Non-cell-autonomous microRNA165 acts in a dose-dependent manner to regulate multiple differentiation status in the Arabidopsis root

Cell fate determination by positional cues is common to both animal and plant development. In animals, some positional cues activate a particular differentiation program in recipient cells regardless of their concentration, whereas another class of positional cues confer multiple differentiation programs in a dose-dependent manner (Tabata and Takei, 2004). In plant development, few, if any, of the intercellular signaling molecules identified thus far appear to control multiple differentiation status in a dose-dependent manner (Bhalerao and Bennett, 2003; Hara et al., 2007; Stahl et al., 2009; Kondo et al., 2010; Matsuzaki et al., 2010; Sugano et al., 2010).

Owing to its simple structure and amenability to genetic analysis, the Arabidopsis root has been used as a model system to study plant tissue patterning. In root cross-sections, different tissue layers are organized in a radially symmetric pattern, with the central stele surrounded by single layers of endodermis, cortex and epidermis (Fig. 1) (Dolan et al., 1993). The endodermis and cortex are together termed the ground tissue. The stele consists of the outermost pericycle layer, which surrounds the inner vascular tissue, in which two xylem cell types, protoxylem and metaxylem, are specified at the peripheral and central positions, respectively, along the single xylem pole (Fig. 1) (Carlsbecker and Helariutta, 2005). Although pattern formation in the Arabidopsis root relies on intimate cell-cell communication mediated both by ligand-receptor interactions (Hirakawa et al., 2008; Stahl et al., 2009; Matsuzaki et al., 2010) and mobile transcription factors (Kurata et al., 2005), these molecules generally impart a certain differentiation status to the recipient cells regardless of their dosage. For example, the GRAS-type transcription factor SHORT-ROOT (SHR) moves from the stele to the adjacent cell layer, where it promotes endodermis differentiation and activates the expression of another GRAS-type transcription factor, SCARECROW (SCR), which promotes periclinal cell division (Helariutta et al., 2000; Nakajima et al., 2001; Levesque et al., 2006). Ectopic expression of SHR confers SCR expression and endodermis differentiation to the recipient cells (Nakajima et al., 2001; Sena et al., 2004).

Recently, it has been reported that SHR and SCR together activate the transcription of two microRNA (miRNA) genes, MIR165A and MIR166B, in the endodermis (Carlsbecker et al., 2010). The products of MIR165A and MIR166B, possibly mature forms of miR165 and miR166, non-cell-autonomously suppress Class III homeodomain leucine zipper (HD-ZIP III) transcripts in the peripheral stele. miRNA-dependent suppression of HD-ZIP III, primarily that of PHABULOSA (PHB), in the peripheral stele is required for protoxylem differentiation (Carlsbecker et al., 2010). Although that study clearly demonstrated the non-cell-autonomous function of endodermis-derived miR165/166 in repressing HD-ZIP III expression in the stele, it is not yet clear whether this suppression is required solely for the differentiation of protoxylem in the peripheral stele or if it has a broader range of functions in root patterning. Furthermore, it is not known whether miRNA165/166 act simply to eliminate their target transcripts in the peripheral stele or if they act in a dose-dependent manner to define a pattern of target transcript distribution.

In this study, we first addressed the role of miR165/166-dependent suppression of PHB in ground tissue patterning and pericycle differentiation, and then analyzed the means by which ground tissue-derived miR165 regulates the PHB expression pattern in the stele. Analyses in an inducible miR165 expression line revealed that ground tissue-derived miR165 acts in a dose-dependent manner to establish a PHB expression gradient across the stele, which in turn is required for the correct differentiation of two xylem cell types and the pericycle.

Arabidopsis mutant/marker lines are listed in Table 1. AGI codes for the genes described in this study are as follows: ACT7, At5g09810; AHP6, At1g80100; ATHB8, At4g32880; AGO1, At1g48410; CNA, At1g52150; CRE1, At2g01830; JKD, At5g03510; MIR165A, At1g01183; MIR165B, At4g00885; MIR166A, At2g46685; MIR166B, At3g61897; MIR166C, At5g08712; MIR166D, At5g08717; MIR166E, At5g41905; MIR166F, At5g43603; MIR166G, At5g63715; PHB, At2g34710; PHV, At1g30490; REV, At5g60690; SCR, At3g54220; SHR, At4g37650; and SKOR, At3g02850. Because all Arabidopsis lines had the Col background except for phb-1d (in Ler) (McConnell and Barton, 1998), phb-1d was crossed twice with Col-0 to reduce possible phenotypic differences between the two ecotypes, and heterozygous phb-1d/+ plants were used throughout the study. Plant growth conditions were as described (Miyashima et al., 2009).

