Introduction

Boron (B), an essential micronutrient, plays a crucial role in plant growth and development as well as sexual reproduction1,2. It is involved in the formation of the primary cell wall3, membrane integrity4,5 and cell-wall structure6 and participates in diverse physiological and biochemical processes, including cell-to-cell signalling7, secondary metabolism8, and gene expression9, in higher plants. However, our knowledge of the exact role of B in intercellular processes is still limited.

Given that the concentration range from deficiency to toxicity is narrower for B than for any other plant essential nutrient10, B-deficiency11 and/or B-toxicity12,13 can easily be observed in diverse plant species, including Citrus species. B-deficiency can be corrected by applying B fertilizers. However, improper application of B fertilizer leads to B-toxicity13,14. In agricultural lands in different regions of the world, B-toxicity often occurs in B-rich soils or in soils where B accumulates due to a continual inflow of desalinated irrigation waters or industrial pollutants that are rich in B, resulting in a low yield and poor quality of many crops and horticultural plants12,15. However, the molecular basis of B-toxicity in Arabidopsis thaliana was not reported until recently, by Aquea et al.16. These authors found that B-toxicity induced the expression of genes related to ABA signalling but repressed the expression of genes encoding water transporters, triggering a water-stress response associated with root growth inhibition.

Plant tolerance to B-toxicity varies widely among species and/or cultivars12,13,17,18. However, the mechanisms underlying tolerance to B-toxicity in plants are still controversial. The most widely accepted theory, suggested for herbaceous plant species, is that there is a reduction of intercellular B levels through B efflux transporters. In barley (Hordeum vulgare), for instance, up-regulation of the B efflux transporter BOR1 in the roots decreases the absorption of B from soils15,19,20, whereas up-regulation of BOR2, which is homologous to BOR1, in the leaves leads to transport of B out of symplasts and into apoplasts17,18. Similarly, down-regulation of another efflux transporter, nodulin-like intrinsic protein (NIP) in roots can also decrease B transport from roots to shoots, thus reducing B accumulation in aboveground plant parts21. However, a recent physiological study by Landi et al.22 indicated the involvement of photoprotection by foliar anthocyanins in the tolerance mechanisms of sweet basil (Ocimum basilicum) to B-toxicity.

In Citrus, B-toxicity mainly occurs in the leaves and has little phenotypic effect on the roots. Although similar total B levels have been observed in the leaves or roots of tolerant and intolerant species, free B levels are significantly lower in the leaves of tolerant species than in those of intolerant species23. Nevertheless, transcriptome24 and proteome analyses25 have not yet identified any differentially expressed genes homologous to either BORs or NIPs from B-toxic Citrus leaves or roots. Further anatomical studies showed that the effects of B-toxicity on leaf structure are mainly limited to leaf veins. B-toxicity results in intensive exocytosis, accompanied by cell-wall thickening in the phloem of tolerant C. sinensis but induces cell death of phloem tissue through autophagy in intolerant C. grandis23. Our previous studies suggested that other mechanisms of tolerance to B-toxicity might exist in woody Citrus plants.

MicroRNAs (miRNAs) are 21–24 nucleotide (nt) small RNAs generated from non-coding RNA genes. As important post-transcriptional regulators, miRNAs have been shown to negatively regulate gene expression by directing the slicing of target mRNAs in the central portion of complementary regions26 or through inhibition of target mRNA translation27,28. Evidence demonstrates that plant miRNAs play a crucial role in the regulation of plant growth and development as well as the response to biotic and abiotic stresses29,30,31,32. For example, in A. thaliana, miR399 has been experimentally proven to maintain phosphorus (P) homeostasis by regulating UBC24 transcript levels33, whereas miR395 mediates the regulation of sulfate (S) accumulation and allocation by targeting APS and SULTR2;134, respectively. Up-regulation of such miRNAs under P- or S-deficiency suggests that miRNAs are involved in the adaptive responses of plants to nutrient stress. Recently, two additional reports, by Ozhuner et al.35 for barley and Lu et al.36 for Citrus, indicated that a number of miRNAs are up- or down-regulated under either B-toxic or B-deficient conditions, suggesting that miRNA expression could also respond to micronutrient-related stress. Nevertheless, little information about B-toxicity-responsive miRNAs is available in woody plants.

With the objective of exploring the mechanism of tolerance to B-toxicity in woody plants, we report high-throughput sequencing data (Illumina) in addition to a bioinformatic analysis of small RNAs from the leaves of B-tolerant and B-intolerant Citrus species treated with different B levels. Predicted targets of the candidate miRNAs were then verified using modified 5′-RACE and quantified via qRT-PCR. On this basis, two validated target genes that function in cell-wall metabolism were anatomically verified. Our results indicated that miR397a plays a pivotal role in the tolerance of woody Citrus to B-toxicity by targeting genes that are responsible for secondary cell-wall synthesis.

