Introduction

The walnut family (Juglandaceae) containing ca. 60 extant species belonging to ca. 10 genera, a woody family in the order Fagales, is mainly distributed in subtropical to tropical forests [1,2,3]. Members of this family play an important role in local forest ecosystems, and some of them are important nut, timber, and medicinal trees. According to APG IV (2016), Juglandaceae is grouped into three subfamilies: Rhoipteleoideae, Engelhardioideae, and Juglandoideae [4]. Among the three subfamilies of Juglandaceae, species of the Juglandoideae subfamily were very common in temperate deciduous forests in the Northern Hemisphere, while species of the Engelhardioideae subfamily and the Rhoipteleoideae subfamily were mainly distributed in subtropical and tropical forests [2].

In the Engelhardioideae subfamily, species of Engelhardia Lesch. ex Blumewidely distribute in the tropical and subtropical regions of eastern Asia [5], which are widely used in wood and tea, and also play a significant role in the ecosystem [6, 7]. There are about 9 Engelhardia species in China, which mainly occur in the southwest, south to southeast [7]. The Engelhardia species are deciduous or semi-evergreen trees or evergreen tree, often with even-pinnate compound leaves, monoecious or dioecious, fruit nut-like, when the fruit is ripe, the bracts grow, membranous, and connate with the fruit to form a nutlet with 3-lobed wings [8]. In previous studies on the species delimitation of Engelhardia [7, 9] and the phylogeography of two trees species (i.e., E. roxburghiana and E. fenzelii) [10], combined plastid regions (psbA-trnH, trnL-trnF, rps16, trnS-trnG, and rpl32-trnL), one nuclear ribosomal internal transcribed spacer (nrITS), and Microsatellite (nSSR) data were used. Yet to date, except in individual cases [3, 11, 12], comparative analyses of multiple Engelhardia plastome are still lacking.

Plastomes, as critical organelles, play a pivotal role in plant cells, underscored by their widespread application in evolutionary and phylogenetic studies [13, 14]. Their importance is attributed to maternal uniparental inheritance and a highly conserved structure, making them valuable in dissecting plant evolutionary histories and relationships across a wide array of studies [15, 16]. Plastome is usually a closed circular tetrad structure composed of DNA double-stranded molecules, including a large single copy region (LSC), a small single copy region (SSC) and a pair of inverted repeat region (IRa/IRb) [17]. Early phylogenetic analyses used partial plastomes DNA sequences. However these fragments did not have enough information to distinguish closely related plant species, while whole plastid genomes can provide in-depth information to improve our understanding of species evolution [18]. The complete plastomes had made great progress in elucidating the relationship in monocot [19], and also explained the relationship between several major lineages of angiosperms [20]. Meanwhile, plant plastome gene and genome evolving during the process of speciation can help us understand how species adapts to diverse ecological habitats [14].

In the study of phylogeny and evolution of species, fossils play a crucial role in determining the time of species differentiation [21]. For example, the well preserved Rhynie Chert fossils were helpful to provide insights into the life cycle of early land plants [22], and the usage of fossil correction and molecular clock methods can well support the pre-Cretaceous origin of angiosperms [23]. Previous studies based on fossil evidence have greatly promoted our understanding of Juglandaceae as a whole [2, 24], but for subfamily Engelhardioideae, generally only showing the differentiation time of E. roxburghiana due to insufficient sampling. Therefore, the divergence times of the Engelhardioideae species remain unsolved. What is the divergence time within the Engelhardia? Are the results based on different fossil calibration points consistent with those of predecessors? It is necessary to increase the species sampling of the Engelhardia to explore the phylogenetic relationship, divergence and orgin of Juglandaceae.

In this study, a total of 14 individuals from eight species of Engelhardia (E. anminiana, E. fenzelii, E. hainanensis, E. roxburghiana, E. serrata, E. spicata, E. spicata var. rigida, and E. villosa) and one outgroup species Rhoiptelea chiliantha were newly sequenced (Table 1), and 42 plastome sequences from 36 Juglandaceae species and six outgroup species were downloaded from GenBank. The whole plastomes were used to explore evolution and the deep phylogenetic relationship among the species of Engelhardia and (or) other genera, subfamilies, even the whole Juglandaceae family. Our specific goals were as follows: (1) to compare the plastomes and identify the variation in Engelhardia; (2) to identify genomic structural variation across Juglandaceae plastomes; (3) to deepen the understanding on the codon usage bias and gene evolution in Juglandaceae plastomes; (4) to infer and test the phylogenetic relationship and divergence time among the genera and subfamilies of Juglandaceae using plastome data.

Table 1 Taxa, voucher and GenBank accession numbers of Engelhardia species and Rhoiptelea chiliantha sequenced in this study
Full size table

Results

Characteristics of Engelhardia plastomes

The lengths of the complete plastomes of Engelhardia were slightly different, ranging from 161,069 bp to 162,336 bp, exhibiting a quadripartite structure with a large singlecopy (LSC) region (89,927–91,637 bp), dual inverted repeat (IR) regions (25,813–26,016 bp), and a small single-copy (SSC) region (18,790–19,203 bp) (Fig. 1, Table 2). There were a total of 134 genes were identified in the newly sequenced plastomes, including 88 protein-coding genes (CDS), two pseudogenes (Ψycf1、Ψrps19), 37 transfer RNA (tRNA) genes and eight ribosomal (rRNA) genes (Table 2). The ycf1 in the IRb region (Ψycf1) and the rps19 in IRa region (Ψrps19) of all Engelhardia species were identified as pseudogenes (Table S2). Among these genes, there were 18 intron-containing genes, of which three genes rps12, clpP and ycf3, had two introns, and the rest contained a single intron (trnA-UGC, trnG-UCC, trnI-GAU, trnK-UUU, trnL-UAA, trnV-UAC, rpl2, rpl16, rps16, rpoC1, atpF, ndhA, ndhB, petB, and petD) (Table S2). These newly-generated Engelhardia plastomes were deposited in GenBank (Assession Number showed in Table 1).

