The bottle gourd variety “BG-54” used in this study was obtained commercially from Zhongziku APP (http://www.zhongziku.cc/). The plants were cultivated under controlled conditions at the Zhuanghang Comprehensive Experimental Station(E 121°28′, N 30°57′) of the Shanghai Academy of Agricultural Sciences. The photon flux density ranged from 650 to 850 W.m^-2 with temperatures between 10–25°C and relative humidity of 50–70%. Fresh leaves were frozen in liquid nitrogen and stored at -80°C. Total DNA was isolated following the protocol for the Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA) and the Oxford Nanopore PromethION (Oxford Nanopore Technologies, Oxford Science Park, UK).

Raw data from second-generation sequencing were filtered using fastp software (version 0.20.0, https://github.com/OpenGene/Fastp) (Chen et al., 2018). Third-generation sequencing data of mitochondrial reads were filtered using Filtlong (version 0.2.1, https://github.com/rrwick/Filtlong). The filtered third-generation data were aligned to the reference gene sequence using Minimap2 (version 2.1) (Li and Birol, 2018), and sequences with alignment lengths greater than 50 bp were selected. Sequences with overlaps greater than 1 kb and similarity greater than 70% were chosen as seed sequences. The original data were iteratively compared to the seed sequences to obtain all third-generation sequencing data of the mitochondrial genome. The third-generation assembly software Canu (Koren et al., 2017) was used to correct the obtained data. The corrected sequences were then aligned with the second-generation data using Bowtie2 (v2.3.5.1), and Unicycler (v0.4.8) was used to assemble the second- and third-generation data. Due to the complex physical structure of the mitochondrial genome, including subrings and non-circular forms, the corrected third-generation sequencing data were manually compared with the contigs obtained in the second step using Minimap2 to determine the branching direction and obtain the final assembly result ( Figure 1 ).

Protein-coding genes and rRNA sequences were annotated by comparing them with published plant mitochondrial sequences using BLAST, followed by manual adjustments based on related species. Transfer RNA (tRNA) genes were annotated using tRNAscan-SE (http://lowelab.ucsc.edu/tRNAscan-SE/) (Chan and Lowe, 2019). Open Reading Frames (ORFs) were identified using the Open Reading Frame Finder (http://www.ncbi.nlm.nih.gov/gorf/gorf.html), with the minimum length set to 102 bp. Redundant sequences and those overlapping with known genes were excluded. Sequences with alignments longer than 300 bp were annotated against the nr library. Potential RNA editing sites in the protein-coding genes (PCGs) of bottle gourd were predicted using the online Predictive RNA Editor for Plant Mitochondrial Genes (PREP-Mt) suite (http://prep.unl.edu/) (Mower, 2005). The physical circular map of the mitochondrial genome was generated using the Organellar Genome DRAW (OGDraw) v1.2 program (https://chlorobox.mpimp-golm.mpg.de/OGDraw.html). Relative synonymous codon usage (RSCU) was calculated using the CAI Python package developed by Lee (Lee, 2018),and codon frequencies were determined using the Codon Usage tool in the Sequence Manipulation Suite (bioinformatics.org/sms2/codon_usage.html) (Stothard, 2000).

Three types of repeats—simple sequence repeats (SSRs), tandem repeats, and dispersed repeats—were identified in the bottle gourd mitochondrial genome. SSRs were detected using the MIcroSAtellite identification tool (v1.0, parameters: 1-10 2-5 3-4 4-3 5-3 6-3) implemented in a Perl script (Thiel et al., 2003). Tandem repeats (>6 bp repeat units) were identified using Tandem Repeats Finder v4.09 (trf409.linux64, parameters: 27 7 80 10 50 2000 -f -d -m) (http://tandem.bu.edu/trf/trf.submit.options.html) (Benson, 1999). Dispersed repeats were detected using BLASTn (v2.10.1) with the following parameters: -word_size 7 and E-value 1e-5.

Gene sequences were aligned using MAFFT V7.310 (https://mafft.cbrc.jp/alignment/software/), and the nonsynonymous-to-synonymous substitution ratio (Ka/Ks) was calculated using the Ka/Ks Calculator V2.0 (https://sourceforge.net/projects/kakscalculator2/). The MLWL method was employed for the calculations.

Nucleotide diversity (Pi) was used to assess sequence variation among different species, with regions of high variation serving as potential molecular markers for population genetics. Homologous gene sequences from different species were globally aligned using MAFFT software (v7.427, –auto mode). The aligned sequences were concatenated, trimmed using trimAl (v1.4.rev15, parameter: -gt 0.7), and analyzed with DNAsp5 to calculate Pi values for each gene.