Results

Plant growth

As we previously reported, in C. grandis, B-toxicity leads to progressive basal-to-top development of toxic symptoms in the leaves, beginning with tip yellowing, followed by marginal and interveinal chlorosis, and finally, senescence and abscission. However, no visible symptoms were observed in C. sinensis leaves under the same conditions, suggesting that C. sinensis is more tolerant to B-toxicity than C. grandis.

Primary data analysis of sequences from small RNA libraries

The total numbers of raw reads generated from B-sufficient and B-toxic libraries via Solexa deep sequencing were 20,680,130 and 16,889,149 for C. grandis and 19,886,678 and 13,833,254 for C. sinensis, respectively. After removing adaptor sequences and filtering out low-quality and ‘N’-containing reads, 11,438,525 (55.31%) and 8,350,155 (49.44%) clean reads were retained from the B-sufficient and B-toxic libraries of C. grandis and 11,188,651 (56.26%) and 8,185,253 (59.17%) from those of C. sinensis, respectively. The numbers of unique clean reads generated from the libraries of the same species subjected to different B treatments were 1,988,267 (9.61%) and 1,350,479 (8.00%) in control and B-toxic C. grandis leaves and 1,593,866 (8.01%) and 1,214,103 (8.77%) in control and B-toxic C. sinensis leaves, respectively. The annotation of the sRNAs is shown in Table 1.

Table 1 Summary of sRNAs from the leaf samples of C. sinensis and C. grandis treated with sufficient- and toxic-B level.

Based on the constructed library datasets, it was observed that the unique clean reads exhibited an uneven length distribution, with the majority (~85%) ranging from 18 to 24 nt in length (Supplementary Fig. 1). Among the unique clean reads, sRNAs of 21 nt were the most abundant, followed by those of 24, 20 and 22 nt. Further analysis revealed that on average, up to 8.23% of the 21 nt sequences, but only 1.25% of the 24 nt sequences, in the four libraries mapped to the C. clementine genome and were assigned to miRNAs, suggesting that 21 nt-long sRNAs are the most important miRNAs in Citrus plants. These results were consistent with previous reports by Song et al.37 for C. trifoliate and Yin et al.38 for soybean [Glycine max (L.) Merr.]. One of the important characteristic features of miRNAs is their initial 5′-uridine39. Here, 81.26 ± 3.11% of the 21 nt-long sRNAs in our four libraries started with a 5′ uridine, whereas 55.36 ± 3.82% of the 24 nt-long sRNAs began with a 5′ adenosine (Supplementary Fig. 2). Nevertheless, B-toxicity resulted in more 21 and 24 nt reads in C. grandis and fewer 20, 21 and 24 nt reads in C. sinensis.

Identification of conserved and non-conserved miRNAs using sRNA libraries and bioinformatic analysis

To identify known miRNAs, the sequences of the clean reads from the four sRNA libraries were aligned with known miRNAs from other plant species in miRBase 21. A total of 750,280 perfectly matched unique sequences were found to be orthologues of 221 reported miRNA families. These conserved miRNAs were assigned to 387 putative precursors, which showed dramatic variations in copy number ranging from 2 to 29,727, indicating a significant discrepancy in their accumulation in Citrus (Supplementary Dataset S1). It is worth mentioning that many conserved miRNAs in the Citrus genome have more than one locus (Supplementary Fig. 3); such loci can be located on different chromosomes containing different precursor sequences, and mature miRNAs can arise from either the 5′ or 3′ arms of precursors (Fig. 1a). In addition, we identified a number of single-base variants of miRNAs in the four sRNA libraries. Such miRNAs exhibited moderate read counts; hence, they may not be derived from occasional sequencing errors, indicating that single nucleotide polymorphisms (SNP) of miRNAs might exist in plants, as reported previously39.

Figure 1: Bioinformatic analysis of miR156/157 members in Citrus.
figure 1

(a) Secondary structure of miR156/157 isoforms. miRNA precursor sequences were predicted using MTide (Zhang et al. 73) and folded by RNAfold (Hofacker, 2003). Positional information of each miRNA precursor refers to the start position on the respective chromosome (C. clementine genome v1.0). Cloned mature miRNA sequences were highlighted in red and miRNA* sequences in blue. (b) Alignment of mature miRNA isoforms using DNAMAN (version 7); ath, Arabidopsis; ghr, Gossypium hirsutum. (c) Polymorphism of miR156/157 members and their target (Ciclev1001392m) binding sites. miR-SNPs in miR156/157 are unlikely to change the target gene except for those in Scaffold_3_470286, which increased the mismatch positions, resulting in loss of its target. Watson-Crick pairing (vertical dashes), G:U wobble pairing (o) and mismatch (×) are indicated (the same below).