Fig. 1
figure 1

Gene map of Engelhardia plastomes. The species name is displayed in the upper left corner, and the genome map includes 5 tracks. From the inside out, the first track (A) displays forward and reverse repeats connected by red and green arcs. The second track (B) shows the tandem repeats, represented by blue line segments. The third track (C) displays the microsatellite sequences, represented by green and yellow line segments. The fourth track (D) displays large single copy (LSC), small single copy (SSC), and inverted repeat (IRa and IRb). The fifth track (E) displays the GC content of the genome. The genes are distributed in the outermost circle (F), the optional codon pusage bias is displayed in parentheses after the gene name. The genes shown inside and outside of the circle are transcribed in clockwise and counterclockwise directions, respectively. Genes from different functional groups are shown in different colors

Full size image
Table 2 Complete features of 13 newly assembled Engelhardia plastomes and one Rhoiptelea chiliantha plastome
Full size table

The overall GC content of Engelhardia plastomes was 35.8%–36.0% (Table S3), and the GC contents of coding sequence (CDS) regions was 37.2%–37.3%. We found that the GC content of LSC (33.2%–33.6%) and SSC (29.3%–29.6%) region were lower than those of IR regions (42.6%–42.7%) (Table S3).

Comparative analysis of Engelhardia plastomes

Multiple plastomes comparison among all the Engelhardia species using mVISTA and Mauve alignment showed high degree of collinearity. It was found that the composition and sequence of genes in Engelhardia were highly consistent, and no inversion or translocation of DNA fragment rearrangement was detected in the sequences (Fig. S2). The regions with relatively low identity were rps16_trnQ-UUG, trnS-GCU_trnG-UCC, trnT-GGU_psbD, trnF-GAA_ndhJ, ndhK_ndhC, accD_psaI, petA_psbJ, and ndhF_trnL-UAG (Fig. S1). Most of the DNA sequence variation in Engelhardia species occurred in non-coding regions such as gene spacer region and gene intron region, and the sequence differentiation between LSC and SSC region was significantly higher than that in IR region (Figs. S1-2).

By analyzing the boundary differences of LSC, SSC, IRa and IRb sequences in plastomes of Engelhardia, it was found the inner boundary differences were small. There was no large regional expansion and shortening of spacer regions occurred, which was consistent with the conserved character of plastomes within this genus (Fig. 2). The ycf1 gene in all species spanned the SSC/IRa region, with the length of ycf1 in SSC is 4623 bp–4729 bp, and that in IRa is 1004 bp–1104 bp. The pseudogene (Ψycf1) was formed at the corresponding position near the IRb/SSC boundary, and the extension of the short Ψycf1 fragment into the SSC region was observed in all Engelhardia species. The overlap** of Ψycf1 and ndhF has only been detected in E. anminiana, E. spicata and E. villosa. The rps19 gene spanned the LSC/IRb regions in all the Engelhardia species, and it formed a pseudogene (Ψrps19) at the IRa/LSC boundary (Fig. 2).

Fig. 2
figure 2

Comparison of the border positions of SSC, LSC, and IR regions among 13 Engelhardia plastomes. Genes close to or spanned the boundaries were shown in yellow boxes

Full size image

Repetitive sequences in Engelhardia plastomes

Plastid genome repeats include dispersed repeats and tandem repeats. Dispersed repeats are further divided into four types: forward, reverse, complement and palindromic repeats. REPuter software identified 2,368 repeated sequences, including 24–47 forward repeats, 7–16 reverse repeats, 21–31 palindromic repeats, 1–4 complement repeats, and 89–163 tandem repeats, in the 13 Engelhardia plastomes (Table S4, Fig. 3). Most of the tandem repeats existed in non-coding regions such as IGS and introns (Table S4, Fig. S3). Overall, tandem repeats were more prevalent in Engelhardia, accounting for about 60.52% of all repeat types. On the contrary, the complement repeats were relatively small, accounting for 1.01% (Table S4, Fig. S4).

Fig. 3
figure 3

Analysis of repeated sequences in 13 Engelhardia plastomes. A Statics of dispersed repeat sequences. B Statics of simple sequence repeats (SSR). C Statics of different types of SSR. D Statics of the overall proportion of different types of SSR

Full size image

In this study, statistical analysis of SSR was performed by MISA online software, and a total of 1530 SSR loci were detected in the 13 Engelhardia plastomes. The total number of SSRs varied little among individuals, ranging from 111 (E. roxburghiana_JFL02) to 127 (E. villosa). The majority of these plastid SSR (ptSSR) were mono-nucleotide repeats, accounting for 71.24% of all SSR, followed by di- (13.07%), tri- (5.69%), tetra- (4.97%) and trinucleotide repeats (4.64%), while hexanucleotide repeats was the least, accounting for only 0.39% (Table S4, Fig. 3). The A/T type mononucleotide was the most abundant SSR, accounting for 98.44%, and only 17 G/C single nucleotide repeats were detected, which also resulted in the enrichment of A and T in the plastomes. Most SSRs were located in the LSC region (72.88%), and a smaller percentage of SSRs were distributed in the SSC (19.67%) and IR (7.45%) regions, respectively. Furthermore, most of the SSRs (87.84%) were distributed in the IGS and introns, while only 12.16% in the coding sequences (Table S4, Fig. 4).

Fig. 4
figure 4

The distribution of simple sequence repeats (SSR) in 13 Engelhardia plastomes. A Statics of number of SSRs in the LSC, SSC, IR regions and in all the CDS. B Statics of the overall proportion of SSRs detected in different regions. C Statics of the overall proportion of SSRs detected in CDS and non-coding sequences

Full size image

Comparative analysis of genomic variation in the Carya, Engelhardia, and Juglans plastomes

Comparative analysis of nucleotide polymorphisms of Carya, Engelhardia, and Juglans plastomes, the variability in Engelhardia was higher than that of Carya and Juglans (Fig. 5). There were eighteen hypervariable regions in Engelhardia with Pi > 0.010 were trnH-trnK, trnK-rps16, rps16-psbK, trnG-atpI, rpoB-trnT, trnT-psbD, psbC-trnM, rps4-trnT, trnL-ndhJ, ndhC-trnV, petA-psbJ, psbE-rpl33, rps11-rps8, rps3-rpl2, trnN-ndhF, ndhF-ccsA, ndhA and ndhH-ycf1, while there were only seven and eight hypervariable regions in Carya and Juglans, respectively. Among them, trnK-rps16, ndhF-rpl32, and ycf1 were common high variation hotspots in these three genera (Fig. 5).