For quantitative real-time RT-PCR (qRT-PCR), RNA was extracted from the root meristem region (the distal ~1 cm of the root tip) using the RNeasy Plant Mini Kit (Qiagen, Valencia, CA, USA), and reverse-transcribed using Super Script II reverse transcriptase (Invitrogen, Carlsbad, CA) with oligo(dT) primers (Invitrogen). qRT-PCR was performed using Cyber Premix ExTaq (Takara Bio, Shiga, Japan) and the Light Cycler System (Roche Diagnostics, Basel, Switzerland). Expression levels were normalized to that of ACTIN 7 (ACT7). For the measurement of mature miR165/166, pulsed reverse transcription (Tang et al., 2006) was performed with oligo(dT) and miR165UPL primers. The primers used are listed in Table S1 in the supplementary material.

Excised roots were fixed with 4% (w/v) paraformaldehyde in phosphate-buffered saline (PBS, pH 7.0) for 12 hours at 4°C, dehydrated in ethyl and a tertiary butyl alcohol series and embedded in Paraplast Plus (McCormick Scientific, St Louis, MO, USA). Sections (6-8 μm) were mounted on MAS-coated glass slides (Matsunami Glass, Osaka, Japan). For preparation of the PHB riboprobe, 936 bp PHB coding sequence was amplified by RT-PCR from Col-0 root RNA with the primers listed in Table S1 in the supplementary material, and cloned into the pGEM-T vector (Promega, Madison, WI, USA). The PHB riboprobe was transcribed with the DIG RNA Labeling Kit (Roche Diagnostics) and partially hydrolyzed to ~300 bp. In situ hybridization was performed as described (Kouchi and Hata, 1993). Signals were detected with the DIG Nucleic Acid Detection Kit (Roche Diagnostics).

Primers used for DNA construction are listed in Table S1 in the supplementary material. For the construction of endoplasmic reticulum (ER)-localized GFP reporters for MIR165/166, JKD and SKOR (pMIR165/166-GFPer, pJKD-GFPer and pSKOR-GFPer), the 5′ upstream region of each gene was amplified by PCR and fused with the GFPer coding region kindly provided by Jim Haseloff (Cambridge, UK).

For the construction of PHB-GFP and PHBmu-GFP, an 8.5 kb PHB genomic fragment including 3.6 kb promoter and 0.6 kb 3′ regions was amplified from Col-0 genomic DNA with primers Xba-PHB-(−)3570 and Kpn-PHB-(+)5015R, digested with XbaI and KpnI, and cloned into the pAN19 vector (modified pUC19) to produce pAN-PHB. miRNA-resistant mutations were introduced by inverse PCR amplification of pAN-PHB with primers PHB(+)miR1445F and PHB(+)miR1444R, followed by self-ligation to produce pAN-PHBmu. In order to insert a GFP coding sequence at the C-termini of the PHB and PHBmu coding regions, a 1.1 kb BglII-KpnI fragment spanning the C-terminal coding region and the 0.6 kb 3′ region of PHB was excised from pAN-PHB and cloned into pAN19 to produce pAN-PHB-Cter. A HpaI restriction site was created immediately before the PHB stop codon of pAN-PHB-Cter by inverse PCR amplification with primers HpaI-PHBCterF and PHB(+)4443R, followed by self-ligation. A blunt-ended 3xGly-GFP(S65T) coding sequence (Morita et al., 2002) was inserted at the HpaI site of pAN-PHB-Cter to produce pAN-PHB-Cter-GFP. A BglII-KpnI fragment containing PHB-Cter-GFP-PHB-3′ was used to replace the corresponding regions of pAN-PHB and pAN-PHBmu to give pAN-PHB-GFP and pAN-PHBmu-GFP, respectively. Finally, the PHB-GFP and PHBmu-GFP fragments were inserted into pBIN30 (modified pBIN19 with Basta resistance).

For the construction of MIR165Amu, a 4.6 kb MIR165A genomic fragment including 3.9 kb promoter and 0.6 kb 3′ regions was amplified from Col-0 genomic DNA by PCR with primers miR165a(−)3927 and miR165a(+)722R and cloned into the pAN19 vector to give pAN-MIR165A. Mutations enabling miR165 to suppress PHBmu were introduced in the same way as for PHBmu-GFP described above, with primers miR165a(+)93mu and miR165a(+)70R. The MIR165Amu fragment was then inserted into the pBIN40 binary vector (modified pBIN19 with hygromycin resistance).