As small functional units, the biological functions of miRNAs may be altered by single nucleotide shifts in their precursor sequences, and in mature miRNA sequences in particular, driving the evolution of new miRNAs40. Elucidating the sequence variations of conserved miRNA isoforms could be quite helpful in the investigation of miRNA evolution in plants. Sequence alignment revealed that both SNPs and length differences of miRNAs (miR-LDs) could be observed (Fig. 1b; supplementary Fig. 4). In most miRNAs, the mature sequences of a given family exhibit a divergence of 1–3 nucleotides at both ends of the sequence. Seed sequences generally remain highly conservative, and SNPs are mainly located at mismatch positions. For example, the SNP sites observed in miR156/157 were identical or quite similar, causing diverse members of this miRNA family to target identical or similar genes, thus showing functional conservation (Fig. 1b,c). In contrast, the situation for the miR170/171 family was found to be complicated, with multi miR-LDs accompanied by divergent SNP sites potentially causing mismatches of miRNA-mRNA binding sites, resulting in changes in or loss of their target genes (supplementary Fig. 4) and even potential generation of new miRNAs39. These phenomena are consistent with previous reports on the miRNAs of both plants and animals39,40,41,42.

The remaining non-annotated sequences (538,928 reads; Supplementary Dataset S1) were assigned using MTide software (http://bis.zju.edu.cn/MTide/) for novel miRNA prediction. Finally, 312 candidate novel miRNAs were identified, and their characteristic hairpin structure formation ability was predicted. Considering that the availability of less stable anti-sense miRNA (miRNA*) sequence, indicating precise excision of a discrete miRNA/miRNA* duplex from the stem-loop precursor, is an important criterion for the validation of novel miRNAs43, we also searched for miRNAs* in our datasets; 261 miRNAs* (83.65%) for new Citrus miRNA candidates were found, providing evidence supporting them as new miRNAs. The remaining 51 predicted miRNAs without corresponding miRNAs* can be considered potential novel miRNAs based on the dominance of their read counts from one arm of the predicted stem-loops43 and the low detectability of corresponding miRNAs* for a large number of candidate miRNAs that exhibited only a few reads39,44. However, further verification approaches will be needed.

Differential expression of miRNAs under B-sufficient and B-toxic conditions

IDEG6 was applied for the analysis of differentially expressed miRNAs45 because we used mixed samples for sRNA library construction and sequencing, without biological replicates. Our results showed species specificity of the miRNA expression patterns in the two Citrus species, even under different B levels (Fig. 2). However, our real interest was in miRNAs that were either shared by or specific to the two Citrus species, or were differentially expressed in response to different B treatments. We compared the miRNA expression levels in each species after treatment with different B levels and identified 51 (25 up- and 26 down-regulated) and 20 (16 up- and 4 down-regulated) differentially expressed miRNAs in C. grandis and C. sinensis, respectively, after B-toxic treatment. Among the differentially expressed miRNAs, one conserved miRNA (miR5156) expressed in C. grandis was species specific but was repressed under B-toxic conditions. Another novel miRNA (ci-miRN4) exhibited opposite expression patterns in the two species, whereas the rest were significantly up- or down-regulated only in either C. grandis or C. sinensis (Supplementary Dataset S2).

Figure 2: DEG clusters of differentially expressed miRNAs in Citrus leaves treated with different B levels.
figure 2

The red arrowhead showed miRNAs which specifically expressed in C. sinensis, whereas the black one indicated those in C. grandis.

Validation of miRNA expression patterns through Illumina sequencing

To confirm the consistency of the deep sequencing results and comparative analyses, we verified the expression patterns of 20 miRNAs using stem-loop qRT-PCR. Among the differentially expressed miRNAs, miR395a and miR397a were the most significantly up-regulated in B-toxic C. grandis leaves, but both were down-regulated in B-toxic C. sinensis leaves (Fig. 3). Overall, the qRT-PCR results were similar to those from direct sequencing in 82.5% of cases. The exceptions were miR847, miR395a and miR160a in C. sinensis and miR6232a, miR2622.2, miR160a and a non-conserved ci-miRN16 in C. grandis (Fig. 3, Supplementary Dataset S2). Among these miRNAs, ci-miRN16 was down-regulated under B-toxicity when assayed via qRT-PCR, whereas it was up-regulated according to direct sequencing. This discrepancy may have resulted from the adaptor linkage efficiency in deep sequencing or PCR bias before sequencing38.

Figure 3: Relative abundance of miRNAs in Citrus leaves under B-sufficient and B-toxic conditions.
figure 3

MiRNA expression in B-sufficient and -toxic leaves was determined using stem-loop qRT-PCR. Results represent mean ± SD (n = 3). “*”, “**” and “***” indicate significant difference at P < 0.05, 0.01 and 0.001 level, respectively (the same below). All the values were expressed relative to the control.

