Soaked the seeds in deionized water for 3–4h, then placed the seeds at room temperature to fully dry them for 3–4h, soaked the seeds in 1% NaClO solution for 10min in the Vertical Flow Clean Bench, and then washed them with sterile water three times. About 20–30 sterilized seeds were inoculated per petri dish in MS solid medium (Murashige and Skoog, 1962) for germination and then maintained in a culture room under white fluorescent lamps (1,900 lux), with a 16h light/8h dark cycle at 25°C for 4–7days. Individual germinated seedlings were transferred to a sterile tissue culture bottle containing MS medium supplemented with 30g/L sucrose and kept in a culture room under white fluorescent lamps (1,900 lux), with a 16h light/8h dark cycle at 26°C for 4–7weeks.

Total genomic DNA was used to build sequence libraries (Illumina Inc., San Diego, CA, United States), and it was extracted from leaves using the cetyltrimethylammonium bromide (CTAB) method. The experimental procedures were performed in accordance with the standard protocol provided by Illumina, and sequence libraries were prepared with the NEB Next Ultra DNA Library Prep Kit. An Illumina NovaSeq sequencer was used to sequence paired-end (PE) sequencing libraries with an average 350bp insert length at the Wuhan Benagen Tech Solutions Company Limited. From these data, over 20 million clean reads with a 150bp read length were passed through quality control. Then, we performed nanopore sequencing to facilitate the connection of high-quality contigs into longer complete sequences in subsequent genome assembly.

Since the original sequencing data may contain low-quality sequences, adaptor sequences, etc. (Ewing and Green, 1998), to ensure the reliability of the information analysis results, the original sequencing data need to be filtered to obtain clean reads, which are stored in FASTQ format (Cock et al., 2009). This project used the filtering software SOAPnuke (version: 1.3.0) to remove reads with an N base content of more than 5%, low-quality (quality value less than or equal to 5) reads with a base number of 50%, and reads with adapter contamination.

UniCycler (0.4.8) software was used to assemble the filtered reads (Wick et al., 2017). First, high-accuracy Illumina data (Q30>85%) were assembled to obtain a high-quality cp genome skeleton (contig), and then nanopore data were used to connect the high-quality contig into a longer complete sequence. The accuracy of the cp genome assembled by this strategy is determined by the accuracy of the Illumina data, and it can effectively avoid the sequence contamination introduced by the error of nanopore data splitting. Finally, Pilon software was used to further correct the Illumina data for errors in the assembled genome, resulting in a genome with higher final accuracy.

Unicycler software was aligned with the close reference genome (NC_043882.1) by blastn (version: BLAST 2.2.30+; parameter: -evalue 1e-5), and the best candidate sequence was selected based on the alignment. The chloroplast genome connection relationship was determined based on the sequence sequencing depth and read alignment and by comparison with closely related species. The connection relationship and whether it forms a loop need to be further experimentally verified.

Chloroplast genome function annotation includes coding gene prediction and noncoding RNA annotation [ribosomal RNA (rRNA) and transfer RNA (tRNA) annotation]. CPGAVAS2¹ was used for gene annotation and drawn based on the results of the cp genome annotation (Shi et al., 2019).

CodonW (Version: 1.4.4) was used to analyze codon usage bias and statistically estimate the frequency of relative synonymous codon usage (RSCU). Codons with RSCU values >1 were defined as high-frequency codons. We conducted a codon usage bias analysis using the following gene sequences: “ATG”, “TTG”, “CTG”, “ATT”, “ATC”, “GTG”, and “ATA”, which were used as start codons; and “TGA”, “TAG”, and “TAA”, which were used as stop codons; in addition, protein-coding genes shorter than 300bp in length were excluded from the base composition analysis to avoid sampling bias (Wright, 1990).

MISA (Sebastian et al., 2017) was used to perform a SSR analysis of the E. breviscapus chloroplast [version: 1.0; default parameters: the corresponding minimum number of repetitions of each repeating unit (unit size) are: 1–8, 2–4, 3–4, 4–3, 5–3, and 6–3, with 1–8 meaning that when a single nucleotide is used as the repeat unit, the number of repeats must be at least 8 to be detected; see http://pgrc.ipk-gatersleben.de/misa/misa.html for details].

The software vmatch² was used to find scattered long repetitive sequences in the E. breviscapus cp genome (Jean-Simon et al., 2010). The long repeat sequences included three types: forward (F), palindrome (P), and tandem (T) repeats.

