PlantCircRNA: a comprehensive database for plant circular RNAs

Abstract

Circular RNAs (circRNAs) represent recently discovered novel regulatory non-coding RNAs. While they are present in many eukaryotes, there has been limited research on plant circRNAs. We developed PlantCircRNA (https://plant.deepbiology.cn/PlantCircRNA/) to fill this gap. The two most important features of PlantCircRNA are (i) it incorporates circRNAs from 94 plant species based on 39 245 RNA-sequencing samples and (ii) it imports the original AtCircDB and CropCircDB databases. We manually curated all circRNAs from published articles, and imported them into the database. Furthermore, we added detailed information of tissue as well as abiotic stresses to the database. To help users understand these circRNAs, the database includes a detection score to measure their consistency and a naming system following the guidelines recently proposed for eukaryotes. Finally, we developed a comprehensive platform for users to visualize, analyze, and download data regarding specific circRNAs. This resource will serve as a home for plant circRNAs and provide the community with unprecedented insights into these mysterious molecule.

Graphical Abstract

Graphical Abstract.

Introduction

Circular RNAs (circRNAs) represent recently discovered novel regulatory non-coding RNAs that are present in many eukaryotes (1–5). Traditional RNA detection and analysis methods require free 5′ and 3′ ends, but this classical method completely ignores and underestimates the expression and biological importance of circRNAs (6). The study of circRNAs has crucial research value. First, circRNAs are important components of the regulatory network mediated by non-coding RNA (7,8): They competitively bind to microRNAs (miRNAs), preventing miRNAs from binding to and negatively regulating their target; have special functions (9,10); play an important regulatory role in messenger RNA (mRNA) expression (11,12); and could be an important link to clarify the expression of regulatory network components (13–15). Second, circRNAs have great potential application. Indeed, the strong tissue specificity, structure, stability in the cytoplasm, and ability to play a long-term cell regulatory function (16,17) of circRNAs mean that they have a wide range of application prospects in agriculture (18) and medicine (19–21).

In 2011, Salmena et al. (22) proposed the ‘competitive endogenous RNA’ hypothesis, that is, mRNAs and long non-coding RNAs communicate with and regulate each other by using miRNA response elements (MREs) as a ‘universal language’. Specifically, various RNAs, including circRNAs, influence the expression levels of other RNAs by competing for limited miRNA resources (23–26). Since 2011, a number of studies have demonstrated the importance of circRNA expression, and circRNAs have become a research hot spot (27–30). In 2012, Salzman et al. (31) claimed to identify and experimentally verify the widespread existence of circRNAs in human cells by using computational methods and high-throughput sequencing of the full transcriptome. Later, two independent groups (13,14) confirmed that circRNAs can bind competitively to miRNAs, thus affecting the expression of the target mRNAs, and clarified that circRNAs affect the biological function of miRNAs. In 2022, Breuer et al. (32) and Meganck et al. (33) reported the design and expression of synthetic circRNAs affecting biological functions. Recently, Huang et al. (21) reported that noncanonical translation of circRNAs resulted in anti-tumor immunity.

There are several circRNA databases, including circbase (34), CircNet (35), circBank (36) and CIRCpedia (37). In addition, numerous plant circRNA studies have been published, and PlantCircBase (21 species) (38) and PlantCircNet (8 species) (39) have been developed from experimental validation and public resources. In 2016, our team developed AtCircDB (40) targeting Arabidopsis and CropCircDB (41) targeting crops, which have attracted wide attention from the community. In PubMed, we identified only 81 plant circRNA articles (related to 30 species) that have been published since 2012, which is quite a small number. A large comprehensive plant circRNA database is urgently needed for the research community. To fill this gap, we analyzed 39 245 total RNA-sequencing (RNA-seq) samples from 1928 independent studies to extract circRNAs. We found 673 443 circRNAs for 94 plant species. Using these sequences, we generated PlantCircRNA, which also includes the information from AtCircDB and CropCircDB and is a much more comprehensive resource. In addition, we incorporated tissue, stress information, as well as previous publications into the platform, which should help users to track tissue- and stress-specific changes in circRNAs. We also annotated the circRNAs with a new naming system recently proposed for eukaryotes (42). Finally, we built a user-friendly website to access PlantCircRNA to speed up the research of plant circRNAs. This comprehensive database should provide unprecedented insights into plant circRNAs.

