Identification of exosome-like nanoparticle-derived microRNAs from 11 edible fruits and vegetables

Isolation and purification of ELNs

Eleven edible plants (blueberry, coconut, ginger, grapefruit, Hami melon, kiwifruit, orange, pea, pear, soybean, and tomato) were randomly chosen for this study. All EPDELN samples were isolated and purified by differential centrifugation as described previously (Ju et al., 2013; Mu et al., 2014; Wang et al., 2015; Zhuang et al., 2015). Fruits and vegetables were purchased from a local market and washed three times. Apart from coconut, juice was extracted in two different ways. Fruits and vegetables with abundant juice (blueberry, grapefruit, kiwifruit, orange, and pear) were peeled, wrapped in gauze, and squeezed by hand, while less juicy ones (ginger, Hami melon, pea, soybean, and tomato) were ground with phosphate-buffered solution (PBS) in a mixer. The collected juice was sequentially centrifuged at 1,200 × g for 20 min, 3,000 × g for 20 min, and 10,000 × g for 60 min at 4 °C in a Sorvall Lynx 6000 centrifuge (Fisher Scientific, Shanghai, China) to remove large particles and cellular debris. The supernatant was filtered through a 1-μm membrane filter (Millipore, Bedford, MA, USA) and then centrifuged at 150,000  × g for 90 min at 4 °C in an LE-80 ultracentrifuge (Beckman Coulter, Palo Alto, CA, USA) to obtain EPDELNs. EPDELNs were resuspended in 250 μl PBS.

Atomic force microscopy (AFM) and nanovesicle particle size analysis

To morphologically characterize the isolated membrane fraction, specimens were diluted 1:1,000 in PBS, and absorbed onto freshly cleaved mica sheets for 20 min. To provide a surface coated with formulations to a suitable density, the mica sheets were rinsed three times with deionized water and then dried with filter paper before detection. Surface morphology was examined under an atomic force microscope (Asylum Research MFP-3D-Bio; Digital Instruments, Santa Barbara, CA, USA) as described by Mu et al. (2014). The particle size distribution of EPDELNs was evaluated using a Light Scattering System (Brookhaven BI-200SM) as previously described (Mu et al., 2014). Measurements were made in PBS at pH 7.0 at 25 °C after appropriate dilution of each EPDELN sample.

Small RNA sequencing and data analysis

Total RNAs from the 11 EPDELN samples were separately obtained using TRIzol LS reagent (Invitrogen, Waltham, MA, USA) according to the manufacturer’s instructions. After resuspending each RNA sample in 30 μl RNase-free water (Takara Bio, Shiga, Japan), the RNA quality was examined by 1% agarose gel electrophoresis and on an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Samples were stored at −80 °C. To construct each library, small RNA ranging in size from 14 to 36 nt was purified by polyacrylamide gel electrophoresis and ligated using proprietary adaptors. The modified small RNA was then reverse-transcribed into cDNA and amplified by PCR. Finally, the 11 libraries were sequenced on an Illumina HiSeq 2500 platform.

The resulting sequence reads were subjected to a series of stringent filters, including the removal of low-quality reads, repeat sequences, and adaptor sequences, to generate clean data. A bioinformatics pipeline (MIRPIPE) for miRNA discovery and profiling was then applied as previously described (Kuenne et al., 2014). MIRPIPE was also applied in a subsequent analysis. In brief, the filtered reads were identified against known plant mature miRNAs in miRBase21.0 (November 2016; http://www.mirbase.org/index.shtml) using the criterion of a maximum of two mismatches. The identified reads were treated as conserved miRNAs, and the number of reads per miRNA were normalized to reads per million (RPM). The miRNA profiling data have been deposited in NCBI Gene Expression Omnibus (GEO: GSE116095) and are accessible through the GEO Super Series.

Prediction and functional annotation of miRNA target genes

TargetScan (http://www.targetscan.org/vert_71/) was used to annotate miRNA target gene (Wei et al., 2017). The generated list of target genes was subsequently uploaded to the Database for Annotation, Visualization and Integrated Discovery (DAVID) bioinformatics resource (Huang da, Sherman & Lempicki, 2009) (https://david.ncifcrf.gov/) for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses. DAVID GO enrichment was regarded as significant using a criterion of P < 0.05. Sequence information on human mRNAs was collected from the NCBI database (http://www.ncbi.nlm.nih.gov/). The NCBI database and RNAhybrid (https://bibiserv.cebitec.uni-bielefeld.de/rnahybrid/) were used in combination to search for human mRNAs containing potential binding sites for plant miRNAs.