Fig. 5
figure 5

Nucleotide diversity and variation distribution of plastomes of Carya, Engelhardia and Juglans. The curved line depicts the fluctuation of π values across the genome alignment (dotted line marked the π values at 0.010), while the boxes below the curve represent the distribution of SNVs (top), Deletions and Insertions (bottom). The shadow layers in grey indicate the approximate range of IRs regions

Full size image

Using R. chiliantha as a reference, we characterized genomic variations including single nucleotide variants (SNVs), insertions and deletions (InDels) in the plastomes of Juglandoideae and Engelhardioideae and found that they were very different among different species (Table S5a). A total of 115,213 SNVs, 9502 insertions (1–274 bp) and 10,428 deletions (1–2,468 bp) were identified in all the collected species (Table S5d). The number of SNVs, deletions and insertions per kb varied at the plastid genome level, with the average values of 15.03, 1.36 and 1.24 in Juglandaceae, 11.84, 1.04 and 0.93 in Carya, 17.48, 1.48 and 1.64 in Engelhardia, and 17.20, 1.70 and 1.28 in Juglans, respectively. In terms of these three types of genomic variation, the IR regions presented the fewest numbers per kb, with the average values of 1.71, 0.15 and 0.09 in Juglandaceae, 1.81, 0.13 and 0.11 in Carya, 1.66, 0.20 and 0.10 in Engelhardia, and 1.70, 0.14 and 0.05 in Juglans, respectively. The LSC regions exhibited the maximum numbers of SNVs, deletions and insertions per kb with the average values of 9.04, 0.97 and 0.93 in Juglandaceae, 6.17, 0.70 and 0.63 in Carya, 10.62, 1.01 and 1.23 in Engelhardia being, and 11.30, 1.30 and 1.04 in Juglans (Table S5b). These results collectively indicate that the IR regions were more conserved than the single-copy regions.

All genomic structural variations were mapped onto a phylogenetic tree constructed based on the plastid genome, with very different times of insertion events and deletion events occurring in Carya (insertion events: 132–199 times; deletion events: 135–236 times), Engelhardia (186–364; 155–311), and Juglans (149–230; 192–331). So,the structural variation of Carya was less than that of Engelhardia and Juglans. The range of structural variation among Engelhardia species was relatively large, especially in E. serrata and E. villosa, with 329 insertions and 306 deletions in E. serrata, and 364 insertions and 311 deletions in E. villosa (Fig. S5).

The corresponding genomic positions of these identified InDels were mapped and located into these plastomes of Juglandoideae and Engelhardioideae. It was found that 90% of InDels were found in intronic (35%) and intergenic regions (55%) for Juglandaceae, 92% of InDels were found in intronic (43%) and intergenic regions (49%) for Carya, 88% of InDels were found in intronic (33%) and intergenic regions (55%) for Engelhardia, and 91% of InDels were found in intronic (31%) and intergenic regions 60%) for Juglans (Table S5c; Fig. 6).

Fig. 6
figure 6

The average number of SNVs, Deletions and Insertions located on the gene intervals, exons, introns, and RNA genes of plastomes of Carya, Engelhardia and Juglans

Full size image

Codon usage analysis of Juglandaceae plastomes

Comparing the coding sequences of 50 genes with a length over 300 bp of Juglandaceae plastomes, it was found that there were two codons with RSCU value of 1, which were AUG and UGG encoding methionine (Met) and tryptophan (Trp)(Table S6a). There were 29 codons with RSCU > 1, of which 16 end with U and 12 end with A, which was the same in Engelhardioideae, Juglandoideae and Rhoipteleoideae (Table S6a). The codons ending with U or A were the preferred codons in the plastomes of these three subfamilies (Fig. S6). There was no significant difference in the codon bias of the genes of most Juglandaceae plastomes (Fig. S6). However, the A/T content of the third base in the coding sequence was significantly higher than the G/C content, T3s (0.4748–0.4782) > A3s (0.4399–0.4438) > G3s (0.1695–0.1722) > C3s (0.1613–0.1649) (Table S6b). We found that Carya ovata and Carya palmeri had the highest values of ENC, GC3s and GC, while Platycarya strobilacea had the lowest values. No significant differences in codon preferences within genera were detected among these three subfamilies (Table S6b).

The codon usage pattern parameters, ENC, Fop, CBI and CAI of the coding genes of three subfamilies were further calculated and plotted (Table S6c). The CAI values were between 0.09 and 0.31, psbA, rbcL and psbD with the highest CAI value, and rpl20, rpl18 and rps8 with the lowest one. Most of the CBI values ranged from -0.23 to 0.23, the highest were psbA, psbD and rbcL, and the lowest were ndhF, ndhG and rps14. Most of the Fop values were between 0.26–0.55, the highest were psbA, psbD and rbcL, and the lowest were ndhG, ndhF and petD. Most of the ENC value were concentrated between 35.71 and 60.6, the highest were ycf3, ycf2 and rpl2, and the lowest were rps18, petD and rps14 (Table S6c).The highly expressed genes in the plastid genome of three subfamilies were ycf2, rpoC1 and rpoC2, and the low expressed genes are rps18, petD and rps14 (Table S6d). Combined with the 29 high-frequency codons with RSCU value > 1 in Table S4a, 10 common optimal codons were finally determined, which were CUU, GUU, UCU, UCA, CCU, CCA, GCU, AAU, CGA, GGA, and all end with A or U (Table S6d).

There was a positive correlation between the codon preference index (CBI) and the optimal codon usage frequency (Fop), and the highest correlation coefficient of 0.97 (Table S6e). The correlation coefficients between CAI and CBI and between CAI and Fop were also higher, which were 0.72 and 0.76, respectively, showing a positive correlation. In addition, there were negative correlation between T3s/C3s, T3s/A3s, T3s/G3s, T3s/GC3s, T3s/GC, C3s/A3s, C3s/G3s, A3s/G3s, A3s/CAI, A3s/CBI, A3s/Fop, A3s/ENC, A3s/GC3s, A3s/GC, G3s/CAI, G3s/CBI, G3s/Fop, CAI/GC, etc. Among them, A3s/CAI showed the highest degree of negative correlation, with a correlation coefficient of -0.57 (Fig. S7). The results of three subfamilies were similar to the whole Juglandaceae family, with the highest correlation coefficients being CBI and Fop, followed by the correlation coefficients between CAI and CBI, as well as between CAI and Fop (Fig. S7). The ENC values were positively correlated with T3s, C3s, G3s and GC3s. However, ENC values were negatively correlated with A3s. Our results showed that the content of the third base of synonymous codons was closely related to gene expression levels, and T3s, C3s and G3s were positively correlated with gene expression, while A3s was negatively correlated with gene expression (Table S6e, Fig. S7).