Homologous sequence analysis between chloroplast and mitochondrial genomes were conducted using BLAST, with a similarity threshold of 70% and an E-value of 1e-5. The results were visualized using Circos v0.69-5.

Phylogenetic analysis was conducted using the mitochondrial genomes of bottle gourd and 32 other species representing 24 families. Sequences from different species were aligned using MAFFT software (v7.427, –auto mode). The aligned sequences were concatenated, trimmed with trimAl (v1.4.rev15, parameter: -gt 0.7), and the best-fit evolutionary model (GTR) was determined using jModelTest-2.1.10. A maximum likelihood phylogenetic tree was constructed using RAxML V8.2.10 (https://cme.h-its.org/exelixis/software.html) under the GTRGAMMA model with 1,000 bootstrap replicates.

Collinearity analysis of the bottle gourd mitochondrial genome was performed using two methods. The first method involved comparing genomes using nucmer (4.0.0beta2) with the –maxmatch parameter to generate dot-plot diagrams. The second method utilized BLASTn (v2.10.1+) with parameters set to -word_size 7 and E-value 1e-5. Fragments with alignment lengths greater than 300 bp were screened, and collinearity maps were generated by comparing the assembled species with selected species.

The Illumina MiSeq and Nanopore sequencing produced 29,675,595 and 1,406,000 reads, respectively, with a mean read length of 7,433 bp for Nanopore sequencing. The complete mitochondrial genome of bottle gourd is a circular sequence of 357,496 bp with a GC content of 45.03% ( Figure 2 ). The sequence has been submitted to the GenBank database (accession number: PP727017). The mitochondrial genome contains 63 genes, including 34 mRNAs, 24 tRNAs, 4 rRNAs, and 1 pseudogene ( Table 1 ). Notably, three copies of the nad1 and nad5 genes were identified. Additionally, three tRNA genes located in repeat regions were found in two or three copies (trnp-TGG, trnM-CAT, and trnW-CCA) ( Figure 2 ).

In the mitochondrial (mt) genome of bottle gourd, the protein-coding genes (PCGs) can be categorized into 10 functional groups. These include ATP synthases (5 genes), cytochrome C biogenesis accessory proteins (4 genes), ubiquinol cytochrome C reductases (1 gene), cytochrome C oxidases (3 genes), maturases (1 gene), transport membrane proteins (1 gene), NADH dehydrogenases (9 genes), ribosomal proteins (LSU) (2 genes), ribosomal proteins (SSU) (6 genes), and succinate dehydrogenases (2 genes). Most PCGs utilize the typical ATG start codon, while cox1, nad1, and nad4L begin with ACG, likely due to C-to-U RNA editing at the second codon position ( Table 1 ). Four types of stop codons were identified: TGA, TAG, TAA, and CGA. RNA editing from C to U was observed in the stop codons of atp9 and sdh4 ( Table 1 ). The usage frequencies of these stop codons were 26.47% (TGA), 17.65% (TAG), 50% (TAA), and 5.88% (CGA), with TAA being the most frequently used stop codon.

The coding sequence (CDS) length of the bottle gourd mitochondrial genome is 30,212 bp, encoding 10,104 codons. Among these, 31 codons exhibited a relative synonymous codon usage (RSCU) value greater than 1, indicating a higher usage frequency compared to other synonymous codons. Analysis of RSCU for 24 PCGs in the bottle gourd mitogenome revealed that all NNT and NNA codons had RSCU values exceeding 1.0, except for the termination codon TGA (0.97) and the alanine codon GCA (0.98) ( Figure 3 ). Codon usage in the bottle gourd mitogenome showed a strong bias toward A or T(U) at the third codon position, a pattern commonly observed in the mitochondrial genomes of land plants.

In this study, a total of 497 RNA editing sites were predicted across 34 protein-coding genes (PCGs) in the mitochondrial (mt) genome of Lagenaria siceraria ( Table 2 ; Figure 4 ). Among these, the genes rps19, rps7, and sdh3 had the fewest predicted editing sites, with only 2 each. In contrast, ccmFn and nad4 contained the highest number of predicted editing sites, with 38 each. Following RNA editing, the hydrophobicity of 60.76% of the amino acids remained unchanged. However, 7.85% of hydrophobic amino acids were converted to hydrophilic, while 30.99% of hydrophilic amino acids became hydrophobic. All RNA editing events in the bottle gourd mt genome involved C-to-U conversions, with editing occurring at both the first and second positions of the triplet codon. This resulted in the conversion of proline (CCC) to phenylalanine (TTC or TTT). Notably, RNA editing in the coding genes atp9 and sdh4 led to premature termination of the coding process.

