Whole-Genome Sequencing of Acer catalpifolium Reveals Evolutionary History of Endangered Species

Abstract Acer catalpifolium is an endangered species restricted to remote localities of West China. Understanding the genomic content and evolution of A. catalpifolium is essential to conservation efforts of this rare and ecologically valuable plant. Here, we report a high-quality genome of A. catalpifolium consisting of ∼654 Mbp and ∼35,132 protein-coding genes. We detected 969 positively selected genes in two Acer genomes compared with four other eudicots, 65 of which were transcription factors. We hypothesize that these positively selected mutations in transcription factors might affect their function and thus contribute to A. catalpifolium’s decline-type population. We also identified 179 significantly expanded gene families compared with 12 other eudicots, some of which are involved in stress responses, such as the FRS–FRF family. We inferred that A. catalpifolium has experienced gene family expansions to cope with environmental stress in its evolutionary history. Finally, 109 candidate genes encoding key enzymes in the lignin biosynthesis pathway were identified in A. catalpifolium; of particular note were the large range and high copy number of cinnamyl alcohol dehydrogenase genes. The chromosome-level genome of A. catalpifolium presented here may serve as a fundamental genomic resource for better understanding endangered Acer species, informing future conservation efforts.


Introduction
The genus Acer (maple) belongs to the family Sapindaceae and consists of nearly 200 species (Piao et al. 2020). Maples form some of the largest broad-leaved deciduous tree population in eastern Asia, North America, and Europe (Wolfe and Tanai 1987;van Gelderen et al. 1994). Maple trees are key resources for commercial products; their bark is used for sugar and dye, ashes for soap, and trunks for pulpwood and lumber (Gelderen and Gelderen 1999). Species in the Acer genus have a diverse range of leaf morphologies, as well as complex reproductive systems with unisexual, polygamodioecious, or androdioecious flowers (Shang et al. 2012;Rosado et al. 2018). Thus, these species are ideal specimens for the study of variation in leaf morphology and plant sexual diversity. Populus trichocarpa (black cottonwood), a well-known deciduous broadleaf tree, is usually used to make crates, pallets, and paper; comparatively, some maples (like bigleaf maple and silver maple) have different mechanical properties, including larger values for modulus of elasticity and modulus of rupture, which means that they perform better than P. trichocarpa in terms of both wood strength and stiffness (Ross 2010). According to the U.S. Department of Agriculture, a large proportion of maple lumber is further manufactured into a variety of products, including flooring and furniture (Betts 1959), making Acer an ideal genus to study properties of wood.
Maple trees mainly grow in temperate and subtropical areas of the Northern Hemisphere, and there are some species that are native to Asia. For example, in China, some Acer species are widely distributed (A. amplum, A. longipes, A. mono, and A. truncatum), whereas others are endangered (A. catalpifolium, A. miaotaiense, and A. yangjuechi) (Bi et al. 2016). Acer catalpifolium is a deciduous tree that is under threat of extinction (Yu et al. 2020) and is narrowly distributed in the rainy zone of western China (28 30 0 -33 N, 102 30 0 -104 E). Acer catalpifolium has a straight tree trunk and umbrella-shaped crown and usually grows up to 25 m tall ( fig. 1A). It has papery leaves with ovate or ovate-oblong leaf blades, yellow-green flowers, yellowish, and glabrous samaras with spreading wings (fig. 1B-E) (Xu et al. 2008). Because of A. catalpifolium's beautiful appearance, local people in the western Sichuan province sometimes plant them along roadsides (FOC, http://www.iplant.cn/, last accessed December 13, 2021).
The population size of A. catalpifolium has decreased steadily. In 2009, it was listed in the Wild Plants with Extremely Small Populations (WPESP) Rescue and Protection Plan (The State Forestry Administration of The People's Republic of China 2009). Scientists have carried out many studies on A. catalpifolium to understand this population decline. The possible negative effects of decreased seed germination, shade from neighboring dominant species, and severe human disturbance on A. catalpifolium growth and biomass accumulation have been researched Song and Liu 2020). Moreover, population structure analysis shows that A. catalpifolium has a tendency to convert from a stable population to a decline-type population (Zhang et al. 2018;Xu and Liu 2019). Therefore, it is imperative to understand why the population of A. catalpifolium continues to decrease and how we might be able to intervene, interrupt, or even reverse this population decline. Traditional studies on plant phenotypes and the surrounding environment have begun to tackle these questions, and to continue this research it is essential to assemble a high-quality genome. This will allow us to investigate the possible genetic causes of population decline in A. catalpifolium and inform future research to recover the population.
Here we report a high-quality, chromosome-level genome assembly of A. catalpifolium using long-and short-read sequencing methods. The genome data allowed us to understand the molecular evolutionary history of A. catalpifolium, especially genome structure variation and gene family evolution. Comparisons between the genome of A. catalpifolium and the previously published A. yangbiense genome (Yang et al. 2019) provided us an opportunity to directly identify genomic differences between the two endangered species and hypothesize about the impact of these differences on population size. Some positively selected genes in A. catalpifolium contained mutations in functional domains, which may cause gene functional deficiency. Because a highquality genome of A. catalpifolium is now available, additional omics analyses can be performed to help improve the survival rate of the tree and save this endangered plant from extinction, which is important for the conservation of biodiversity on Earth.