Prediction of targets for differentially expressed miRNAs through GO analysis

Here, a total of 751 genes were predicted for 257 differentially expressed conserved and non-conserved miRNAs among the four libraries using bioinformatics. Because of the species and/or lineage specificity of miRNA expression in diverse plants and animals46,47,48,49,50,51, miRNAs that are specifically expressed in either C. grandis or C. sinensis and consistently expressed under B-sufficient and B-toxic conditions might be involved in the differences between the two Citrus species under B-toxic stress. However, the majority of such species-specific miRNAs are non-conservative and have no predicted targets. To better understand how miRNAs function in the response/adaptation of Citrus to long-term B-toxic stress, we therefore focused on target genes predicted from miRNAs that showed a discrepancy between B-sufficient and B-toxic conditions in each species. Finally, 102 (71 for C. grandis and 31 for C. sinensis, respectively) genes were obtained from the 71 (51 and 20 for C. grandis and C. sinensis, respectively) differentially expressed miRNAs identified in leaves (Supplementary Dataset S3). It is noteworthy that one miRNA might have several potential targets, which may either belong to the same family (e.g., genes targeted by miR397a) or have distinct predicted functions (e.g., genes targeted by miR847); conversely, a single gene can be targeted by several miRNAs, some of which are up-regulated, whereas others are down-regulated (e.g., Ciclev10010074m was predicted to be targeted by both miR6232a and ci-miRN16, which were down- and up-regulated, respectively, in C. grandis).

GO categories were then assigned to all the predicted targets according to the cellular component, molecular function and biological process categories (Fig. 4). The results indicated that noticeable changes in the sample frequencies of the GO terms, in which the potential targets of miRNAs that were differentially expressed under B-toxic stress were enriched, appeared between the two Citrus species in 5 categories. These categories were macromolecular complexes, extracellular regions and the membrane-enclosed lumen, which were classified as cellular components, and transport activity and structural molecule activity, which were classified as molecular functions.

Figure 4: GO analysis of target genes predicted from differentially expressed miRNAs.
figure 4

Green rectangle indicated significant alternations of sample frequencies (Blue bars) of GO terms. Red bars displayed background frequencies of GO terms, in which the putative B-toxic-adaptation-relevant miRNA targets were enriched.

qRT-PCR relative expression analysis of target genes

Eighteen genes targeted by 11 differentially expressed miRNAs (seven conserved and four novel miRNAs) were assayed via qRT-PCR. Only six genes exhibited the expected changes in mRNA levels, suggesting that they might be regulated via miRNA-mediated cleavage under B-toxic stress. Two genes targeted by miR164a and one targeted by miR160a were not detected in control leaves, despite being detected at extremely low levels relative to actin in B-toxic leaves. The expression of three other target genes was positively correlated with the levels of their corresponding miRNAs. The remaining six target genes maintained a relatively stable expression level (Table 2, Fig. 5a,b). Overall, our results are consistent with previous reports in Citrus36 and in Arabidopsis52, validating the low efficiency of expression profiles for predicting miRNA target genes.

Table 2 Relative expression of miRNA targets under B-toxicity in Citrus using qRT-PCR.
Figure 5: QRT-PCR expression profiles and experimental validation of the predicted mRNA targets for miR160a (a) and miR397a (b).
figure 5

The mRNA cleavage sites were determined by RLM-5′-RACE. The targeted mRNA sequences and mature miRNA sequences were shown. Vertical arrows indicated the 5′ termini of miRNA-mediated cleavage products, as identified by 5′-RACE, with the frequency of clones shown. ND, Not detected in controls.

Verification of potential ci-miRNA target genes with RLM-RACE

To investigate the confidence of the predicted targets of miRNAs, we experimentally verified the cleavage of selected targets through modified 5′-RACE analysis. The RLM-5′-RACE procedure was successfully used to map the cleavage sites in six predicted miRNA target genes. Ciclev10011194m, Ciclev10030860m, Ciclev10000695m and Ciclev10027901m were confirmed as real targets of miR160a, and Ciclev10038090m and Ciclev10011400m were confirmed as real targets of miR397a because all the 5′-ends of the mRNA fragments mapped to the nucleotide that paired with the tenth nucleotide of each miRNA with a higher frequency than was observed for each pairing oligo (Fig. 5). Ciclev10011194m and Ciclev10030860m are homologous to the Oryza sativa auxin response factor (ARF) gene ARF18, whereas Ciclev10000695m and Ciclev10027901m are similar to AthARF17and AthARF10, respectively. Ciclev10038090m and Ciclev10011400m are similar to isoforms of Laccase-17 (LAC17) and LAC4, respectively.

