MicroRNA482/2118 is lineage‐specifically involved in gibberellin signalling via the regulation of GID1 expression by targeting noncoding PHAS genes and subsequently instigated phasiRNAs

Summary

MicroRNA482/2118 (miR482/2118) is a 22‐nt miRNA superfamily, with conserved functions in disease resistance and plant development. It usually instigates the production of phased small interfering RNAs (phasiRNAs) from its targets to expand or reinforce its silencing effect. Using a new high‐quality reference genome sequence and comprehensive small RNA profiling, we characterized a newly evolved regulatory pathway of miR482/2118 in litchi. In this pathway, miR482/2118 cleaved a novel noncoding trans‐acting gene (LcTASL1) and triggered phasiRNAs to regulate the expression of gibberellin (GA) receptor gene GIBBERELLIN INSENSITIVE DWARF1 (GID1) in trans; another trans‐acting gene LcTASL2, targeted by LcTASL1‐derived phasiRNAs, produced phasiRNAs as well to target LcGID1 to reinforce the silencing effect of LcTASL1. We found this miR482/2118‐TASL‐GID1 pathway was likely involved in fruit development, especially the seed development in litchi. In vivo construction of the miR482a‐TASL‐GID1 pathway in Arabidopsis could lead to defects in flower and silique development, analogous to the phenotype of gid1 mutants. Finally, we found that a GA‐responsive transcription factor, LcGAMYB33, could regulate LcMIR482/2118 as a feedback mechanism of the sRNA‐silencing pathway. Our results deciphered a lineage‐specifically evolved regulatory module of miR482/2118, demonstrating the high dynamics of miR482/2118 function in plants.

Introduction

Small RNAs (sRNAs), a class of short regulatory RNAs of 20–24 nucleotides (nt) in length, are essential for plant development and environmental adaption mainly via post‐transcriptionally gene silencing (PTGS) (Borges and Martienssen, 2015; C. Chen et al., 2018). According to the feature of biogenesis and function, plant sRNAs are classified as microRNA (miRNA) and short interfering small RNA (siRNA) (Axtell, 2013). MiRNAs pair with their target genes and induce gene silencing through either transcript cleavage or translation inhibition (Achkar et al., 2016; Voinnet, 2009; Yu et al., 2017). In some cases, after the induced cleavage, miRNAs induce the generation of secondary phased siRNAs (phasiRNAs) from their target genes (Allen et al., 2005; Liu et al., 2020; Yoshikawa et al., 2005). PhasiRNAs are involved, at the post‐transcriptional level similar to miRNAs, in diverse biological processes in plants (Liu et al., 2020). On one hand, phasiRNAs can act in cis to cleave their precursors and homologs to enhance the silencing effect of miRNAs; on the other hand, they can function in trans to regulate genes other than the parental genes to further expand the functions of miRNAs (Jiang et al., 2020; Liu et al., 2020; Tamim et al., 2018; Zhang et al., 2020).

MircroRNA482/2118 (miR482/2118), a 22‐nt miRNA superfamily consisting of two subfamilies, miR482 and miR2118, targets a large range of genes and triggers the production of 21‐nt phasiRNAs, which is widespread and conserved in seed plants (Komiya, 2017; Zhang et al., 2021). It mainly targets hundreds of NUCLEOTIDE BINDING SITE‐LEUCINE‐RICH REPEATS (NBS‐LRRs), generating phasiRNAs to regulate disease defence in gymnosperms and eudicots, while triggers the production of phasiRNAs from hundreds of long noncoding RNAs (lncRNAs) involved in plant reproduction in monocots (Xia et al., 2015; Zhai et al., 2011; Zhang et al., 2021). In eudicots, in addition to targeting NBS‐LRRs, some species‐ or lineage‐specific functions of miR482/2118 have been evolved (Canto‐pastor et al., 2019; Tang et al., 2020). For example, a tomato specific lncRNA TAS5 endows miR2118b a subfunction in disease defence (Canto‐pastor et al., 2019). And in litchi, miR2118 negatively affects fruit storability via targeting an energy‐related gene encoding calcium ATPase (Ca²⁺‐ATPase) (Tang et al., 2020). Recently, it was found that miR2118 could, in several eudicots such as flax, strawberry and columbine, induce the biogenesis of abundant 21‐nt phasiRNAs in pre‐meiotic anther, likely regulating reproductive development as it does in monocots (Pokhrel et al., 2021).

The hormone gibberellin (GA) is a prominent hormone in modulating diverse plant developmental processes including seed germination and development, flower development and fruit initiation (Arnaud et al., 2010; Gao and Chu, 2020; Tyler et al., 2004). When bioactive GAs are present, GAINSENSITIVE DWARF1 (GID1) perceives GAs leading to a conformational change, which induces the interaction with negative regulator DELLA and results in DELLA degradation by a 26S proteasome, releasing the downstream GA‐activated genes (Gao and Chu, 2020). In Arabidopsis, AtGID1A played a major role during fruit set and growth, whereas AtGID1B and AtGID1C had specific roles in seed development and pod elongation, respectively (Gallego‐Giraldo et al., 2014). It has been shown that GA signalling was post‐transcriptionally regulated by several miRNAs, including miR156 and miR159 (Yu and Wang, 2020). miR156 and miR159 mediated GA signalling through their targeted SPLs and GAMYBs interacted with DELLA proteins; once DELLA degradation induced by GA perception, releasing of SPLs promoted the floral transition while GAMYBs affected flower development, fruit parthenocarpy and development such as fruit set (da Silva et al., 2017; Tsuji et al., 2006; Wang et al., 2018; Yu et al., 2012; Yu and Wang, 2020; Zhao et al., 2018).

Litchi (Litchi chinensis Sonn.) is an economically important fruit tree in South Asia. Seed size is a critical agronomic trait, with smaller seed or seedless being a desirable feature for commercial breeding. Some litchi varieties produce a proportion of aborted seeds and thus have a higher flesh recovery (Wang et al., 2017). How this seed abortion occurs remains elusive. It is well‐documented that GA is an essential hormone for seed formation and development in plants (Kim et al., 2005; Kozaki and Aoyanagi, 2022; Serrani et al., 2007). The GA content of normal‐seeded cultivars is often higher than the cultivars with aborted seed, and the ratio of promoting hormone (IAA + GAs + CTKs) and inhibiting hormone ABA is higher in normal seed varieties, indicating that the disorder of endogenous hormones may be the cause of litchi seed abortion (Huang, 2001; Qiu et al., 1998). Recently, we found, in litchi, miR482/2118 has gained a variety of new functions via regulating many litchi‐specific target genes. One of these genes showing great sequence similarity to the GA receptor GID1 produces ample phasiRNAs (Ma et al., 2018). It implies that miR482/2118 might be involved in GA signalling in litchi directly by the adoption of a new target gene. In this study, taking advantage of our newly published high‐quality litchi genome and multi‐omics data, we characterized two miR482/2118 targeted noncoding genes, which produced phasiRNAs to target GID1 in litchi. Our experimental evidence demonstrated that this newly evolved miR482/2118‐TASL‐GID1 pathway was involved in seed development and this regulatory module was under a feedback regulation of GA signalling in litchi.