Genome Assembly and Annotation of A. catalpifolium
By integrating PacBio Sequel long reads ($34.39 Gb), Illumina short reads ($34.40 Gb), and Hi-C paired-end reads ($45.55 Gb), we assembled a chromosome-level reference genome of A. catalpifolium (supplementary tables S1 and S2, Supplementary Material online). The genome size of A. catalpifolium was estimated to be about 650.47 Mb based on 19-mer frequency distribution (supplementary fig. S1, Supplementary Material online). PacBio long reads were assembled de novo, and assembled contigs were polished with Illumina short reads (supplementary fig. S2). The resulting genome assembly comprised 3,112 contigs with a contig N50 value of $0.66 Mb. The sequences were further scaffolded with Hi-C reads. Ultimately, a genome assembly of 654.51 Mb in length was produced, consisting of 1,978 scaffolds (with a scaffold N50 of 37.58 Mb) (  Tracks I through V represent the name and size of the 13 chromosomes, gene density, repeat density, GC content density, and collinear blocks, respectively, which were calculated in 500-kb windows. To understand the evolutionary history of Acer, we performed inter-and intragenomic comparisons among A. catalpifolium, A. yangbiense, and Vitis vinifera. The synonymous substitutions per synonymous site (Ks) distributions of the orthologs and paralogs were investigated. With intragenomic comparisons, we detected 3,311, 2,445, and 2,288 syntenic paralogs in A. catalpifolium, A. yangbiense, and V. vinifera, respectively. The Ks distribution plot showed significant peaks at approximately 1.4 for paralogous gene pairs in both A. catalpifolium and A. yangbiense, and at 1.1 for V. vinifera paralogs. The peak for Ks distribution of A. catalpifolium and V. vinifera orthologs was at approximately 0.9, and the peak for A. catalpifolium and A. yangbiense orthologs was at approximately 0.05 ( fig. 2A). The Ks analyses showed that the most recent whole-genome duplication (WGD) event in A. catalpifolium, as well as in A. yangbiense, was the wellstudied gamma event shared by core eudicots (Jaillon et al. 2007), providing no evidence for further lineage-specific WGD in Acer. To further investigate this, we also performed intergenomic syntenic comparisons using MCScanX with default parameters ). We found a 1:1 syntenic depth ratio between V. vinifera to A. catalpifolium and A. yangbiense, which supported the findings from the Ks analyses; these results indicated that there was not a lineagespecific WGD after the divergence of A. catalpifolium, A. yangbiense, and V. vinifera ( fig. 2C and D). Therefore, the genome of A. catalpifolium that lacks further WGD events since the origin of core eudicots may be representative of the ancestral genome of eudicots, facilitating investigation of their subsequent genome evolution.
Furthermore, A. catalpifolium and A. yangbiense genomes were compared with the previously reconstructed ancestral eudicot karyotype (AEK) genome (Murat et al. 2017). In A. catalpifolium, 5,153 genes had a syntenic relationship with AEK, whereas in A. yangbiense, 6,869 syntenic genes were identified. Although there were fewer A. catalpifolium genes in syntenic regions compared with A. yangbiense, a clearer 3:1 syntenic depth ratio was inferred in A. catalpifolium-AEK compared with A. yangbiense-AEK ( fig. 2B). Then, 20,505 orthologous gene pairs were plotted, indicating a high similarity of genome structures between A. yangbiense and A. catalpifolium. Chromosome-level macrosynteny further suggests a high level of conserved genome structure between the two species and few small-scale genomic rearrangements ( fig. 2C and D). This is likely due to the relatively short time of divergence between the two Acer species, discussed below. We also identified 3,299 and 3,653 tandem duplications in A. catalpifolium and A. yangbiense, respectively. When compared with V. vinifera, A. catalpifolium shared 25,855 syntenic gene pairs, and A. yangbiense shared 23,242 syntenic gene pairs. After filtering, 9,631, 9,200, and 15,529 best "one-to-one" syntenic matchings were obtained in V. vinifera-A. catalpifolium, V. vinifera-A. yangbiense, and A. catalpifolium-A. yangbiense, respectively. We integrated "one-to-one" reciprocal best BLASTP hits between A. catalpifolium and A. yangbiense with the previous results, and found that 395 genes were lost specifically in A. catalpifolium, whereas 848 genes were lost specifically in A. yangbiense. The majority of the genes that were well conserved amongst the three species provide an estimation of the ancestral eudicot gene content. The different numbers of lost genes between the two Acer species indicated different histories of genomic sequence evolution.