The pJ0571-GVG UAS-tdTomatoEr UAS-PHBmu-GFP transgenic plants were generated by sequentially transforming Arabidopsis plants with the pJ0571-GVG UAS-tdTomatoEr and UAS-PHBmu-GFP constructs. For construction of pJ0571-GVG UAS-tdTomatoEr, the T-DNA insertion site in the enhancer trap line J0571 (Jim Haseloff) was identified between At4g39900 and At4g39910 by TAIL-PCR (Liu et al., 1995) and subsequent co-segregation analysis. A 0.8 kb ground tissue-specific enhancer sequence (pJ0571) was amplified from J0571 genomic DNA with the primers listed in Table S1 in the supplementary material. The GAL4:VP16:GR (GVG) coding sequence was constructed by fusing the GAL4:VP16 (GV) coding region amplified from an enhancer trap line (Jim Haseloff) with the rat GR coding sequence (Lloyd et al., 1994). tdTomatoEr coding sequence was generated by fusing a signal peptide (MKTNLFLFLIFSLLLSLSSAEL) and ER-retention signal (HDEL) sequences to the 5′ and 3′ ends, respectively, of the tdTomato coding sequence (Shaner et al., 2004). The tdTomatoEr coding sequence was then inserted between the 5xUAS-TATA (gift from Jim Haseloff) and the nopaline synthase terminator (NosT) to produce UAS-tdTomatoEr-NosT. The resulting pJ0571-GVG and UAS-tdTomatoEr-NosT were assembled in the pBIN19 vector. For UAS-PHBmu-GFP construction, a DNA fragment containing the PHBmu-GFP coding region was amplified from pAN-PHBmu-GFP with primers PHB(−)287 and PHB-4546R, and inserted between the 5xUAS-TATA and NosT in the pBIB vector (Becker, 1990).

The indMIR165Amu/PHBmu-GFP plants (genotype pJ0571-GVG UAS-tdTomatoEr UAS-MIR165Amu PHBmu-GFP) were obtained by crossing plants harboring the pJ0571-GVG UAS-tdTomatoEr UAS-MIR165Amu with plants homozygous for PHBmu-GFP and heterozygous for MIR165Amu, and maintained as a line heterozygous for MIR165Amu and homozygous for the other transgenes. Progeny lacking MIR165Amu can be easily identified by their phb-1d-like appearance and were used for transcriptional activation experiments. For the construction of pJ0571-GVG UAS-tdTomatoEr UAS-MIR165Amu, the MIR165Amu transcribed regions were amplified with primers Hind-MIR165A(−)33 and Bam-MIR165A(+)135R and inserted between 5xUAS-TATA and NosT. The resulting UAS-MIR165Amu-NosT fragment was inserted into the vector containing pJ0571-GVG and UAS-tdTomatoEr described above. The control indMIR165A/PHBmu-GFP plants (expressing wild-type miR165, genotype pJ0571-GVG UAS-tdTomatoEr UAS-MIR165A PHBmu-GFP) were generated similarly using a construct harboring MIR165A.

pCRE1-MIR165A and pAHP6-GFPer were gifts from Yka Helariutta (University of Helsinki, Finland) and pPHB-GFPer was a gift from Ji-Young Lee (Cornell University, NY, USA).

All binary plasmids were introduced into Agrobacterium tumefaciens strain GV3-101 and used to transform Arabidopsis plants by floral dip (Clough and Bent, 1998).

Confocal laser-scanning microscopy (CLSM) and preparation of root cross-sections were carried out as described (Miyashima et al., 2009). Fluorescence intensity was measured with ImageJ (NIH) and LAS AF (Leica Microsystems, Wetzlar, Germany) software. For auxin treatment, seeds were germinated on agar media containing 10 μM N-1-naphthylphthalamic acid (NPA). Three days post-germination, seedlings were transferred to agar media containing 10 μM naphthalene acetic acid (NAA), grown for 20 hours and then observed by differential interference contrast (DIC) microscopy with a Nikon E1000 microscope (Nikon, Tokyo, Japan) after tissue clearing with 8:1:1 (w:v:v) chloral hydrate:glycerol:water. The xylem cell wall was visualized by Safranin (Kubo et al., 2005) or propidium iodide staining followed by CLSM or DIC microscopy after tissue clearing. Lateral root primordia were also observed by DIC microscopy after tissue clearing.

RNA samples were prepared from Col-0, scr-3 and ago1-101 roots (Miyashima et al., 2009) as described above. Two-color microarray hybridization was performed for the pairs scr-3/Col and ago1-101/Col using a 44K Arabidopsis 3 microarray (Agilent Technologies, Palo Alto, CA, USA). Dye-swap replication was performed for each experiment. Microarray data have been deposited at GEO with accession number GSE16460.

Loss-of-function argonaute 1 (ago1) mutants exhibit root radial pattern defects in the ground tissue (Miyashima et al., 2009). In a search for ago1-dependent genes in the Arabidopsis root by microarray analysis, we found that three HD-ZIP III genes, PHB, PHAVOLUTA (PHV) and REVOLUTA (REV), were upregulated in ago1 (see Fig. S1 in the supplementary material). Expression of HD-ZIP III genes is suppressed by the action of miR165 and miR166 (McConnell et al., 2001; Prigge et al., 2005; Ochando et al., 2008). To examine whether the radial pattern defects of ago1 roots were caused by upregulation of HD-ZIP III genes, we analyzed the root radial pattern of a dominant phb-1d mutant that expresses miRNA-resistant PHB transcripts as a result of an insertion in its miR165/166 target site (McConnell et al., 2001; Mallory et al., 2004). Serial cross-sections from the root meristem region revealed that the phb-1d stele is composed of fewer cell files (25.3±3.7, n=8) than wild-type stele (50.1±3.1, n=8), and that most (9 of 11) phb-1d roots contained an extra layer in one or two cortex cell files (compare Fig. 2A with 2B).