Materials and methods

Data sources

We collected sequencing data as follows: On 26 May 2023, we searched the NCBI SRA database (43) (https://www.ncbi.nlm.nih.gov/sra) with the following keywords: species Latin name, ‘total RNA-seq’ and ‘Illumina’. We removed files with a size <1 G or polyA selection. On 2 October 2023, we further searched the database, and incorporated the second batch of data for further analysis. The final selected files had to meet the following criteria: (i) polyA-selected files were removed from the ‘LibrarySelection’ option; (ii) file size > 1 G and (iii) file must be ‘transcriptomics’ for ‘LibrarySource’ and ‘RNA-seq’ for ‘LibraryStrategy.’ On 1 December 2023, we searched the database with one additional keyword from tissue (‘root’, ‘stem’, ‘leaf’, ‘flower’ and ‘seed’) and abiotic stress (‘drought’, ‘heat’, ‘cold’ and ‘salt’) for the third batch of data. More detailed information about the sequencing samples is available to download under ‘Additional data’ section at the ‘Help’ page.

We searched the PubMed database (https://pubmed.ncbi.nlm.nih.gov/) with keywords (species name and ‘circular RNA’) to collect all related journal articles. We filtered the results to limit the publications to those published since 2012, the year the first circRNA was first reported. We manually looked through each article to find the related circRNAs. We incorporated these circRNAs into PlantCircRNA.

circRNA identification

Many studies have reported that the prediction accuracy and sensitivity of the current circRNA tool requires improvement (44). Hence, to achieve a robust and comprehensive prediction for circRNAs, we followed the strategy we employed for AtCircDB. First, we used the fastq-dump command with the ‘-R pass’ parameter to control the sequencing quality. Second, we used two state-of-the-art software programs, CIRI2 (45), and find_circ (13), to detect potential circRNAs as suggested by Vromman et al. (46). These two tools will achieve high accuracy, and complementarily increase the sensitivity. Third, we improved the metric (the detection score) (40) to measure the robustness of circRNAs to rank the back-splicing sites. This updated metric includes the robustness of circRNAs detected by different algorithms and the consistency of detection in different sequencing samples:

where n_i is the number of samples with detected circRNAs by CIRI2 or find_circ, and n is the total number of sequencing samples. At least two independent samples were treated as the detected samples. We normalized this value—min represents the minimum of the raw score, and max represents the maximum of the raw score—to help compare this score across species. The final metric measures the consistency of detecting this circRNA in various samples. A high value indicates a high possibility to detect this circRNA. Thus, circRNAs with the highest occurrence have a detection score of 100, while circRNAs with lowest occurrence have a detection score of 0.

Tissue and stress annotation

To annotate the tissue and stress information, we searched the NCBI SRA database with the following keywords: species, tissue (leaf, stem, root, flower, fruit, seed) and circRNAs. Then, we downloaded all of the related sample information. After comparison with the sample ID downloaded for the analysis, we annotated each sample with tissue information. Simultaneously, we searched the NCBI SRA database with the following keywords: species name and abiotic stress (drought, heat, cold and salt). Then, we downloaded the related sample IDs, compared them with the sample IDs of the sequencing samples for analysis, and annotated the database with the stress information.