Gene Family Evolution and Gene Expansion
To better understand the gene content and gene family evolution in A. catalpifolium, we used OrthoMCL v14.137 with default parameters (Li et al. 2003) to reconstruct the gene families of 13 angiosperms, including nine eudicots (A. catalpifolium, A. yangbiense, Arabidopsis thaliana, Aquilegia coerulea, Citrus sinensis, Dimocarpus longan, P. trichocarpa, Solanum lycopersicum, V. vinifera), three monocots (Ananas comosus, Musa acuminata, Oryza sativa), and an outgroup of one basal angiosperm, Amborella trichopoda. A total of 29,299 orthologous groups (orthogroups) were identified, 666 of which were specific to A. catalpifolium and 377 of which were specific to A. yangbiense. In addition, we found 781 gene families specific to Acer and 158 specific to Sapindales.
To understand the patterns of gene family gains, losses, expansions, and contractions following the diversification of angiosperms, we first constructed a phylogenetic tree using 873 single-copy gene families from the 29,299 aforementioned orthogroups ( fig. 3A). The single-copy genes were gathered to generate a supermatrix and put into RAxML (Stamatakis et al. 2005) for phylogeny analysis. Next, we estimated the divergence time using r8s (Sanderson 2003) with four fossil calibrations (see Materials and Methods for details). We found the divergence time between A. catalpifolium and A. yangbiense to be approximately 6.9 million years ago (Ma) and the time of divergence between the Acer genus and D. longan to be approximately 29.5 Ma ( fig. 3A), which was consistent with previous research (Yang et al. 2019). We also extrapolated the divergence of the two Acer species to be approximately 5.  (Han et al. 2013). Of these 220 orthogroups, 179 were significantly expanded and 41 were significantly contracted (P value <0.01).
Based on GO enrichment analysis using clusterProfiler (Yu et al. 2012), 2,156 A. catalpifolium genes in the 179 expanded orthogroups were highly enriched in "ATPase activity, coupled to transmembrane movement of substances" (P value ¼ 4.34EÀ35) and "calcium ion transmembrane transport" (P value ¼ 6.53EÀ23) ( fig. 3B; supplementary figures S4, S5 and table S7, Supplementary Material online). It is important to note that ATPase and calcium ion transport are functionally related to phototropin-related light signal transduction pathways (Dodd et al. 2010;Inoue and Kinoshita 2017). Moreover, 159 Acer-specific A. catalpifolium genes in the significantly expanded orthogroups were highly enriched in "response to cadmium ion" (P value ¼ 4.87EÀ11) and "response to stimulus" (P value