Because the number of root ground tissue layers is controlled by the actions of SHR and its downstream target SCR (Nakajima et al., 2001; Levesque et al., 2006; Cui et al., 2007), we analyzed the distribution of SHR:GFP and GFP:SCR fusion proteins in phb-1d. In wild-type roots, SHR:GFP is localized to both the cytoplasm and nucleus of the stele, the site of their synthesis (Nakajima et al., 2001). SHR:GFP is also detected in a single layer adjacent to the stele, where it is localized exclusively to the nucleus, and no GFP fluorescence is detected in external cell layers (Fig. 2C) (Nakajima et al., 2001). GFP:SCR is localized exclusively to the nucleus of the single layer adjacent to the stele (Fig. 2D) (Gallagher et al., 2004). In phb-1d roots, nuclear-localized SHR:GFP and GFP:SCR were detected not only in a single cell layer adjacent to the stele, but also in the supernumerary ground tissue layers (arrows in Fig. 2E,F).

It has been reported that loss-of-function mutations in a putative zinc-finger transcription factor gene, JACKDAW (JKD), result in patches of supernumerary ground tissue layers owing to unrestricted intercellular movement of SHR towards the cortex (Welch et al., 2007). Consistent with this report, expression of a pJKD-GFPer reporter and the level of endogenous JKD transcripts measured by qRT-PCR were reduced in phb-1d (Fig. 2G-I). Quantification of GFP fluorescence intensity indicated that the downregulation of JKD is more pronounced in the ground tissue than in the quiescent center (QC) (Fig. 2J), suggesting that the moderate decrease in the JKD transcript level measured in root extracts was due to selective downregulation of JKD in the phb-1d ground tissue or to the severe reduction in the number of stele cell files, which increases the proportion of ground tissue-derived RNA in whole root extracts. These results suggest that miRNA-dependent suppression of PHB is required to maintain JKD expression in the ground tissue, which in turn regulates ground tissue patterning by restricting SHR and SCR proteins to the endodermis.

Specification of xylem cell types is controlled by the dosage of HD-ZIP III activities: a high dosage specifies metaxylem and a low dosage specifies protoxylem (Carlsbecker et al., 2010). HD-ZIP III activity in the peripheral stele is suppressed by the miR165/166 derived from the endodermis, which in turn is controlled by SHR- and SCR-dependent transcription. As a result, ectopic metaxylem is formed at the protoxylem position in phb-1d, scr-3 and shr-2, where the PHB expression domain is expanded (Carlsbecker et al., 2010) (see Fig. S2 in the supplementary material). To examine whether the expansion of the PHB expression domain affects other cell types in the root stele, we analyzed the differentiation status of the pericycle in these mutants.

We examined our microarray data, comparing the transcript profiles of scr-3 and wild type, as well as the previously published expression map of the Arabidopsis root meristem (Birnbaum et al., 2003). We found that several genes known to function in the root stele were downregulated in scr-3 (see Table S2 in the supplementary material). ARABIDOPSIS HISTIDINE PHOSPHOTRANSFER PROTEIN 6 (AHP6) is expressed in the protoxylem precursors and abutting pericycle cells (Mähönen et al., 2006). In order to visualize the expression pattern of AHP6, we introduced the pAHP6-GFPer reporter (Mähönen et al., 2006) into scr-3 and phb-1d. In wild-type root, AHP6 is expressed in a pair of three-cell domains comprising one protoxylem and two abutting pericycle cells (Fig. 3A) (Mähönen et al., 2006). In phb-1d roots, AHP6 expression was lost completely or detected in only one cell of the three (88.5%, n=26; Fig. 3B). In scr-3, AHP6 expression was lost in the pericycle cells and expression attenuated in the protoxylem (100%, n=10; Fig. 3C). We also generated a GFPer reporter for the STELAR K⁺ OUTWARD RECTIFIER (SKOR) gene involved in the release of potassium to the xylem sap (pSKOR-GFPer) (Gaymard et al., 1998). SKOR is normally expressed in the pericycle, near the xylem pole and above the differentiation zone (Fig. 3D,G). In both phb-1d and scr-3, SKOR expression was severely reduced, indicating that pericycle function is at least partly affected in these mutants (Fig. 3E,F,H,I). Aberrant function of the pericycle in scr-3 and phb-1d was also suggested by reduced periclinal division in response to external auxin (Himanen et al., 2002; Fukaki and Tasaka, 2009) (Fig. 3J-L). These results indicate that miRNA-dependent suppression of PHB is required not only for protoxylem specification, but also for the correct differentiation of the pericycle.