ENC values for all screened gene coding sequences ranged from 35.71 to 60.6. The ENC frequencies were calculated using the formula (ENCexp-ENCobs)/ENCexp, ranging from -0.25 to 0.28. There were 2051 ENC frequencies in the range of -0.1 to 0.1, which were close to the expected ENC values (Table S6f). Based on the standard curve formula ENC = 2 + GC3 + 29/[GC32 + (1 − GC3)2], we took ENC as the ordinate and GC3s as the abscissa to draw a scatter plot (Fig. 7). It was found that most of the genes were located on or near the standard curve (Fig. 7A). However, we also found the observed ENC values of six genes (rpl16, rps18, cemA, psbA, rps14, and ycf3) in all species deviated significantly from the standard curve (Fig. 7A, B). Among all genes, ycf3 showed the highest ENC value, while rps18 and rpl16 showed the lowest ENC value (Fig. 7B; Table S6f).

Fig. 7
figure 7

ENC and PR2-plots of protein-coding genes in plastomes of 50 Juglandaceae species. A ENC plots showing observed and expected and ENC values vs. GC3s values of protein-coding genes in these plastomes. B Comparison of ENC differences in two different climate zones. C PR2-plots showing the base composition characteristics of protein-coding genes in the 50 Juglandaceae plastomes. Red, genes of species from the Engelhardioideae species; Green, genes of species from the Juglandoideae species; Blue, genes of species from the Rhoipteleoideae species

Full size image

PR2-plot is used to analyze the composition of the four bases at the third position of the codon encoding the amino acid, plotting with G3/(G3 + C3) and A3/(A3 + T3) as the horizontal and vertical coordinates. The results showed that A/T and G/C (pyrimidine versus purine) were used slightly differently at the third codon position in the protein-coding sequence of Juglandaceae (Fig. 7C). The PR2-plot showed that there was a slight imbalance in the use of A/T and G/C at the third codon of the CDSs of 36 Juglandaceae plastomes, especially the four CDSs (psbA, rpl20, rpl16 and rps8) (Fig. 7C).The number of genes in the third and fourth quadrants was more than that in the first and second quadrants, and the number of genes distributed in the fourth quadrant was greater than the number of genes distributed in the other three quadrants, so G and T were used most frequently (Fig. 7C).

Selective pressure analysis of CDS in Juglandaceae

To analyze the evolutionary pressure among the protein coding sequences of eight Juglandaceae species, the Ka/Ks value of 80 protein coding sequences (CDS) were calculated. The results showed that the Ka/Ks values of 78 genes was almost all less than 1, with only rpl22 and psaI showing Ka/Ks > 1 We also found that rps16 was only subjected to positive selection in Juglandaceae and Engelhardioideae. For all Juglandaceae samples, the Ka/Ks values of photosynthesis-related genes were significantly lower than those of self-replication-related and other genes (Fig. 8A, Table S7b). For functional classification genes, except for differences in photosynthesis related genes between Engelhardioideae and Juglandoideae, other Ka/Ks values showed no significant differences (Fig. 8C, Table S7c).

Fig. 8
figure 8

Analyses of evolutionary pressure on plastid gene homologues in 50 Juglandaceae species. A A comparison of Ka/Ks values among photosynthesis-correlated genes, self-replication-correlated genes, and other protein coding genes in the three subfamilies. B A comparison of Ka/Ks values among gene homologues from the three subfamilies for photosynthesis-correlated genes, self-replication-correlated genes, and other protein coding genes. *, p < 0.05; **, p < 0.01; ***, p < 0.001,; NS, p > 0.05. C A heatmap showing the Ka/Ks values of CDS genes within the Juglandoideae, Engelhardioideae and Rhoipteleoideae subfamily

Full size image

Phylogenetic analysis of Juglandaceae

In this study, Quercus rubra (Fagaceae) was used as the outgroup, ML and BI trees of Juglandaceae species based on the complete plastomes (excluding one copy of the inverted repeat) showed nearly identical topologies (Fig. S8). The phylogenomic results indicated that Juglandaceae family was mainly divided into three groups, including Juglandoideae, Engelhardioideae and Rhoipteleoideae subfamilies, with a very high support rate (BS = 100%, PP = 1). The phylogenomic tree further supported 7 major branches, corresponding to 7 genera, namely the monophyletic Carya, Juglans, Pterocarya, Cyclocarya, Platycarya, Engelhardia and Rhoiptelea.

There were two main clades in Juglandoideae, clade I was Carya, clade II were Juglans, Pterocarya, Cyclocarya and Platycarya. In the ML tree, the support rate inside clade I (BS = 63–100%) was lower than that of clade II (BS = 66–100%). The species of Carya were divided into two sects, C. hunanensis, C. kweichowensis, C. sinensis, C. polianei, C. tonkinensis, and C. cathayensis were grouped together, while the remaining 12 species were grouped together. The Juglans were divided into three sects, namely the Sect. Juglans or Dioscaryon, Sect. Cardiocaryon, and the Sect. Rhysocaryon. The Sect. Juglans/Dioscaryon included J. regia and J. sigillata, while the Sect. Cardiocaryon included J. mandshurica, J. ailanthifolia and J. hopeiensis, the Sect. Rhysocaryon included J. cinerea, J. nigra, J. hindsii, J. major and J. microcarpa. The Pterocarya were divided into two sects, one included P. fraxinifolia, P. stenoptera and P. hupehensis, the other included P. macroptera var. insignis and P. tonkinensis. Cyclocarya paliurus was a single species of Cyclocarya, which was closely related to Pterocarya according to the phylogenetic relationships.

The species of the Engelhardioideae were closely related and were further divided into two main clades, which was consistent with the Sect. Engelhardia (Clade I) and Sect. Psilocarpeae (Clade II), with a very high support rate (BS = 100%, PP = 1). The Clade I included E. spicata, E. spicata var. rigida, E. hainanensis, E. serrata, E. anminiana, and E. villosa. The clade II included E. roxburghiana and E. fenzelii, which were sister species. The Rhoipteleoideae subfamily only included R. chiliantha, a single genus and single species.