Results

Revisiting the miR482/2118 family and their targets in litchi

We have previously characterized seven members of miR482/2118 in litchi based on a draft scaffold‐level genome (Ma et al., 2018). To get a complete view of the functional diversity of miR482/2118, here, in combination with a new high‐quality chromosomal‐level reference genome sequence (Hu et al., 2022) and more small RNA sequencing data from different tissues, a total of 12 miR482/2118 members were identified from 13 loci of the litchi genome (Table S1), with LcMIR482g and LcMIR482f producing mature miRNAs with identical sequences (Figure S1). According to the classification of miR482/2118 members in a previous study (Shivaprasad et al., 2012), there are six miR482‐type members (starting with ‘UC’) and six miR2118‐type (with a 2‐nt shift backward) in litchi (Figure 1a). Ten MIR482/2118 gene loci generated mature miRNA from the 3′ arm of their precursor, while three mature miRNAs (Lc‐miR482b, Lc‐miR482d, Lc‐miR482e) were generated from the 5′ arm (Table S1, Figure S1). Mature miR482/2118 sequences are more conserved compared to the star sequences, and there is a conserved bulge (asymmetrical mismatch) at the 10th position of the miR482‐type sequence, which confers the mature sequence of miR482/2118 a length of 22‐nt (Figure 1b). In addition, miR482/2118s are distributed on six different chromosomes, with a clear cluster phenomenon on chromosomes 6, 7 and 14, especially on chromosome 14 (Figure 1c).

Figure 1.

Revisiting of miR482/2118 family and their targets in litchi. (a) Alignment of mature sequences for all litchi miR482/miR2118 family members. (b) Structure and sequence conservation of the miR482/2118 duplex. (c) The distribution of LcMIR482/2118 and 23 PHAS loci targeted by miR482/2118. (d) Venn diagram showed target genes and PHAS loci targeted by miR482/2118. (e) Annotation of 23 PHAS genes targeted by miR482/2118.

To further explore the functions of miR482/2118 family, we re‐analysed its target genes using the newly updated genome. A total of 47 annotated target genes were identified, among which 19 genes belonged to the NBS‐LRR family, 13 genes of unknown functions, and others including genes homologous to kinases, hormone response proteins, calcium transport proteins, etc. (Table S2). As miR482/2118 is capable of triggering downstream phasiRNA production, we also characterized 43 PHAS loci targeted by miR482/2118 and generating 21‐nt phasiRNAs (Phasing score ≥ 10) (Table S3). Among them, 23 PHAS loci overlap protein‐coding genes, with most PHAS genes belonging to disease‐resistance protein (Figure 1d,e; Table S3). Distribution of these 23 PHAS target genes also showed clusters phenomenon on chromosomes (Figure 1c), for instance, a supercluster of seven target genes, all of which are disease‐resistance and stress‐resistance related genes, was present on chromosome 4. Both the miR482/2118s and phasiRNAs generated from the aforementioned 23 loci showed organ‐preferential enrichment, as previously reported Ma et al. (2018), with many of them highly expressed in vegetative tissues, while a few miRNA and PHAS loci specifically accumulated in other organs, like flower and fruits (Figure S2).

Neofunctionalisation of Lc‐miR482/2118 in the repression of gibberellin signalling receptor LcGID1 via phasiRNA derived from noncoding PHAS genes

In our previous study (Ma et al., 2018), we proposed that miR482/2118 has evolved some lineage‐ and species‐specific functions in litchi via new target gene acquisition, in addition to the target genes related to disease and stress resistance (Figure 1e; Tables S2 and S3). When we further checked these litchi‐specific PHAS genes targeted by miR482/2118, the PHAS19 gene (LITCHI007943) aroused our interest. PHAS19 is located on chromosome 11 with five exons, and produced profuse phasiRNAs after the cleavage directed by miR482/2118 (Figure 2a). Consistent with the 2‐nt shift of miR482 in contrast to miR2118 type members, two distinct cleavage signals with a 2‐nt interval could be detected in both degradome data and 5′ RLM‐RACE results, suggesting both miR482‐type and miR2118‐type members can target the PHAS19 gene (Figure 2b).

Figure 2.

Lc‐miR482/2118s target an lncRNA LcTASL1 ensuing phasiRNA regulating LcGID1. (a) miR482/2118 targets LITCHI007943 leading to the production of abundant phasiRNAs. (b) The miR482/2118‐mediated cleavage on LcTASL1 was validated by both degradome data and 5′ RLM‐RACE. Red and blue arrows indicated the cleavage site of the miR482 and miR2118 type member, respectively. Grey arrow indicated others. (c) Sequence similarity shown by alignment of LITCHI007943 and two LcGID1 genes. PhasiRNA generated in 21‐nt phase beginning with miR482 type member sliced site. (d) LITCHI007943 is a long noncoding RNA LcGID1B. (e) and (f) Represent LcTASL1‐derived phasiRNA targeted LcGID1C validated by degradome sequencing data.

The transcript of LITCHI007943, 1306‐bp in full‐length verified by 5′‐ and 3′‐RACE amplifications, showed a high level of sequence similarity to the GA receptor gene LcGID1 at the nucleotide level (Figure 2c). However, it can hardly encode a meaningful protein sequence, with the longest open reading frame (ORF) translating into 84 amino acids (Figure 2d) and a low protein‐coding potential score evaluated by CPC2 analysis (0.172, http://cpc2.gao‐lab.org/), indicating that LITCHI007943 is likely a long noncoding RNA (lncRNA). The longest predicted ORF region is of a great homology to the GID1 domain (Figure 2d), overlapping with the phasiRNA production region in LITCHI007943 (Figure 2a,c; Figure S3). We found that these phasiRNAs generated from LITCHI007943 were able to cleave LcGID1 homologs in trans in litchi (Figure 2e,f; Figures S4 and S5). The gene LITCHI007943 acted like the trans‐acting gene (TAS) in Arabidopsis, which is noncoding and generates tasiRNAs to regulate downstream genes in trans, thus, we named LITCHI007943 a TAS‐like 1 gene (LcTASL1) in litchi. Additionally, LcTASL1‐derived 21‐nt phasiRNAs also acted in cis on LcTASL1, which likely amplified the silencing effect of Lc‐miR482/2118 (Figures S4 and S5).