The Arabidopsis genome contains nine MIR165/166 genes (MIR165A, MIR165B and MIR166A-G) that potentially regulate the expression of the HD-ZIP III genes including PHB (Reinhart et al., 2002; Tang et al., 2003). Carlsbecker et al. analyzed the expression of eight MIR165/166 genes (all but MIR165B) and found that MIR165A and MIR166B are expressed specifically in the root endodermis in an SHR- and SCR-dependent manner (Carlsbecker et al., 2010). We independently constructed GFPer reporters for all nine MIR165/166 genes and found that, in addition to MIR165A and MIR166B, MIR166A is also expressed in the endodermis (see Fig. S3A,D,G,J-L in the supplementary material). Our reporter analysis also suggested that MIR166A and MIR166B are expressed in the QC (see Fig. S3D,G in the supplementary material). Similar to MIR165A and MIR166B, expression of MIR166A was also dependent on SCR (see Fig. S3B,E,H in the supplementary material). The extent of the SCR dependence varied among the three MIR165/166 genes: with increased detector sensitivity, considerable expression of the GFPer reporter was detected for MIR166B, whereas it was barely detectable for MIR165A and MIR166A (see Fig. S3C,F,I in the supplementary material). Consistent with these observations, total levels of mature miR165/166 were reduced to 20-40% of wild-type levels in scr and shr mutants (see Fig. S3M in the supplementary material). SHR- and SCR-dependent expression of the three MIR165/166 genes was also confirmed by introducing each reporter line into a pSCR-SHR background. GFPer fluorescence was detected throughout the supernumerary ground tissue layers, where both SHR and SCR are expressed (see Fig. S4 in the supplementary material) (Nakajima et al., 2001). These results indicate that the single cell layer, composed of the endodermis and QC, acts as the sole source of miR165/166 in the Arabidopsis root meristem.

To correlate the PHB expression pattern and the level of miR165/166 in the endodermis, we generated reporter lines for PHB expression (PHB-GFP) in which the GFP coding sequence was inserted into the genomic context of PHB (Fig. 4A). In the wild-type background, PHB:GFP showed a graded expression pattern in the stele, with a peak at the center and gradually decreasing toward the periphery (Fig. 4B). Introduction of silent mutations into the miR165/166 target site of PHB-GFP (Mallory et al., 2004) (PHBmu-GFP) resulted in GFP fluorescence that was distributed throughout all cell layers (Fig. 4C), with nearly uniform intensity across the stele (compare bottom panels of Fig. 4B,C). This expression pattern of PHBmu:GFP is likely to reflect the transcription pattern of PHB, as a pPHB-GFPer transcriptional reporter showed a similar distribution of GFPer fluorescence to PHBmu-GFP (see Fig. S5 in the supplementary material) (Lee et al., 2006).

We then crossed the PHB-GFP reporter line with scr-3. In scr-3, PHB:GFP still showed a graded distribution pattern, as in wild type, but its expression domain was expanded to include the whole stele (Fig. 4D). The expression pattern of PHB:GFP in shr-2 was very similar to that in scr-3 (Fig. 4F). This expression pattern was somewhat intermediary between those of PHB-GFP and PHBmu-GFP (compare Fig. 4B-D,F), and consistent with the observation that expression of miR165/166 in the endodermis is reduced in scr-3 and shr-2, but not totally abolished (see Fig. S3I,M in the supplementary material).

The fidelity of the PHB-GFP and PHBmu-GFP reporters in monitoring the expression patterns and functions of endogenous PHB was confirmed by three experiments. First, in situ hybridization revealed the same distribution patterns of endogenous PHB transcripts as those of PHB:GFP and PHBmu:GFP proteins in wild type and phb-1d, respectively (compare Fig. 4B with 4G and Fig. 4C with 4H). The expression pattern of PHB:GFP in scr-3 was also consistent with the distribution of endogenous PHB transcripts in scr-3 (compare Fig. 4D with 4I). Second, transcription of PHBmu-GFP by the ground tissue-specific J0571 promoter resulted in the expression of PHBmu:GFP protein exclusively in the ground tissue, indicating that PHBmu:GFP (which is identical to PHB:GFP in amino acid sequence) does not move from the sites of its production (see Fig. S6 in the supplementary material). Third, PHBmu-GFP plants were indistinguishable from strong phb-1d mutants in every morphological aspect: both phb-1d and PHBmu-GFP bore narrow cotyledons and filamentous leaves due to the loss of abaxial-adaxial polarity (see Fig. S7A,C in the supplementary material) (McConnell and Barton, 1998; McConnell et al., 2001). They both had supernumerary ground tissue layers in roots (see Fig. S7B,D in the supplementary material). Most notably, PHBmu-GFP roots had the same cell differentiation defects in the stele as observed for phb-1d, with metaxylem formed with the expense of protoxylem (see Fig. S2B,E,I in the supplementary material), and the expression of pSKOR-GFPer was severely attenuated in PHBmu-GFP roots (see Fig. S7G in the supplementary material). All these observations indicate that the expression patterns and phenotypes obtained from the PHB-GFP and PHBmu-GFP reporter lines reflect the endogenous PHB expression pattern and the morphological defects caused by the loss of miRNA-dependent suppression.