The divergence time and historical diversification of Juglandaceae

By using multiple fossil correction points to estimate the differentiation time of the Juglandaceae family, the results showed that the crown nodes of Juglandaceae were approximately 97.69 Mya (95% highest posterior density (HPD): 95.49 Mya–100.58 Mya), which differentiated from Myricaceae in the Early Cretaceous period (Fig. 9). The three subfamilies, namely Rhoipteleoideae, Engelhardioideae and Juglandoideae, successively diverged at 89.28 Mya (95% HPD: 85.6–92.96 Mya; Middle Cretaceous) and 73.59 Mya (95% HPD: 69.01–78.13 Mya; Late Cretaceous) (Fig. 9).

Fig. 9
figure 9

The time-calibrated phylogenetic tree of Juglandaceae based on 80 protein-coding genes of plastomes. Estimated mean divergence times using a relaxed molecular clock model with 4 fossil priors (red stars). The blue bar on the node represents the estimated 95% HPD intervals around the mean divergence time. Nodes are numbered by ages. The genera and subfamilies of Juglandaceae are shown in the figure

Full size image

Most genera of Juglandaceae were diverged between 46.20 and 73.59 Mya. The differentiation time of the two clades in the Engelhardoideae subfamily was approximately between 27.64 and 46.11 Mya, mainly occurring from the Early Eocene to the Middle Oligocene. The phylogeny and divergence times didn’t support rapid radiation occurred in the evolution history of Engelhardia. In the Juglandoideae subfamily, the crown age of Carya was estimated at 64.98 Mya (95% HPD: 60.49–69.70 Mya), Platycarya at 60.51 Mya (95% HPD: 56.32–64.91 Mya) during the Late Paleocene, and Cyclocarya paliurus at 54.10 Mya (95% HPD: 50.84–57.41 Mya). The divergence of Pterocarya and Juglans was estimated at 46.29 Mya (95% HPD: 43.43–49.63 Mya) during the Middle Eocene. Most genera of Juglandoideae subfamily were diverged between 46.29 and 64.98 Mya in the relatively warm and dry climate of the Middle Paleocene to the Early Eocene (Fig. 9).

Discussion

Comparative analysis of Engelhardia plastomes

In this study, the plastomes of 13 individuals from eight species Engelhardia were newly sequenced, annotated and compared. The results showed that all species of the Engelhardia had a typical tetrad structure, and the genome sizes were similar, about 161 kb (161,069 bp–162,336 bp), and the GC contents of the plastomes were 35.8%–36.0%, which were similar to the sequence length and composition of the previously obtained Juglandaceae plastomes [1, 25,Codon usage bias and gene evolution in Juglandaceae plastomes

The codon usage pattern of the Juglandaceae plastomes plays an important role in exploring its evolutionary process [43]. In our study, codon bias of Juglandaceae plastomes were investigated. The frequency of usage of multiple synonymous codons encoding the same amino acid is not equal, a phenomenon known as codon usage bias [44]. The relative synonymous codon usage (RSCU) can directly reflect the preference of codon usage [45]. So RSCU values of all selected plastomes were calculated. We found that the third base of most codons ends with A or U (Table S6a), this result is consistent with the results of studies such as Crataegus [46], Pisum [47], and Miscanthus [48], indicating that the third base in the plant plastomes may have similar usage patterns [49]. By constructing high and low expressed gene sets, codons with RSCU > 1 and ΔRSCU > 0.08 were defined as optimal codons. Then, nine optimal codons of Juglandaceae plastomes were determined, all ending with A or U (Table S6d). In general, G and C (or A and T) are distributed proportionally at the third codon base, indicating that the codon usage bias of the species is affected by mutational pressure [50]; if disproportionately distributed over the third base of a codon, it indicates that the codon usage bias is influenced by natural selection pressure [51]. Therefore, it was speculated that the codon bias in the plastid genome sequence of Juglandaceae was not only affected by base mutations, but also affected by natural selection pressure. Genes from Engelhardioideae, Juglandoideae and Rhoipteleoideae species were presented in different colors in ENC and PR2-plots (Fig. 7). There was no significant potential difference in the main driving forces of codon use bias among these three subfamily plants (Fig. 7; Table S6).

We found a positive correlation between the codon preference index (CBI) and the optimal codon usage frequency (Fop), with the highest correlation coefficient of 0.97, indicating that the codon usage pattern in the plastomes of Juglandaceae may be determined by the optimal codon usage frequency during evolution [43]. In the Juglandaceae and two subfamilies (Juglandoideae, Engelhardioideae), the ENC values were positively correlated with T3s, C3s, G3s and GC3s. However, ENC values were negatively correlated with A3s. ENC values can be used to determine the relative expression levels of genes [52], so we inferred that the content of the third base of synonymous codons in Juglandaceae and two subfamilies (Juglandoideae, Engelhardioideae) was closely related to gene expression levels, and T3s, C3s and G3s are positively correlated with gene expression, while A3s was negatively correlated with gene expression (Table S6e, Fig. S5). In the Rhoipteleoideae subfamily, C3s, G3s, and GC3s were positively correlated with gene expression, while T3s and A3s were negatively correlated with gene expression (Fig. S6).

There were 50 protein-coding genes with lengths longer than 300 bp in the plastomes of Juglandaceae. The ENC values of these screened gene coding sequences ranged from 35.71 to 60.60. According to the ENC values ranging from 20 (completely biased) to 61 (unbiased) [53], and when the ENC value is less than 35, the codon usage of the gene or genome has a strong bias [54]. Based on these two characteristics, we found that the codon usage bias of protein-coding genes in Juglandaceae plastids was weak. There were 2051 ENC frequency ratios between -0.1 and 0.1 (Table S6f), which were close to the expected ENC value, indicating that the difference between the expected ENC value and the actual values of most genes are small. The results showed that the content of bases at the third position of synonymous codons was closely related to gene expression. The GC content of the third base of codons (GC3s) is considered to be most likely to directly reflect codon usage patterns [55] and may be an important factor leading to codon usage bias. The scatter plot was drawn with ENC as the ordinate and GC3s as the abscissa to explore the main features of codon usage (Fig. 7). When the scatter points are on or near the standard curve, it indicates that the codon preference is affected by mutational pressure, and vice versa, it indicates that the codon usage preference is affected by factors such as natural selection [56]. It was found that most of the scatter points were located on or near the curve (Fig. 7), indicating that the mutation had a greater effect on codon bias. Further ENC-plot analysis showed that the ENC values of most genes were close to the expected value (Fig. 7A), suggesting that the codon usage biases of these genes were related to GC3, and mutation was the main factor influencing factor. In addition, some genes (rpl16, rps18, and rps14) were much lower than the expected curve (Fig. 7), which also confirmed the influence of natural selection on codon preferences of these genes.

