Background

Murraya paniculata (L.) Jack, commonly called orange jessamine and synonymous with Murraya exotica, belongs to the genus Murraya in the family Rutaceae [1] (Additional file 1: Fig. S1). This ornamental plant is commonly used in landscaping and is now cultivated worldwide, including in Africa, America, the Caribbean, South America, Europe, and Oceania. Its leaves and flowers can be used to extract essential oils or for medicinal purposes, and the roots and stems also have a variety of uses [2]. M. paniculata is described in the Chinese pharmacopoeia as having anti-inflammatory, anti-biotic, and analgesic properties. A hot water extract of dried roots or stems can be used as an ecbolic with full-term deliveries for pregnant women in China. The dried bark and fruit are used in South East Asia as an astringent, to reduce fever, and to treat dysentery [3]. The stem bark is used in India for the treatment of coughs, hysteria, and rheumatism, while a paste of the leaf mixed with turmeric powder is applied to soothe fractured bones [4]. The volatiles collected from M. paniculata could attract Diaphorina citri, a vector of the bacterial causative agent of Huanglongbing [5] whereas M. paniculata and M. koenigii are reported to be resistant to the disease [6, 7].

As opposed to its use as a traditional medicine in the past, M. paniculata is now more likely to be sold in nurseries and stores as an ornamental plant. Its flowers have a strong fragrance that enhances its ornamental value. Studies have shown that flowers of M. paniculata follow a nocturnal/crepuscular pattern of blooming [8]. The volatile compounds, which are usually emitted at night, mainly include benzenoids, terpenoids, indoles, phenylacetaldehyde, and methyl palmitate [8]. The flowering time of M. paniculata can be divided into three stages; early, mid, and late, according to the dynamics of compounds emitted during the day. In the early flowering stage, the main components of the flower volatiles are phenolic compounds. At the mid-flowering stage when the flowers are fully open, terpenoids are the main volatiles. At the late-flowering stage, large amounts of phenylacetaldehyde are emitted. Each metabolite in the flower volatiles released during the flower’s life cycle is believed to have a specific function [9, 10]. For example, benzenoids attract pollinators, while terpenoids have the dual effect of attracting and discouraging insects from visiting flowers [11]. The emission of large amounts of phenylacetaldehyde during the daytime can deter possible nectar thieves that visit during the late-flowering stage [8].

The genus Murraya and Citrus both belong to the subfamily Aurantioideae of the family Rutaceae. However, the components of flower volatiles differ among M. paniculata and some Citrus plants, including C. maxima, C. sinensis, C. limon, and C. aurantium [8, 12, 13]. The volatiles of most Citrus plants mainly consist of linalool, β-myrcene, α-myrcene, limonene, (E)-ocimene, methyl anthranilate, and indole, but benzaldehyde and phenylacetaldehyde are missing in Citrus flower volatiles while they are enriched in the flower volatiles of M. paniculata [8, 12, 13]. The molecular basis for the biosynthesis of these compounds and the reason for the significant differences in volatiles between Citrus and Murraya are still unknown.

Here, we report the first chromosome-level reference genome of M. paniculata with a total assembly length of 216.87 Mb of 9 chromosomes and a contig N50 of 18.25 Mb. Comparative genomic analyses with other published Aurantioideae genomes revealed that transposon expansion is responsible for the different genome sizes between M. paniculata and the other Aurantioideae species whose genome size were reported to be in a range of 301.39 Mb (C. clementina) to 406.06 Mb (C. medica) [14,15,16,17,18,19,20] (http://citrus.hzau.edu.cn/statistics.php#genomeInfo). Furthermore, the flower volatiles analysis and transcriptome analysis in M. paniculata and C. maxima at different developmental periods showed how transposon polymorphism has affected the expression of genes encoding phenylacetaldehyde synthase. The reference genome of M. paniculata is a valuable genomic resource for further research on Rutaceae plants. Studies on floral volatiles and comparative genomics analysis reveal that transposon insertion may affect plant volatile biosynthesis and provide insights into the molecular mechanism of floral volatile biosynthesis in M. paniculata.