LcTASL2 enhanced the function of miR482/2118‐LcTASL1 on LcGID1 via cascade phasiRNA production

In the list of identified PHAS loci, we noticed that there was one adjacent to LcTASL1, overlapped with an annotated gene LITCHI007944 (Figure 3a). Sequence analyses revealed that it was also a long noncoding RNA (lncRNA, CPC of 0.197) and possessed great sequence similarity to LcGID1 and LcTASL1 (Figure 3b; Figures S6 and S7). In contrast to LcTASL1, target site of miR482/2118 was absent in LITCHI007944 (Figure 3a), suggesting that LITCHI007944 was not directly targeted by miR482/2118. Its phasiRNA production was likely triggered by phasiRNAs generated from LcTASL1 because of their high sequence similarity (50.03%) (Figure 3b; Figures S7 and S8). Given the sequence similarity between LITCHI007944 and LcGID1 (Figure 3b; Figure S7), phasiRNAs produced from LITCHI007944 also targeted LcGID1 in trans, similar to LcTASL1 (Figure S9); therefore, we named LITCHI007944 as LcTASL2. Additionally, both LcGID1 genes also produced siRNAs without obvious miRNA cleavage signal, mainly from the 2nd exon which possesses great sequence similarity to LcTASL1 and LcTASL2 (Figure 3c,d). Like LcTASL1, these phasiRNAs produced from LcTASL2 and two LcGID1 genes could act in cis on their precursor (Figure S4).

Figure 3.

LcTASL1‐derived phasiRNA triggering phasiRNA production from LcTASL2 and cascade regulating LcGID1. (a) PhasiRNA distribution along LITCHI007944 (LcTASL2). (b) Sequence similarity shown by alignment of two GID1‐like genes LcTASL1 and LcTASL2 with LcGID1C. (c, d) PhasiRNA distribution along LcGID1B (c) and LcGID1C (d). (e) The newly identified pathway of miR482/2118‐TASL‐GID1 in litchi. (f) Presentation and absence of TASL, GID1 and DELLA homologs in some representative species.

Taken together, we discovered, in litchi, miR482/2118 evolved a new role via targeting a noncoding gene (LcTASL1) and generating phasiRNAs to regulate LcGID1B and LcGID1C, which are involved in GA signalling pathway, meanwhile LcTASL2 targeted by LcTASL1‐derived phasiRNAs produce phasiRNAs to reinforce the silencing effects on LcGID1 genes (Figure 3e). In addition, we also checked the presence/absence of the miR482/2118‐TASL‐GID1 pathway in species close to litchi. Comparative analyses revealed that TASL1 and TASL2 homologs were present only in certain species from Sapindaceae. Of the five species checked, only three, litchi, longan (Dimocarpus longan) and rambutan (Nephelium lappaceum), have these two TASL genes, and the other two species distant to litchi, balloon‐vine (Cardiospermum halicacabum) and yellowhorn (Xanthoceras sorbifolia) do not have them (Figure 3f), indicating that the miR482/2118‐TASL‐GID1 pathway was evolved before the split of the three close species (litchi, longan and rambutan) from the common ancestor (Figure 3f).

The miR482/2118‐TASL‐GID1 pathway was likely involved in seed development

To explore the biological function of the newly identified pathway, the expression pattern of LcTASL1, LcTASL2, LcGID1 genes and resulting phasiRNAs were analysed in various organs in litchi (Figure S10). PhasiRNAs from these four 21‐nt PHAS loci were highly enriched in young fruit (Figure S10), as were the expression level of LcTASL1 and LcTASL2 (Figure S10). However, LcGID1B and LcGID1C were relatively lowly expressed in young fruit, but with the highest accumulation in roots (Figure S10). The lower expression of LcGID1B and LcGID1C in fruit was likely due to the repressive effect of the phasiRNAs generated from LcTASL1 and LcTASL2. The fruit‐preferential expression of LcTASL1 and LcTASL2 and the prominent accumulation of phasiRNAs in young fruit suggested that this newly evolved pathway mediated by miR482/2118 might play an important role in litchi fruit development.

To further dissect the detailed function of the pathway in litchi fruit, we collected litchi fruit samples from the cultivar ‘Huaizhi’ (‘HZ’) during the whole developmental process from unpollinated ovary to mature fruit (Figure 4a). Transcriptome profiling revealed that LcTASL1 and LcTASL2 had higher expression in pericarp and seeds than the aril, in which only low expression was detected in the early stages (Figure 4b), suggesting that these two LcTASL genes may be more important in pericarp and seed development. In spite of overall comparable expression levels, LcTASL1 were of distinct expression trends in pericarp and seed. Its expression increased gradually until reaching the peak level at 31 DAP in pericarp, but it peaked much faster at 3 DAP and decreased slowly in seed (Figure 4b). For LcTASL2, it accumulated to a much higher level in seed than in pericarp (Figure 4b). Compared with the two LcTASL genes, LcGID1 genes showed remarkable expression in all tissue samples tested. Overall, the level of LcGID1B was higher than LcGID1C in all samples (Figure 4c), and both of them showed higher accumulation at early stages of seed development (Figure 4b). All these results suggested that the pathway likely played a more important role in seed than other parts of the litchi fruit.

Figure 4.

Expression pattern of key genes and associated sRNAs in the miR482/2118‐TASL‐phasiRNA‐GID1 pathway during litchi cv. ‘HZ’ fruit development. (a) Phenotypes of litchi (cv. ‘HZ’) fruit development. Samples of ‘HZ’ fruit were arranged in the black box. Red lines under the fruit samples indicated 0.25 mm. (b) Comparison of the expression levels (Fragments Per Kilobase Million, FPKM) of two LcTASL and LcGID1 genes during ‘HZ’ fruit development. The green, orange and grey panels indicated the pericarp, seed and aril sampled separated from litchi fruit, respectively. (c) The expression level of miR482/2118s in seed of ‘HZ’. (d) The abundance of phasiRNAs (reads per 10 million, RPTM) generated from LcTASL1, LcTASL2, LcGID1B and LcGID1C during ‘HZ’ seed development. Multiple comparisons were conducted within group at different time points. Error bars indicate SEs from three replicates and the letters above bars indicate a significant difference (P < 0.05).