The expression of miR397a affects cell-wall polysaccharide deposition in Citrus leaf veins

Given that both LAC4 and LAC17 contribute to secondary cell-wall synthesis in Arabidopsis stems and that LAC17 is involved in the deposition of G lignin units in fibres53, we examined the cell-wall polysaccharides of vascular bundles in the two Citrus species treated with different B levels. In B-tolerant C. sinensis, the xylogens of vessel elements under B-toxic conditions were similar to those under B-sufficient conditions. Moreover, secondary deposition of cell-wall polysaccharides could often be observed in regions near the pits of vessel elements (Fig. 6a,c). Intriguingly, the components of the deposited secondary cell walls were so sensitive to the silver reagent that they appeared to be extremely condensed under transmission electron microscopy (Fig. 6c), suggesting a significant difference in their polysaccharide structure or biochemical components from those of xylogens. However, B-toxic treatment resulted in poorly developed vessel elements in C. grandis, with fewer silver particles deposited in the secondary cell wall (Fig. 6b,d).

Figure 6: Crossed sections of the leaf mid-veins stained using PASH produce.
figure 6

B-toxic treatment led to the deposition of secondary cell-wall components in the xylem parenchyma cells (XP) near the pits of vessel elements (VE) in C. sinensis (c), arrowheads); but restrained the development of VEs in C. grandis (d). Chl, chloroplast; M, mitochondria; PCW, primary cell walls.

Discussion

As important post-transcriptional regulators, miRNAs have been extensively studied in the past several years. A great number of studies have demonstrated that miRNAs are involved in the adaptive responses of plants to various biotic and abiotic stresses54,55,56,57,58. However, only a few miRNAs that respond to B-toxic stress have been identified in barley35 thus far, and there are no reports of B-toxicity-responsive miRNAs in woody plant species. Systematic evaluation of miRNAs in Citrus under B-toxicity will provide insights into the mechanisms underlying the tolerance to and/or the molecular basis of B-toxicity in woody plants.

Here, through Solexa deep sequencing, a total of 699 (387 conserved and 312 novel) candidate miRNAs were identified from Citrus leaves treated with different B levels. Among the identified miRNAs, 51 (23 conserved and 28 novel) were differentially expressed in the B-toxicity-treated leaves of intolerant C. grandis and 20 (6 conserved and 14 novel) in those of tolerant C. sinensis (Supplementary Dataset S2), demonstrating that miRNAs play an important role in Citrus responses to B-toxicity.

Sequence variations of miRNAs in Citrus

Variation in the seed sequences of miRNAs may be an important factor driving the evolution of miRNAs in plants39. As reported for Arabidopsis40,41,42, both miR-SNPs and miR-LDs appeared in identified miRNAs that were either conserved or non-conserved in Citrus (Supplementary Dataset S1). In most cases, the members of a miRNA family are located in clusters on different chromosomes, with identical SNP sites mostly being located beyond the seed sequences (i.e., miR156/157, Fig. 1b). However, some miRNA family members were found to show divergence in their locations and sequences (i.e., miR170/171, Supplementary Fig. 4), possibly due to the high heterozygosity of the examined species, which might have arisen as a result of long-term selection via cross-breeding in Citrus. Interestingly, some of the miRNAs (i.e., miR156c and miR171h) that exhibited extensive variations in sequences beyond the seed region (Fig. 1b; Supplementary Fig. 4) were expressed in a species-specific manner at extremely low levels (Supplementary Dataset S1) and presented no predicted targets, thus allowing plants to accumulate mutations under selective pressure for miRNA evolution, indicating that species specificity might be involved in miRNA evolution.

Interactions between miRNAs and their predicted targets

At present, the most efficient way to assess and define the putative functions of a given miRNA in plants is target prediction using bioinformatics, which is based on the high degree of homology between miRNAs and their target sequences59. Using computational techniques, a total of 102 genes were predicted from 26 B-toxicity-responsive miRNAs in Citrus (Supplementary Dataset S3). As noted by Meyers et al.43, it is very difficult to validate the true targets of a given miRNA, and the outcomes of both our pRT-PCR and 5′-RACE assays were disappointing: only 6 of 18 (33.33%) selected target genes exhibited the expected changes in mRNA levels when examined via qRT-PCR (Table 2; Fig. 5a,b; Supplementary Dataset S1), and only 6 of 29 (20.69%) genes were confirmed as true targets cleaved via miRNA-mediated cleavage when RLM-5′-RACE analysis was applied.

The contradictions in our results might arise from the diversity of miRNA regulatory mechanisms as well as the complexity of the gene regulation network at a genomic scale. It has been known for quite some time that plant miRNAs regulate their specific target mRNAs via cleavage based on their near-perfect complementarity to their target genes26. However, it was not until very recently that Li et al.28 experimentally demonstrated translation inhibition of targets by miRNAs in Arabidopsis. These two theories of miRNA regulation, together, might explain most of our failures to validate miRNA targets using either the qRT-PCR or 5′-RACE approach.

It is interesting that some miRNA targets, such as Ciclev10027901, that were confirmed as real targets of miRNAs according to the 5′-RACE approach and were therefore expected to be negatively regulated at the transcription level were instead found to be positively expressed as their miRNA levels were increased according to qRT-PCR (Figs 3 and 5a). Considering that a single gene can be targeted by several miRNAs simultaneously and that another potential miR160a cleavage site, located approx. 60 bp downstream from the verified exact cleavage site, was identified in our 5′-RACE experiments, this contradiction might be attributed to the complex gene regulation network.