Results

Genome assembly and annotation

The genome size of M. paniculata is estimated to be about 256.76 Mb, with a heterozygosity rate of 0.44% based on K-mer analysis (K = 21) (Additional file 1: Fig. S2) using 8.74 Gb NGS short reads. To obtain a high-quality, chromosome-level genome assembly, 34.14 Gb ONT long reads (142 × coverage depth), 18.57 Gb paired-end NGS short reads (73 ×), and 41.78 Gb paired-end Hi-C reads (163 ×) were generated by different sequencing platforms (Additional file 2: Table S1). Firstly, the ONT long reads and NGS short reads were used for primary assembly, error correction, and polishing, finally generating an assembly of 216.87 Mb with a contig N50 of 18.25 Mb. Using 208.27 million Hi-C reads (Additional file 2: Table S1), contigs in the polished assembly were successfully ordered, oriented, and clustered into 9 pseudomolecules (Additional file 1: Fig. S3) (Additional file 2: Table S2). About 64.06 Mb repeat sequences were identified, accounting for 29.54% of the assembly, including a large amount of long terminal repeat (LTR) retrotransposons with a numerous of Ty3-Gypsy (7.29%) and Ty1-Copia (5.37%) (Fig. 1; Table 1; Additional file 2: Table S3). Using a combination of de novo prediction, homology alignment prediction, and transcriptome-based prediction, we predicted 23,548 protein-coding genes in the M. paniculata genome (Fig. 1; Table 1; Additional file 2: Table S4). Approximately 70.93% (16,703) of the protein-coding genes were functionally annotated by Swiss-Prot (Additional file 2: Table S4). In addition, 163 miRNAs, 246 rRNAs, and 481 tRNAs were identified (Table 1; Additional file 2: Table S5). The completeness was 98.24% and 98.50% when BUSCO was run in the genome and protein mode, respectively (Additional file 2: Table S6 and Table S7).

Fig. 1
figure 1

M. paniculata genomic landscape. (a) chromosomes, (b) gene density, (c) repeat density, (d) Copia distribution, (e) Gypsy distribution, (f) GC content, (g) collinear blocks. The density showing in heatmaps and histograms was calculated using 500 kb sliding windows. The darker the colour in the heatmaps, the higher the density

Full size image
Table 1 Summary of genome assembly and annotation for the M. paniculata
Full size table

Comparative genomics analysis

To investigate the phylogenetics of M. paniculata, its protein sequences and those of C. clementina, C. reticulata, C. sinensis, C. medica, C. maxima, Poncirus trifoliata, Atalantia buxifolia, and Zanthoxylum armatum as the outgroup were collected for comparative genomics analysis. In total, 4995 single-copy gene families were identified and used for phylogenetic tree construction and estimation of divergence time (Fig. 2A). The phylogenetic tree revealed that M. paniculata diverged from the common ancestor approximately 25 (17.5–30.7) million years ago (MYA) (Fig. 2A). Expansion and contraction of gene families was also estimated from the tree. A total of 372 significantly expanded gene families (P < 0.05) and 1663 contracted gene families (P < 0.05) were detected in M. paniculata. KEGG pathway enrichment analysis using the genes in expanded gene families showed that they are enriched in some metabolic pathways such as “amino acid metabolism”, “phenylpropanoid biosynthesis”, and “sesquiterpenoid and triterpenoid biosynthesis” (Additional file 3). The homologous gene pairs in M. paniculata were collected for whole genome duplication (WGD) analysis. The synonymous substitutions per site (Ks) distance of these gene pairs revealed that M. paniculata has not undergone any species-specific WGD events after the shared ancient WGD in M. paniculata and C. sinensis (Ks 1.4–1.5) (Fig. 2B; Additional file 1: Fig. S4) [19]. These genomes of M. paniculata, C. sinensis, C. medica, and P. trifoliata show high collinearity (Additional file 1: Fig. S5). However, the genome size and transposable elements (TEs) length of M. paniculata are smaller than the other three species (Additional file 1: Fig. S6; Additional file 2: Table S8 and Table S9).