Next, we checked the expression of sRNA in litchi seed, including both miR482/2118 family and phasiRNAs generated from two LcTASLs and LcGID1s (Figure 4c,d). We found that most miR482/2118s did not show a consistent expression pattern, agreeing with the diverse function of miR482/2118s. Many of them had higher accumulation at late stages (38–45 DAP, Figure 4c). LcTASL/LcGID1‐derived phasiRNAs were significantly up‐regulated after pollination and peaked at 24 DAP, and phasiRNAs derived from LcTASL2 were of the greatest abundance (Figure 4d), agreeing well with its higher expression level in seed (Figure 4b). This high abundance of phasiRNAs in litchi seed, especially the high accumulation of LcTASL2‐derived ones, further confirmed the role of this PHAS pathway in seed development.

As seed size is a critical phenotypic trait of litchi fruit, fruits with smaller seed are often of greater commercial value. We also profiled the expression of these genes and sRNAs in fruits of ‘Nuomici’ (‘NMC’), a cultivar producing small seed resulted from premature seed abortion (Figure S11). We found that generally the expression of LcTASLs and LcGIDs in ‘NMC’ is similar to ‘HZ’ which produces well‐developed large seed. Regarding the downstream phasiRNAs, their accumulation, especially for those derived from LcTASL1 and LcTASL2, was noticeably accumulated earlier with less abundance in large‐seed ‘HZ’ than aborted‐seed ‘NMC’ (Figure S11). All in all, given the high expression of LcTASLs and LcGID1s and the profuse accumulation of phasiRNAs in litchi seed, and the important role of GA in plant seed development (Kim et al., 2005; Kozaki and Aoyanagi, 2022; Serrani et al., 2007), we proposed that the miR482/2118‐TASL‐GID1 pathway is likely vital for the seed development in litchi (Huang, 2001; Qiu et al., 1994; Ye et al., 1992).

Constructed miR482/2118‐TASL‐GID1 pathway affected flower and silique development in Arabidopsis

As a perennial fruit tree, litchi lacks a robust transgenic system to valid gene function in vivo, therefore, we managed to verify the biological functions of this pathway by constructing miR482/2118‐TASL‐phasiRNA pathway in Arabidopsis thaliana, as the GA signalling pathway is conserved between litchi and Arabidopsis. Comparative genomic analyses revealed that there is no homolog of two LcTASLs in Arabidopsis, while phasiRNA from two LcTASL genes could pair well with and target the GID1 genes of Arabidopsis (Figure 3f; Figures S12 and S13). We firstly performed transformation of single gene of LcMIR482a, LcTASL1 and LcTASL2, and then crossed together LcTASL1 with LcTASL2 (LcTASL1 × LcTASL2), LcMIR482a with LcTASL1 (LcTASL1 × LcMIR482a) and LcTASL1 × LcTASL2 × LcMIR482a to examine the effect of TASL‐derived phasiRNAs on the expression of AtGID1 genes and the phenotype of transgenic Arabidopsis.

At least three homozygous lines of each transgenic gene were obtained with significantly higher expression than wild type (WT) (Figure S14a–d). Among these homozygous lines, plants with overexpression of LcTASL1 showed a relatively low expression, which was caused by the cleavage of At‐miR472 in Arabidopsis with great sequence similarity to Lc‐miR482/2118s (Figures S14c and S15), suggesting that At‐miR472‐LcTASL1‐AtGID1 pathway was built up in the LcTASL1 transgenic plants. Thus, LcTASL1‐derived phasiRNA could be triggered by At‐miR472 in LcTASL1 and LcTASL1 × LcTASL2 lines, while in LcTASL1 × MIR482a and LcTASL1 × LcTASL2 × LcMIR482a lines by both At‐miR472 and the transformed Lc‐miR482a (Figure S14e). In LcTASL1 × LcTASL2, elevated LcTASL1 caused an increased expression of secondary phasiRNAs (Figure S14e), which significantly reduced the expression of LcTASL2 (Figure S14d) and AtGID1 (Figure S14g). Additionally, in LcMIR482a × LcTASL1, over‐accumulated miR482 resulted in the decrease of LcTASL1 expression (Figure S14c) and then the expression of AtGID1 (Figure S14g). Compared to LcTASL1 × LcTASL2, introducing the miR482 led to a decrease in the expression of LcTASL1 (Figure S14c) and LcTASL2 (Figure S14d) and these phasiRNAs produced from them (Figure S14e,f). All these results showed that phasiRNAs produced by LcTASL1 could act in trans to trigger the production of phasiRNAs from LcTASL2 and all these LcTASL1‐ and LcTASL2‐derived phasiRNAs could regulate the AtGID1s, which confirmed our proposed regulatory network (Figure 3e).

In these transgenic Arabidopsis plants with the build‐up of miR482a‐TASL‐phasiRNA pathway, we found that the filament length was significantly reduced in both OX‐LcTASL1 and OX‐LcTASL1 × LcTASL2 (Figure 5a,b), and the siliques of all transgenic lines containing LcTASL1 showed different degrees of abnormalities including seed abortion and smaller pod size (Figure 5c–e; Figure S16). The observation that overexpressed LcTASL1 was negatively associated with the silique length and the ratio of seed number/silique length (Figure 5d,e) suggested the constructed pathway affected seed and fruit development in Arabidopsis. These defects in transgenic lines are quite similar to the phenotype of Arabidopsis gid1 mutant, which is of shorter filament and shorter siliques with less seeds (Gallego‐Giraldo et al., 2014; Griffiths et al., 2006; Iuchi et al., 2007). Thus, these results proved that the miR482/2118‐TASL‐GID1 pathway could affect flower and seed formation, and eventually fruit development via the regulation of GA signalling.

Figure 5.

miR482/2118‐TASL‐GID1 pathway affected flower and fruit development in Arabidopsis. (a) Flower phenotype in different transgenic Arabidopsis lines. (b) Filament length at the eighth node of Arabidopsis plants. Error bar indicated the SE from 10 replicates and each column represented the average value of the length of all filaments of the eighth flowers. Different letters above bars indicate a significant difference (P < 0.05) using multiple comparisons. (c) Inflorescence phenotype in different transgenic Arabidopsis lines. (d) Silique phenotype in different transgenic Arabidopsis lines. (e) Maternal phenotype of different Arabidopsis lines compared with wide type (WT) Arabidopsis. Mature fruits were individually harvested and fruit length and seed number were measured. Ratio of seed number to silique length was normalized to the WT. Error bar indicates the SEs from at least five replicates and the asterisk above bar represented the significance level by the Student's t‐test. Signal asterisk indicated P < 0.05, while double asterisk indicated P < 0.01.