Due to the influence of natural selection and base mutation, PR-plot drawing analysis can show the preference of coding genes in the genome in the use of the third codon base. If the base mutation occurs in the third codon, the proportion of synonymous codons AT and CG in the gene or genome is equal. Conversely, if there is selective pressure, some codons "preferred" for translation will be used more frequently [57]. The PR-plot analysis of the Juglandaceae and three subfamilies showed that the selection of A/T and G/C at the third base of the protein-coding sequence was different, and the frequency of using G and T (purine) bases was higher(Fig. 7C), indicating that it was mainly influenced by selection pressure. Based on ENC-plot analysis and PR-plot analysis, natural selection and mutation jointly affect the codon usage patterns of Juglandaceae plastomes, with mutational pressure playing a major role, which is consistent with the results of Oncidium Gower Ramsey [58].

Synonymous and non-synonymous nucleotide substitution patterns are valuable for gene evolution studies [59]. Due to the effect of purification selection, the substitution rate of non-synonymous nucleotides is lower than that of synonymous nucleotides, so the ratio of Ka/Ks is less than 1 in most cases [60]. In order to gain a clear understanding of the adaptive evolution of the Juglandaceae plastomes, we calculated the Ka/Ks ratios of protein-coding genes [41]. Our results showed that only the the Ka/Ks ratio of ycf1 was greater than 1, and the Ka/Ks ratios of the remaining 79 gene was less than 1, indicating a strong purification selection pressure (Table S7a). We also noted that rps16 was only in positive selection in Juglandaceae and Engelhardioideae. As a self-replication related gene in plant plastid organelles, rps16 is essential for plant survival [61]. The positively selected rps16 gene may play a key role in the process of adaptation of Engelhardioideae species. There were differences in photosynthesis-related genes between Engelhardioideae and Juglandoideae subfamily (Fig. 8C, Table S7b), which may be due to differences of photosynthetic adaptation between temperate subfamily Juglandoideae and tropical subfamily Engelhardioideae [2]. The distribution of these genes to the plastomes indicated that most genes contained in the SSC and LSC regions experienced greater selection pressure than other plastid genomic regions, while the IR regions were more conserved. In addition, genes with different functions evolve with different rates, and the selection pressure of genes involved in photosynthesis in plastome is often lower than that of genes related to self-replication and other functions, resulting in differences in gene expression and function [62] (Fig. 8).

Phylogenetic relationships of Juglandaceae

Plant taxonomy are traditionally based on morphological characteristics, but the morphology is often affected by factors such as environment and parallel evolution [63], so molecular evidence is also needed. Based on nuclear genes and plastid gene fragments, predecessors have carried out related research on the phylogenetic relationship of Engelhardia [9, 10], but these plastid gene fragments do not have enough information to distinguish closely related species. In our study, ML and BI phylogenetic trees were constructed based on 50 accessions of Juglandaceae and 6 species from Myricaceae, Betulaceae and Fagaceae (Fig. S8). The phylogenetic tree constructed based on two different algorithms, ML and BI, presented almost identical topological structures.

The Juglandaceae was divided into three groups, including Juglandoideae, Engelhardioideae and Rhoipteleoideae subfamily [64], and had a very high support rate (BS = 100%, PP = 1) (Fig. S8). First, the five main branches of Juglandoideae subfamily correspond exactly to five genera, namely Carya, Juglans, Pterocarya, Cyclocarya, and Platycarya, all of which had high support rates (BS = 100%, PP = 1). According to the fruit morphology, these five genera were divided into two categories, including winged and wingless, namely Pterocarya, Cyclocarya, and Platycarya belong to winged types, while Carya and Juglans belong to wingless types [64]. According to the phylogenetic tree results, it was found that the phylogenetic relationship between the Juglans and Pterocarya was closer. Although the fruit morphology of the two genera was completely different, the fruit morphology of Carya and Juglans was similar, the phylogenetic relationship is distant [21, 65, 66]. Second, the species of Engelhardioideae subfamily were closely related and were divided into two main clades, which is consistent with the Sect. Engelhardia (Clade I) and Sect. Psilocarpeae (Clade II) [67] and get strongly supported (BS = 100%, PP = 1). The Clade I included E. spicata, E. spicata var. rigida, E. hainanensis, E. serrata, E. anminiana, and E. villosa. The clade II included E. roxburghiana and E. fenzelii, which were sister species. Third, R. chiliantha, the only species in Rhoipteleoideae subfamily, was located at the base of Juglandaceae in phylogenetic relationships and was also an endangered endemic species in China [3, 9, 68, 69].

Exploring the origin and evolutionary relationship of Juglandaceae

In previous studies, the crown age of Juglandaceae based on fossil data was approximately 84 Mya during the Cretaceous [70, 71]. Our results indicated that the divergence time of Juglandaceae was approximately 97.69 Mya (95% HPD: 95.49–100.58 Mya) using an older fossil time node from a fossil plant Budvaricarpus serialis (ca. 85 Mya) [72, 73]. The three subfamilies Rhoipteleoideae, Engelhardioideae, and Juglandoideae, successively differentiated at 89.28 Mya (95% HPD: 85.61–92.96 Mya) and 73.59 Mya (95% HPD: 69.01–78.13 Mya) (Fig. 9).

