Plant materials
Plant materials for Yaghooti grapes from Sistan were obtained from cloned grapevines (V. vinifera) in gardens of the Agricultural Research Institute of Zabol University. Three stages of cluster formation were sprayed with gibberellin (Merck, Germany) at a concentration of 40 mg/L 24 hours before sampling. Gibberellin concentrations were determined according to previous studies (Casanova, Casanova et al. 2009; Amkha, Saengkai et al. 2017). Control samples were sprayed with distilled water simultaneously. The first (April 2, 2016), second (April 16, 2016), and third (April 29, 2016) sampling coincided with the formation of the first clusters, the time of berry formation, and the final size of the clusters, respectively. Control samples were also taken at the same intervals.
RNA extraction, library construction and sequencing
RNA was extracted from six samples including three untreated (control group) and three treated with gibberellin (treatment group), based on the Japelaghi technique (Japelaghi, Haddad et al. 2011). The quality and quantity of RNA were assessed using Bioanalyzer 2100, and Agilent technology (2200 TapeStation). Ultimately, the samples with high RNA quality and quantity were used for library construction. Total RNA from six samples was sequenced by Illumina HiSeq 2500 method (Macrogen, Seoul, South Korea) using a commercial kit (TruSeq Stranded Total RNA sample preparation kits with Ribo-Zero Plant) along 101 bp paired-end reads.
Identification of circRNAs in total RNA-Seq data
In this study, the CirComPara integrated pipeline (Gaffo, Bonizzato et al. 2017) with eight detection tools was used to identify and quantify circRNAs from grape rRNA-depleted RNA-Seq (Ribominus-Seq). Quality control of raw reads and read statistics were performed using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Adapter sequences and low-quality raw reads were removed in the pipeline using Trimmomatic (Bolger, Lohse et al. 2014). The circRNA identification tools based on default parameters were:
Find-circ
Find-circ (Memczak, Jens et al. 2013) is the first circRNA detection tool that uses the Bowtie2 (Langmead and Salzberg 2012) read mapper. All reads are first mapped to the reference genome. The reads are continuously aligned along their entire length thus BSJ is undetectable. To find anchor regions for the remaining reads, a further step is to independently align short flanking sequences from the 5' and 3' ends of the reads. To find the exact position of the breakpoint, the read fragments are expanded if the flanking sequences can be mapped in reverse orientation (Jakobi and Dieterich 2019). Finally, a list of circRNA candidates is generated.
CircRNA finder
CircRNA_finder (Westholm, Miura et al. 2014) is a de novo circRNA prediction method, i.e. no gene annotations or exon-intron structures are required. circRNA_finder is written in Perl and awk scripts and uses a STAR(Dobin, Davis et al. 2013) aligner. This tool can accept paired-end and single-end data. After mapping, the data are post-processed by the circRNA_finder, which creates multiple tables with the final coordinates for each circRNA candidate.
DCC
DCC (Cheng, Metge et al. 2016) is a Python-based program that works with Chimeric.out.junction files generated by STAR that contain discovered chimeric reads, thus eliminating the need to parse the alignment files from SAM. The software can increase its sensitivity by using paired-end datasets. It also has several filtering steps to reduce the number of circRNA candidates falsely identified as positive. In addition, DCC also notifies users of potential new cases of alternative linear splicing.
CIRI
A robust tool for extracting circRNAs from RNA-seq reads is CIRI (Gao, Wang et al. 2015). Unlike other algorithms based on annotation or circRNA enrichment, this approach uses a novel algorithm based on signal detection by paired chiastic clipping (PCC) in the BWA-MEM(Li 2013) sequence alignment/map (SAM), together with systematic filtering to eliminate false positives.
CircExplorer2
CircExplorer2 (Zhang, Wang et al. 2014) is a multi-mapper tool that uses STAR, segemehl (Hoffmann, Otto et al. 2009) and TopHat (Trapnell, Pachter et al. 2009) Aligner. When specificity is critical, this multi-mapper approach can be useful, and the results of multiple variants of the same CircExplorer2 analysis can be used to identify a core set of circRNAs that are detected by all methods. When using the TopHat-based approach, the software was also improved in its ability to harness information from reads that were not originally mapped. After a de novo assembly step, such reads are now used to locate potential new exons and splicing events (Jakobi and Dieterich 2019).
Testrealign
Testrealign (Hoffmann, Otto et al. 2014) is a comprehensive, un-biased algorithm for analyzing single-end read data for gene splicing, trans-splicing, and fusion events. The approach is integrated with the mapping tool segemehl and is based on algorithms for concatenation, dynamic programming, and suffix arrays. This tool uses segemehl aligner
The V. vinifera genome and annotation files (GTF) were downloaded from the Ensemble Plant database (Bolser, Staines et al. 2016). Only circRNAs expressed with at least two back-splice reads and identified together by at least two methods were considered reliable circRNAs for downstream analyses (Gaffo, Buratin et al. 2022). The expression levels of the identified circRNAs were normalized using reads per million mapped (RPM).
Bioinformatic approach and functional predictions of circRNAs
For each developmental stage of the clusters, 150 cricRNAs from each up-regulated and down-regulated level were selected separately under gibberellin treatment. The key miRNAs, that showed interactions with DE circRNAs, were finally identified using miRBase version 21.0 (Kozomara and Griffiths-Jones 2014) and psRNATarget (Dai and Zhao 2011). Only miRNAs with an expectation coefficient of ≤ 5 in psRNATarget were selected for subsequent analysis. For circRNAs from intergenic regions, some key miRNAs were also analyzed separately at each stage. Target genes for miRNAs were identified using miRNEST (Szcześniak and Makałowska 2014), TarDB (Liu, Liu et al. 2021), PsRobot (Wu, Ma et al. 2012), and articles (Ding, Zhou et al. 2012; Meng, Shi et al. 2014). The co-expression network of target genes for miRNAs was reconstructed using Cytoscape 3.7.2. software (Shannon, Markiel et al. 2003). The topology of the network was also identified using the Network Analyzer 2.7 plugin (Assenov, Ramírez et al. 2008). The GO was analyzed using the BiNGO 3.0.3 plugin (Maere, Heymans et al. 2005) in Cytoscape software. The interaction of genes in the co-expression network was calculated based on co-expression, co-occurrence, and text mining parameters using the STRING database (Szklarczyk, Gable et al. 2021).