There is no report on the extraction method of chloroplast of E. breviscapus before, therefore, we have developed a method suitable for extracting chloroplast DNA of E. breviscapus, which is based on the modified high – salt low – pH method (Shi et al., 2012). Borax was not used in this method, which greatly guaranteed the safety of the operators. In addition, we removed the step of “add 1.5ml of 5mol/L KAc (PH 5.2) and continue freezing for 30min. Then centrifuge at 10,000g for 15min, and discard the precipitate”, this saved extraction time and effectively simplified the extraction step. After extracting high-quality chloroplast genomic DNA of E. breviscapus, regular PCR was performed as follows: a 30s initial denaturation step at 98°C, followed by 35cycles of denaturation for 10s at 98°C, annealing for 30s at 55–63°C, extension for 30s per kbp sequence length at 72°C, and a final 7min extension step at 72°C. Phusion High-Fidelity DNA Polymerase (New England Biolabs, China) and 2×EasyTaq PCR SuperMix (+dye; TransGen Biotech, China) were utilized for DNA cloning and PCR analysis, respectively. The sequence information of the primers for amplification was provided in Supplementary Table S1. The pEASY-Blunt Zero Cloning Kit (TransGen Biotech, China) was utilized for seamless cloning of the selected amplified fragments.

The synthetic operon constructs for chloroplast transformation (pBtEa1t, pBtEa2t, pBtEat3, and pBtEIat) were based on the empty vector pBluescript SK (+). They were all constructed using the pEASY-Uni Seamless Cloning and Assembly Kit (TransGene Biotech, China), which can be used to recombine the vector linearized by any method and the PCR fragment with 15–25bp overlapping region with both ends of the linearized vector. All the fragments needed for chloroplast expression vector construction were then assembled into synthetic operons as follows:

For the first round (Round I) of gene assembly, The empty vector pBluescript SK (+) was digested with Not I and Hind III, and then Eb-PpsbA, aadA gene and Eb-TpsbA were cloned into linearized vector and generating construct pBa1, Eb-Prrn, aadA gene, and Eb-TrbcL were cloned into linearized vector and generating construct pBa2, Eb-Prrn, aadA gene, and Eb-Trps16 were cloned into linearized vector and generating construct pBa3, all the recombinant vectors were verified by Sanger sequencing for the next round of gene assembly.

For the second round (Round II) of gene assembly, the vectors pBa1, pBa2, pBa3, and pBluescript SK (+) were digested with Sal I, and then Eb-PpsbA, EGFP gene, and Eb-TpsbA were cloned into linearized vector pBa1 and generating construct pBEa1, Eb-Prrn, EGFP gene, and Eb-TrbcL were cloned into linearized vector pBa2 and generating construct pBEa2, Eb-Prrn, EGFP gene, and Eb-Trps16 were cloned into linearized vector pBa3 and generating construct pBEa3, Eb-Prrn, EGFP gene, Eb-IEE, aadA gene, and Eb-Trps16 were cloned into linearized vector pBluescript SK (+) and generating construct pBEIa, all the recombinant vectors were verified by Sanger sequencing for the next round of gene assembly.

For the third round (Round III) of gene assembly, the vectors pBEa1, pBEa2, pBEa3, and pBEIa were digested with Kpn I, and then the homologous recombination fragment Eb-trnI was cloned into linearized vectors and generating construct pBtEa1, pBtEa2, pBtEa3, and pBtEIa, respectively and verified by Sanger sequencing. Finally, the vectors pBtEa1, pBtEa2, pBtEa3, and pBtEIa were digested with Sac I, and then the homologous recombination fragment Eb-trnA was cloned into linearized vectors and generating construct pBtEat1, pBtEat2, pBtEat3, and pBtEIat (Supplementary Figure S1) and verified by Sanger sequencing.

After PCR identification and Sanger sequencing, four verified vectors (pBtEat1, pBtEat2, pBtEat3, and pBtEIat) and the empty vector pBluescript SK (+) were transformed into the E. coli expression strain Transetta (DE3). Then, the transformed E. coli strains were grown in Luria-Bertani (LB) medium at 37°C unless otherwise indicated. Spectinomycin and ampicillin were used for selection at appropriate concentration.

After culturing in 20ml of LB medium for 14–16h at 16°C (150–180rpm), EGFP gene expression in transformed E. coli strains was examined using a Thermo Scientific Varioskan LUX Multimode Microplate Reader (Thermo Fisher Scientific, MA, United States). The loading volume was 200μl, and the “Fluorescence” protocol was executed in Skanlt Software 6.1 RE, which included an excitation wavelength at 485nm and an emission wavelength at 520nm. Subsequently, the fluorescence intensity of E. coli strains containing different vectors was counted. Each sample was repeated for three times to ensure the credibility of the data, in order to compare the effects of different combinations of chloroplast regulatory elements on the expression level of green fluorescent protein in E. coli, the expression level of fluorescent protein in each sample was compared by unpaired t-test (p<0.05), the results are represented as a bar graph.