Evidence of Positive Selection Across the Acer Genus at the Whole-Genome Level
In domesticated species, it has been shown that decreased effective population size has led to an increase in the    The top ten GO terms were selected for display after using Benjamini-Hochberg multiple test correction for P value adjustment (adjusted P value <0.05). (C) Visualization of the results from the GO enrichment analysis of 159 Acer genus-specific significantly expanded genes. The top ten GO terms were selected for display after using Benjamini-Hochberg multiple test correction for P value adjustment (adjusted P value <0.05).
frequency of slightly deleterious variants near sites of positive selection (Liu et al. 2017). Given the declining trend of A. catalpifolium's population size (Xu and Liu 2019), we investigated whether this association between positive selection and deleterious mutations exists in A. catalpifolium. The number of nonsynonymous substitutions per nonsynonymous site (Ka) was divided by the number of synonymous substitutions per synonymous site (Ks) to calculate x. A super matrix was generated from single-copy orthogroups in the 13 selected species; the super matrix was then used to estimate x using the "free-ratio" model of codeml in PAML with default parameters (Yang 2007). Our results clearly showed that the x values of Acer species were generally larger than 0.23, whereas the x values of other eudicot species were generally smaller than 0.20 (supplementary fig. S7, Supplementary Material online). In addition, we performed the same x analysis with six additional eudicots (A. thaliana, P. trichocarpa, C. sinensis, D. longan, A. catalpifolium, and A. yangbiense) using Arabidopsis as an outgroup (Xu et al. 2017). The results revealed a similar pattern in which Acer species, in particular A. yangbiense, had larger x values than other species ( fig. 4A). Lastly, we estimated the x values for each of the 6,116 modified single-copy orthogroups (see Materials and Methods for details) in the same six eudicots mentioned above using the "branch" model in PAML. Notably, 5,261 of the 6,116 orthogroups showed a larger x ratio in the Acer lineages, 2,300 of which were statistically significant (P value < 0.05, v 2 test) ( fig. 4B). Our results showed that x is elevated in the Acer clade relative to the other examined species.
Both positive selection and relaxation of negative selection can lead to a higher x ratio (Xu et al. 2017). To test if A. catalpifolium or A. yangbiense has experienced positive selection, the 6,116 aforementioned orthogroups in six eudicots were used to perform "branch-site" model analysis. In total, 969 orthogroups were identified that harbor codons with x > 1 in the Acer clade (P value <0.05), 683 of which showed statistically significant evidence of positive selection (P value <0.01) (supplementary  Six genes belonged to one orthogroup annotated as the NAC gene family. One asterisk above the alignment means that the site was possibly under positive selection in A. catalpifolium (P value <0.05). Two asterisks above the alignment means that the site showed highly significant evidence of positive selection in A. catalpifolium (P value <0.01).
selected sites in the protein domain, representing different amino acids in A. catalpifolium or A. yangbiense compared with the other five species. The NAC (also named NAM, "No Apical Meristem") family is one of the largest plant-specific TF families and is involved in embryonic, floral, and vegetative development, as well as defense and abiotic stress (Olsen et al. 2005). In the NAC gene family, the Arabidopsis gene NAC DOMAIN CONTAINING PROTEIN 20 (NAC020, AT1G54330.1) has been proposed to be an early regulator of phloem sieve element (SE) differentiation (Kondo et al. 2016). The NAC020 orthologous gene in A. catalpifolium (EVM0008503.1) was predicted to have six positively selected sites in the N-terminal of the NAM domain ( fig. 4D). We compared the predicted 3D protein structure of these two genes using the trRosetta webserver (Du et al. 2021) and found a missing helix structure in

Evolutionary History of FRS-FRF Gene Family
Considering that the natural habitat of A. catalpifolium has high canopy coverage and limited access to sunlight (Zhang et al. 2018), we specifically examined the gene families that function in shade avoidance syndromes. It has been reported that the FAR1-RELATED SEQUENCE (FRS) and FRS-RELATED FACTOR (FRF) gene families are involved in multiple shadeavoidance responses Xie et al. 2020). In Arabidopsis, FRS-FRF genes are reported to be regulated by far-red light (Lin and Wang 2004;Lin et al. 2007). We identified FRS-FRF family members in A. catalpifolium and A. yangbiense based on sequence similarity and protein domain conservation using the 18 FRS-FRF genes in Arabidopsis as references (Ma and Li 2018). In total, we identified 93 and 68 members of the FRS-FRF gene family in the A. catalpifolium and A. yangbiense genomes, respectively.
Amino acid sequences of the FRS-FRF family members in Arabidopsis, A. catalpifolium, and A. yangbiense

Lignin Biosynthesis
A previous study in poplar showed that low-lignin wood has reduced strength and stiffness (Voelker et al. 2011). Therefore, we investigated the genetic basis of lignin biosynthesis in P. trichocarpa and A. catalpifolium, with the aim of understanding why these two trees have different wood traits. Following a previous study on lignin biosynthesis (Vanholme et al. 2019), we used the genome sequence of A. catalpifolium together with P. trichocarpa and A. yangbiense to investigate the genes encoding ten key enzymes in the lignin biosynthesis pathway (fig. 6). The Ensemble Enzyme Prediction Pipeline (E2P2) was used to predict lignin biosynthesis pathway genes in the three plants (Schlapfer et al. 2017). A total of 109 genes were identified in the A. catalpifolium genome as participating directly in lignin biosynthesis (supplementary fig. S15 and table S13, Supplementary Material online), including 12 phenylalanine ammonia-lyase (PAL) genes, two cinnamate 4-hydroxylase (C4H) genes, four 4-coumarate: CoA ligase (4CL) genes, ten cinnamoyl-CoA reductase (CCR) genes, 37 cinnamyl alcohol dehydrogenase (CAD) genes, 22 p-hydroxycinnamoyl-CoA: quinate/shikimate (HCT) genes, three p-coumarate 3-hydroxylase (C3H) genes, five caffeoyl-CoA o-methyltransferase (CCoAOMT) genes, eight ferulate 5-hydroxylase (F5H) genes, and six caffeic acid o-methyltransferase (COMT) genes. There were 99 such genes in P. trichocarpa and 96 in A. yangbiense (supplementary tables S14 and S15, Supplementary Material online).
Notably, eight CAD genes in A. catalpifolium were found in the significantly expanded orthogroups, whereas only two 4CL genes in P. trichocarpa were found in the significantly expanded orthogroups, and no gene was found in A. yangbiense. The expansion of CAD family may contribute to the higher lignin concentration in A. catalpifolium compared with P. trichocarpa. Because lignin performs important roles in stem integrity, hydraulic conductivity, and biotic/abiotic defenses, altering both lignin content and syringyl/ guaiacyl (S/G) composition are important for the adaptation of trees to the environment (Porth et al. 2013). The predicted genes encoding F5H and COMT enzymes in A. catalpifolium make the modulation of the S/G ratio possible (Boerjan et al. 2003), which may further improve the adaptation ability of A. catalpifolium. In addition, these findings provide an opportunity to improve wood traits with lignin engineering.

