ALTERED MERISTEM PROGRAM1 regulates leaf identity independently of miR156-mediated translational repression

Plant life histories are underpinned by a series of developmental transitions, the correct timing of which are crucial for plant survival and reproductive success (Huijser and Schmid, 2011). Early in their development, plants transition from a juvenile to an adult phase of vegetative growth (vegetative phase change, VPC). This transition is regulated by a decline in the abundance of two miRNAs, miR156 and miR157, which results in an increase in the expression of their targets, SQUAMOSA PROMOTER BINDING PROTEIN-LIKE (SPL) transcription factors. These transcription factors promote the expression of adult vegetative traits, in part by promoting the expression of miR172 (Wu and Poethig, 2006; Wu et al., 2009). VPC is thus regulated by inverse gradients of expression of two miRNAs, miR156 and miR172.

ALTERED MERISTEM PROGRAM1 (AMP1) encodes a putative carboxypeptidase (Helliwell et al., 2001), and was identified in a genetic screen for juvenilized mutants over 20 years ago (Conway and Poethig, 1997). Mutations in AMP1 produce a large number of small round leaves that lack abaxial trichomes (juvenile leaves) and have a higher rate of leaf initiation (Telfer et al., 1997). An initial study suggested that this phenotype was not associated with a change in the timing of VPC, implying that the timing of VPC is regulated independently of leaf number (Telfer et al., 1997). However, this result conflicts with more recent studies showing that pre-existing leaves promote VPC (Yang et al., 2011, 2013; Yu et al., 2013). The phenotype of amp1 is also surprising given the evidence that AMP1 is required for miRNA-mediated translational repression (Li et al., 2013). miR156 regulates the expression of its targets primarily by promoting their translational repression (He et al., 2018). Consequently, amp1 mutants are expected to have a reduced number of juvenile leaves due to elevated SPL gene expression, which is the exact opposite of the amp1 phenotype.

To resolve these issues, we investigated the interaction between AMP1 and the miR156-SPL module. Our results indicate that AMP1 promotes adult leaf traits in parallel to, or downstream of, the miR156-SPL module. We also found no evidence that AMP1 is required for translational repression by either miR156 or miR159. This latter result suggests that the mechanism by which miRNAs repress translation in plants is different for different transcripts.

amp1-1 (hereafter, amp1) mutants resemble plants with reduced SPL gene expression in having an increased rate of leaf initiation, an increased number of rosette leaves, an enlarged shoot apical meristem (SAM), and small, round rosette leaves that lack abaxial trichomes (Fig. 1A-F) (Chaudhury et al., 1993; Huang et al., 2015; Telfer et al., 1997; Yang et al., 2018). To determine whether this phenotype is attributable to a reduction in SPL activity, we constitutively expressed the miR156 target site mimic MIM156 in amp1. MIM156 sequesters endogenous miR156 and leads to a de-repression of SPL gene activity (Franco-Zorrilla et al., 2007). 35S::MIM156 plants have a relatively slow rate of leaf initiation, have enlarged and somewhat elongated rosette leaves, produce abaxial trichomes unusually early in shoot development, and have a relatively small SAM (Fig. 1A-F). amp1; 35S::MIM156 plants had a vegetative phenotype intermediate between that of the two parental genotypes but which was more similar to amp1 than to 35S::MIM156. The rosette leaves of amp1; 35S::MIM156 were approximately the same size as amp1 leaves but were similar in shape to 35S::MIM156 (Fig. 1A,B). amp1 plants rarely produced rosette leaves with abaxial trichomes (although abaxial trichome production on cauline leaves was unaffected; Fig. S1), whereas about 25% of amp1; 35S::MIM156 produced rosette leaves with abaxial trichomes late in shoot development. In contrast, all 35S::MIM156 plants produced rosette leaves with abaxial trichomes by plastochron 3 (Fig. 1C). Similarly, the rate of leaf initiation (Fig. 1D) and total rosette leaf number (Fig. 1E) in amp1; 35S::MIM156 were closer to that of amp1 than 35S::MIM156. Finally, the SAM of amp1; 35S::MIM156 was more similar in size to amp1 than to 35S::MIM156 (Fig. 1F). To confirm that the amp1 phenotype does not require miR156, we expressed a miR156-resistant version of SPL9 (rSPL9) in amp1. The phenotypes of amp1; rSPL9 and amp1; 35S::MIM156 plants were highly similar (Fig. S2). Taken together, these results suggest that the phenotype of amp1 is not a consequence of miR156-mediated repression of SPL activity, implying that AMP1 acts either downstream of, or in parallel to, the miR156-SPL module. This conclusion is consistent with the observation that amp1 plants flower earlier than wild type (Fig. 1G) despite remaining in the juvenile phase (Fig. 1C).