The Juglandoideae subfamily differentiated approximately between 69.01 and 78.13 Mya, from Cretaceous to Paleogene. The Northern Tropical Hypothesis [74, 75] provides a reasonable explanation for the origin and diversity of the Juglandoideae subfamily, that is during the warm Paleocene and Eocene periods, species of the Juglandoideae subfamily formed and rapidly diversified, spreading from North America to Europe and Asia through the North Atlantic Road Bridge and the Bering Land Bridge [76]. However, the global cooling that occurred after the extreme heat period of the Paleocene Eocene led to the extinction of most species [77,78,79]. The Cylocarya and Platycarya were endemic to East Asia, while Pterocarya were mainly distributed in the Caucasus and East Asia regions of southern Russia. Carya and Juglans have a wide distribution range in Eurasia, possibly due to their nutty fruit morphology, which facilitates animal transportation and transmission [3]. According to our results, the divergence time of Carya and Juglans was about 64.98 Mya, of Juglans and Pterocarya was about 46.29 Mya, of Pterocarya and Cyclocarya was about 54.10 Mya. Therefore, we inferred that differentiation events within the Juglandoideae subfamily occurred a long time ago and had undergone a long evolutionary process [21, 66].

The divergence time of the two clades in the Engelhardioideae subfamily was approximately between 27.64 and 46.11 Mya, mainly occurring from the Early Eocene to the Middle Oligocene. The earliest fossil record of the fruit of Engelhardia existed in South America and North America, and the oldest Alatonucula ignis fossil had been found in the early Eocene strata of Argentina [65]. At the same time, a fossil was found in the Miocene strata of Alaska, USA (Palaeocarya olsoni) [70]. This means that these taxonomic groups were widely present in parts of the Northern and Southern Hemispheres during the Eocene. Perhaps due to the high temperatures in the Paleogene, the species of Engelhardia were widely distributed in high latitude areas. Based on the discovery of the earliest fruit fossil of Palaeocarya in China in the Late Eocene strata of Hainan Island (Palaeocarya sp.) [80], it indicates that Engelhardia plants began to occupy tropical Asia in the late Eocene, while species diversity emerged in the Oligocene Miocene.

In short, our study accurately estimated the divergence time of Juglandaceae species using 80 coding sequences (CDs) from plastomes. And we found that Juglandaceae species had a complex evolutionary history and species diversity, which may be influenced by geographical changes, climate changes, and animal coevolution during the evolutionary process.

Conclusion

This study analyzed characteristics of the plastid genome of newly sequenced species of eight Engelhardia species, and clarified that the basic structure of the plastid genome was a typical tetrad structure. Three mutation hotspot regions were found and they can be used as potential molecular markers for inferring phylogenetic analysis and species identification. InDels may be an important driving factor for the plastome evolution of Juglandoideae and Engelhardioideae. Natural selection and mutation jointly affected the codon usage patterns of Juglandaceae and three subfamilies, with mutation pressure playing a major role. Phylogenetic results fully supported Engelhardia as a monophyletic group including two sects and the division of Juglandaceae into three subfamilies. Divergence time analysis revealed that Engelhardia originated in the later Cretaceous and diversified in the later Eocene, and Juglandaceae originated in the Early Cretaceous and differentiated in Middle Cretaceous. Overall, this study demonstrated that the plastome sequences displayed variable information to resolve phylogenetic relationships and were helpful to understand how species adapts to diverse ecological habitats.

Materials and methods

Plant materials and DNA extraction

In this work, a total of 13 individuals from all currently recognized eight species of Engelhardia (E. anminiana, E. fenzelii, E. hainanensis, E. roxburghiana, E. serrata, E. spicata, E. spicata var. rigida, and E. villosa) and one outgroup species Rhoiptelea chiliantha were collected in the tropical and subtropical Asia. The materials were identified by Yong-Hua Zhang, Hong-Hu Meng and Pan Li. The fresh leaves from each accession were dried with silica gel for further DNA extraction. Total high-quality genomic DNA was extracted from all plant materials using Plant DNAzol Reagent (Hangzhou Lifefeng Biotechnology Co., Ltd, Hangzhou, China). The detailed information of taxon, voucher number, collector and GenBank accession number are listed in Table 1.

DNA resequencing, plastomes assembly and gene annotation

High-quality genomic DNA from each sample was used for the whole genome sequencing (WGS) to obtain paired-end 100 bp raw reads on the BGISEQ-500 platform (BGI, Shenzhen, China) according to the manufacturers’ procedures. The quality of raw reads was controlled by removing the Phred score lower than 30, remaining the high-quality sequences for genome assembly using the GetOrganelle software [81]. The command lines used for the assembly were as follows: get_organelle_reads.py -1 forward.fq -2 reverse.fq -o plastome_output -R 15 -k 21,45,65,85,105 -F plant_cp. All targeted plastomes sequences were concatenated and manually edited with Geneious Prime 2021 software (http://www.geneious.com/), using the plastome sequences of Carya sinensis (MN892516) and Rhoiptelea chiliantha (MT701585) as the reference genomes. At the same time, the CPGAVAS2 web server (http://www.herbalgenomics.org/cpgavas) was used to predict the types and structures of all the protein-coding and non-coding genes in the plastomes. The final plastomes annotation was determined by comparing the results of Geneious Prime 2021 and CPGAVAS2. Finally, CPGView [82] were used to visualize the plastome map. The 13 newly generated complete plastome sequences were deposited in GeneBank (Accession numbers were listed in Table 1). Plastomes of 43 other species were downloaded from NCBI GenBank repository and re-annotated using the earlier method, the GenBank accession numbers are shown in Table S1.

Comparative analysis of plastome structure features

We used these newly-sequenced Engelhardia individuals to study the genomic variation in Engelhardia. Two methods were used for comparative genomic analysis: (1) The comparison of the plastomes sequence identity was using MAVUE and mVISTA [83]. Sequence rearrangement detection was performed on 13 plastomes using the Mauve alignment plugin in Geneious Prime 2021 software, and 13 sequences were aligned using the LAGAN model in the online software mVISTA. (2) The comparison of the expansion and contraction of IR regions was presented. The IR boundary regions were visualized by using the online website IRScope (https://irscope.shinyapps.io/irapp/).

Repeat sequences detection

The genome of organisms, especially higher organisms, contains a large number of repeat sequences, which can be divided into dispersed repeat sequence (DRS) and tandem repeat sequence [84] according to their distribution patterns in the genome. First, the DRS in the plastomes of eight Engelhardia species were predicted by the REPuter software [85]. The forward, reverse, palindromic and complement repeat sequences were identified using the following parameters: length of repeat unit ≥ 30 bp, sequence consistency ≥ 90% (Hamming distance = 3). Then, the TRS in the plastomes were predicted by using the Tandom Repeats Finder (TRF) web server (https://tandem.bu.edu/trf/trf.html). Finally, the MISA software was used to identify simple sequence repeats (SSR), setting the minimum repetition threshold values for mono-, di-, tri-, tetra-, penta-, and hexa-nucleotide were set to 10, 5, 4, 3, 3, 3, respectively.