Annotation of circRNAs and functional elements

To understand the relationship between circRNAs and genes, exons, and introns, we annotated circRNAs by following the previous pipeline for AtCircDB (40) and CropCircDB (41). First, we sorted the alignment files with SAMTools v1.9 (47), and indexed them for later analysis. We utilized Bioconductor packages (GenomicRanges v1.32.6 (48), GenomicAlignments v1.16.0 (48), and Biotrings v2.48.0 (https://bioconductor.org/packages/Biostrings/)) to analyze the sorted and indexed alignment bam files. We used SplicingTypesAnno v1.0.2 (49), developed by our team, to take annotation files (GTF or GFF) as the input and compared the detailed circRNA information, including the start position; end position; strand information; and gene, exon and intron information. We classified circRNA locations into three possible categories based on the reference genome—exon, intron and intergenic region—and we recorded antisense information based on the host genes. We used R packages (ggplot2 v3.3.5 (50) and lattice v0.20-38 (51)) for data visualization.

To compare our results with extrachromosomal circular DNAs (eccDNAs) as well as collected publications, we download the eccDNA data for Arabidopsis from http://deepbiology.cn/circDNA/ (52). We used R for the comparison. To reconcile the difference of bed and gtf format utilized in different publications, we allowed at most a 2-nucleotide difference at the start or end position.

Integration of the AtCircDB and CropCircDB data

The original AtCircDB hosts 84 685 circRNAs of Arabidopsis from 622 RNA-seq samples in 87 independent studies. The original CropCircDB covers 38 785 circRNAs in maize from 244 RNA-seq samples, and 63 048 circRNAs in rice from 288 RNA-Seq samples. In this database, we incorporated all of the above mentioned circRNAs. To annotate the previous circRNAs, we inserted ‘AtCircDB’ or ‘CropCircDB’ in the source column.

Website development

We constructed the platform by utilizing R Shiny, a web application framework for R, enabling the creation of interactive web apps directly from R (https://shiny.posit.co/), and incorporated various additional technologies to enhance functionality and user experience. For interface elements such as online buttons, we employed shinyWidgets (https://dreamrs.github.io/shinyWidgets/), alongside shinyjs (https://deanattali.com/shinyjs/) and Bootstrap's CSS styles (https://getbootstrap.com/) to manage the appearance of the website. In terms of genomic data visualization, we integrated multiple specialized packages: For genome browsing capabilities, we implemented JBrowse (53) (http://jbrowse.org/) to make track file visualization accessible. For aesthetically pleasing statistical graphics in an interactive fashion, we utilized ggplot2 (50) (http://ggplot2.tidyverse.org/), plotly (https://plotly.com/r/) (54) and echarts4r (https://echarts4r.john-coene.com/respectively). For other plotting functionalities, we used ggbio (https://bioconductor.org/packages/ggbio/) (55) and karyoploteR (https://bernatgel.github.io/karyoploter_tutorial/) (56). We employed MySQL (https:/mysql.com/) to store the related data. We developed all codes by using R Studio (https://posit.co/products/open-source/rstudio/). Finally, the platform is hosted on a CentOS Linux system with Shiny Server (https://posit.co/products/open-source/shinyserver/) to provide stable web access.

Results and discussion

Database overview

PlantCircRNA is an extension and combination of AtCircDB (40) and CropCircDB (41). It includes all the circRNAs in AtCircDB and CropCircDB. In addition, we systematically analyzed total RNA-seq datasets available at the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA), covering 94 plant species (Figure 1A, Table 1). The raw data come from 39 245 RNA-seq samples belonging to 1928 independent studies that were deposited into the NCBI SRA from 2012 to 2024 (Figure 1B, C). In total, we identified 673 443 circRNAs (Figure 1D). To understand the stress and tissue effects, we also manually curated 7107 tissue-related sequencing samples (2561 root, 300 fruit, 834 stem, 372 flower, 2246 leaf and 794 seed samples) and 1376 stress-related sequencing samples (362 drought, 238 salt, 325 cold and 451 heat samples) (Figure 1E). Furthermore, we analyzed 81 original publications and their related circRNAs to annotate the findings (Figure 1C). With these efforts, we have produced a comprehensive circRNA resource.