To explore the relationship between AMP1 and the miR156-SPL module in more detail, we examined the effect of amp1 on the abundance of miR156 and SPL transcripts. qRT-PCR analysis of the shoot apices of plants grown in short days (SD) showed that amp1 had no consistent effect on the level of miR156 or miR157 (Fig. 2A), or the transcripts of three direct targets of these miRNAs: SPL3, SPL9 and SPL13 (Fig. 2B). To test whether amp1 affects SPL expression independent of miR156 and miR157, we measured the transcripts of these genes in 35S::MIM156 and amp1; 35S::MIM156 plants. As expected (He et al., 2018), all three SPL transcripts were significantly elevated in 35S::MIM156 relative to wild type. All three transcripts were elevated to a much lesser extent in amp1; 35S::MIM156 than 35S::MIM156 (Fig. 2B). Together, these results suggest that AMP1 may transcriptionally regulate SPL expression, but only in the absence of miR156/miR157. The reduced accumulation of SPL transcripts in amp1; 35S::MIM156 relative to 35S::MIM156 plants could also be explained either by silencing of the MIM156 transgene, or by suppression of miR156 sequestration by MIM156 in the amp1 background. However, we found no evidence that MIM156 is silenced in amp1 (Fig. S2G), and the similarity between the amp1; 35S::MIM156 and amp1; rSPL9 phenotypes (Fig. S2A-F) suggests that miR156-sequestration is functional in amp1; 35S::MIM156.

We then examined the expression of these genes in successive rosette leaf primordia (LP) of plants grown in SD. Because amp1 initiates leaves more rapidly than wild type, LP were grouped according to the time of harvest rather than position on the shoot. Both the level and rate of decline of miR156 were almost identical in wild type and amp1 (Fig. 2C). miR157 was elevated in all LP, but declined at approximately the same rate as in wild-type plants. SPL9 and SPL13 transcripts were also elevated in the LP of amp1 relative to wild type (Fig. 2D), but these differences were relatively modest (twofold or less) and not statistically significant. Taken together, these data suggest that the vegetative phenotype of amp1 is not caused by increased expression of miR156/miR157 or decreased expression of SPL genes. It is possible that AMP1 regulates SPL expression independently of miR156 (Fig. 2B). However, the observation that amp1 does not have a significant effect on SPL9 and SPL13 expression at 20 days after germination (DAG) (Fig. 2D), when the levels of miR156 and miR157 are very low (Fig. 2C), suggests that this is unlikely.

To determine whether AMP1 regulates the expression of genes downstream of the miR156-SPL module, we examined the expression of two AP2-like transcription factors, TOE1 and TOE2, which are targets of the SPL-regulated miRNA miR172. The transcripts of both genes were generally elevated in amp1 (Fig. 2E), and this effect was not attributable to a change in the level of miR172 because amp1 had no effect on this miRNA (Fig. 2C). TOE1 inhibits trichome production on the abaxial side of the leaf by repressing the transcription of GLABRA1 (GL1) in association with the abaxial specification gene, KANAD1 (KAN1) (Wang et al., 2019; Xu et al., 2019). In wild-type plants, GL1 expression increased dramatically at 13-14 DAG and 20 DAG, consistent with the increase in abaxial trichome production over this period. GL1 had a similar temporal pattern in amp1 but was almost completely absent in the earliest LP, and was present at much lower levels than wild type in LP harvested at 20 DAG (Fig. 2F). In contrast, the expression of TRANSPARENT TESTA GLABRA1 (TTG1), which promotes trichome initiation in parallel to GL1 (Pesch et al., 2015), was actually elevated in amp1 (Fig. 2G). These results suggest that AMP1 promotes abaxial trichome formation via GL1, not TTG1, and that it acts downstream of miR156-SPL. However, whether the effect of amp1 on trichome production depends on TOE1 remains to be determined.

The juvenilized phenotype of amp1 was originally attributed to the increased rate of leaf initiation in this mutant (Telfer et al., 1997). However, this interpretation is inconsistent with more recent studies showing that pre-existing leaves promote the transition to the adult vegetative phase by repressing miR156 (Yang et al., 2011, 2013; Yu et al., 2013). To determine the basis of this discrepancy, we characterized the effect of CLAVATA3 (CLV3) and CLV1 mutations on VPC. We chose these mutations because they resemble amp1 in having an enlarged SAM and an accelerated rate of leaf initiation (Clark et al., 1995; Leyser and Furner, 1992).