To visualize the efficiency of PHB suppression by endodermis-derived miR165 outside the endodermis, we constructed a mutated MIR165A gene, the product of which targets PHBmu-GFP transcripts, in the context of the genomic MIR165A sequence including its own promoter (hereafter MIR165Amu for the transgene and miR165mu for its mature miRNA product) (Fig. 4A), and co-expressed it with the PHBmu-GFP reporter. In the presence of MIR165Amu, the distribution pattern of PHBmu:GFP became indistinguishable from that of the miRNA-sensitive PHB:GFP in wild-type roots (compare Fig. 4B with 4E). Moreover, MIR165Amu rescued all morphological defects caused by the PHBmu-GFP transgene (see Fig. S7E,F in the supplementary material), as well as protoxylem differentiation (see Fig. S2E,F,I in the supplementary material) and pSKOR-GFPer expression (see Fig. S7H in the supplementary material). These results indicate that miR165mu produced in the endodermis allows PHBmu-GFP transcripts to recapitulate the wild-type distribution pattern of PHB, and that the PHBmu-GFP and MIR165Amu transgenes closely mimic the relationship between the endogenous PHB and MIR165A genes.

Having confirmed the reliability of PHBmu-GFP and miR165mu in reporting the interaction between endogenous PHB and miR165, we modified this system to address quantitative aspects of the capacity of ground tissue-derived miR165 to suppress PHB and cell differentiation in the stele. The transcribed region of MIR165Amu was placed downstream of the binding site for the steroid hormone-inducible GVG transcriptional activator (Aoyama and Chua, 1997), and the resulting gene, UAS-MIR165Amu, was assembled with the ground tissue-specific pJ0571-GVG driver and the UAS-tdTomatoEr reporter. This construct was then introduced into the PHBmu-GFP reporter line. The resulting line (hereafter indMIR165Amu/PHBmu-GFP) was capable of ground tissue-specific and steroid hormone-inducible expression of miR165mu and tdTomatoEr, allowing us to analyze quantitatively the effects of ground tissue-derived miR165mu on the expression of PHBmu-GFP.

When indMIR165Amu/PHBmu-GFP plants were germinated in the presence of different concentrations of the inducer dexamethasone (DEX), expression of PHBmu-GFP was suppressed in the stele to a variable extent (Fig. 5C-E). This effect was caused by the sequence-specific suppressive function of miR165mu toward PHBmu-GFP, as control plants (indMIR165Awt/PHBmu-GFP) in which wild-type miR165 was expressed in the same manner did not affect the PHBmu-GFP expression pattern (Fig. 5F). We measured the fluorescence intensity of PHBmu:GFP in the peripheral and central stele, as well as the intensity of tdTomatoEr in the endodermis, which reflects the MIR165Amu transcription level (see Fig. S8 in the supplementary material). The intensity of PHBmu:GFP in the peripheral stele correlated inversely with the intensity of tdTomatoEr in the endodermis (r=0.70; s.d.=15.8), suggesting that endodermis-derived miR165mu suppressed the PHBmu-GFP level in the stele in a dose-dependent manner (Fig. 5A).

We then classified these roots into three categories based on their PHBmu:GFP expression pattern. Plants in category 1 showed PHBmu:GFP fluorescence throughout the root and the plants of this category were exclusively found among those grown without DEX (Fig. 5C). Plants in category 2 showed a PHBmu:GFP pattern similar to that observed for PHB-GFP in scr-3 and shr-2 backgrounds, with GFP expanded to the outermost stele but no further (Fig. 5D, compare with Fig. 4D,F). Plants in category 3 showed PHBmu:GFP only in the central stele, similar to the PHB-GFP and PHBmu-GFP/MIR165Amu roots (Fig. 5E, compare with Fig. 4B,E). We also analyzed the xylem differentiation status of each plant and found that most of the plants in categories 1 and 2 (92%, n=12) had metaxylem in the peripheral stele (Fig. 5G,H), whereas those in category 3 (89%, n=19) had protoxylem in the peripheral stele (Fig. 5I). Expression of AHP6 and SKOR, as well as the capacity of the pericycle to form lateral root primordia, were also restored by inducing miR165mu, although lateral root primordium development was slightly retarded as compared with wild type (see Fig. S9B-D in the supplementary material) (Malamy and Benfey, 1997), indicating that miR165-dependent formation of the PHB expression gradient is required for pericycle function.

A comparison of the ratio of PHBmu:GFP intensities between the central and peripheral stele with the intensity of tdTomatoEr in the endodermis (Fig. 5B) revealed that the PHBmu:GFP ratio increases with increasing levels of MIR165Amu transcription in the endodermis, up to the miR165mu expression level roughly corresponding to ~50 units of tdTomatoER fluorescence intensity (horizontal axis in Fig. 5B) that was used to estimate transcriptional activation by GVG. This suggests that the level of MIR165Amu transcription in the endodermis has a quantitative effect on the formation of the PHBmu:GFP gradient in the stele. Taken together, these results suggest that, in wild-type roots, a certain level of MIR165/166 transcription in the endodermis allows PHB expression to form a gradient across the stele, which is required for the correct arrangement of two xylem cell types and for pericycle differentiation.