Figure 1.

Data collection and process pipeline for PlantCircRNA. (A) RNA-seq Sample distribution of 94 plant species. (B) Scatterplot of total number of RNA-seq samples versus released date. (C) Data collection versus release date. (D) circRNA distribution of 94 plant species. (E) Venn diagram of abiotic stress and tissue. (F) Process workflow of plant circRNAs.

Table 1.

A representative list for 94 plant species. Note: full list is available at the website

Class	Species	Annotation Version	Annotation source	Taxon ID	Source
monocot	Ananas comosus	F153	Ensembl	4615	plants.ensembl.org/Ananas_comosus
	Asparagus officinalis	Aspof.V1	Ensembl	4686	plants.ensembl.org/Asparagus_officinalis
	Echinochloa crus-galli	ec_v3	Ensembl	90 397	plants.ensembl.org/Echinochloa_crusgalli
	Eleusine coracana	v1.1	JGI	4511	phytozome-next.jgi.doe.gov/info/Ecoracana_v1_1
	Oryza sativa	IRGSP-1.0	Ensembl	39 947	plants.ensembl.org/Oryza_sativa
	Spirodela polyrhiza	v2	MIPS	29 656	phytozome-next.jgi.doe.gov/info/Spolyrhiza_v2
	Triticum aestivum	IWGSC	Ensembl	4565	plants.ensembl.org/Triticum_aestivum
	Zea mays	Zm-B73-REFERENCE-NAM-5.0	Ensembl	4577	plants.ensembl.org/Zea_mays
dicot	Arabidopsis thaliana	TAIR10	Ensembl	3702	plants.ensembl.org/Arabidopsis_thaliana
	Arabis alpina	A_alpina_V4	Ensembl	50 452	plants.ensembl.org/Arabis_alpina
	Brassica juncea	ASM1870372v1	Ensembl	3707	plants.ensembl.org/Brassica_juncea
	Brassica napus	AST_PRJEB5043_v1	Ensembl	3708	plants.ensembl.org/Brassica_napus
	Brassica rapa	Brapa_1.0	Ensembl	3711	plants.ensembl.org/Brassica_rapa
	Capsella rubella	v1.1	JGI	81 985	phytozome-next.jgi.doe.gov/info/Crubella_v1_1
	Capsicum annuum	ASM51225v2	Ensembl	4072	plants.ensembl.org/Capsicum_annuum
	Citrullus lanatus	Cla97_v1	Ensembl	3654	plants.ensembl.org/Citrullus_lanatus
	Corymbia citriodora	v2.1	JGI	34 329	phytozome-next.jgi.doe.gov/info/Ccitriodora_v2_1
	Cucumis melo	Melonv4	Ensembl	3656	plants.ensembl.org/Cucumis_melo
	Cucumis sativus	ASM407v2	Ensembl	3659	plants.ensembl.org/Cucumis_sativus
	Cynara cardunculus	CcrdV1	Ensembl	59 895	plants.ensembl.org/Cynara_cardunculus
	Daucus carota	ASM162521v1	Ensembl	79 200	plants.ensembl.org/Daucus_carota
	Eucalyptus grandis	Egrandis1_0	Ensembl	71 139	plants.ensembl.org/Eucalyptus_grandis
	Eutrema salsugineum	Eutsalg1_0	Ensembl	72 664	plants.ensembl.org/Eutrema_salsugineum
	Ficus carica	UNIPI_FiCari_1.0	Ensembl	3494	plants.ensembl.org/Ficus_carica
	Glycine max	Glycine_max_v2.1	Ensembl	3847	plants.ensembl.org/Glycine_max
	Gossypium raimondii	Graimondii2_0_v6	Ensembl	29 730	plants.ensembl.org/Gossypium_raimondii
	Helianthus annuus	HanXRQr2.0-SUNRISE	Ensembl	4232	plants.ensembl.org/Helianthus_annuus
	Iberis amara	v1.1	JGI	190 884	phytozome-next.jgi.doe.gov/info/Iamara_v1_1
	Ipomoea triloba	ASM357664v1	Ensembl	35 885	plants.ensembl.org/Ipomoea_triloba
	Juglans regia	Walnut_2.0	Ensembl	51 240	plants.ensembl.org/Juglans_regia