Analysis of nucleotide polymorphism and mutation sites

We analysed nucleotide polymorphisms based on the Pi values of Carya, Engelhardia, and Juglans, respectively. The plastomes were aligned using mafft alignment with default settings in Geneious Prime 2021. The Pi of protein-coding genes, noncoding genes, and the intergenic regions extracted from the plastome was calculated using DnaSP v6.0 [86] to show the nucleotide diversity at genus level. In order to eliminate interference from different individuals of the same species, we only chosen E. hainanensis_HN01, E. fenzelii_TTD01 and E. roxburghiana_BPZ11 accessions on behalf of E. hainanensis, E. fenzelii, and E. roxburghiana, respectively. The parameters were set as: window length = 600 bp, step size = 200 bp. After that, the corresponding loci were located and counted in Generous Prime 2021 software, and sequence fragments with Pi values greater than 0.01 were used as candidate high variation regions.

In order to conduct comprehensive comparison of genomic variation of Carya, Engelhardia, and Juglans plastomes, we calculated their total number, length, and percentage of SNVs and insertion and deletions (InDels) sites located in gene intervals, exons, introns, and RNA genes. For Engelhardia species, we only kept one individual from the same species to conduct nucleotide polymorphisms analysis. To draw density bar graphs of SNVs, InDels data, we used the Genome Varscan plugin in TBtools, and the detection parameters were set as: the number of threads (CPU) to 2, the genome sequence divergence standard (Diff) to OneIn Thousand, VarRange from 0 to 1,000,000. Carya, Engelhardia, and Juglans were aligned with R. chiliantha, which was selected as the reference sequence, then variation site information was outputted.

Codon usage bias

Different species use different codons with different frequencies, and there will be certain preferences [87, 88]. Studying the differences in codon usage patterns among families or genera can help us effectively understand the genetic evolution patterns of species. In addition, exploring the codon usage patterns of plant plastomes is beneficial to explore the adaptive mechanisms of plants under different evolutionary patterns [45]. In addition to the overall analysis of Juglandaceae, we also analyzed Engelhardioideae, Juglandoideae, and Rhoipteleoideae. The consensus genes encoding proteins with a sequence length longer than 300 bp and the start codon of ATG were used to analyse codon bias. T3s, C3s, A3s, G3s, CAI, CBI, Fop, ENC and GC values were calculated using CodonW software (http://codonw.sourceforge.net/). And plot according to the calculated correlation values: (I) Plot with ENC as the y-axis and GC3s as the x-axis to evaluate the influence of base composition on codon usage bias, the observed ENC value was compared with the expected ENC value using the following equation: ENC = 2 + GC3s + 29/[GC3s2 + (1—GC3s)2]; (II) Using [A3/(A3 + T3)] as the y-axis, with [G3/(G3 + C3)] as the x-axis, draw a coordinate map to assess the effects of genetic mutation and natural selection on codon usage preferences. All the screened genes were sequenced according to ENC value as a whole. The upper and lower 5% gene samples were selected and defined as low expression group and high expression group, and the RSCU value of each group was calculated. The RSCU difference between low expression group and high expression group was calculated. The codon with RSCU > 1 and △RSCU > 0.08 is defined as the optimal codon.

Phylogenetic relationships of Juglandaceae

Plastomes were aligned using MAFFT v 7.308 [89] implemented in Geneious Prime 2021. Maximum Likelihood (ML) analysis and Bayesian Inference (BI) were employed for the phylogenomic reconstruction of Juglandaceae. The best-fit nucleotide substation model for ML and BI analysis was determined by Modeltest v3.7 [90], in which the complete plastomes data was GTR + I + G and BI analyses were performed using the RAxML-HPC v8.1.11 and MrBayes v3.2.3 online tools available from the CIPRES Science Gateway web server [91, 92]. The ML analysis was conducted with default settings with 1000 bootstrap replicates. BI trees were produced with the setting of 5,000,000 generations, under GTRGAMMA model with one cold and three incrementally heated Markov Chain Monte Carlo (MCMC) run simultaneously [92] in two parallel runs sampling every 1000 generations. The first 25% of the trees were discarded as burn-in. The remaining trees were used for generating the consensus tree.

Evolutionary analysis of Juglandaceae plastomes

We observed Ks (synonymous), Ka (non-synonymous) substitution, and Ka/Ks ratio using pairwise alignment of protein-coding sequences of Morella rubra and other 50 selected species of Juglandaceae. We used M. rubra as the reference in each pair of alignments to make pairwise alignments with every gene. Geneious Prime 2021 was used to extract 80 common protein-coding genes, and DnaSP v6.0 was used to calculate the Ka and Ks substitutions. In addition, in order to detect the selection pressure on whole plastid genes with different functions, CDS genes were divided into photosynthesis-related, self-replication-related, and other functional genes (Table S1). Finally, we plotted boxplot graphs of the Ka/Ks values of CDS genes based on different functional classifications or taxonomic groups, and marked the significance of the differences between the groups. All analyses were conducted in R version 4.3.0 (https://www.R-project.org/).

Divergence-time estimation and fossil calibration

We estimated the divergence time of Juglandaceae species based on 80 coding sequences (CDSs), combined with the calibration of four fossils (Table S8) [24, 70,71,72,73, 93,94,95,96]. The nucleotide substitution model is the same as the MrBayes parameters mentioned above. Before setting to Yule Process Specialty Tree model, set the molecular clock to log normal relaxation Molecular clock. For the MCMC program, the chain length was 5 × 108 generations, sampling every 10,000 generations. All options were set in BEAUTi v1.10.4, exported as an XML file, and run in BEAST v1.10.4 [97]. We checked the convergence of the Markov chains in Tracer v.1.6 (http://beast.bio.ed.ac.uk/Tracer/) and combined the chains after removing a burn-in of the first 50% generations. The effective sample sizes (ESS) exceeded 200 for all parameters. The program FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/) was used to visualize mean node ages and highest posterior density (HPD) intervals at 95% (upper and lower) for each node and to estimate branch lengths and divergence times.