Nevertheless, our data and those of others37,38,39,60 support the prevailing model indicating that miRNAs regulate specific targets through cleavage in plants.

MiRNAs associated with the response to B-toxicity

High-throughput sequencing technology has recently been successfully applied to identify miRNAs on a genomic scale in plants under either B-toxic (in barley leaves and roots)35 or B-deficient (in Citrus roots)36 stress. Unfortunately, the published expression profiles of miRNAs responsive to B-stress have little in common with ours. This difference might be due to the species specificity and spatio-temporal specificity of miRNA expression in plants. It could also be caused by different expression patterns of miRNAs in the leaves of barley and Citrus under B-toxicity, which might indicate different tolerance mechanisms between herbaceous and woody plants.

Among these differentially expressed miRNAs, miR395a and miR397a were confirmed by qRT-PCR to be the most significantly up-regulated and exhibited negative expression patterns between C. grandis and C. sinensis (Fig. 3) when exposed to B-toxicity. In Arabidopsis, miR395 targets genes belonging to the ATP sulfurylase gene family61 as well as sulfate transporters55 to modulates sulfate accumulation and allocation34. In Citrus, miR395a was predicted to target Ciclev10004931m, described as 3-ketoacyl-CoA thiolase 2, which is presumed to respond to wounding or positively regulate the abscisic acid-activated signalling pathway. However, its predicted cleavage site was confirmed to be false via the 5′-RACE approach. Considering its high expression level and the fact that B-toxicity resulted in leaf senescence in C. grandis, miR395a might regulate its target Ciclev10004931m via translation inhibition.

Using 5′-RACE, we confirmed two genes (LAC4 and LAC17) as real targets of miR397a. In Arabidopsis, in vivo expression analysis indicated that both miR397 and its targets (LAC4 and LAC17) are mainly expressed in vascular tissues53,62; over-expression of miR397b causes a reduction of the lignification of vascular and interfascicular tissues62. Similarly, an increase in the expression of miR397a was accompanied by down-regulation of both LAC4 and LAC17 after B-toxicity treatment, leading to poor development of vessel elements in the vascular bundles of C. grandis (Fig. 6d), suggesting that both miR397a and its two targets might also be specifically expressed in vascular bundles in Citrus. Given that LAC4 and LAC17 cluster in different groups based on multiple amino acid sequence alignment63,64, they might have different biological functions. In a study by Wang et al.62, LAC4 was found to be involved in lignin biosynthesis and the seed yield, whereas loss-of function single mutants of LAC17 displayed no significant morphological changes. In the current study, LAC17 remained at the same level, whereas LAC4 was up-regulated approx. 2-fold in C. sinensis when treated with toxic levels of B. A PASH procedure performed on B-toxicity-treated leaf veins of C. sinensis revealed that secondary cell-wall polysaccharides in xylem parenchyma cells were deposited in regions near the pits of vessel elements but reacted differently to the silver reagent than did xylogen (Fig. 6a,c), suggesting that LAC4 might have an additional function in secondary cell-wall synthesis. Thus, the modification of cell-wall structures and components in the xylem resulting from the modulation of both LAC4 and LAC17 by miR397a might play an important role in the Citrus response to long-term B-toxicity

Both miR160a and miR164a were significantly down-regulated in B-toxic C. grandis leaves (Fig. 3). Using bioinformatics, 6 ESTs homologous to the ARF gene family were predicted to be targeted by miR160a (Supplementary Dataset S3), among which Ciclev10000693m, Ciclev10000695m and Ciclev10000696m might be copies of the same gene with different chromosomal loci based on DNA sequence alignment and ORF prediction. Therefore, only Ciclev10000695m and the other 3 potential target genes were included in the subsequently experiments, and all were confirmed as real targets of miR160a. Based on the extensive documentation of the involvement of ARFs in auxin signalling65,66, miR160a might play a role in auxin signalling in the Citrus response to B-toxicity; however, it is difficult to draw a conclusion because of the opposite expression patterns of the four ARF genes (Fig. 5a). For miR164a, two genes, encoding a NAC domain protein (Ciclev10008619m) and polygalacturonase (Ciclev10033386m), were predicted as targets. In plants, polygalacturonase is responsible for pectin solubilization and depolymerization67. Down-regulation of miR164a in response to B-toxicity might increase polygalacturonase activity, resulting in cell-wall modifications. This notion was supported by a recent report by Ghanati and Heidarabadi6, who found that an increased B supply led to a significant decrease in the cell or cell-wall dry weight and a reduction of the relative amounts of major wall components, including pectin, cellulose and hemicellulose B.

Most of the other differentially expressed miRNAs identified in the present study were predicted to target diverse genes related to transcription factors that remain to be determined and will be further discussed elsewhere.