	Lactuca sativa	Lsat_Salinas_v7	Ensembl	4236	plants.ensembl.org/Lactuca_sativa
	Lupinus albus	v1	JGI	3870	phytozome-next.jgi.doe.gov/info/Lalbus_v1
	Lupinus angustifolius	LupAngTanjil_v1.0	Ensembl	3871	plants.ensembl.org/Lupinus_angustifolius
	Papaver somniferum	ASM357369v1	Ensembl	3469	plants.ensembl.org/Papaver_somniferum
	Phaseolus coccineus	v1.1	HA	3886	phytozome-next.jgi.doe.gov/info/Pcoccineus_v1_1
	Pistacia vera	PisVer_v2	Ensembl	55 513	plants.ensembl.org/Pistacia_vera
	Pisum sativum	Pisum_sativum_v1a	Ensembl	3888	plants.ensembl.org/Pisum_sativum
	Populus trichocarpa	Pop_tri_v4	Ensembl	3694	plants.ensembl.org/Populus_trichocarpa
	Prunus avium	PAV_r1.0	Ensembl	42 229	plants.ensembl.org/Prunus_avium
	Quercus lobata	ValleyOak3.0	Ensembl	97 700	plants.ensembl.org/Quercus_lobata
	Quercus rubra	v2.1	HA	3512	phytozome-next.jgi.doe.gov/info/Qrubra_v2_1
	Quercus suber	CorkOak1.0	Ensembl	58 331	plants.ensembl.org/Quercus_suber
	Salix purpurea	v5.1	JGI	77 065	phytozome-next.jgi.doe.gov/info/Spurpurea_v5_1
	Schrenkiella parvula	v2.2	LSU	98 039	phytozome-next.jgi.doe.gov/info/Sparvula_v2_2
	Solanum lycopersicum	SL3.0	Ensembl	4081	plants.ensembl.org/Solanum_lycopersicum
	Solanum tuberosum	SolTub_3.0	Ensembl	4113	plants.ensembl.org/Solanum_tuberosum
	Spinacia oleracea	Spov3	UCD	3562	phytozome-next.jgi.doe.gov/info/Soleracea_Spov3
	Vigna radiata	Vradiata_ver6	Ensembl	3916	plants.ensembl.org/Vigna_radiata
	Vitis vinifera	PN40024.v4	Ensembl	29 760	plants.ensembl.org/Vitis_vinifera
others	Ceratopteris richardii	v2.1	JGI	49 495	phytozome-next.jgi.doe.gov/info/Crichardii_v2_1
	Dunaliella salina	v1.0	JGI	3046	phytozome-next.jgi.doe.gov/info/Dsalina_v1_0
	Galdieria sulphuraria	ASM34128v1	Ensembl	130 081	plants.ensembl.org/Galdieria_sulphuraria
	Marchantia polymorpha	Marchanta_polymorpha_v1	Ensembl	3197	plants.ensembl.org/Marchantia_polymorpha
	Physcomitrium patens	Phypa_V3	Ensembl	3218	plants.ensembl.org/Physcomitrium_patens
	Volvox carteri	v2.1	JGI	3067	phytozome-next.jgi.doe.gov/info/Vcarteri_v2_1

We followed the protocol for AtCircDB (40) and CropCircDB (41) to set up the analysis pipeline and to systematically extract all potential circRNAs in plants (Figure 1F). To make the results easily accessible to the research community, we designed an updated web platform based on AtCircDB and CropCircDB. We collected additional data, including circRNAs, host genes, stress and tissue, publications, and sequencing samples, and added them to the database. Furthermore, we provided Search, Browser, Tool and Statistics functions for data visualization and meta-analysis. We designed the Search and Browser functions for individual-level circRNAs, and the Tool and Statistics functions for group- or species-level circRNAs. There is also a Download function so that researchers can use these resources under a CC BY-NC license.