Fig. 2
figure 2

Comparative genomic analysis of M. paniculata. A Phylogenetic tree constructed using coding sequences of single-copy gene families among 9 Aurantioideae species. Divergence times were estimated by MCMCTree. Support rates were calculated by RAxML-NG with GTR + G model and 200 bootstraps. All nodes have a bootstraps support percentage of 100. The green and red numbers on the tree or next to the species label indicate the expansion and contraction gene family counts. B Density distributions of collinear block Ks (synonymous substitutions per synonymous site) median of M. paniculata

Full size image

To further investigate the difference in the repeat sequence among these species, the TEs identified in these four species were classified in more detail. The retrotransposons in M. paniculata included 28.24 Mb of LTR sequences, mainly Gypsy and Copia, accounting for approximately 12.66% of the genome; while the DNA transposons mainly consisted of the terminal inverted repeat (TIR) superfamilies CACTA, Mutator, PIF-Harbinger, Tc1-Mariner, and hAT [21]. The number of LTRs in the P. trifoliata, C. maxima and C. sinensis genomes was 1.42-times, 2.54-times, and 1.71-times than that in the M. paniculata genome, respectively (Additional file 1: Fig. S7). We calculated the distribution density of LTRs and TIRs in the 10-kb range upstream and downstream of genes’ coding regions with 100-bp sliding windows (Additional file 1: Fig. S8). The percentage of genes with LTRs inserted nearby was significantly lower in the genome of M. paniculata than in the genomes of the other three species (Additional file 1: Fig. S8). The percentage and distribution of TIRs differed slightly among the four species (Additional file 1: Fig. S8). A significant expansion of the transposase gene family was detected in the genomes of C. sinensis and C. maxima (Additional file 1: Fig. S9). Next, genes from these four species with LTR insertions in the 3-kb upstream region were collected for KEGG pathway enrichment analysis separately (Additional file 4), which showed that the genes from different species with LTR inserted nearby may be similarly involved in several metabolites biosynthesis pathways. These results imply that the LTR insertions may affect the metabolite biosynthesis in these species.

Differences in flower volatile components between M. paniculata and C. maxima

To explore the volatiles differences of flowers between Murraya and Citrus, we identified and quantified the flower volatiles of M. paniculata and C. maxima at different flowering stages (F1: early flowering, F2: mid flowering and F3: late flowering) (Fig. 3A) by headspace solid-phase microextraction (HS-SPME) and gas chromatography-mass spectrometry (GC–MS) (Additional files 5 and 6). As the principal component analysis (PCA) showed, the volatile compounds at the F2 and F3 flowering stages were separated from those at the F1 flowering stage in M. paniculata, whereas the flower volatiles at all three flowering stages of C. maxima were clustered together (Fig. 3B). There were five shared volatile compounds at all flowering stages of M. paniculata and C. maxima, but more compounds unique to M. paniculata flowers (Fig. 3C). In the F1 flowering stage of M. paniculata, the main volatile compounds were linalool (48.0%) and beta-myrcene (17.5%). However, we found a dramatic change in volatiles composition in the F2 and F3 flowering stages of M. paniculata, with benzaldehyde (47.7% and 46.2%, respectively) and phenylacetaldehyde (19.3% and 25.7%, respectively) being the main volatile components (Fig. 3D). The main volatile components of C. maxima flower volatiles were linalool and D-limonene at all flowering stages (Fig. 3E). The comparison of the flower volatile components between C. maxima and M. paniculata showed that C. maxima lacks phenylacetaldehyde and benzaldehyde, which are the main volatile compounds in flower volatiles of M. paniculata at the F2 and F3 stages. Considering that many genes involved in metabolites biosynthesis pathways have TE insertions nearby (Additional file 4), we speculated that the volatiles difference between M. paniculata and C. maxima may be related to TE insertion around genes in relevant pathways.