The statistical evaluation was performed using SPSS 19.0 software (IBM Corp., United States). All data were presented as the means±SE of at least three replicates. ANOVA was followed by a t-test. Mean values were considered significantly different at p<0.05.

This project adopted the whole genome shotgun strategy to construct a 350bp library from the DNA samples of E. breviscapus and then used the Illumina NovaSeq sequencer to perform PE sequencing on this library, and the reads were 150bp in length. In this sequencing, the raw data sequencing volume of the E. breviscapus sample was 4.2Gbp. After data filtering, the clean data volume of the sample was 4.1Gbp (Table 1). The original sequencing statistics of the third-generation nanopore sequencing were shown in Supplementary Table S2.

The E. breviscapus cp genome was completely assembled into a single molecule of 152,164bp, the assembly result of the cp genome of E. breviscapus contained 0 Gap, the GC content was 37.20, the assembly coverage was 332.55X, and the sequencing quality value (Phred) was above 30 (Figure 1), indicating that the sequencing quality was high.

The conserved size and quadripartite structure of the complete E. breviscapus cp genome was similar to the cp genomes of previously published data. These genomes were usually 152kb in size and included two inverted repeat regions (IRs) and two single-copy regions, namely, a large single copy and small single copy (LSC and SSC, respectively) region.

The cp genome of E. breviscapus consisted of two single-copy regions isolated by two identical IR regions of 24,699bp each, one LSC region of 84,657bp and one SSC region of 18,109bp. The proportions of LSC, SSC, and IR sizes in the entire cp genome were 55.6, 11.9, and 32.5%, respectively (Figure 2; Table 2). The GC contents of the LSC, SSC, and IR and whole cp genome were 35.06, 31.00, 43.13, and 37.20%, respectively, which were consistent with the published E. breviscapus cp genomes (Supplementary Table S3). The cp genome of E. breviscapus was compared to previously published data, no structural differences were found among all compared data, it demonstrated that gene order, gene content, and entire genome structure were conserved in E. breviscapus cp genome.

The cp genome of E. breviscapus consisted of 127 coding regions made up of 83 protein-coding genes, 36 tRNAs, and eight rRNAs, of which 95 genes were unique and 16 genes were repeated in two inverted regions consisting of five protein coding genes, seven tRNAs, and four rRNAs (Figure 2; Table 3). Among these 95 unique genes, one gene crossed different cp genome boundaries: rps12 crossed two IR regions, the LSC region and the SSC region (two 3′ end exons repeated in IRs and 5′ end exon situated in LSC; Figure 2). Of the remaining 94 genes, 82 were situated in the LSC, including 61 protein-coding genes and 21 tRNAs, and 12 were situated in the SSC, including 11 protein-coding genes and one tRNA.

The E. breviscapus cp genome was composed of protein coding genes, tRNAs, rRNAs, and intronic and intergenic regions. Noncoding DNA accounted for 68,539bp (45.0%) of the whole E. breviscapus cp genome, protein-coding genes accounted for 71,850bp (47.2%), tRNA accounted for 2,725bp (1.8%), and rRNA accounted for 9,050bp (6.0%).

Most of the protein-coding genes contained only one exon, while 13 genes contained at least one intron, of which eight genes were distributed in the LSC, one was distributed in the SSC and four were distributed in both IRs (Supplementary Table S4). Among them, three genes (rps12, clpP, and ycf3) contained two introns, while 10 genes (trnA-UGC, trnE-UUC, trnL-UAA, rpl2, ndhA, ndhB, petB, atpF, rpoC1, and trnK-UUU) contained one intron. The longest intron of trnK-UUU was 2,543bp, and it included the 1,515bp encoding the matK gene (Gu et al., 2016). The rps12 gene was predicted to be trans-spliced with a repeated 3′ end duplicated in the two IRs and a single 5′ end exon in the LSC (Redwan et al., 2015).

Relative synonymous codon usage is an assessment of synonymous codon usage bias. This value is equal to the ratio of the actual observed value of synonymous codons to the expected value of the average use of synonymous codons. If there is no bias for codon usage, then the RSCU value will be 1; if the codon is used more frequently than other synonymous codons, then its RSCU value will be greater than 1; otherwise, the RSCU value will be less than 1 (Nie et al., 2012; Liu et al., 2020).

The results of the RSCU value analysis (Supplementary Table S5) of various amino acids showed that there were 31 codons with RSCU≥1, of which 12 codons, such as UUA, AGA, CAA, and GGA, end in A; 16 codons, such as UCU, GCU, ACU, and UAU, end in U; and the only ones ending with G are UUG, UGG, and AUG. These findings indicated that the codons ending with A and U in the chloroplast genome of E. breviscapus appeared more frequently and are preferred codons, while the codons ending with C and G were non-preferred codons.