Characteristics of plant circRNAs

In this study, we investigated 94 plant species, including 71 dicots, 13 monocots and 10 other species (e.g. bryophytes and thallophytes). After processing all of the samples, we identified 673 443 circRNAs for these species, which we deposited in the database. The genome size of the 94 plant species ranges from 13.7 Mb (Galdieria sulphuraria) to 14 547 Mb (Triticum aestivum) (Figure 2A). Interestingly, the genome size of wheat is twofold larger than that of any other plant species included in the database. Arabidopsis has a relatively small genome (119.7 Mb) but the highest circRNA annotation percentage (57%) (Figure 2B), followed by Oryza sativa (375 Mb, 31% annotation), Gossypium raimondii (761 Mb, 18% annotation), Zea mays (2182 Mb, 8% annotation), Glycine max (978 Mb, 23% annotation), Brassica napus (848 Mb, 36% annotation), and Triticum aestivum (14 547 Mb, 2.6% annotation).

Figure 2.

Characteristics of plant circular RNAs. (A) Annotated genome size versus reference genome. (B) Percent of annotated genome size versus reference genome without outlier (Triticum aestivum). (C) Boxplot of circRNA length of three types of plant species. (D) circRNA length of 8 plant species. (E) Number of circRNAs versus number of exons. (F) Annotation of start and end position of circRNAs for 8 plant species. (G) Number of genes versus number of hosted circRNAs. (H) Percent of circRNAs as antisense for 8 species. (I) Number of circRNAs detected in different number of samples for 17 species. (J) Heatmap of abiotic-specific and tissue specific circRNAs for 8 species.

The circRNA lengths vary across species. Interestingly, more than 75% of the circRNAs are smaller than 1000 base pairs (bp) for both dicots and monocots (Figure 2C). Specifically, 76% and 77% of Arabidopsis and rice circRNAs, respectively, are smaller than 500 bp (Figure 2D). We also evaluated the relationship between circRNAs and exons. Surprisingly, most circRNAs come from one exon (Figure 2E). As the number of related exons increases, the number of circRNAs decreases dramatically. These results are in line with our previous finding: ‘circular RNAs are generally generated from fewer exons’ (40).

We also evaluated the differences between linear transcripts (mRNAs) and circRNAs by comparing their start and end positions. Arabidopsis and rice have a similar pattern of start and end positions regarding their locations in exons, introns, and intergenic regions. However, this pattern changes for other species, such as G. max and Brassica rapa (Figure 2F). We noted a higher percentage of start and end positions located in the intergenic region, which may be the result of imprecise annotation.

We next assessed the relationship between host genes and circRNAs following our previous work (57). Overall, we found that most genes host only one circRNA (Figure 2G). For example, 65% of all genes in Arabidopsis are related to circRNAs, of which 18% host only one circRNA, 4% host five circRNAs, and only 1.4% host 10 circRNAs. In rice, 52% of all genes are related to circRNAs, with 16% hosting one circRNA, 3.1% hosting five circRNAs, and 0.96% hosting 10 circRNAs. We also searched for circRNAs related to antisense transcripts given that the first reported human circRNA is antisense to the CDR1 gene (13,14). In plants, we also identified a large number of circRNAs that are antisense transcripts. In particular, 12% and 28% of Arabidopsis and rice circRNAs, respectively, are antisense (Figure 2H). In addition, we deposited circRNAs detected in at least two samples in the database. We detected most circRNAs with 2–5 independent samples (Figure 2I). Of note, we detected many Arabidopsis and rice circRNAs in more than six samples because much more sequencing data are available for these species.