Fig. 3
figure 3

Analysis of flower volatiles of M. paniculata and C. maxima. A Flower morphology of M. paniculata (left) and C. maxima (right) at different flowering stages. B Principal component analysis (PCA) of flower volatile components at different flowering stages. Scale of the axis is relative distance. Dots with different colours represent different groups of samples. C Venn diagram showing shared or unique types of volatiles among different flowering stages of M. paniculata and C. maxima. D and E Relative percentage content of volatiles out of total flower volatiles at different flowering stages of M. paniculata (D) and C. maxima (E). Data for volatile contents statistics were obtained from three replicates of flowers with similar growth status. Area of each sector represents the relative abundance of corresponding volatile compound calculated by GC–MS peak area (Additional files 5 and 6)

Full size image

TEs insertions may lead to differential PAAS expression in M. paniculata and C. maxima

To validate the hypothesis and find evidence for the effect of TE insertions on, the genes involved in the benzaldehyde and phenylacetaldehyde biosynthesis and metabolism pathway were identified in M. paniculata and C. maxima, respectively (Fig. 4A). These genes belonged to 11 gene families. Compared with C. maxima, M. paniculata had one extra member in the PAL, PAAS, 4CL, and AAO gene family. Three tandemly repeated PAAS genes in M. paniculata, namely Me2G_2379, Me2G_2381, and Me2G_2382, caught our attention, whose relative expression levels were significantly higher than those of their two homologues in C. maxima (Cg1g029630 and Cg1g029640) (Fig. 4A and B; Additional file 7). Genome structural analysis revealed that there are no LTR insertions in the upstream region of the three PAAS genes of M. paniculata, while insertions with the same LTR fragment were observed in the upstream promoter regions of Cg1g029630 and Cg1g029640 (Fig. 4C and D; Additional file 1: Fig. S10), which may affect the expression levels of Cg1g029630 and Cg1g029640, thus leading to the difference of phenylacetaldehyde content in the flowers of M. paniculata and C. maxima. Similar LTR insertions were also observed in the promoter regions of corresponding PAAS genes in C. sinensis (Additional file 1: Fig. S10 and Fig. S11).

Fig. 4
figure 4

Analysis of genes involved in benzaldehyde and phenylacetaldehyde biosynthesis and metabolic pathway. A Genes involved in benzaldehyde and phenylacetaldehyde biosynthesis and metabolic pathway with heatmaps showing the relative expression levels in flowers from three flowering stages, leaves, and stems. TPMs were obtained from the mean values of three replicates (Additional file 7). Gene expression levels are normalised and represented as log2(TPM + 1). Blue, low expression levels; red, high expression levels. B Protein neighbour-joining tree of PAAS genes in f M. paniculata and C. maxima with outgroups of RhybPAAS (Rosa hybrid; Uniprot entry: Q0ZS27), PhybPAAS (Petunia hybrida; Uniprot entry Q0ZQX0), and AthaPAAS (A. thaliana; Uniprot entry: Q8RY79) constructed with 1000 bootstrap replicates. C Gene collinearity analysis between Me2G_2379, Me2G_2381, and Me2G_2382 in M. paniculata and Cg1g029630, Cg1g029640 in C. maxima. D Overview of LTR insertions upstream PAAS genes Cg1g029630 and Cg1g029640 in C. maxima. Black line represents genome sequence; green bold arrow represents length and orientation of CDS; red bold arrow represents length and orientation of LTR fragment. TSS: transcription start site; TTS: transcription termination site

Full size image

Biosynthesis of phenylacetaldehyde by three master-effect PAAS genes in M. paniculata

We identified a total of 9 PAAS genes in M. paniculata, but only some of them were highly expressed in the flower in the flowering stages. Thus, we verified and compared the catalytic activity of the products of these PAAS genes to clarify their role in phenylacetaldehyde biosynthesis. The optimised full-length CDS of these PAAS genes (Additional file 2: Table S10) were transferred into the Escherichia coli expression strain Rosetta (DE3), and the phenylacetaldehyde content in the assay mixture was determined by GC–MS. The proteins encoded by most PAAS genes were able to catalyse the production of phenylacetaldehyde (Fig. 5; Additional file 8). Furthermore, kinetic characterisation of purified recombinant PAAS revealed an apparent Km for L-phenylalanine (L-Phe) of 1.13 ± 0.31 mM − 36.59 ± 4.35 (mean ± S.E., n = 3). The PAAS Kcat were 0.96 ± 0.26 × 10−3 − 34.60 ± 0.2612.63 × 10−3 min−1 (mean ± S.E., n = 3) (Additional file 1: Fig. S12; Additional file 2: Table S11). The proteins encoded by Me2G_2379, Me2G_2381, and Me2G_2382 showed very strong catalytic activity to produce phenylacetaldehyde. Combined with the results of transcriptome analysis, these three genes likely encode the main enzymes involved in the biosynthesis of phenylacetaldehyde in M. paniculata flowers.