Discussion
A. catalpifolium is endangered, but because genomic information was previously unavailable, conservation efforts for this tree have been limited to traditional methods such as natural interventions (Song and Liu 2020). In this study, we generated a high-quality genome of A. catalpifolium, thus accelerating future research and aiding in conservation of this tree. With the genomes of two endangered Acer species now available, we estimated their evolutionary divergence using two methods. One method was construction of a phylogenetic tree with estimated branch lengths and some fossil evidence for calibration. The other method was to utilize the formula k ¼ Ks/2T to estimate divergence time, supposing that A. catalpifolium, A. yangbiense, and V. vinifera shared

S)
A c y a n 0 2 G 0 1 4 5 9 0 0 .1 the same mutation rate with their common ancestor, which experienced the gamma WGD event. In our study, the second method estimated a younger divergence time (5.2-5.9 Ma) than the first (6.9 Ma), which means these two methods were based on different rates of nucleotide substitution.
Previously, researchers found that a biodiversity hotspot in the Hengduan Mountains region has uplifted in the last eight million years (Xing and Ree 2017), which might have facilitated the divergence of A. catalpifolium and A. yangbiense and their speciation. Given that both A. catalpifolium and A. yangbiense are endangered, we reconstructed the demographic history of A. catalpifolium and A. yangbiense using a pairwise sequentially Markovian coalescent (PSMC) model (Li and Durbin 2011), and found that populations of   6.-Comparison of the copy number of genes encoding ten key enzymes involved in lignin biosynthesis. In every reaction depicted, the name above the arrow represents the short name of the enzyme, and the histogram beneath the arrow represents the copy number of genes encoding the enzyme.
A. catalpifolium and A. yangbiense show roughly the same trend: decreasing steadily after the Last Glacial Maximum (LGM) (supplementary fig. S16, Supplementary Material online). More information such as survival rate, climate change data, and plant-environment interactions should be included in a population fluctuation simulation. Furthermore, studies on population genomics and Genome-Wide Association Study (GWAS) are needed to investigate the genetic structure and specific genetic variations of A. catalpifolium, which may help to identify key factors that affect the effective population size.
The PSMC result is partially confirmed by our identification of PSGs in both A. catalpifolium and A. yangbiense, some of which had positively selected sites in the protein domain and led to altered 3D protein structure. Previous research has shown that abnormal expression of AT1G54330.1 (orthologous gene of the PSG EVM0008503.1) caused severe root growth defects and discontinuous SE differentiation in roots (Kondo et al. 2016). We hypothesize that protein structural changes would affect biological functions of some PSGs in both trees. This might have adverse impacts on tree growth and development, affecting the survival of the endangered A. catalpifolium and A. yangbiense. We additionally performed GO enrichment analysis on genes in the 41 orthogroups that were significantly contracted in A. catalpifolium. Notably, they were highly enriched in catalytic activity and binding functions, such as "organic phosphonate transmembrane-transporting ATPase activity" (P value ¼ 5.71EÀ08) and "transition metal ion binding" (P value ¼ 2.31EÀ05) (supplementary fig. S17, Supplementary Material online). This suggests that some genes related to basic molecular functions were lost after the divergence of the two Acer trees, which might also contribute to the declining populations of A. catalpifolium. Further genetic and transgenic approaches are needed to experimentally test these hypotheses.
A. catalpifolium is located in the rainy zone of western China (Yu et al. 2020). As part of the Hengduan Mountains, this region has a special climate and unique vegetation types, with perennial rain and insufficient sunshine duration (Zhuang and Gao 2002). Previous studies have shown that low illumination intensity has adverse effects on A. catalpifolium sapling growth (Zhang et al. 2018). Because FRS genes play an important role in shade-avoidance responses, the significant expansion of the FRS-FRF gene family in A. catalpifolium might have helped the tree to adapt to the rainy zone of western China. Therefore, experiments such as overexpressing specific FRS genes are needed to test whether these genes have the potential to raise the survival rate of A. catalpifolium, which could save the tree from extinction. Future studies could combine mutant phenotype analysis with transcriptomic analysis to identify additional genes that are involved in important biological processes such as seed germination, photosynthetic efficiency, lignin content engineering, and biotic/abiotic stress responses.
With the genome of A. catalpifolium now available, biological methods such as CRISPR-Cas9 genome editing (Chen et al. 2019) can be used to transplant novel genes into A. catalpifolium that may improve the tree's capacity for adaptation.
A. catalpifolium exhibits no further lineage-specific WGD events after the well-acknowledged gamma event shared by core eudicots (Jaillon et al. 2007). WGD events may serve as important factors in improving adaptive evolution by creating a large amount of raw genetic material (Van De Peer et al. 2017). Therefore, we hypothesized that the extremely small population of A. catalpifolium could be, at least in part, associated with the lack of recent WGDs. Previous studies have shown that synthesized allohexaploid wheat displays polyploid growth vigor and adaptation from cytological, genetic, and epigenetic perspectives (Li et al. 2015). Other synthesized allopolyploids also go through rapid genomic reorganization after WGD, like Brassica and Tragopogon (Song and Chen 2015). Because polyploidy is a major force in shaping plant biodiversity, it is possible to acquire genetic novelty for environmental adaptation by synthesizing allopolyploids or autopolyploids using A. catalpifolium as the progenitor (Leitch and Leitch 2008).