Key features of PlantCircRNA

We incorporated two key factors into PlantCircRNA: tissues (roots, stems, leaves, flowers, and seeds) and abiotic stresses (drought, heat, cold, and salt), which have important effects on the biological function of circRNAs. For example, in Arabidopsis, we detected 110 drought-specific circRNAs, 250 heat-specific circRNAs, 324 cold-specific circRNAs, and 30 salt-specific circRNAs (Figure 2J). Simultaneously, we found 996 root-specific circRNAs, 116 stem-specific circRNAs, 433 leaf-specific circRNAs, 1620 flower-specific circRNAs and 1379 seed-specific circRNAs (Figure 2J). We manually curated 81 studies (Supplementary Data S1). Among them, 47 studies provided data, and 27 studies had data that could be used for further annotation (58–84). These studies comprised 30 species, suggesting that a large number of plant species lack circRNA information. We will keep track of new research and deposit the relevant new data into PlantCircRNA.

To help users to understand the existence of circRNAs in various samples, we updated our metric to provide a detection score to measure consistency. This updated metric consists of two components, namely the number of sequencing samples and the detection robustness of the algorithm. We normalized this score, with a minimum of 0 and a maximum of 100, to facilitate comparison across species. For example, in Arabidopsis, we found one circRNA (1:30059886–30064646_-) with a detection score of 100, which is present in 11461 samples. In addition, two publications (58,59) also reported this circRNA, which provides external support. In rice, we found one circRNA (4:19391425–19401494_-) with a detection score of 100. This circRNA is included in CropCircDB and has also been detected in the root, stem, leaf, and seed, and in samples under drought, heat, and salt stress.

In the previous databases, we designed a circRNA system that follows the naming system from humans and animals (40). This system combines the species abbreviation, chromosome, start, end, strand, and gene names together. Recently, Chen et al. (42) proposed a novel naming system for eukaryotes. This system includes exon and intron information to clearly differentiate the origin of the transcript and the gene. In this work, we followed this new guideline, and updated our system using ‘circName’ with a similar architecture. In addition, to maintain consistency between different versions, we have kept the ‘circID’ (chromosome, start, end, and strand information) as our own unique key to trace across different studies.

To understand the relationship between circRNAs and extrachromosomal circular DNAs (eccDNAs), we compared the circRNAs of Arabidopsis in the databases to the eccDNAs detected in our previous study (52). We found that the locations of 47 circRNAs share the same genome locus as eccDNAs (Supplementary Data S2). This finding further supports the hypothesis proposed by Iparraguirre et al. (85): ‘Transcription events across the junction of circular DNAs would result in a transcript with a junction similar to those present in circRNAs’. Because the detection algorithm depends on the back-splicing junction to identify circRNAs, these 47 molecules are treated as a circular structure. In reality, they may come directly from eccDNAs and function as linear transcripts.

Database description

The web interface of PlantCircRNA hosts the following main functions: Browser, Search, Download, Tools, Statistics, and Help. The detailed user guide for each main function is available for users on the Help page. Contact information is provided for users to provide suggestions to improve the website and to incorporate more circRNA information. The website can be accessed at https://plant.deepbiology.cn/PlantCircRNA/ (Figure 3A).

Figure 3.

Web interface for PlantCircRNA. (A) Main functions of the database. (B) Species visualization provided in Browser page, including scatterplot for genome size, circRNA distribution in chromosomes, and phylogenetic trees. (C) Annotation of each circRNA in Search page with genomic information as well as structure view. (D) All resources available at Download page under CC BY-NC license.

