11, GATA23, E2FA and SHY2/IAA3 genes) in pNF-YA2:HA-NF-YA2mut lines when compared with WT (Fig. S13). These latter results clearly suggest that NF-YA2 and/or the NF-Y complex act indirectly on LR density and root architecture.
The regulatory miR169defg/NF-YA2 node acts at least in pericycle cells
In order to explore miR169defg expression patterns, we transformed WT plants with constructs that contained the 2-kb genomic sequences upstream of the miR169d or miR169e stem-loop structures fused to GUS-GFP reporter genes. Indeed, in our culture conditions, we were never able to detect miR169 f and g precursors. Faint histological GUS staining was detected for miR169d and miR169e promoters in several independent transgenic lines (Fig. 7a). In our conditions, we never observed a strong signal in the stele for miR169e, as previously described by Li et al. (2010). Root transversal sections showed GUS staining at phloem poles for miR169d and miR169e promoters (with an additional expression in pericycle cells for miR169e), whereas NF-YA2 and NF-YA10 were expressed in the whole pericycle (Fig. 7b) and faintly in the vasculature. These results are in agreement with data derived from the analysis of small RNAs libraries which detected miR169 in phloem cells at least in Arabidopsis and Brassica napus (Buhtz et al., 2010).
In order to visualize root cells where miR169 can post-transcriptionally regulate NF-YA2, we analysed crosses between MIM169defg or WT and pNF-YA2:GFP-NF-YA2wt plants. In WT background, we were not able to detect any GFP signal (Fig. 8a). By contrast, in MIM169defg background (Fig. 8b) where miR169 were sequestered, or in pNF-YA2:GFP-NF-YA2mut lines, where NF-YA2 was overexpressed (Fig. 8c), a clear GFP signal was observed in specific cell files, likely pericycle cells although the faint fluorescence did not allowed us to do a 3D deconvolution. A faint expression of NF-YA2 gene has also been observed in the stele that may be linked to miR169d and miR169e expression in phloem. Notwithstanding this, it has been shown – at least for miR165/166 – that miRNAs can circulate through plasmodesma from pericycle to stele (Carlsbecker et al., 2010) in an opposite manner to certain TF or target transcripts. One alternative could also be that miR169 moves from the phloem to pericycle cells to inhibit its targets.
Modulation of root architecture, by PR root growth and LR development, allows plants to cope with abiotic stresses such as drought or nutrient availability. Both miRNAs and transcription factors are known to play crucial roles in plant development processes and in response to environmental conditions but few have been linked to root growth and architecture (Khan et al., 2011). Here, by using a mimicry approach and deregulation of the miR169 targets NF-YA2 and NF-YA10, we have shown that genes regulated by the miR169defg isoform were involved in primary root growth and/or LR initiation. In Arabidopsis, miR169 has previously been involved in drought and nitrate responses. Indeed, overexpression of the miR169a isoform, which principally targets NF-YA5, induced drought sensitivity or altered nitrogen responses (Li et al., 2008; Zhao et al., 2011) and miR169defg isoform is induced by nitrate starvation in Arabidopsis roots (Liang et al., 2012). Moreover, in roots, a clear upregulation of NF-YA2 and NF-YA10 has been observed in response to phosphate starvation (Woo et al., 2012). Recently, functional and transcriptional analysis of Arabidopsis NF-YA gene family also revealed a clear impact of NF-Y complexes on carbohydrate metabolism and cell wall modifying enzymes (Leyva-González et al., 2012). These results, all together, are consistent with the fact that miR169 target genes, the NF-YA family, could act directly or indirectly in developmental responses to environmental signals such as the modification of root architecture induced by abiotic stresses. Indeed, all of these stresses can affect, at different levels, auxin and sucrose homeostasis and transport (Kazan, 2013; Lemoine et al., 2013) two well-known operators of growth and development in plants. The impact of the miR169/NF-Y complex on root architecture identified here and more particularly the regulation of LR density could be linked to the regulation observed in M. truncatula – nodule development by miR169 (Combier et al., 2006). The formation of nodules is actually a reprogramming of root cells into a new organ to allow legume plants to cope with nitrogen starvation.
In Arabidopsis, promoter:GUS reporter lines (Li et al., 2010) have previously shown that many miR169 genes are mostly expressed in vascular tissues, and some genes have presented distinct patterns in these tissues. The maximum level of GUS activity in roots has been observed for miR169e. However, in roots or seedlings all miR169 precursors have been detected in similar amounts (Li et al., 2008; Zhao et al., 2011). Using cell-sorting technology on cell types of three root regions followed by deep sequencing, Breakfield et al. (2012) have observed that the miR169 gene family was moderately expressed and that miR169 isoforms presented the lowest variance in quantity across root tissues. We have limited our investigation to the miR169d and miR169e genes, which produce the miR169defg isoform and whose expression was detected in phloem pole cells in transversal root sections. This is consistent with both Li et al. (2010) and the existence of an abundant quantity of mature miR169 in small libraries from phloem cells (Buhtz et al., 2010). Taken together, these data suggest that miR169 genes could be principally expressed in the stele. The homogenous accumulation of miR169 in different cell files in the root reported by Breakfield et al. (2012) may thus be explained by different mechanisms such as changes in miR169 stability in certain cell types or even movement between cell files. Alternatively, the 2-kb promoter length used in this study may not accurately reflect the expression of miR169d and miR169e genes.