Fig. 5
figure 5

Phenylacetaldehyde analysis using PAAS heterologously expressed in Escherichia coli. A GC–MS results showing that Me1G_2960, Me2G_2381, Me3G_0317, Me3G_1222, Me3G_2284, Me4G_0535, and Me5G_1302 encode enzymes that are functional when expressed in Escherichia coli. Extract ion chromatogram shows products produced by strains expressing target genes or the empty vector control (CK). Peaks at retention time about 8 min represent phenylacetaldehyde. X-axis represents retention time; Y-axis represents relative abundance. B Mass spectrum of phenylacetaldehyde (molecular ion m/z 91). C Concentration of phenylacetaldehyde produced by products of 9 PAAS genes. Data (mean values ± SEs) were obtained from three replicate experiments (Student’s t-test, *P < 0.05, **P < 0.01)

Full size image

Discussion

Flower volatile compounds have various roles, they attract insects for pollination, improve plant reproduction and adaptability, and act as a language to communicate and interact with the surrounding environment [22, 23]. As ornamental plants with strong flower aroma, the flower volatiles of M. paniculata differ from those of Citrus plants in Aurantioideae. The genetic basis of the diversity of specialised metabolites can be illustrated by high-quality genome sequencing [24]. Most of the Rutaceae genomes published so far are in the genus Citrus, including C. clementina, C. reticulata, C. maxima, C. sinensis, and C. medica, but genomic information was not available for the genus Murraya. In this study, we presented a high-quality reference genome of M. paniculata, which is the first chromosome-level genome assembly in Murraya plants. M. paniculata diverged from Rutaceae about 25 MYA. No large-scale variation in chromosome structure was detected in the M. paniculata genome compared with that of C. sinensis. Besides the core eudicot shared γ triplication event, M. paniculata has not undergone any additional species-specific WGD events, consistent with C. sinensis [19]. Genomes of M. paniculata and other published Citrus plants have a high degree of genetic collinearity, implying that different characters between M. paniculata and Citrus may be not due to large-scale chromosome structural variations.

TEs are major component of plant genomes, their activity and diversity can influence the genome size and structure [25, 26]. Thus, considering the differences in genome size among M. paniculata and some representative species in Aurantioideae subfamily, we analysed the TEs in their genomes. The results showed that there are significantly more TEs in the C. sinensis and C. maxima genomes than in the M. paniculata genome. In particular, the numbers of LTRs distributed near protein-coding genes are about two-fold higher in C. sinensis and C. maxima than in M. paniculata. Several studies have shown that LTRs tend to be inserted near genes with functions in resistance and development, driving divergence among related species [27, 28]. We found that genes with TEs insertions in their upstream 3-kb regions are mainly enriched in several types of metabolites biosynthesis pathways in M. paniculata and C. maxima whereas the GC–MS analysis revealed notable differences in flower volatile compounds between them in three flowering stages. Thus, we hypothesised that the differences in flower volatiles may be related to TE insertions in the vicinity of genes encoding biosynthetic or metabolic enzymes in these plants [23].