In particular, the Browser and Search pages help users to query the database and to review the detailed circRNA information. For example, the Browser page provides the following information: species name, sample number, circRNA number, genome size, annotated genome size (Figure 3B). For each circRNA, the following details are provided: chromosome, host gene ID, genomic coordinates of the back-splicing sites, strand information, start/end annotation, length, antisense, detection algorithm, publications, source, total sample number, and junction reads (Figure 3C). This service gives users a clear picture of the targeted circRNAs. To take full advantage of the detection score, we also inserted a range search for users to review related circRNAs within certain ranges.

The Download page allows users to download our data and use it under a CC BY-NC license (for non-commercial usage) (Figure 3D). The species name, sequencing sample number, file size, and the updated data are annotated for each file. A user can either download the circRNAs for all species or each species depending on the research purpose.

In addition, we designed the Tool and Statistics pages to visualize circRNAs at the group or species level. The Tool page includes two features. (i) JBrowse (53) is adapted to visualize circRNAs in the genome browser. A user can easily visualize circRNAs, exons, transcripts and genes. (ii) The Chromosome map tool is available for users to visualize a group of circRNAs in the chromosome view. Users can add circID or circName as well as gene ID to the chromosome to visualize the full picture of these targeted molecules. The Statistics page provides the analysis of all circRNAs at the species level.

The entire platform works as follows. The Browser and Search functions fetch the query from user, search the MySQL database, and return the circRNA information at the individual level. The Download, Statistics, and Tool functions are set up for users to access, visualize, and analyze the circRNAs at the group or species level. In addition, to take full advantage of this database, we will regularly update the new circRNAs from sequencing samples and up-to-date publications.

Application cases

Below are two examples that demonstrate the usage of PlantCircRNA.

Example 1 (Supplementary Figure S1): AT1G21580 is a gene that encodes a zinc-finger protein. Using the Search function, we input AT1G21580 in the gene field and found 17 circRNAs. Interestingly, we found one circRNA (1:7558475–7560288_-) in flowers and leaves, but under drought and heat stresses. In addition, we found that this circRNA exists in 220 RNA-seq samples. Notably, it had also been reported in one independent publication.

Example 2 (Supplementary Figure S2): To understand the rice circRNAs under salt stress, we used the ‘Advanced Search’ option. First, we chose ‘Oryza sativa’ in the ‘Select Species’ option, and then in the ‘Advanced Search,’ we selected ‘salt’ under the ‘stress_specific’ option. Hence, we searched for specific circRNAs with certain criteria. We found 151 candidates, half of which are related to roots. We selected circID 11:3489631–3489681_- and found that this circRNA had been detected in 18 independent samples. It represents a good target for further analysis.

Summary and future directions

PlantCircRNA is a comprehensive platform for 94 plant species, including 71 dicots, 13 monocots, and 10 other species. This new resource not only provides an integrated resource for circRNAs extracted from 39 245 RNA-seq samples, but also imports the original knowledge from AtCircDB and CropCircDB. It also provides circRNAs from several tissues and under different stresses. In addition, each circRNA is annotated with a detection score and experimental validation (i.e. findings from independent studies) to assess its consistency. We developed a comprehensive platform for the research community to visualize, analyze, and download data related to targeted circRNAs. In the next step, we will continue to collect new circRNAs from diverse plant species. This resource should provide the community with unprecedented insights regarding the relatively mysterious circRNAs in plants.

Data availability

PlantCircRNA is a comprehensive database for plant circular RNAs (https://plant.deepbiology.cn/PlantCircRNA/). More details are available under Download and Help menu at the website.

Supplementary data

Supplementary Data are available at NAR Online.

Funding

National Natural Science Foundation of China [32070684, 31571306]; Project of Shandong Province Higher Educational Program for Introduction and Cultivation of Young Innovative Talents in 2021. Funding for open access charge: National Natural Science Foundation of China [32070684, 31571306]; Project of Shandong Province Higher Educational Program for Introduction and Cultivation of Young Innovative Talents in 2021.

Conflict of interest statement. None declared.