Possible tolerance mechanisms of Citrus to long-term B-toxicity

The most widely accepted theory of B transport from the roots to the shoots of plants involves passive transport via mass flow within the transpiration stream68. Upon reaching the leaf, free B in the transpiration stream will first flow into the phloem, before being transported into mesophyll tissue, either passively or actively, in dicotyledons. Therefore, we have reason to believe that B may first accumulate in the phloem, generating toxic effects, which has been confirmed by the finding that B-toxic treatment results in limited toxic effects in leaf veins through triggering programmed cell death of phloem tissue in intolerant C. grandis23.

In the present study, B-toxic treatment resulted in secondary deposition of cell-wall components in regions near the pits of vessel elements in tolerant C. sinensis and poorly developed vessel elements in intolerant C. grandis (Fig. 6), indicating that modifications of cell-wall structures and components in the xylem might be involved in mitigating B-toxicity to the phloem.

It has been demonstrated that in plants, B forms the most stable diesters with cis-diols on furanoid rings69, participating in cell-wall assembly by cross-linking polysaccharide monomers into dimers70. Given that B exists as boric acid under physiological conditions and can form esters and complexes with a wide variety of hydroxyl-rich compounds71, cell-wall polysaccharides as well as other cell-wall components, such as hydroxyl-proline-rich glycoproteins4, could be considered candidates for B binding, even with weak bonds. Therefore, alterations of cell-wall structures and components in the xylem might restrict the inflow of free B from the xylem into the phloem in tolerant C. sinensis. Indeed, the expected lower level of free B in B-toxicity-treated leaves of C. sinensis has been detected23.

Above all, the results reported herein illustrate that through the modulation of LAC17 and LAC4, miR397a plays a pivotal role in secondary cell-wall biosynthesis in vascular bundles, thereby contributing to the adaptation of woody Citrus to long-term B-toxicity.

Materials and Methods

Plant culture and treatments

Plant culture and B treatments were performed as previously reported23. Briefly, 11-week-old seedlings were sand cultured with a nutrient solution containing 10 (control) or 400 (B-toxic) μM H3BO3 every other day for 15 weeks. Fully expanded leaves excised at 1/3 the height of the plants were immediately frozen in liquid N2 at the end of the culture experiment and stored at −80 °C until RNA extraction. Mid-vein samples were directly fixed with 2.5% glutaraldehyde in 0.1 M phosphate buffer (PBS, pH 7.4) at 4 °C for 24 h for anatomical analyses.

RNA isolation and quality assessment

Total RNA was extracted using the TRIzol reagent (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions, with minor modifications (3 M sodium acetate was used for the removal of polysaccharides). The quantity and purity of the total RNA were analysed using a NanoDrop 2000 spectrophotometer (Thermo scientific, USA) and an Agilent 2100 bioanalyser (USA) with the RNA Integrity Number (RIN) > 8.0. The total RNA was divided, stored at −80 °C, and used for high-throughput sequencing, qRT-PCR and RACE amplification.

Small RNA library construction and high-throughput sequencing

Approximately 1 μg of mixed total RNA from five replicates was used to prepare a small RNA (sRNA) library with the TruSeq Small RNA Sample Prep Kit (Illumina, USA) according to the manufacturer’s protocol. The sRNA libraries were gel-purified in 6% PAGE gels, and their concentration was then diluted to 2 ng/μl. The insert size of the sRNA libraries was tested with an Agilent 2100 bioanalyser. After accurate quantification via qRT-PCR, single-end sequencing was performed on an Illumina HiSeq 2500.

Data processing, sRNA annotation and miRNA identification

The raw reads obtained from Illumina sequencing were subjected to the Illumina pipeline filter, and the datasets were further processed to remove adapter dimers, junk and low-complexity sequences, common RNA families (rRNA, tRNA, snRNA, snoRNA) and repeats, as reported previously72. The remaining sequences were subjected to BLAST searches against miRBase 21 (http://www.mirbase.org/) to identify known miRNAs. Non-annotated sequences of 18–30 nucleotides (nt) in length were mapped to the Citrus clementina genome (JGI version 1.0, http://www.phytozome.org/clementine.php) with the MTide program (http://bis.zju.edu.cn/MTide) to identify novel sequences with stem-loop precursors, as described by Zhang et al.73. MiRNA prediction was performed according the key criteria previously used by Meyers et al.43.

Differential expression analysis of miRNAs

For the analysis of miRNA expression, both the fold-change and P-value of each identified miRNA in the control and B-toxic libraries were calculated as previously reported by Lu et al.36. The P-value was then adjusted to the false discovery rate (FDR) using the Benjamini Hochberg Method. A 2-fold cut-off was set to determine up- and down-regulated miRNAs, in addition to a FDR of less than 0.01.

Prediction of miRNA target genes

The prediction of genes targeted by differentially expressed miRNAs was performed with TargetFinder based on rules suggested by Allen et al.61 and Schwab et al.74.