Simple sequence repeats are sequences with motifs from 1 to 6bp in length repeated multiple times (see section “Materials and Methods” for cutoff criteria), and they are distributed throughout the cp genome and often used as markers for breeding studies, population genetics, and genetic linkage mapping (Grassi et al., 2002; Timme et al., 2007).

A total of 223 SSRs were found in the E. breviscapus cp genome (Figure 3; Supplementary Table S6). These SSRs included 142 complex repeat-type SSRs (63%), 50 dinucleotide SSRs (22%), 12 trinucleotide SSRs (5%), 14 tetranucleotide (6%), four pentanucleotide SSRs (1%), and one hexanucleotide SSR (0.003%; Figure 3A; Supplementary Table S6). Among the 223 SSRs, 87% of SSRs (195) were the AT type, with copy numbers from 8 to 23 (Supplementary Table S6). Forty-eight SSRs were detected in protein-coding genes, 19 SSRs were detected in introns, and 156 were detected in intergenic regions (Figure 3C). In relation to the quadripartite, 151 SSRs were situated in the LSC, while 32 and 40 were identified in the IR and SSC, respectively (Figure 3B).

Long repetitive sequences may promote the rearrangement of the cp genome and increase population genetic diversity (Cavalier-Smith, 2002). A total of three unique repeats were found in the E. breviscapus cp genome, the size of two palindromic repeats were 48bp and 24,699bp, and the forward repeat was 39bp in size. The palindromic repeat which was 48bp in size had two same starting sites in the chloroplast genome sequence, these two starting sites were located at 74,576bp, another palindromic repeat’s two starting site were located at 84,657 and 127,465bp respectively, the first starting site of the forward repeat located at 43,603bp and the second starting site located at 119,435bp. In addition, forward repeats are usually caused by transposon activity (Gemayel et al., 2012), which will increase under cell stress (Voronova et al., 2014). However, the origin and production mechanism of long repetitive sequences are not yet fully understood. Previous studies have shown that slipped-strand mispairing and inapposite recombination of repetitive sequences may lead to genome rearrangement (Timme et al., 2007). Moreover, forward repeats can cause changes in genome structure; therefore, they can be used as markers in phylogenetic studies.

Before cloning the target sequence, we first obtained the specific sequence information of the target sequence through previous research, and this information included the sequence length, putative transcriptional elements, and secondary structure. Then, through a sequence alignment, secondary structure analysis, etc., the specific information of these target sequences was initially obtained from the cp genome of E. breviscapus (Table 4).

Using E. breviscapus genomic DNA as a template, the target gene fragment was amplified with the primers shown in Supplementary Table S1. As shown in Figure 4, all amplified bands were the same size as the target gene fragment. Eb-Prrn was a fragment containing the rRNA operon promoter from E. breviscapus which was PCR amplified with a Shine-Dalgarno (SD) sequence derived from the chloroplast rbcL gene, Eb-psbA was a fragment containing the E. breviscapus psbA promoter and the E. breviscapus psbA leader. Eb-TpsbA, Eb-TrbcL, and Eb-Trps16 were the terminator of psbA, rbcL and rps16 gene, respectively. Eb-IEE was an intercistronic expression element conferring intercistronic RNA processing and, in this way, enhancing expression of downstream cistrons of the operon. Eb-trnI was a fragment containing the trnI gene and part of the flanking sequence of this gene and Eb-trnA was a fragment containing the trnA gene and part of the flanking sequence of this gene. The cloned chloroplast regulatory element sequence can be used for subsequent vector construction.

The verified vectors pBtEat1, pBtEat2, pBtEat3, and pBtEIat were transformed into the E. coli expression strain Transetta (DE3). As shown in Figure 5, after incubation, E. coli containing the empty vector could not grow on LB medium containing ampicillin and spectinomycin (Figure 5A). However, E. coli containing the recombinant vector could be grown on LB medium containing ampicillin and spectinomycin (Figures 5B–E), indicating that the selected regulatory elements had prokaryotic characteristics and could mediate the correct expression of foreign genes.

As shown in Figure 6, the fluorescence intensity of the E. coli expression strain containing the recombinant vector was higher than that of the control test group. Among them, the expression strain containing the recombinant vector pBtEat3 (containing the Prrn-EGFP-Trps16 expression cassette) had the highest fluorescence intensity, which showed that the promoters Prrn and 3′ UTR Trps16 could mediate the high-efficiency expression of foreign genes in E. coli. In this study, the EGFP gene was efficiently expressed in E. coli transformed with the E. breviscapus chloroplast expression vector, suggesting that the EGFP gene may also be highly expressed in E. breviscapus chloroplasts.