Like amp1 (Telfer et al., 1997), clv3 and clv1 produced smaller rounder rosette leaves, and more leaves without abaxial trichomes (Fig. 3A-C). However, this increase in the number of juvenile-like leaves was not associated with a delay in the juvenile-to-adult transition. Instead, clv3 mutants produced leaves with abaxial trichomes 1 day earlier than wild-type plants (Fig. 3D). To determine whether the phenotype of clv1 and clv3 is dependent on miR156, we introduced 35S::MIM156 into these mutants. This transgene was epistatic to clv1 and clv3 with respect to their effect on leaf shape (Fig. 3A,B) and abaxial trichome production (Fig. 3C), suggesting that their effect on these traits requires miR156.

We then examined the effect of clv3 and clv1 on the expression of miR156 and its targets, SPL9 and SPL13, in shoot apices (Fig. 3E) and LP (Fig. 3F). qRT-PCR revealed that clv1 and clv3 have slightly reduced levels of miR156, and elevated levels of SPL9 and SPL13, although these differences were only statistically significant in a few cases. If these relatively small differences in miR156 and SPL gene expression are functionally significant, they would be expected to promote the appearance of adult traits, not repress the expression of these traits as is the case in clv1 and clv3. To explore this inconsistency, we examined the effect of clv3 on the expression of miR156-sensitive (sSPL9) and miR156-resistant (rSPL9) versions of the SPL9::SPL9-GUS reporter (Xu et al., 2016). There was no obvious difference in the expression of these reporters in the presence or absence of clv3 (Fig. 3G), supporting the conclusion that the effect of clv3 on leaf identity is not attributable to a change in the level of miR156 or its targets.

Instead, the effect of clv3 and clv1 on leaf identity is primarily attributable to their effect on the rate of leaf initiation. Specifically, clv3 and clv1 appear to increase the number of juvenile leaves by accelerating the rate of leaf production during the period when miR156 levels are high. This conclusion is supported by the observation that 35S::MIM156 is epistatic to these mutations with respect to their effect on leaf identity (Fig. 3A,B), i.e. miR156 is required for their leaf identity phenotypes. Consistent with the evidence that leaves promote the juvenile-to-adult transition by repressing miR156 (Yang et al., 2011, 2013; Yu et al., 2013), clv3 and clv1 have slightly reduced levels of miR156 and slightly elevated levels of SPL9 and SPL13 (Fig. 3E,F). However, this relatively small effect is apparently insufficient to interfere with the function of these genes in specifying juvenile leaf identity.

The increased number of juvenile leaves in amp1 is also partly attributable to its higher rate of leaf initiation (Telfer et al., 1997). However, amp1 differs from clv3 and clv1 in having a much more significant effect on leaf identity. In addition, the phenotype of amp1 is less sensitive to a reduction in miR156 than the phenotype of clv3 and clv1 (Fig.  1A-E; Fig. 2A-D). This observation, and the effect of amp1 on the expression of genes involved in abaxial trichome production (Fig. 2E,F), suggest that AMP1 operates independently of miR156 to regulate genes involved in leaf identity. A direct effect of AMP1 on leaf identity genes would explain why amp1 has a more severe vegetative phenotype than clv3 and clv1, and why the phenotype of amp1 is relatively insensitive to changes in the level of miR156.

Given the role of AMP1 in translational repression (Li et al., 2013), it is possible that the abundance of SPL transcripts in amp1 (Fig. 2B,D) does not accurately reflect their biological activity. To determine whether AMP1 is required for the post-translational regulation of SPL genes, we first measured the amount of SPL9 and SPL13 transcript cleavage in wild-type and amp1 plants. Consistent with a previous study on miR156-mediated cleavage (He et al., 2018), the rate of transcript cleavage for both SPL9 and SPL13 declined during vegetative development in wild-type plants (Fig. 4A). For SPL9, this happened at a slower rate in amp1, presumably in part because of the higher level of miR156 in the amp1 13-14 DAG sample compared with wild type (Fig. 2C), and the threshold-dependence of miR156 activity (He et al., 2018). However, later in development, transcript cleavage for both SPL9 and SPL13 in amp1 was similar to wild type (Fig. 4A). This demonstrates that miR156 is functional in amp1 and confirms the observation that AMP1 is not required for transcriptional cleavage (Li et al., 2013).