The insertion of TEs in genomic regions can seriously affect the regulation of neighbouring genes’ expression, resulting in altered traits. In Arabidopsis thaliana and A. lyrata, TEs reduce the average expression levels of adjacent genes because of interference with cis-regulatory elements, and these effects differ between the two species [27]. Genomic studies on apple and Sicilian blood orange revealed LTR insertions influence the expression of adjacent genes [29,30,31]. In this study, we found LTR inserted upstream C. maxima PAAS genes (Cg1g029630 and Cg1g029640) (Fig. 4D), which may be associated with their relatively low expression levels in flowers. Three homologous M. paniculata PAAS genes (Me2G_2379, Me2G_2381, and Me2G_2382) that have no LTR insertions nearby show higher transcript levels during flower development. The proteins of these three genes were shown to have high phenylacetaldehyde synthesis activity (Fig. 5C). Considering the fact that phenylacetaldehyde is one of the main components of the flower volatiles of M. paniculata, the difference between M. paniculata and C. maxima flower volatiles may be largely caused by these TE insertions adjacent to C. maxima PAAS genes, which results in the lack of phenylacetaldehyde in C. maxima flowers. More experimental evidence and bioinformatic analysis are required to further investigate the molecular mechanisms underlying the differences in volatiles.

Conclusions

The study presented a high-quality chromosome-level genome assembly of M. paniculata. Obvious differences in transposon contents were detected in the genomes among M. paniculata and other Aurantioideae species, especially in the upstream regions of genes, which may affect metabolites biosynthesis. Furthermore, three genes (Me2G_2379, Me2G_2381, and Me2G_2382) with strong synthetic activity for phenylacetaldehyde were newly identified in M. paniculata, whereas the expressions of two homologues of these genes in C. maxima may be affected by LTR insertions which led to the lack of phenylacetaldehyde in flowers of C. maxima. The study proposed how transposons impact on the composition of flower volatiles among species in Murraya and Citrus. These genomic resources of M. paniculata will contribute to further research on species in Rutaceae.

Methods

Plant materials

Fresh healthy leaves were harvested from an individual plant of M. paniculata at Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences (21°55′11″N, 101°15′27″E) in April 2021 and immediately frozen in liquid nitrogen, then stored at − 80 °C until DNA extraction. Leaves, stems, and flowers from M. paniculata were also sampled for RNA-seq. Flowers at three different flowering stages were collected from different M. paniculata and C. maxima individuals for volatile identification and transcriptome analysis at Xishuangbanna Tropical Botanical Garden in March 2022.

Library construction and sequencing

For ONT sequencing, high-quality genomic DNA was extracted, size-selected to prepare the ONT library, and sequenced on the PromethION platform (Oxford Nanopore Technologies, Oxford, UK). For whole genome sequencing on the NGS platform, a paired-end library with an insert size of 350 bp was constructed according to the manufacturer’s protocol. For Hi-C sequencing, fresh young leaves were fixed in formaldehyde (1 g leaf material per 100 mL). Chromatin was cross-linked and digested using the restriction enzyme DpnII. Then the DNA sample was biotin labelled and ligated. After shearing and size-selecting, fragments containing biotin were captured to construct the paired-end library. For RNA sequencing, the total RNA was extracted from mixed leaf, stem, and flower tissues using RNAiso Plus (Qiagen, Hilden, Germany) according to the manufacturer’s protocol and libraries were prepared using the TruSeq RNA Sample Preparation kit. The quantifications of these paired-end libraries were performed using a Qubit 3.0 fluorometer (Thermo Fisher Scientific Inc., Waltham, MA, USA) and the quality assessments were carried out by an Agilent 2100 instrument (Agilent, Palo Alto, CA, USA). Qualified libraries were then sequenced using the DNBSEQ-T7 platform (MGI, Shenzhen, China) with a layout of 150 bp. All paired-end raw reads were filtered by fastp v0.23.2 [32] (-f 5 -t 5 -n 0 -l 140).

Genome assembly and pseudomolecule construction

Raw reads from ONT sequencing were error corrected and trimmed using Canu v2.1.1 [33] (-nanopore -correct -trim) and the longest 40 × reads were retained to generate the draft assembly by NextDenovo v2.4.0 [34] and polished with ONT and NGS reads using NextPolish [35] (task = best). Redundancies were removed using Purge_Dups [36]. Bwa-mem2 v2.2 [37] was used to align the clean Hi-C reads to the draft assembly. A valid Hi-C links file was obtained using juicer v1.6 [38]. Then the links file was processed by 3D-DNA v201008 [39] to order, orient, and cluster the contigs automatically followed by manual validation and adjustments by juicebox v1.11.08 [40]. The Hi-C contact heatmap was generated using the 3D-DNA visualization module and juicebox.