Cells, reagents, and oligonucleotides

The HeLa cell line obtained from the Chinese Academy of Sciences cell bank (SGST.CN) was cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, Waltham, MA, USA) supplemented with 10% fetal bovine serum (Gibco) and incubated at 37 °C in a humidified atmosphere containing 5% CO₂. Synthetic miRNA molecules with 2′-O methylation at the 3′ end were purchased from Gene Pharma (Shanghai, China).

Dual-luciferase reporter assay

To experimentally validate miRNA–target interactions, a dual-luciferase reporter assay system was used. Wild-type (WT) or mutant (MUT) binding sites were cloned into a pmirGLO vector (100 ng per well). The vector harboring WT or MUT binding sites combined with either synthetic MIR-168c (50 nM), synthetic MIR-8155 (50 nM), or an miRNA control (50 nM) were then co-transfected into HeLa cells grown in 96-well plates at a density of 5  × 10³ cells/well using Lipofectamine 3000 (0.3 μL per well) (Invitrogen, Waltham, MA, USA). Cells were collected after 48 h, and Renilla and firefly fluorescence levels were assayed using the Dual-Luciferase Reporter Assay System kit (Promega, Madison, WI, USA).

Statistical analysis

Data were presented as means ± SEM. Statistical analyses were performed using the Student’s t-test, with differences considered significant at p < 0.05.

Isolation and characterization of EPDELNs

To verify our hypothesis that miRNAs are enriched in EPDELNs and can potentially mediate interspecies intercellular communication, we randomly selected 11 commonly consumed fresh fruits and vegetables (Fig. 1 and Fig. S1). EPDELNs were isolated by differential centrifugation (Mu et al., 2014; Wang et al., 2013; Zhuang et al., 2015) and identified at the nanometer scale by AFM. The 11 different edible fruits and vegetables were found to contain a substantial quantity of round or oval vesicles (Fig. 1 and Fig. S1). These vesicles exhibited a morphological ultrastructure similar to that of exosomes from mammalian bodily fluids (Zhuang et al., 2015). To further characterize the EPDELNs, a dynamic light scattering (DLS) analysis was also performed. As shown in Fig. 1 and Fig. S1, the size distribution of EPDELNs isolated from the 11 fresh fruits and vegetables ranged from 100 to 1,000 nm, consistent with values obtained for other EPDELNs in previous studies (Ju et al., 2013; Mu et al., 2014; Wang et al., 2014).

Small RNA sequencing results

To investigate the expression profile of miRNAs in EPDELNs, total RNA of EPDELNs was extracted and subjected to capillary electrophoresis analysis. The EPDELNs were found to contain many small RNAs shorter than 25 nt in length, but little or no 18S and 28S ribosomal RNA, thus confirming the presence of small RNAs in EPDELNs (Fig. S2). Eleven small RNA libraries were then constructed (Table 1) and sequenced to generate a total of 202,560,279 raw reads. After applying a series of stringent filters, the remaining 141,076,120 reads (69.65% of raw reads) from all libraries were considered to be reliable miRNA candidates (Table 1). Because pre-miRNA sequences were unavailable in miRBase for certain edible plants and as genome sequence information was lacking for some plants in this study, the filtered reads were directly mapped to all known plant mature miRNAs in miRBase 21.0. Matching miRNAs were considered to be conserved miRNAs. In total, we identified 418 conserved miRNAs from the 11 EPDELN samples (Table S1), 32–127 per species (Fig. 2A). ELNs in ginger had the fewest miRNA species (n = 32) and those in soybean had the most (n = 127). Regarding the size distribution of the generated reads, a clear peak was observed at 20–22 nt (Fig. 2B), consistent with findings from previous studies on various plants in which the majority of small RNAs were 20–24 nt in length (Liu et al., 2015).