Although miR156 induces transcript cleavage, it represses the expression of its targets primarily by promoting translational repression (He et al., 2018). To examine the effect of amp1 on this process, we crossed sSPL9 and rSPL9 GUS-reporter constructs into amp1. There was no obvious difference in the staining intensity of these reporter proteins in wild type and amp1 (Fig. 4B). This impression was confirmed by spectrophotometric intensity measurements of the sSPL9-GUS reporter in LP of wild type and amp1 harvested at 20 DAG (Fig. 4C,D), a time point at which transcript cleavage was nearly equivalent in these genotypes (Fig. 4A). These results indicate that amp1 has no effect on the repression of SPL9 by miR156. To test whether translational repression in amp1 is affected in other SPL-sequence and expression contexts, we introduced a miR156-sensitive UBIQUITIN10::GUS-SPL3 3′UTR reporter construct into wild-type and amp1 backgrounds (Fig. 4E). There was no obvious difference in the expression of this reporter in these genotypes, providing further support for the conclusion that AMP1 is not required for miR156 activity.

To determine whether miR156 is uniquely insensitive to amp1, we examined the effect of amp1 on the expression of MYB33, a transcription factor that also regulates shoot identity (Guo et al., 2017) and is translationally repressed by miR159 (Li et al., 2014). miR159-sensitive and miR159-resistant versions of MYB33-GUS (Millar and Gubler, 2005) were crossed into amp1, and wild-type and amp1 plants were stained for GUS activity 1 week after germination, and at flowering. MYB33-GUS was repressed in a miR159-dependent fashion in leaves and floral organs of wild-type plants, and amp1 had no obvious effect on this expression pattern (Fig. 4F,G). We conclude from these results that AMP1 is not universally required for the translational repression of miRNA-targets.

Whether or not AMP1 functions in translational repression might depend on the subcellular localization of this process. AMP1 has been shown to colocalize with the key silencing component ARGONAUTE1 (AGO1) on the endoplasmic reticulum (ER) (Li et al., 2013). However, AGO1 also localizes to processing bodies (p-bodies), which are cytoplasmic mRNA-ribonucleoprotein complexes involved in translational silencing (reviewed by Chantarachot and Bailey-Serres, 2018). Loss of the p-body protein SUO leads to a reduction in the translational repression of the miR156-target SPL3 (Yang et al., 2012), suggesting that p-bodies are also important sites of miRNA-mediated translational repression. These results suggest that miRNA-mediated translational repression occurs in distinct subcellular compartments in a sequence-specific manner, and that unique sets of proteins function in each compartment (e.g. AMP1 on the ER, SUO in p-bodies). This model is supported by the observation that isoprenoid biosynthesis, which is necessary for the formation of membrane-bound compartments, is required for miRNA activity (Brodersen et al., 2012).

How different miRNA-mRNA complexes are targeted to different cytoplasmic compartments is unclear. AMP1-dependent miRNAs appear to have an increased number of bulges in the precursor hairpin relative to AMP1-independent miRNAs (Fig. S3), although it is perhaps unlikely that any such signals would persist during miRNA processing. The strength of target complementarity is known to affect silencing efficacy (Li et al., 2014), and could also drive subcellular distribution, but there is also no trend in target mismatch number between the AMP1-dependent/independent classes of miRNA (Table S1). Given the overlapping expression domains of a number of these miRNAs and AMP1 (Fouracre and Poethig, 2016; Vidaurre et al., 2007), it is unlikely that the site of translational repression is either developmentally regulated or determined by AMP1 localization. At the cellular level, there is evidence to suggest that miRNA sequences include signals that control the specificity of intercellular mobility (Skopelitis et al., 2018). It will be fascinating to see if the same signaling mechanisms determine the destination of miRNAs within cells.

All stocks were in a Col-0 background. The following genetic lines have been described previously: amp1-1 (Chaudhury et al., 1993); SPL9::sSPL9-GUS, SPL9::rSPL9-GUS (Xu et al., 2016); 35S::MIM156 (Fouracre and Poethig, 2019); clv1-4 (Clark et al., 1993); and MYB33::sMYB33-GUS, MYB33::rMYB33-GUS (Millar and Gubler, 2005). clv3-10 (CS68823) was obtained from the Arabidopsis Biological Resource Center (Ohio State University, OH, USA). Seeds were sown on fertilized Farfard #2 soil (Farfard) and kept at 4°C for 3 days prior to transfer to a growth chamber, with the transfer day counted as day 0 for plant age (0 DAG). Plants were grown at 22°C under a mix of both white (USHIO F32T8/741) and red-enriched (Interlectric F32/T8/WS Gro Lite) fluorescent bulbs in either long day (LD; 16 h light/8 h dark; 80 μmol m⁻² s⁻¹) or SD (10 h light/14 h dark; 120 μmol m⁻² s⁻¹) conditions.