GA‐responsive factor GAMYB feedback regulates MIR482/2118

Our findings have demonstrated that the newly evolved miR482/2118‐TASL‐GID1 pathway could regulate litchi seed development via GA signalling. Given the broad regulatory role of GA in fruit development, whether any of these elements in the new regulatory module are under the regulation of GA signalling aroused our curiosity. We then conducted promoter sequence analyses on LcMIR482/2118s, LcTASLs and LcGID1s, and found that a few genes had at least one GA response element (Table S4). The promoters of LcMIR482d and LcMIR482e contained a GARE‐motif, which has been reported to be bound by GA‐responsive MYB proteins (GAMYB) that are critical components of GA signalling pathway (Achard et al., 2004). Four GAMYBs were identified in litchi, in which the expression of LcGAMYB33 was higher than the other three members in all seed samples, suggesting the importance of LcGAMYB33 (Figure 6a; Figure S17).

Figure 6.

LcGAMYB33 regulated the transcription of LcMIR482e and LcMIR482f. (a) Expression of four LcGAMYB genes in different development stages of ‘HZ’ seed. Error bars indicate SEs from three biological replicates and different letters above bars indicate a significant difference (P < 0.05) using multiple comparisons. (b) Analysis of LcGAMYB33‐enriched regions in the DAPseq assay. Pie chart shows the percentage distribution of LcGAMYB33‐binding peaks in each category. (c) LcGAMYB33 binding motif identified by DAPseq analysis. (d, e) The binding peaks of LcGAMYB33 in the promoter of LcMIR482e (d) and LcMIR482f (e). LcGAMYB33 bound LcMIR482e via two binding motifs (‘a’ and ‘b’), while just one binding motif showed on LcMIR482f. The left panels shown the binding site of LcGAMYB33 on LcMIR482e (d) and LcMIR482f (e), while the right column indicated the relative enrichment comparing with the empty vector pIX‐Halo (Mock). The asterisk above bar represented the significance level by the Student's t‐test. Signal asterisk indicated P < 0.05, while double asterisk indicated P < 0.01. (f, g) Electrophoretic mobility shift assays (EMSAs) showing the binding ability of LcGAMYB33 with the promoters of LcMIR482e (f) and LcMIR482f (g) in vitro. The shifted bands indicated by arrows suggest the formation of DNA–protein complexes. ‘+’ and ‘−’ represent presence and absence, respectively. Triangles indicate increasing amounts of mutant or unlabelled probes used for testing the specificity of binding and competition. Probes without biotin labels were loaded as unlabelled competitors. GST protein alone was used as the negative control. In f, ‘Mutant probe (a)’ and ‘Mutant probe (b)’ signify the sequence mutation of motif a and b, respectively, while ‘mutant (ab) probe’ denotes the mutations made in both binding motifs simultaneously. (h) LcGAMYB33 activated the expression of LcMIR482e and LcMIR482f in vivo, as shown by transient dual‐luciferase reporter assays. Both effector and reporter vectors were cotransformed into tobacco leaves. After incubation of 72 h, the ratio of LUC to REN was detected. Error bars indicate SEs from six replicates. Different letters indicate a significant difference (P < 0.05) using multiple comparisons.

To profile the binding target genes of LcGAMYB33, we performed a DNA affinity purification sequencing (DAP‐seq) for LcGAMYB33 to unravel its genome‐wide binding sites. In total, most (58%) LcGAMYB33 binding peaks were distributed in the genic regions, while 24% was in the intergenic, 11% was in the upstream (<2k from TSS) and only 7% was in the downstream (<2k from TTS) (Figure 6b). A motif with a core sequence of ‘TAACTGACT’ (e‐value = 2.0e‐212) was characterized from 70.2% of the LcGAMYB33 binding regions (Figure 6c). For the miR482/2118‐TASL‐GID1 module, LcGAMYB33 could bind to the promoter or gene coding regions of LcMIR482e, LcMIR482f, LcTASL2, LcGID1C and three LcDELLA genes (Figure 6d,e; Figure S18), The LcMIR482e has a binding motif of ‘AGTTAGTTAG’ at around 2 kb before the MIRNA precursor region, while the binding motif of ‘AAGTTGTTGA’ is much closer to the precursor of LcMIR482f (Figure 6d,e). These bindings of LcGAMYB33 detected by DAP‐seq were further validated by electrophoretic mobility shift assay (EMSA) (Figure 6f,g; Figure S19). The band shift reduced in a competition experiment with increased concentration of unlabelled probes (cold probe) (Figure 6f,g; Figure S19) and mutation of the binding element showed a weak binding band (Figure 6f; Figure S19). We also confirmed the binding ability of another motif ‘TCAACAGACT’ in LcMIR482e, belonging to GARE‐motif, which reported to be a binding site of GAMYB (Gubler et al., 1999; Figure 6f; Figure S19). To examine whether LcGAMYB33 could regulate LcMIR482e and LcMIR482f, dual‐luciferase reporter assays were performed in vivo. A pGreenII 0800 vector containing an LUC reporter gene driven by the LcMIR482e and LcMIR482f promoter was cotransformed with an empty pGreenII 62‐SK vector or pGreenII 62‐SK‐LcGAMYB33 vector into N. benthamiana leaf epidermal cells. Compared with the empty vector samples, the cells expressing LcMIR482e and LcMIR482f exhibited a significantly higher LUC/REN ratio (Figure 6h). Collectively, these findings support the notion that LcGAMYB33 act as transcriptional activator of LcMIR482e and LcMIR482f by directly binding to their promoters. Thus, LcMIR482e and LcMIR482f were under the transcription regulation of LcGAMYB33, form a new feedback regulatory loop in litchi.

Discussion

In this study, we found that Lc‐miR482/2118s gained a novel function by targeting a long noncoding RNA gene, LcTASL1, to induce the production of abundant 21‐nt phasiRNAs, which negatively regulates the expression of LcGID1s to get involved in GA signalling. This silencing effect of LcTASL1‐derived phasiRNAs can be reinforced by another phasiRNA‐generating noncoding RNA gene LcTASL2 (Figure 7). This pathway of miR482/2118‐TASL‐GID is likely associated with the seed development in litchi. We also found a critical component of the GA signalling, LcGAMYB33, regulated MIR482/2118 expression at the transcriptional level, functioning as a feedback loop (Figure 7). These results provide the first demonstration of the direct involvement of miRNA in GA signalling via a newly evolved regulatory circuit of phasiRNAs.

Figure 7.