The ability of the ground tissue-derived miR165mu to form a graded PHBmu:GFP distribution in the stele suggests that, in wild-type roots, endodermis-derived miR165 forms an activity gradient that decreases toward the central stele and thereby confers an inverse gradient of PHB transcripts across the stele. As a prerequisite to construct such a model, it is necessary to determine whether the ectopic production of miR165 within the stele efficiently suppresses PHB transcripts throughout the stele. It has been reported that transcription of MIR165A from the stele-specific CYTOKININ RESPONSE 1 (CRE1) promoter (pCRE1-MIR165A) results in ectopic protoxylem formation in the central stele and reduced PHB expression levels (Carlsbecker et al., 2010). The expression pattern of PHB in the pCRE1-MIR165A roots, however, has not been analyzed.

We crossed pCRE1-MIR165A and PHB-GFP plants and found that the graded distribution of PHB:GFP was abolished in the stele of F1 plants, with very weak PHB:GFP fluorescence distributed in a broad domain in the stele (Fig. 6A). As reported previously, these roots have ectopic protoxylem in the central stele (Fig. 6C) (Carlsbecker et al., 2010). By contrast, when pCRE1-MIR165A was crossed with plants homozygous for both PHBmu-GFP and MIR165Amu, the graded distribution of miR165-resistant (and functional) PHBmu:GFP was maintained in the stele (Fig. 6B). Concomitantly, these roots restored the normal xylem arrangement even in the presence of pCRE1-MIR165A (Fig. 6D), indicating that the PHB expression gradient alone is sufficient for correct xylem patterning. Taken together, these results suggest that the miR165/166-mediated suppression from the outer cell layer is crucial for the formation of the PHB expression gradient and hence for correct xylem patterning in the stele.

By restoring MIR165A transcription in the ground tissue of shr and scr using heterologous promoters, Carlsbecker et al. demonstrated that miR165 produced in the ground tissue non-cell-autonomously restricts PHB expression to the central stele (Carlsbecker et al., 2010). This experiment, however, did not address quantitative effects of miRNA on the spatial expression pattern of PHB in the stele. To visualize the expression pattern of PHB in relation to various levels of miR165/166 expression in the ground tissue, we generated two reporter lines, PHB-GFP and PHBmu-GFP, which faithfully recapitulated the distribution patterns and morphological phenotypes of endogenous PHB transcripts and its miR-insensitive mutant form, respectively (Fig. 4 and see Fig. S7 in the supplementary material). Consistent with the reduced miR165/166 levels in shr and scr mutants, PHB-GFP expression in shr and scr was moderately expanded as compared with that in wild type. In situ hybridization revealed similar expansion of endogenous PHB expression in scr. Furthermore, introduction of the MIR165Amu transgene, which targets the PHBmu-GFP transcripts, restricted PHBmu-GFP expression to the central stele, making it indistinguishable from that of PHB-GFP in wild type. Concomitantly, MIR165Amu suppressed most, if not all, of the phb-1d-like phenotype of PHBmu-GFP plants. These results indicate that the PHBmu-GFP and MIR165Amu transgenes closely mimic the relationship between the endogenous PHB and MIR165A genes. Therefore, plants harboring these transgenes provide a reliable tool for correlating the level of miRNA production in the ground tissue to the PHB expression pattern and xylem differentiation phenotype in the stele, under the conditions in which the effects of endogenous miR165/166 can be excluded.

We modified this system to explore the quantitative effects of miR165mu produced in the ground tissue, by placing MIR165Amu transcription under the ground tissue-specific and DEX-dependent transactivation system. The results indicated that the level of miR165mu in the ground tissue has a dose-dependent effect on the PHB expression pattern and on xylem cell type specification in the stele (Fig. 5). Production of miR165 within the stele abolished the PHB expression gradient (Fig. 6), providing developmental significance to the non-cell-autonomous regulation of PHB by miR165. These results strongly suggest that endogenous miR165 (and possibly miR166 as well) expressed in the endodermis can move towards the central stele, forming an activity gradient that decreases toward the central stele. Such an miR165/166 activity gradient is then translated into an inverse gradient of PHB transcripts across the stele (Fig. 7A). The PHB gradient thus formed specifies xylem cell types in a concentration-dependent manner, with a high dosage specifying metaxylem and a low dosage protoxylem (Fig. 7A) (Carlsbecker et al., 2010). This model explains the PHB expression patterns and cell differentiation defects observed in scr, shr and phb-1d mutants (Fig. 7B,C).

In wild-type roots, the distribution of PHB transcripts in the stele is somewhat biased toward the xylem pole (Fig. 4G), whereas transcription of either PHB or MIR165/166 does not exhibit this bias (see Fig. S3J-L and Fig. S5 in the supplementary material). This suggests the existence of as yet unknown mechanisms that modulate the effect of miR165 within the stele. It has been reported that ATHB8 and CNA, two of the HD-ZIP III genes targeted by miR165/166, are preferentially transcribed along the xylem pole (Lee et al., 2006; Carlsbecker et al., 2010). In these cells, less miR165/166 might be available for PHB suppression. Other mechanisms, such as differential miRNA mobility and/or the distribution of RNA-induced silencing complex (RISC) activity, are also conceivable.