Plant Materials and DNA Extraction
Acer catalpifolium seeds were collected from a 20-year-old tree grown in the West China Subalpine Botanical Garden, Dujiangyan, Sichuan Province (31 00 0 33 00 N, 103 37 0 00 00 E). In 2017, seeds were germinated in pots filled with organic substrate and silica sand. Seedlings were transplanted in the Beijing Forestry University climate chamber with constant 60% moisture, 20 C air temperature, and 12 h light supply. Total genomic DNA was extracted from fresh leaves using the CTAB method (Doyle and Doyle 1987).

Genome Sequencing
We used three platforms for DNA sequencing: Illumina, PacBio, and Hi-C. The Illumina Hiseq X Ten platform was used for 150 bp paired-end sequencing with average library insert sizes of 350 bp, generating 34.40 Gb reads. The library preparation was performed using ultrasonication to fragment DNA sequences, then DNA polymerase was added to repair fragmented DNA into dsDNA at the flat terminal. T4 polynucleotide kinase phosphorylated the 5 0 terminal, and a poly-A tail was added to the 3 0 end of dsDNA. SMRT libraries were constructed following the Quail protocol (Quail et al. 2012). Libraries were sequenced using the PacBio Sequel platform and 34.39 Gb reads were generated. Hi-C libraries were prepared according to the following steps: 1) formaldehyde was used to fix samples, cross-linking intracellular proteins, and DNA to preserve their interaction relationships. 2) Restriction enzymes were used to cut DNA to produce sticky ends on both sides of cross-linking. 3) Biotin-labeled bases were introduced to facilitate subsequent DNA purification. 4) DNA fragments cross-linked with proteins were amplified by PCR. 5) DNA was cross-linked, purified, and broken into 300-700 bp fragments. Strand affinity magnetic beads were used to capture and remove DNA fragments cross-linked with proteins from downstream sequencing.

RNA Extraction and Sequencing
Total RNA was extracted from the leaf, stem, and root tissues of A. catalpifolium using the modified CTAB method (Yang et al. 2008). RNA quality was determined by estimation of the ratio of absorbance at 260 nm/280 nm. Sequencing libraries were prepared according to the following steps: 1) mRNA was enriched with VAHTS mRNA-seq V3 Library Prep Kit for Illumina (Vazyme). 2) mRNA was randomly interrupted using fragmentation buffer. 3) Using mRNA as the template, the first cDNA strand was synthesized with random hexamers. The buffer, dNTPs, RNase H, and DNA Polymerase I were then added to synthesize the second cDNA strand. 4) AHTS DNA Clean Beads (Vazyme) were used to purify cDNA. 5) The cDNA library was obtained through PCR amplification. The paired-end sequencing libraries were sequenced on an Illumina HiSeq X Ten platform.

Genome Synteny and WGD Analysis
The proteomes of A. yangbiense and V. vinifera were downloaded from public resources, namely GigaDB (Sneddon et al. 2012) and Phytozome v12 (Goodstein et al. 2012). Sequence similarity was analyzed between the three species (A. catalpifolium, A. yangbiense, and V. vinifera) using BLASTP (ÀE 1eÀ10). MCScanX ) was then used with default settings to identify syntenic blocks and colinear gene pairs. To obtain the best "one-to-one" syntenic matchings, we eliminated colinear blocks with fewer than nine genes, then eliminated gene pairs without macrosynteny information support. Finally, the gene in a block containing the largest number of genes was retained. The proteins of those homologous gene pairs were first aligned using MUSCLE v3.8.31 (Edgar 2004) with default parameters, then converted to CDS alignment using PAL2NAL v14 (Suyama et al. 2006) with default parameters. The Ks values were calculated with custom scripts using the Yang-Nielsen (YN) model in PAML (Yang 2007).