In the mitochondrial (mt) genome of bottle gourd, we identified a total of 260 interspersed repeats with lengths of 29 bp or greater. Among these, 123 were forward repeats, and 137 were palindrome repeats. The longest forward repeat sequence measured 2,349 bp, while the longest palindrome repeat sequence was 1,689 bp. As illustrated in Figure 5 , forward repeats were most abundant in the 30–39 bp range, whereas palindrome repeats were most abundant in the 40–49 bp range.

A total of 100 simple sequence repeats (SSRs) were detected in the bottle gourd mitogenome. These included 32 (32%) mononucleotide repeats, 25 (25%) dinucleotide repeats, 9 (9%) trinucleotide repeats, 30 (30%) tetranucleotide repeats, and 4 (4%) pentanucleotide repeats ( Table 3 ). Mononucleotide, tetranucleotide, and dinucleotide repeats were the most abundant types. Further analysis of SSR repeat units revealed that 90.63% of mononucleotide repeats consisted of A/T bases, and 72% of dinucleotide repeats were AT/TA. The high AT content of these SSRs contributes to the overall AT richness (54.97%) of the bottle gourd mitogenome. Additionally, as shown in Table 4 , a total of 9 tandem repeats, ranging in length from 12 to 39 bp and with a match degree greater than 80%, were identified in the bottle gourd mitogenome.

In genetics, the nonsynonymous-to-synonymous substitution ratio (Ka/Ks) is a key metric for understanding the evolutionary dynamics of genes. The Ka/Ks ratio helps determine whether a protein-coding gene (PCG) is under selective pressure during evolution. Under neutral selection, Ka = Ks, resulting in a Ka/Ks ratio of 1. If Ka > Ks (Ka/Ks > 1), it indicates positive selection, whereas if Ks > Ka (Ka/Ks < 1), it suggests negative (purifying) selection. In this study, the Ka/Ks ratio was calculated for 38 PCGs shared among L. siceraria, C. lanatus, C. sativus, L. acutangula, and C. maxima. As shown in Figure 6 , when comparing the mitochondrial (mt) genome of bottle gourd with that of C. lanatus, 16 PCGs exhibited Ka/Ks values < 1. In comparison to C. sativus, 24 PCGs had Ka/Ks values < 1, while 7 PCGs had Ka/Ks values > 1. Relative to L. acutangula, 15 PCGs showed Ka/Ks values < 1. When compared to C. maxima, 29 PCGs had Ka/Ks values < 1, and 3 PCGs had Ka/Ks values > 1. Notably, nearly all Ka/Ks ratios were less than 1.0, indicating that most PCGs were under stabilizing (purifying) selection during evolution. In contrast, two genes (atp8 and rps10) had Ka/Ks ratios > 1.0, suggesting they underwent positive selection. Additionally, three genes (atp4, rpl10, rpl2, rps19, and rps4) had Ka/Ks ratios close to 1.

Nucleotide diversity (Pi) was calculated for 37 genes to assess sequence variation. A total of 1,338 polymorphic sites were identified ( Supplementary Table S1 ). Among these, the maximum Pi value was 0.05028, corresponding to 65 polymorphic sites, while the minimum Pi value was 0.00594, associated with 4 polymorphic sites ( Figure 7 ).

We identified 45 homologous fragments between the mitochondrial (mt) and chloroplast (cp) genomes, with a total length of 40,579 bp, accounting for 11.35% of the mt genome ( Figure 8 , Table 5 ). These homologous fragments included 8 annotated genes, of which 6 were tRNA genes (trnL-CAA, trnM-CAT, trnN-GTT, trnD-GUC, trnP-TGG, and trnV-GAC) and 2 were ribosomal protein (SSU) genes (rps7 and rps12).

Phylogenetic trees were constructed using the maximum likelihood method to explore the evolutionary relationships between the bottle gourd mt genome and the published mt genomes of 32 plant species. The selected species and their details are listed in Table 6 . The results revealed that C. maxima, C. sativus, C. lanatus, and L. acutangula were closely clustered with bottle gourd ( Figure 9 ).

Based on the phylogenetic tree, the 32 plant species were grouped into three major clusters: angiosperms, gymnosperms, and spore plants. The clustering pattern in the phylogenetic tree aligns with the traditional taxonomic relationships at the family and genus levels, demonstrating the reliability of mt genome-based phylogenetic analysis.

Dot plot analysis revealed only sporadic collinear regions between C. sativus and L. siceraria, indicating poor collinearity ( Figure 10B ). In contrast, C. maxima, C. lanatus, and L. acutangula exhibited better collinearity with L. siceraria ( Figures 10A, C, D ). These findings were further supported by BLASTn collinearity analysis ( Figure 10E ).