The miR169defg genes and their related NF-YA2 and NF-YA10 targets are both expressed in root stele and vascular tissues. In MIM169defg lines, high over-expression of the two latter genes (eventually together with minor changes in other NF-YA genes) was responsible for a reduction of root length and RAM size which correlated with a decrease in cell numbers in the RAM. Thus, sequestration of the miR169defg isoform could reduce cell division or increase cell recruitment at the meristem transition zone. However, NF-YA2 and NF-YA10 expression patterns, over the transition zone, suggested that these genes do not act directly on cell cycle. Transcriptomic analyses of PXVE:NF-YA lines (inducible OE lines) by Leyva-González et al. (2012) showed that many strongly downregulated genes encoded enzymes involved in cell wall modifications. Cell wall composition changes could affect cell shape or size and may disturb the function or transport of vascular tissues or affect apoplastic pathways of nutrients, and consequently impact the primary root meristem. This could partly explain the phenotype observed for MIM169defg lines at the cellular level.
Our promoter-GUS studies showed NF-YA2 and NF-YA10 expression in pericycle cells, whereas miR169d and miR169e are expressed in phloem poles. As LRs arise from pericycle cells close to the xylem poles, we can speculate first that miR169 could migrate from phloem to specific pericycle cells in front of these poles to strongly inhibit NF-YA2 expression or secondly that, in these specific cells, only miR169 f and g genes are expressed.
When the post-transcriptional regulation of NF-YA2 transcripts by miR169, was disrupted (pNF-YA2:HA-NF-YA2mut lines) the size of all aerial organs was smaller than that of soil-grown WT plants. In contrast to this aerial phenotype observed in older plants, in 9-DAG plants, the number of LRs was increased when compared with WT yet there was no impact on the rosette leaves size. As previouly mentioned, sugars play an important role in meristem activity and can leave phloem through either an apoplastic or a simplastic pathway (Patrick, 1997). One possibility is that NF-YA2 overexpression in pNF-YA2:HA-NF-YA2mut lines specifically affects transport into sink tissues. Indeed, an increase in sucrose content has been detected in p35S:NF-YA2 plants by Leyva-González et al. (2012). Alternatively, OE of NF-YA2 alone can induce a stress response known to promote root growth. Curiously, IAA12, one key gene of the LR initiation module, has been detected as a direct target of NF-YA2 by Leyva-González et al. (2012). By using the NPA/NAA LR induction system, we showed that the increase of LR in pNF-YA2:HA-NF-YA2mut lines did not have a direct effect on LRI but instead an indirect effect likely mediated through the vascular tissues. This indirect effect observed here could be linked to sucrose action as signal molecule or nutrient, as it can increase the number of LR (Roycewicz & Malamy, 2012). Alternatively, an indirect action on auxin homeostasis or transport could be involved as ten NF-YA2 potential direct targets are involved in the auxin pathway. In any way, the indirect action on root architecture could also result from differential gene expressions induced by modifications or deposition of histones by NF-Y complex, which encompass NF-YA2, as has been proposed in animals (Gatta & Mantovani, 2011).
MiR169 and their NF-YA targets have been proposed to control many plant responses to environmental stresses. Leyva-González et al. (2012) constructed a model in which the NF-Y complex inhibits an early abiotic stress response to allow the expression of a late response that confers resistance. By monitoring the precise action of miR169defg and NF-YA2 in specific cells, we showed that this regulary node indirectly connects the NF-YA2/NF-YA10 TFs to root growth and architecture, a trait very sensitive to the soil environment. This node controls root growth and development and is a new element to be considered for regulating root architecture.
We thank Marco Todesco, Ignacio Rubio-Somoza, Javier Paz-Ares and Detlef Weigel for making the MIM material available before publication and Samuel Mondy for help with statistical analysis and Fig. 4. This work benefited from the Tournesol and COFECUB exchanges Programs and from facilities and expertise of Imagif in Gif-Sur-Yvette (www.imagif.cnrs.fr), supported by the Conseil Général de l'Essonne. Marie Declerck received a grant from the French Ministry of Education. Work in M.C.'s lab is supported by the ROOT KBBE project and the LabEx SPS (https://www6.inra.fr/saclay-plant-sciences). Work in T.B.'s lab is partly supported by the Interuniversity Attraction Poles Programme (IAP VI/33 and IUAP P7/29 ‘MARS’) from the Belgian Federal Science Policy Office.