Orthogroup Inference
We selected 13 representative plants including A. catalpifolium to determine orthogroups. Protein sequences were downloaded from two sources: A. yangbiense from GigaDB (Sneddon et al. 2012) and the others from Phytozome v12 (Goodstein et al. 2012). We separately ran all-versus-all protein sequence alignments in these species using BLASTP (ÀE 1eÀ5). Orthogroups were generated by OrthoMCL v14.137 (Li et al. 2003) with default parameters using the BLASTP results.

Phylogenetic Analysis and Divergence Time Estimation
We selected 873 single-copy orthogroups from 13 plant species, and amino acid sequences of each orthogroup were included in a multiple sequence alignment using MAFFT v7.427 (Katoh et al. 2002) with default parameters. The aligned protein sequences were transformed into codon alignments using PAL2NAL v14 with default parameters. The codon alignments were concatenated as a supermatrix and the tree was inferred using RAxML (v8.2.12; parameters: -f am GTRGAMMA) with 1,000 bootstrap replicates (Stamatakis et al. 2005). The resulting species tree was used as input to infer the divergence time of 13 plants using r8s v1.81 (Sanderson 2003) with default settings except for setting four fossil records at the key nodes of the tree: "mrca Origin_rosids Acat Vvin; mrca Crown_Eudicots Acoe Acat; mrca Crown_Monocots Macu Osat; mrca Crown_Angiosper ms Atri Macu; constrain taxon¼Origin_rosids max_age¼125; constrain taxon¼Crown_Eudicots min_age¼125; constrain taxon¼Crown_Monocots min_age¼113; constrain taxon¼ Crown_Angiosperms min_age¼132.9 max_age¼260." These fossil calibrations were set with reference to one published paper with minor modifications (Li et al. 2019). The other method to infer the timing of divergence between two Acer trees was based on the assumption of sharing the same k (global mutation rate), meaning k ¼ Ks 1 /2T 1 ¼ Ks 2 /2T 2 .
Calculation of x Value (Ka/Ks Ratio) The 13 aforementioned angiosperms were used to calculate x, assuming that each branch of the phylogenetic tree has a different x (Xu et al. 2017). The 873 single-copy orthogroups from the 13 species were aligned using MUSCLE (default parameters) and converted into codon alignments with PAL2NAL (default parameters). The alignments of the 873 single-copy orthogroups were then concatenated into a super matrix and the "free-ratio" model of CODEML in the PAML 4.9j package (Yang 2007) was used to calculate x for each branch of the phylogenetic tree constructed above.
Next, we used six eudicots (A. catalpifolium, A. yangbiense, Arabidopsis, C. sinensis, D. longan, and P. trichocarpa) to calculate x. Orthogroups from the six species were identified using OrthoMCL (Li et al. 2003) with default parameters. Orthogroups with a single copy in each of the six plants were identified, as were orthogroups with one single copy in Arabidopsis (the outgroup) but multiple copies in the other five species. We aligned single copy Arabidopsis genes (using BLASTP) against those multiple copies from other five species to generate best hits as representative orthologs of each plant. Ultimately, we found 6,116 orthogroups with one copy of one plant in each (Xu et al. 2017).
The 6,116 codon alignments were also concatenated into a supermatrix and the "free-ratio" model was used to calculate x in each branch of the phylogenetic tree. Alignments of the 6,116 orthogroups were used as input to calculate x separately using the "branch" model in PAML, hypothesizing that the Acer branch has a different x from the remaining branches. The likelihood ratio test (LRT) (Anisimova et al. 2001) was used to remove false positive results (P value <0.05). Next, the alignments of the 6,116 orthogroups were analyzed using the "branch-site" model in CODEML to detect PSGs ("fix_omega ¼ 1, omega ¼ 1" for the null model, "fix_omega ¼ 0, omega ¼ 2" for the alternative model). Setting the branch of Acer genus as the foreground, the likelihood of the null model (no site in the Acer clade is positively selected) and alternative model (sites exist in the Acer clade that are positively selected) were calculated. The LRT was used to remove false findings and genes under positive selection (x > 1, P value <0.05) using the Bayes Empirical Bayes method were further analyzed (Yang et al. 2005). Furthermore, setting the node of A. catalpifolium or A. yangbiense as the foreground, the likelihood of the null model (no site in A. catalpifolium or A. yangbiense is positively selected) and alternative model (sites exist in A. catalpifolium or A. yangbiense that are positively selected) were also calculated to identify differentially selected genes. Orthogroups were annotated to different TFs based on their annotations in Arabidopsis in PlantTFDB 5.0 (Tian et al. 2020).