GO analysis of miRNA targets

All predicted targets were mapped to GO terms in the GO database (http://www.geneontology.org/), and gene numbers were calculated for each term. The GO results were expressed according to three categories: cellular component, molecular function and biological process75.

MiRNA assay using stem-loop qRT-PCR

To monitor the relative abundance of miRNAs obtained from Illumina sequencing, 20 miRNAs (11 conservative and 9 non-conservative) were selected to perform qRT-PCR using actin (AEK97331.1) as an internal control. Stem-loop RT primers were designed according to Xu et al.76. MiRNA-specific primers (forward) were designed according Chen et al.77, and their quality was assessed using Primer Software (PREMIER Biosoft International, USA, Version 5.0). All the primers employed for qRT-PCR are listed in Supplementary Table 1.

For stem-loop qRT-PCR, approx. 1 μg of total RNA was used to create a reverse transcription pool with the TaqMan® MicroRNA Assay Kit (Takara, Japan) according to the manufacturer’s instructions. qRT-PCR was then performed as previously described by Lu et al.36, using the miRNA-specific forward primer and a Universal Reverse Primer (Supplementary Table 1). The cycling conditions were as follows: 10 min at 95 °C, followed by 40 cycles of 95 °C for 15 s and 60 °C for 60 s. The samples subjected to qRT-PCR were run in three biological replicates with two technical replicates. Relative miRNA expression was calculated using the ddCt algorithm. Actin was employed as an internal standard for the normalization of miRNA expression, and leaves from control plants were used as a reference sample (set as 1). As total RNA was used for reverse transcription, melting curve analysis followed by agarose gel electrophoresis was conducted to determine the purity and fragment size of the amplicons.

Detection of target mRNAs via qRT-PCR

qRT-PCR analysis of target gene expression was performed as previously described78 with the Mastercycler Ep Realplex System (Eppendorf, Germany). Gene-specific forward and reverse primers were designed approx. 200 bp up- and down-stream from potential cleavage sites, to produce an expected fragment of approx. 100–300 bp in size. The quality of all primer pairs was assessed using Primer Software and the primers are provided in Supplementary Table 2.

Verification of miRNA cleavage sites via 5′-RACE

To verify the nature of potential miRNA targets and to examine how the miRNAs regulate their target genes, an RNA ligase-mediated 5′ rapid amplification of cDNA ends (RLM-5′-RACE) experiment was set up. RLM-5′-RACE was carried out following the instructions of the GeneRacer Kit (Invitrogen), as described by Yin et al.38. Briefly, poly(A)+ mRNA was isolated from total RNA using the polyAtract mRNA isolation system III (Promega, Madison, WI). The GeneRacer RNA Oligo adapter was directly ligated to poly(A)+ mRNA without calf intestinal phosphatase and tobacco acid pyrophosphatase treatment, according to the manufacturer’s instructions. The GeneRacer OligodT primer was then used to synthesize RACE-ready first-strand cDNA in a reverse transcription reaction. First-round PCR was performed using the GeneRacer 5′ primer and a gene-specific primer, followed by second-round PCR with the GeneRacer 5′ Nested Primer and a gene-specific nested primer, which was designed using Primer 5.0 software and was located 300–800 bp downstream from the potential cleavage sites. The gene-specific and nested primers used in these assays are listed in Supplementary Table 3. The nested PCR products were gel-purified, T-cloned and sequenced.

Cytochemical localization of cell-wall polysaccharides and electron microscopy

For anatomical analyses, mid-vein samples were post-fixed with 1% osmium tetroxide for 1.5 h at 4 °C. Tissue blocks were rinsed with 0.1 M PBS (pH 7.4), then dehydrated in a graded ethanol series, and the ethanol was subsequently replaced through four changes of acetone. The samples were embedded in an Epon resin mixture. Sections of 75 nm were obtained with a Leica EM UC6 ultra-microtome (Germany) and transferred to fomvar-carbon-coated nickel grids, then stained using a modified Pickett-Heaps procedure (PASH staining) as described by VanDerWoude et al.79. The prepared specimens were examined at 80 kV with a Hitachi 7700 electron microscope (Japan).

Experimental design and statistical analysis

A completely randomized design was applied with 20 pot seedlings per B-treatment. The experiments were performed with 3–5 replicates. Differences between treatments were subjected to independent t-tests using IBM SPSS statistics software (Version 22).

Additional Information

Accession codes: The sRNA sequence data from this study have been submitted to Gene Expression Omnibus (GEO) under accession GSE74434 at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE74434.

How to cite this article: Huang, J.-H. et al. Illumina microRNA profiles reveal the involvement of miR397a in Citrus adaptation to long-term boron toxicity via modulating secondary cell-wall biosynthesis. Sci. Rep. 6, 22900; doi: 10.1038/srep22900 (2016).