Prediction of possible human target genes of plant miRNAs

miRNA expression levels in each EPDELN library varied widely, from a few RPM to thousands, with the top 20 miRNAs contributing more than 92% (92.41%–99.17%) of the total miRNA expression (Figs. 2C–2E and Fig. S3). Evidence is increasing that plant-derived miRNAs can specifically bind target mammalian mRNAs and influence biological processes (Chin et al., 2016; Zhang et al., 2012a). To further explore the potential function of EPDELN-derived miRNAs in EPDELNs, bioinformatics analysis was used to predict relationships between miRNAs and their potential target genes. We predicted the human target genes of miRNAs of each EPDELN sample (Figs. 3A–3C and Figs. S4A–S4H) using TargetScan according to the principle of base complementary pairing between the plant miRNA seed region and target genes. The results of target gene prediction suggested that some miRNAs in EPDELNs potentially target and theoretically regulate mammalian genes (Fig. 3D, Table 2, and Table S2).

Recent evidence of the intake and bioavailability of dietary miRNAs in humans and animals suggests that plant-derived miRNAs possess immunomodulatory or cancer regulatory capacities (Cavalieri et al., 2016; Chin et al., 2016). Our prediction results also indicate that some of these miRNAs can directly target genes encoding inflammatory factors such as interleukin-6 (IL-6), interleukin-2 (IL-2), interleukin-5 (IL-5), and interleukin-1 (IL-1). For example, MIR-5781 in soybean ELNs can directly target IL17A, which plays an important role in inflammatory responses (Table 2). GO and KEGG term enrichment analyses revealed that the target genes of miRNAs in EPDELNs are related to immune cells and are significantly enriched in cancer-related signaling pathways, such as those associated with small-cell lung, endometrial, and colorectal cancers (Fig. 3D).

EPDELN-derived miRNAs bind target genes in HeLa cells

To further validate the prediction results, we performed a dual-luciferase assay to verify the interaction between MIR-168c and its potential target gene (TSC22D3) (Fig. 3E and Table S3). This indicated that MIR-168c binding significantly reduced luciferase activity for the wild-type target gene, whereas binding to the mutant site had no effect on luciferase activity (Fig. 3F). MIR-8155 was similarly analyzed by dual-luciferase assay, but the result was negative (Figs. S4I, S4J and Table S3). Together, these results suggest that plant miRNAs have the potential to target mammalian mRNAs.

Distribution of miRNA species and expression profiles across 11 EPDELN samples

To better compare the expression distribution of miRNAs among EPDELNs, we divided all miRNAs into three types: frequent miRNAs (FMs), moderately present miRNAs (MPMs), and rare miRNAs (RMs), corresponding respectively to miRNAs simultaneously present in 8–11, 4–7, or 1–3 kinds of EPDELNs. FMs, MPMs, and RMs comprised 26 (Fig. 4A), 39, and 353 miRNAs, respectively (Table S1).

Because miRNAs within ancient genes are more broadly and highly expressed than those within young genes (França, Vibranovski & Galante, 2016), we compared the expression levels of the three miRNA types. Notably, we found that the expression levels of FMs accounted for at least 39.46% of total miRNA expression in each EPDELN sample (Fig. 4B). This indicates that FMs have fewer miRNA species than RMs but a significantly higher cumulative expression level (Figs. 4B and 4C). Further inspection revealed that seven miRNAs were present in all 11 libraries (Fig. 4A) and were almost highly expressed miRNAs of each library (Table S1), reflecting the high expression of FMs. Among FMs, MIR-319 mainly acts on a TCP transcription factor underlying the robust and multilayered control of leaf development (Koyama, Sato & Ohmetakagi, 2017). Additionally, the overexpression of MIR-396 was shown to reduce stomatal cell number and stomatal density in Arabidopsis (Liu et al., 2009), and is thus of great importance for regulating plant tissue cell growth (Rodriguez et al., 2010). To screen for miRNAs highly expressed but only detected in single EPDELN samples, we identified 19 miRNAs from among the 353 RMs whose expression levels were in the top 20 of their own species (Table 3), including MIR-1919a (tomato), MIR-858a (pear), and MIR-1078 (ginger). Among these miRNAs, MIR-1078 in ginger ELNs can act on LEP (encoding leptin), and Fairfax et al. showed that leptin in turn associated with lipopolysaccharide-induced IL-6 expression (Fairfax et al., 2010).