Chromosome-level genome of Proserpinaca palustris reveals key KNOX genes related to heterophylly formation Authors: Yanjing Dong, Qian Qin, Fengqin Li, Zhonghua Tu, Chengyi Tang Published: 2025 ABSTRACT BackgroundProserpinaca palustris, a representative heterophyllous plant, represents an emerging model for studying morphological plasticity and adaptive evolution in amphibious environments. However, the absence of a high-quality reference genome has hindered molecular investigations into its heterophylly regulation mechanisms.ResultsHere, we present the first chromosome-level genome assembly of P. palustris. The final assembly spans 385.32 Mb, anchored onto seven pseudochromosomes (1n = 7), and encodes 22,756 predicted protein-coding genes. Benchmarking Universal Single-Copy Orthologs (BUSCO) assessment demonstrated high completeness, with scores of 95.61% for the genome assembly and 96.43% for the gene set. Comparative genomic analyses revealed that P. palustris underwent only the ancient γ whole-genome triplication (γ-WGT) event, with no subsequent whole-genome duplication (WGD). Nevertheless, the genome has experienced extensive re-diploidization, likely facilitated by structural rearrangements. Transcriptomic profiling identified three KNOTTED1-like homeobox (KNOX) genes (i.e., PPA02C1640, PPA03C1159, and PPA01C1293) that were significantly differentially expressed between submerged and terrestrial leaves in P. palustris, suggesting their potential role in regulating heterophylly.ConclusionsThis study provides the first high-quality genomic resource for the genus Proserpinaca (Haloragaceae) and identifies key KNOX genes implicated in heterophylly regulation in P. palustris.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12870-025-07744-8. Background Proserpinaca palustris, commonly known as mermaid weed, is an amphibious angiosperm belonging to the family Haloragaceae (Saxifragales) [1]– [2]. This species exhibits striking heterophylly, a form of phenotypic plasticity wherein leaf morphology varies substantially in response to environmental conditions (Fig. 1) [3]. Terrestrially grown leaves of P. palustris are lanceolate and serrated, featuring pronounced keratinization, expanded mesophyll, and abundant stomata [3]– [4]. in contrast, submerged leaves are non-keratinized, deeply dissected, and consist primarily of a slender central vein with several filiform lobes, accompanied by markedly reduced stomatal density [3]– [4]. This pronounced dimorphism, termed heterophylly, renders P. palustris an compelling model for investigating the molecular mechanisms underlying leaf plasticity. Fig. 1Leaf dimorphism and chromosomal gene density in P. palustris. TLB, terrestrial leaves and buds; SLB, submerged leaves and buds. Scale bar: Mb Leaf dimorphism and chromosomal gene density in P. palustris. TLB, terrestrial leaves and buds; SLB, submerged leaves and buds. Scale bar: Mb In amphibious and aquatic angiosperms, heterophyllous development is regulated by a complex interplay of environmental and endogenous factors, including photoperiod, temperature, CO2 concentration, and phytohormonal signaling [5–11]. In P. palustris, terrestrial leaf development is typically induced under long-day (LD) photoperiods combined with high temperatures, whereas short-day (SD) or LD with low temperatures promote submerged leaf formation [1, 4]. Phytohormones play a pivotal role in this dimorphic plasticity: exogenous application of gibberellin (GA) under SD or low-temperature LD conditions triggers a transition from submerged leaves to terrestrial leaves [1]. Conversely, inhibition of GA biosynthesis under LD conditions induces the development of submerged leaves [1]. Although terrestrial leaves normally develop only on aerial shoots, application of abscisic acid (ABA) can induce the formation of submerged leaves [1]. This ABA-mediated promotion of leaf development is consistent across other heterophyllous species, such as Callitriche palustris, Ranunculus trichophyllus, Ludwigia arcuata, and Marsilea quadrifolia, whereas GA and ethylene often act antagonistically to maintain submerged leaf morphology [5, 7, 10, 12]– [13]. From a genetic perspective, the morphological