Gene Family Evolution
We used the ultrametric tree identified by r8s v1.81 and orthogroups inferred from 13 plant genomes as input to compute the expansion and contraction of gene families. The program Count (Csuros 2010) was implemented using a Wagner parsimony framework (gain penalty ¼ 1.2, which produced the most reasonable number of expanded/contracted gene families after testing with other values) to understand the expanded/contracted gene families along each branch of the ultrametric tree. CAFE v4.2.1 (De Bie et al. 2006) with default parameters was used to identify rapidly evolving gene families.

GO Enrichment Analysis
The R package clusterProfiler v3.12 (pvalueCutoff ¼ 0.05, pAdjustMethod ¼ "BH," qvalueCutoff ¼ 0.05) was used to perform GO enrichment analysis and identify statistically enriched GO terms (Yu et al. 2012). The genes in significantly expanded or Acer genus-specific orthogroups were set as the foreground and the genes annotated with GO terms at the whole-genome level were regarded as the background. The adjusted P values were calculated to obtain significantly enriched GO terms using the Benjamini and Hochberg method (adjusted P value <0.05).

FRS and FRF Gene Family
FRS genes were identified using a combination of several methods. First, protein sequences of A. catalpifolium and A. yangbiense were searched against three Pfam protein families: FAR1 domain (PF03101), MULE domain (PF10551), and SWIM domain (PF04434) using HMMER (v3.3, parameters: -E 0.01 -domE 0.01) (Mistry et al. 2013). Then, extracted sequences containing none of those domains were filtered to remove any false-positive domains using InterproScan v5.39-77.0 (Quevillon et al. 2005). The genes that passed both HMMER and InterproScan were retained for downstream analyses, and the proteins containing all three domains were regarded as the most conserved FRS members. Next, BLASTP was used against the combined database of 18 FRS and FRF proteins in Arabidopsis and the most conserved FRS genes in A. catalpifolium and A. yangbiense to obtain additional putative FRS candidates (ÀE 1eÀ10). False positive candidate proteins were removed by manually inspecting multiple sequence alignments, and the proteins that had conserved amino acid residues in both MULE and SWIM domains were retained (Lin and Wang 2004). Proteins that had only the FAR1 domain were regarded as FRF genes. The protein sequences for FRS and FRF genes from three species (Arabidopsis, A. catalpifolium, and A. yangbiense) were aligned using MAFFT v7.427 (with default parameters) and put into RAxML (v8.2.12; parameters: -f a -m PROTGAMMAAUTO -# 1000) to construct a phylogenetic tree of the FRS and FRF gene family. FRS and FRF gene locations were integrated with MCScanX results and repetitive sequence annotation to determine the number of genes derived from different types of gene duplication.

Lignin Biosynthesis
The lignin biosynthesis pathway was built based on information of PoplarCyc in Plant Metabolic Network (PMN) (Caspi et al. 2016) and a previous study (Vanholme et al. 2019).
These key enzymes that participate in the pathway were predicted with the E2P2 algorithm v3.0 designed by PMN using proteomes of A. catalpifolium and A. yangbiense separately (with default parameters). The candidate genes in the two species were annotated based on the EC (Enzyme Commission) number of the enzyme.

PSMC Analysis
We inferred a demographic history of A. catalpifolium and A. yangbiense by applying the PSMC model to the complete diploid genome sequences. This method reconstructs the history of dynamic changes in population size over time using the distribution of the most recent common ancestor between two alleles in one individual. Paired-end Illumina reads of A. catalpifolium and A. yangbiense were aligned to their reference genomes using BWA-MEM v0.7.10 (Peters et al. 2011). Then, the heterozygous biallelic SNPs were called using SAMtools v0.1.19 (Li et al. 2009), and reads of very low depth (less than a third of the average depth) or of very high depth (twice the average depth) were removed. Bcftools v1.1 (Li et al. 2009) was used to convert the BCF file into VCF format, and vcfutils.pl (https://github.com/lh3/samtools/blob/master/ bcftools/vcfutils.pl, last accessed December 13, 2021) was used to convert the VCF file into the whole-genome diploid consensus sequence. All of the parameters utilized in the PSMC program were set to default with the exception of a per-generation mutation rate (l) of 7.5Â 10 À9 taken from Arabidopsis (Buschiazzo et al. 2012) and 30 years per Acer generation (g) taken from data for A. mono (Liu et al. 2014).

Supplementary Material
Supplementary data are available at Genome Biology and Evolution online.