Genome annotation

A high-quality non-redundant repeat library of the M. paniculata genome was generated using the Extensive de novo TE Annotator (EDTA) pipeline [41]. The TEs were further classified by TEsorter [42] and intact transposase sequences were extracted simultaneously for tree building using iTOL [43]. Based on the.out file, the EDTA pipeline summarised the TEs’ counts and proportions. The same pipeline was used to process the genomes of P. trifoliata, C. maxima, and C. sinensis. The classified repeat library was used to softmask the genome by RepeatMasker (http://www.repeatmasker.org) [44]. The structural annotation for protein-coding genes was based on de novo prediction, homologous protein alignment, and RNA-seq-based predictions using BRAKER v2.1.6 [45] with the softmasked genome. The RNA-Seq reads were mapped to the genome using HISAT2 v2.2.1 [46] with default parameters. The proteins in OrthoDB v10 [47] were used for protein-based gene prediction. Protein hints were generated using ProtHint v2.6.0 [48]. De novo gene prediction was carried out using protein alignment data and RNA-seq data with AUGUSTUS v3.4.0 [49] and GeneMark-ES Suite version 4.69_lic [50]. TSEBRA v1.0.3 [51] was used to merge the prediction results. The annotation file was modified and formatted by MAKER v3.01.04 [52] and EVidenceModeler [53]. Gene function annotation was conducted using both eggNOG-mapper [54] online (http://eggnog-mapper.embl.de/) and BLAST-based methods. The best alignments were obtained by searching against the NR and SwissProt databases using diamond v2.0.11.149. The completeness of genome assembly and annotation was assessed using Benchmarking Universal Single-Copy Orthologs (BUSCO) [55] with eudicots_odb10.

Phylogenetic analysis

Orthogroups were identified using OrthoFinder v2.3.12 [56] with default parameters taking protein sequences from C. clementina, C. reticulata, C. sinensis, C. medica, C. maxima [20], P. trifoliata, A. buxifolia, and Z. armatum as inputs [57, 58]. Protein sequences from single-copy orthogroups were collected and aligned using MAFFT v7.313 with the G-INS-i strategy [59]. The alignments of single-copy orthogroups were concatenated to generate a super alignment matrix, and a maximum-likelihood tree was constructed using the GTR + G model with 200 bootstraps by RAxML-NG. Divergence times were estimated using MCMCTree in the PAML v4.10.6 [60]. Analyses were run for 100,000 generations with a burn-in of 400,000 iterations and the HKY85 model. The fossil calibration times used to estimate divergence time (13.5–30 MYA between Murraya and Atalantia and 25.1–69.9 MYA between Murraya and Zanthoxylum) were obtained from the TimeTree of Life online in April 2023 [61]. All MCMCTree calculations were run twice to ensure convergence. The expansion and contraction of orthogroups were calculated using CAFE5 [62] under a random birth-and-death model using the OrthoFinder results as input. The gene synteny and gene duplication analyses were performed using JCVI with the MCscan pipeline [63]. Syntenic blocks and synonymous nucleotide substitutions (Ks) were determined and calculated from protein sequence alignments using the NG86 model based on the homologous gene pairs identified by WGDI [64]. The WGD and speciation events were inferred from paralogous and orthologous pairs of Ks distribution peaks, respectively.

Flower volatile compounds analysis

The volatile compounds of the flowers were analysed by HS-SPME coupled with GC–MS. Each sample was quickly placed in a wet glass vessel (50 mL) and sealed with paraffin film. The SPME fibre (50/30 μM DVB/CAR/PDMS) was exposed to the inflorescence for 30 min to extract the volatile components in an air-conditioned laboratory at 25 °C ± 3 °C. After the adsorption period, the fibre head was removed and introduced into the heated injector port of the GC for desorption at 250 °C for 1 min. Subsequent analysis was carried out using an Agilent GC–MS (6890 GC-5973N MSD, Agilent, Santa Clara, CA, USA). The carrier gas (He) was supplied at a constant rate of 1.0 mL/min. The mass detector conditions for MS were as follows: source temperature of 250 °C, electronic impact (EI) mode at 70 eV, with a speed of 4 scans/s over the mass range m/z 33–450 amu in a 1-s cycle. Compounds were auto-matched with mass spectra in the NIST98 database through Chem-Station (Agilent). The normalised peak area measurements were used to calculate the percentage of each flower volatile component.