Perspective model of the miR482/2118‐TASL‐GID1 regulatory pathway. Lc‐miR482/2118s targets a species‐specific noncoding gene LcTASL1 and triggers the generation of secondary phasiRNAs to regulate in trans the expression of GA receptor gene LcGID1; this silencing effect on LcGID1 is reinforced by phasiRNAs from another noncoding PHAS gene LcTASL2. This pathway is associated with the fruit development of litchi via the involvement in GA signalling. Lc‐miR482/2118s is under the transcriptional regulation of LcGAMYB33, a critical component of GA signalling, forming a feedback regulatory loop.

Diversification of miR482/2118 functions in litchi

MiR482/2118 superfamily is a 22‐nt well‐studied miRNA family and widespread in seed plants (Zhang et al., 2021). It originated in gymnosperms and has dual functions in gymnosperms, which was selectively preserved in monocots and eudicots of angiosperms. In eudicots, miR482/2118 mainly targets NBS‐LRRs to produce phasiRNAs, playing an important role in disease resistance, while in monocots it mainly targets lncRNAs to produce phasiRNAs involved in plant reproductive development (Komiya, 2017; Zhang et al., 2021) In addition to these conserved well‐known functions, in our study, based on a new high‐quality genome sequence of litchi, in combination with more deep sequencing sRNA data, we have identified more miR482/2118 members and their target genes, suggesting the function of miR482/2118 has expanded greatly in litchi (Hu et al., 2022; Ma et al., 2018). In addition to NBS‐LRRs and other disease‐related genes and the noncoding LcTASL1 gene of great sequence similarity to the gibberellin receptor GID1C, many other target genes have been found in litchi, including those genes homologous to BR‐signalling kinase 1 (BSK1), calcium‐transporting ATPase, SUPPRESSOR OF GENE SILENCING 3 protein, etc. (Table S3). These genes are involved in different biological processes, for instance, BSK1 acts as a positive regulator of brassinosteroid (BR) signalling (Wang and Chory, 2006). Similar to LcTASL1 of miR482/2118, other target genes might be lineage‐ or species‐specific, i.e., most of these target genes are newly evolved. These results provide a great demonstration that miR482/2118 is in a process of rapid gaining and losing of specific target genes in plants, in addition to its conserved function in disease resistance. Its more diverse function renders miR482/2118 great capacity to participate in more regulatory network of complex trait development, for instance, the seed development in litchi.

Crosstalk between miR482/2118s and gibberellin signalling via secondary and tertiary phasiRNAs in plants

PhasiRNAs, a major class of siRNA, are generally induced by 22‐nt miRNA to regulate plant development and environmental adaption (Liu et al., 2020). It could not only act in cis to cleave their precursors to enhance the silencing effect of miRNAs, but also function in trans to regulate genes other than the parental genes to further expand the functions of miRNAs (Liu et al., 2020; Tamim et al., 2018). Zhang et al. (2020) and Jiang et al. (2020) proved that in rice these 21‐nt phasiRNAs generated from miR2118 PHAS loci acted in a target‐cleavage mode to regulate a large variety of genes, eventually affecting rice sterility, suggesting phasiRNAs expanding the functions of miR2118 family. In addition, several studies have found that secondary siRNAs can induce tertiary phasiRNA production. For example, in eudicots, At‐miR173 and other related miRNAs belonging to the super‐miR7122 family regulate a set of pentatricopeptide repeat (PPR) genes via one or two layers of noncoding trans‐acting like genes (TASL), and secondary and tertiary phasiRNAs were generated to act in cis or in trans (Xia et al., 2013).

In this study, we have confirmed that an lncRNA target LcTASL1 of miR482/2118 via cascade 21‐nt phasiRNAs to repress the expression of the gibberellin receptor GID1 genes, suggesting miR482/2118 is directly involved in gibberellin signalling. Firstly, secondary 21‐nt phasiRNAs generated from LcTASL1 by miR482/2118 can directly target and cleave LcGID1 and LcTASL2 mRNA (Figure 2e,f; Figures S5 and S8), and tertiary 21‐nt phasiRNAs from LcTASL2 and LcGID1s themselves can enhance the silence effect on LcGID1 genes (Figures S4 and S9). Meanwhile, several 22‐nt siRNA produced from LcTASL and LcGID1 also might contribute to this pathway for its relatively high proportion (Figure S20). For example, a 22‐nt phasiRNA (sequence: 5′ UUGAGUCCUUCCACGUACGCCA 3′) generated from LcTASL1 had a significant cleavage signal on LcTASL2 (Figure S20b). GID1 perception is the first step of GA response, and GA is a critical hormone in modulating diverse plant developmental processes including seed germination, flower development, fruit initiation and seed development, suggesting miR482/2118‐TASL‐GID1 pathway might participate in multiple biological processes in litchi. These two LcTASL genes were characterized only in litchi, longan (D. longan) and rambutan (N. lappaceum), all of which are important subtropical and tropical economic fruit trees belonging to the Nepheliinae subtribe of Sapindaceae (Figure 3f), indicating that the miR482/2118‐TASL‐GID1 pathway was evolved from a common ancestor of the three close species (litchi, longan and rambutan). Thus, it is reasonable to believe that the miR482/2118‐TASL‐GID1 pathway may be associated with the development of certain traits unique to them, for instance, the formation of freshy aril.

MiR482/2118 regulate the seed development via GA signalling

Recently, it was reported that miR482/2118, in addition to the targeting of NBS‐LRRs in disease resistance, can induce the production of 21‐nt reproductive phasiRNAs, which were enriched in pre‐meiotic stages of anther, in a few eudicots such as flax, strawberry and columbine (Pokhrel et al., 2021; Zhang et al., 2021). Similar to the reproductive phasiRNAs in monocots, these 21‐nt phasiRNAs likely play a crucial role in male reproductive development (Pokhrel et al., 2021). However, how they are involved in plant reproductive development still remains largely unknown. Increasing studies conducted in monocots, especially in rice, found that 21‐nt reproductive phasiRNAs function via targeting a large number of functional genes to reprogram the mRNA transcriptome in early stage of anther development (Araki et al., 2020; Fan et al., 2016; Jiang et al., 2020; Lan et al., 2022; Zhang et al., 2020). Overall, these studies mainly focused on the male part, the anther development. In this study, we found miR482/2118 also participated in plant reproduction, but it is irrelevant to the flower organ, instead, it regulates the fruit development, especially the seed development, via the LcTASL‐LcGID1 cascade and subsequent GA signalling. On the other hand, with in‐depth and broad sRNA profiling, we were not able to detect the type of reproductive phasiRNA found in other eudicots (Pokhrel et al., 2021) from litchi anther, suggested that the miR482/2118‐directed reproductive phasiRNAs are probably dynamically evolved, as the miR2275‐type 24‐nt reproductive phasiRNAs reported in eudicots (Xia et al., 2019). Apart from the role in fruit development, miR482/2118 may serve many other roles which are promising for further investigation, given the importance of the hormone GA in plant growth and development. The feedback regulation of GAMYB on the transcription of MIR482/2118 suggested a complicated crosstalk between the miR482/2118‐mediated gene silencing and the GA signalling network.