In animal development, a signaling molecule emitted from a localized source confers differential gene expression along the field of its gradient thereby specifying different cell fates (Ashe and Briscoe, 2006). If the concentration of such a molecule is directly interpreted by an individual cell without relaying it to a second intercellular signal, the molecule is called a morphogen (Wolpert, 1969; Tabata and Takei, 2004). The mode of miR165/166 action revealed in our study and that of Carlsbecker et al. (Carlsbecker et al., 2010), i.e. they are emitted from a local source, affect neighboring tissues and their dosage is interpreted by each recipient cell for different cell fates, suggests the possibility that miR165/166 act as a novel morphogen. The mode of miR165/166 action, however, is different from that of classical morphogens in that miR165/166 activity is first converted into an inverse gradient of HD-ZIP III levels. Since this conversion takes place in each recipient cell, this difference does not exclude miR165/166 from being considered morphogens. However, it is not clear whether miR165/166 form a concentration gradient across the stele, as is expected for a morphogen system. In situ hybridization for miR166 showed strong staining in the epidermis and cortex, whereas weak and apparently uniform staining was observed in the stele, suggesting that miR165/166 do not accumulate in the stele, where they are presumed to be consumed for PHB suppression (Carlsbecker et al., 2010). Therefore, although our study strongly suggests the presence of an miRNA activity gradient, its concentration gradient remains elusive or might even not exist.

In leaf development, trans-acting small interfering RNA (ta-siRNA) has been proposed to move from the abaxial-most cell layers, the site of its biogenesis, towards the internal tissue, forming an accumulation gradient along the adaxial-abaxial axis (Chitwood et al., 2009). In this case, however, the ta-siRNA gradient appears to sharpen the expression boundary of target transcripts rather than to form an inverse gradient. In a mathematical model, alteration of the parameters for miRNA diffusion rate and strength of miRNA-mRNA interaction affects the steepness of the mRNA gradient (Levine et al., 2007). This suggests that the effects of non-cell-autonomous small RNAs on the spatial expression patterns of their target transcripts should be analyzed experimentally on a case-by-case basis.

Our results revealed that miRNA-dependent suppression of PHB is also required for correct ground tissue patterning and pericycle differentiation (Figs 2, 3 and see Fig. S7 in the supplementary material). At least in the case of ground tissue, the observed defect was due to the ectopic expression of PHB in these cell layers and not to an indirect effect from the defective stele, as targeted expression of functional PHBmu:GFP protein in the ground tissue resulted in the same patterning defects (see Fig. S6 in the supplementary material). The presence of supernumerary ground tissue layers in phb-1d is consistent with the ectopic expression of SHR and SCR proteins in these cells (Fig. 2C-F) (Nakajima et al., 2001) and the downregulation of JKD in the phb-1d ground tissue (Fig. 2H,J). JKD prevents SHR protein movement and SCR transcription from expanding towards the cortex (Welch et al., 2007). Therefore, currently available data suggest a mechanism in which miRNA165/166 eliminate PHB expression from the ground tissue and maintain JKD expression, ensuring sequestration of SHR and SCR within the endodermis, which in turn is necessary for correct ground tissue patterning. The relatively low penetrance of the ground tissue defects might not be due to residual JKD in the phb-1d endodermis, as loss-of-function jkd mutants also show subtle defects in the ground tissue, possibly owing to functional redundancy with homologous genes (Welch et al., 2007).

In wild-type roots, PHB expression seems to be completely suppressed in the ground tissue, and hence it is not clear why PHB is transcribed there. It has been postulated that one of the developmental roles of miRNA-dependent regulation is to clear out key regulatory transcripts from daughter cell lineages (Rhoades et al., 2002). However, there is no known developmental function for PHB in the ground tissue lineage. An alternative explanation is that transcription of PHB is not regulated by tissue-specific factors in the Arabidopsis root, but instead depends on as yet unknown positional cues that emanate from the stele. A similar mechanism has been proposed for HD-ZIP III expression in the shoot (Emery et al., 2003).

We thank Philip Benfey, Yka Helariutta and Ji-Young Lee for critically reading the manuscript and supplying materials; Miyoko Morita-Terao, Jim Haseloff, Roger Tsien, Annelie Carlsbecker, Minoru Kubo, Noriyoshi Yagi and the Arabidopsis Biological Resource Center for technical advice and materials; and Mikio Nishimura, Makoto Hayashi and Mitsue Fukazawa at the National Institute for Basic Biology, Japan, for assistance with the microarray analysis. This work was supported in part by Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (JSPS) to K.N. (21027025) and S.M. (19-11231). S.M. and S.K. were supported by JSPS Research Fellowships for Young Scientists.