Candidate genes identification

Protein BLASTN and Pfam annotation were used to identify genes involved in benzaldehyde and phenylacetaldehyde biosynthesis and metabolic pathways in M. paniculata and C. maxima. The searching queries were the functionally validated proteins including RhybPAAS (Rosa hybrid; Uniprot entry: Q0ZS27), PhybPAAS (Petunia hybrida; Uniprot entry Q0ZQX0), and AthaPAAS (A. thaliana; Uniprot entry: Q8RY79) [65] obtained from Swiss-Prot database. Genes with high homology and similar domains were identified as candidates.

Calculation of gene relative expression level

The raw RNA-seq reads were first filtered by fastp as described above and then mapped to the reference genome using HISAT2 (–very-sensitive –dta). Raw count for each transcript was tallied by featureCounts v2.0.3 [66] and normalised by TPM (Transcripts Per kilobase of exon model per Million mapped reads) values using an R script. The heatmap was generated using TBtools [67].

Enzyme activity analysis

According to the amino acid sequences, codon-optimised PAAS genes from M. paniculata were synthesised and then pET28a-PAAS recombinant plasmids (Additional file 2: Table S12) were transferred into E. coli Rosetta (DE3) competent cells. The control was transformed with the plasmid without an insert. The cells were grown in LB medium containing 50 μg/mL kanamycin at 37 °C to a culture density of OD600 = 0.5. Then, the expression of PAAS proteins was induced by the addition of IPTG to a final concentration of 0.3 mM. After a 13-h incubation at 18 °C, the cells were harvested by centrifugation and resuspended in lysis buffer containing 0.02 M phosphate buffer (pH 7.4), 0.5 M NaCl, 2% (v/v) glycerol, 0.2 mM pyridoxal 5-phosphate (PLP), and 1 mM PMSF. After removing cell debris, proteins were purified by affinity chromatography on nickel nitrilotriacetic acid-agarose (0.5-mL bed volume) [65].

An in vitro reaction was established with the above-described purified proteins. The time dependence was conducted in reactions of PAAS (5 μg), 5 mM L-Phe, 50 mM Tris–HCl (pH 8.5), 0.2 mM PLP, and 0.1 mM EDTA in a final volume of 50 µL, kept at 28 °C for 10 min, 20 min, 30 min, 40 min, 50 min, 60 min, respectively. Kinetic studies were performed with PAAS (4 mg/mL), L-Phe (2.7 mM, 5.4 mM, 10.8 mM, 21.6 mM, 43.3 mM, respectively), 50 mM Tris–HCl (pH 8.5), 0.2 mM PLP, and 0.1 mM EDTA in a final volume of 50 µL, incubating at 28 °C for 30 min. After incubation at 28 °C, the reaction was stopped by adding 5 μL 10 M NaOH, and then 250 μL ethylacetate was added to extract the reaction product for GC–MS detection. Lineweaver–Burk plots were constructed to obtain the Km value and Kcat. All assays were conducted in triplicate.

The GC–MS system used was an Agilent 7890a gas chromatograph/5975c mass selective detector with a 30-m DB-5MS capillary column. The carrier gas was supplied at a constant flow rate of 1 mL/min. The oven conditions were as follows: initial temperature of 40 °C for 2 min, increased by 10 °C/min to 130 °C, held for 5 min, increased by 8 °C/min to 230 °C, held for 3 min. The inlet temperature was kept constant at 250 °C, and the MS transfer line was set at 290 °C. The MS acquisition parameters included scanning from 50 to 550 m/z in the electron impact (EI) mode for routine analysis.