Methods

Plant materials

Litchi samples including L. chinensis Sonn. cv. ‘Huaizhi’ (‘HZ’) and ‘Nuomici’ (‘NMC’) were grown in an orchard located at South China Agricultural University (Guangzhou, China). Three litchi trees with consistent growth potential were selected as three biological replicates. Fruits were collected from different branches of each tree. When possible, fruit of ‘HZ’ and ‘NMC’ were dissected into different tissues including ovary, seed, pericarp and aril before freezing.

Construction and cultivation of transgenic lines

Overexpression vectors of 35S: LcTASL1, 35S: LcTASL2 and 35S: LcMIR482a were generated by cloning the full length of LcTASL1, LcTASL2 and precursor sequence of Lc‐MIR482a into the vector pCAMBIA1302. The precursor sequence of LcMIR482a was amplified including the stem loop and its upstream and downstream 200 bp. The above plasmids were transformed into A. thaliana (L.) ecotype Columbia‐0 (Col‐0) plants via Agrobacterium tumefaciens‐mediated stable transformation based on the floral dip transformation method (Zhang et al., 2006). At least three homozygous lines carrying the overexpression transgene of 35S: LcTASL1 were chosen to cross with the other two transgenic lines (35S: LcTASL2 and 35S: LcMIR482a). The result hybrid offspring, including 35S: LcTASL1#4 × 35S: LcTASL2#2 (LcTASL1 × LcTASL2), 35S: LcTASL1#4 × 35S: LcMIR482a#1 (LcTASL1 × LcMIR482a) and 35S: LcTASL1#4 × 35S: LcTASL2#2 × 35S: LcMIR482a#1 (LcTASL1 × LcTASL2 × LcMIR482a) were used for phenotype and functional analyses. WT and transgenic Arabidopsis plants were grown in a greenhouse, and the growth conditions were 22 °C (day)/19 °C (night) in long days (16 h light/8 h dark).

All experimental samples for the assays were arranged completely at random. There were at least six replicates of each type of Arabidopsis sample for phenotypic measurements. The silique length and seed number present at the eighth node of the principal Arabidopsis inflorescence were quantitatively assessed. The results were expressed as mean ± standard error and P < 0.05 was set as the minimum significant difference by the Student's t‐test.

Rapid‐amplification of cDNA ends (RACE)

To acquire the full‐length cDNA sequence of LcTASL1 and LcTASL2, we performed 5′‐RACE and 3′‐RACE experiments following the manufacturer's instructions accompanying the SMARTer® RACE 5′/3′ kit (TAKARA). To map the cleavage sites in the LcTASL1 transcript, we performed modified 5′‐RNA ligase‐mediated rapid amplification of cDNA ends (5′‐RLM‐RACE) mainly following the method of Shahid et al. (2018). The cDNA templates were amplified through two rounds of PCR with the universal sense or antisense primers and two gene‐specific primers (Table S5). The PCR products were cloned into the pEASY vector (pEASY®‐Blunt Cloning Kit, TransGen Biotech) and then sequenced in Sangon Biotech (Shanghai). Primers used in 5′ and 3′ RACE were provided in Table S5.

Phylogenetic analysis

For phylogenetic analysis, the genome and amino acid sequence of all species used in this study were obtained from Phytozome (https://phytozome‐next.jgi.doe.gov/) and Sapbase (Li et al., 2022). Full length of LcTASL1 and LcTASL2 were used to blast to other genome sequences by TBtools (Chen et al., 2020) to identify homolog genes. Similarly, GID1 homologs were determined by homolog search as well. Finally, a species tree was constructed using the selected species and the number of homologous genes identified in each species was displayed in heatmap by TBtools (Chen et al., 2020).

qRT‐PCR analysis

Quantitative reverse transcription–polymerase chain reaction (qRT‐PCR) was conducted with Promega GoTaq® qPCR Master Mix (A6001) in a BioRad CFX384 Real‐Time PCR Detection System, with three biological replicates and three technical replicates. The LcGAPDH and LcEF1α were used for litchi mRNA sample control, and the AtACT and AtU6 were used as Arabidopsis miRNA sample control, respectively. The relative expression was calculated using the comparative $2^{\left(-\mathrm{\Delta \Delta }{C}_{\mathrm{t}}\right)}$ method (Pfaff, 2001). Details regarding the qPCR primers were provided in Table S5.

RNA‐Seq and data analysis

Total RNA was extracted using PureLink™ Plant RNA purification Reagent (Invertrogen, Code No: 12322012) from ovary, seed and pericarp, while the aril RNA was extracted using a hot‐boron‐based method (Kong et al., 2022); 1 μg high‐quality RNA from 132 samples with high quality was sent to BioMaker (China, Beijing) for mRNA library construction and RNA sequencing (RNA‐seq). In total, 132 mRNA libraries were constructed and sequenced on a NovaSeq 6000 platform.

Quality control of raw data was conducted using FastQC (Andrews, 2010) and MultipleQC (Ewels et al., 2016) to confirm acceptable quality for downstream analysis. Fastp software (S. Chen et al., 2018) was invoked to remove the low‐quality bases present in the sequencing data at the 3′ end of the splice sequence and read segment. All sequence data were compared to the reference genome of litchi using STAR software (Dobin et al., 2013). The expression of litchi genes was calculated using StringTie software (Pertea et al., 2015).

sRNA‐seq and data analysis

Two ‐microgram total RNA of each sample was sent to Novogene (Beijing, China) for sRNA library construction and sequencing (sRNA‐seq). In total, 48 sRNA libraries were constructed and sequenced on an Illumina HiSeq 2000 platform. All sRNA data, including those from the previous study of Ma et al. (2018) were analysed according to our previous published methods (Chen et al., 2021; Feng et al., 2019).