A 650 bp fragment of the UBIQUITIN10 (At4g05320) promoter and a 400 bp fragment downstream of the SPL3 (At2g33810) stop codon (which contains a miR156 target site) were cloned from Arabidopsis genomic DNA and adapted to the Golden Gate cloning system (Engler et al., 2014). These were combined with the β-glucuronidase sequence from Escherichia coli (pICH7511) to generate a UBQ10::GUS-SPL3 3′UTR transcriptional unit in the pAGM4723 binary vector, using the pFAST-R selection cassette as a selectable marker. Cloning and binary vectors were part of the MoClo cloning toolbox provided by Addgene. Cloning primer sequences are provided in Table S2.

Plants were fixed in 90% acetone on ice for 10 min and washed with GUS staining buffer (5 mM potassium ferricyanide and 5 mM ferrocyanide in 0.1 M PO₄ buffer) and stained for between 8 h and overnight (depending on transgene strength) at 37°C in 2 mM X-Gluc GUS staining buffer. For the quantification of GUS staining intensity, ∼1 mm LP were harvested at 21 DAG, stained overnight and images of stained primordia converted from RGB color mode to hue saturation brightness mode as described previously (Béziat et al., 2017). A consistent position in the middle of the leaf lamina, adjacent to the midvein, was used for measurement.

Shoot apices were cleared and imaged using DIC microscopy according to a protocol described previously (Chou et al., 2016).

Tissue [either shoot apices with leaf primordia (≤1 mm attached), or isolated leaf primordia (0.5-1 mm in size) – sample type is detailed in the respective figure legend] were ground in liquid nitrogen and total RNA extracted using Trizol (Invitrogen) as per the manufacturer's instructions. RNA was DNAse treated with RQ1 (Promega) and 250 ng-1 μg of RNA was used for reverse transcription using Superscript III Reverse Transcriptase (Invitrogen). Gene-specific reverse transcription primers were used to amplify miR156, miR157, miR172 and SnoR101, and a polyT primer was used for mRNA amplification. Three-step qPCR of cDNA was carried out using SYBR-Green Master Mix (Bimake). qPCR reactions were run in triplicate and an average was calculated. For analyses of amp1 shoot apices and clv mutants, separate RNA extractions of three biological replicates were carried out. For analyses of amp1 leaf primordia, three reverse-transcription replicates from single RNA extractions were carried out for each sample (at least 60 LP were pooled for each RNA extraction). Two biological replicates were harvested at 8 DAG – once as part of a biological replicate with 13-14 DAG samples, and once as part of a biological replicate with 20 DAG samples. Transcript levels were normalized to snoR101 (for miRNAs) and ACT2 (amp1 shoot apices, clv mutants), or UBQ10 (amp1 leaf primordia) (for mRNAs), and expressed relative to wild type (amp1 shoot apices, clv mutants) and wild-type 8 DAG (amp1 leaf primordia) samples.

For the quantification of transcript cleavage, a modified 5′RACE protocol was followed as described previously (He et al., 2018). The data presented are the average of three ratios from separate reverse transcription replicates (six in the case of amp1 8 DAG – three reverse transcription replicates from two biological replicates). The qPCR primers used in this study are listed in Table S2.

For statistical comparisons between two genotypes, unpaired two-tailed Student's t-tests were carried out. For comparison of multiple samples, to decrease the chance of false positives, a one-way ANOVA, followed by a Tukey's test, was used for multi-way comparisons. Statistical analyses were carried out in R (r-project.org) and Excel (Microsoft).

For figures featuring boxplots, boxes display the interquartile range (IQR) (boxes), median (lines), and values beyond 1.5×IQR (whiskers); mean values are marked by a solid diamond.

We thank Anthony Millar (Australian National University, Australia) for the kind gift of the MYB33-GUS reporters; members of the Poethig lab for useful discussions; Xiang Yu for comments on miRNA secondary structures; and Melissa Morrison for assistance with collecting phenotypic data.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.186874.reviewer-comments.pdf