DAP‐seq and data processing

The DNA affinity purification sequencing (DAP‐seq) experiment was performed mainly according to the previously published protocol (Bartlett et al., 2017; O'Malley et al., 2016). Briefly, 5 μg genomic DNA was extracted from young leaves of ‘FZX’, the litchi cultivar with genome sequenced and normally developing aborted seed (Figure S21) and fragmented into 200‐bp fragments by ultrasonication. The fragments were ligated with Illumina‐based sequencing adaptors to form a DNA library. The LcGAMYB33 coding sequence was cloned into pIX‐Halo vector and translated in vitro using TNT® SP6 High‐Yield Wheat Germ Protein Expression System (L3260) from Promega Corporation. After incubation of a litchi genome DNA library with HALO‐tagged LcGAMYB33, DNA complex (a DNA‐protein complex) was eluted to amplify with indexed primer and then sequenced in Novogene (Beijing, China). Quality control of raw data was processed by FastQC (Andrews, 2010). Clean reads were aligned against litchi reference genome by bowtie2 (Langmead and Salzberg, 2012) and converted into visual files that can be viewed on IGV by Samtools (Li et al., 2009). Peaks calling was performed by Macs2 (Gaspar, 2018) and MEME‐ChIP suite was used to discover binding motifs (Machanick and Bailey, 2011).

Electrophoretic mobility shift assay

The partial sequence of LcGAMYB33 containing the DNA binding domain was cloned into the pGEX4T‐GST vector harbouring a GST tag (Table S5). The resulting plasmid was transformed into Escherichia coli BL21 (DE3) (Tsingke Biotechnology Company, Beijing, China) for the production of recombinant GST‐LcGAMYB33. GST‐LcGAMYB33 was purified by GSTPur Glutathione Kit (SA008K; Smart‐Lifesciences, Changzhou, Jiangsu Province, China) following kit protocol. Protein concentrations were determined using a Nano Drop 2000 spectrophotometer (Thermo Scientific). Electrophoretic mobility shift assays (EMSAs) were performed following the manufacturer's instructions of LightShift® Chemiluminescent EMSA Kit (Thermo Fisher Scientific, 20148). Briefly, biotin‐labelled probes and GST‐LcGAMYB33 were incubated together, while unlabelled and mutated probes and GST protein alone were used as competitors and negative controls, respectively. Thereafter, free and protein–DNA complexes were separated on a 5% native polyacrylamide gel, transferred onto nylonmembrane and detected by a chemiluminescent imaging system (Beijing kcrx bio‐company, K 3000mini).

Dual‐luciferase reporter assays

The promoter fragments (1500 bp) of LcMIR482e and LcMIR482f were constructed in the pGreenII 0800‐LUC to generate reporter constructs, and LcGAMYB33 gene was inserted into the empty pGreenII 62‐SK vector as the effectors. Both effector and reporter constructs were individually introduced into A. tumefaciens GV3101 (pSoup), and the bacteria were injected into tobacco (Nicotiana benthamiana) leaves with needleless syringes for transient expression assays. For coinjection with effector and reporter constructs, the bacteria were resuspended in infiltration buffer (10 mM MES, 10 mM MgCl₂, 150 μM acetosyringone, pH 5.6) at OD₆₀₀ = 0.65 and incubated for 3 h at room temperature. The leaf zones of infiltration were harvested for enzyme activities assays of firefly luciferase and renilla luciferase at 3‐day postinjection using the DualLuciferase® Reporter Assay System (YEASEN, Shanghai, China). At least six independent biological replicates were examined. Details regarding the Dual‐luciferase reporter assays primers were provided in Table S5.

Author contributions

R. Xia and Y. Zhang designed and conducted this study. Y. Zhang performed the experiments with the help of Z. Zeng, H. Hu, M. Zhao, X. Ma, Y. Liu, Y. Hao and J. Xu, while analysed the data with the assistance of C. Chen and G. Li, Y. Zhang and R. Xia wrote the paper.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

Figure S1 Stem‐loop of litchi miR482/2118 superfamily.

Figure S2 Expression of miR482/2118 and phasiRNAs from 23 PHAS genes triggered by miR482/2118 in litchi different organs.

Figure S3 Sequence similarity shown by alignment of LITCHI007943 and two LcGID1 genes.

Figure S4 LcTASL‐ and LcGID1‐derived phasiRNA acted in cis on their precursor.

Figure S5 LcTASL1‐derived 21‐nt phasiRNAs cleaved LcGID1.

Figure S6 LITCHI007944 (LcTASL2) is a long noncoding RNA (lncRNA).

Figure S7 Alignment of two LcTASL genes and LcGID1B.

Figure S8 LcTASL1‐derived 21‐nt phasiRNAs cleaved LcTASL2.

Figure S9 LcTASL2‐derived 21‐nt phasiRNAs cleaved LcGID1.

Figure S10 Expression pattern of key genes and associated sRNAs in the miR482/2118‐TASL‐phasiRNA‐GID1 pathway in various organs of ‘HZ’.

Figure S11 Expression pattern of key genes and associated sRNAs in the miR482/2118‐TASL‐phasiRNA‐GID1 pathway during litchi cv. ‘NMC’ fruit development.

Figure S12 Prediction target site of LcTASL1‐derived phasiRNAs on Arabidopsis AtGID1.

Figure S13 Prediction target site of LcTASL2‐derived phasiRNAs on Arabidopsis AtGID1.

Figure S14 miR482/2118‐TASL‐GID1 pathway constructed in Arabidopsis.

Figure S15 At‐miR472 cleaved LcTASL1 in 35S: LcTASL1 transgenic lines.

Figure S16 Phenotypes of transgenic Arabidopsis plants.

Figure S17 Expression of four LcGAMYB genes in different development stages of ‘NMC’ seed.

Figure S18 Genome browser view of the distribution of the LcGAMYB33 binding peaks in the promoter and gene body of LcTASL2, LcGID1C and three LcDELLA.

Figure S19 Electrophoretic mobility shift assays (EMSAs) showing the binding ability of LcGAMYB33 with the mutant promoters of LcMIR482e in vitro.

Figure S20 LcTASL1‐derived 22‐nt phasiRNA cleaved LcTASL2.

Figure S21 Phenotype of litchi cv. ‘Feizixiao’ (‘FZX’) mature fruit.

Table S1 Revisiting miR482–2118 family in litchi.

Table S2 Revisiting target genes of miR482–2118 family in litchi.

Table S3 Revisiting PHAS loci of miR482–2118 family in litchi cv. ‘HZ’ different organs.

Table S4 Promoter analysis of key genes in the miR482/2118‐TASL‐GID1‐DELLA pathway.

Table S5 Primers and probes used in this study.

Data availability

The sRNA and PARE data sets generated from Ma et al. (2018) study have been deposited in National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) under the accession number GSE98698. All RNA‐seq and sRNA‐seq data generated from this study were deposited in NCBI under the BioProject ID number PRJNA951495.