divergence between terrestrial and submerged leaves is orchestrated by a complex genetic regulatory network in which the KNOX gene family serves as a key component. As members of the TALE homeobox superfamily, KNOX genes are categorized into three subfamilies: Class I, Class II, and Class M [14–16]. Members of Class I and Class II possess two main domains, i.e., the MEINOX and the homeodomain, whereas Class M proteins lack the homeodomain [16]– [17]. Class I members include Shoot Meristemless (STM), BREVIPEDICELLUS (BP/KNAT1), KNAT2, and KNAT6; Class II comprises KNAT3, KNAT4, KNAT5, and KNAT7 [15]. Accumulating evidence underscores the crucial role of KNOX genes in regulating heterophyllous leaf development across various amphibious plants, including Myriophyllum aquaticum (Haloragaceae), Hygrophila difformis, and Rorippa aquatica [18–21]. In M. aquaticum, MaKN1 (GenBank No.: EU203679.1) is expressed in the shoot apical meristem (SAM) and early leaf primordia (stages P0-P8), with expression terminating in terrestrial leaves after stage P8 but persisting in submerged leaves, thereby promoting enhanced lobation [18]– [19]. Similarly, in H. difformis, overexpression of HdSTM (GenBank No.: MZ365290.1) does not alter terrestrial leaf morphology but significantly enhances lobe formation under submerged conditions [20]. Furthermore, in R. aquatica, RaSTM (GenBank No.: AB971253.1) contributes to heterophylly regulation through modulation of GA accumulation [21]. Despite these insights, the regulatory mechanisms of KNOX genes in heterophylly remain insufficiently characterized across most amphibious plant species. Extensive studies in model plants have elucidated the central role of KNOX genes in leaf morphogenesis, mediated through interactions with hormonal pathways including GA, ABA, cytokinin, and brassinosteroids [22]– [23]. Class I KNOX genes are initially expressed in meristematic tissues but are downregulated in incipient leaf primordia to permit normal leaf initiation [24]. Their overexpression in Arabidopsis thaliana induces leaf serration, and in compound-leaved species, such as Solanum lycopersicum, Cardamine hirsuta, and Medicago truncatula, enhances leaf complexity [15, 17]– [18]. In contrast, Class II KNOX genes are typically expressed in mature tissues and often antagonize Class I functions [24–26]. Loss-of-function mutations in Class II genes result in phenotypes resembling Class I overexpression, including enhanced serration, whereas their overexpression suppresses SAM activity and simplifies leaf morphology, mirroring Class I loss-of-function mutants [15, 24, 27]. Nevertheless, the identity, evolutionary classification, and regulatory roles of KNOX genes in the heterophyllous amphibious plant P. palustris remain to be systematically investigated. The absence of high-quality genomic resources has significantly impeded molecular research on P. palustris, thereby constraining its establishment as a model system for investigating heterophylly. To overcome this limitation, we generated a chromosome-level genome assembly for P. palustris, the first genome within the genus Proserpinaca (Haloragaceae), by integrating Illumina short reads, Oxford Nanopore Technology (ONT) long-reads, and high-throughput chromosome conformation capture (Hi-C) data. Utilizing this genomic resource, we performed a comprehensive analysis of the KNOX gene family combined with transcriptomic profiling, which enabled the identification of key KNOX genes implicated in the regulation of heterophylly in P. palustris. These results provide valuable genomic resource and a solid foundation for accelerating P. palustris as a promising model organism for investigating leaf morphological plasticity. Materials and methods Plant materials P. palustris was acquired from commercial sources and subsequently propagated using tissue culture techniques (Figure S1). High-quality genomic DNA was extracted from fresh leaves of a single clonal line maintained in vitro for genome sequencing. For transcriptome sequencing, samples included terrestrial leaves and buds (TLB) and submerged leaves and buds (SLB). SLB were cultivated under submerged conditions for one month, while TLB were maintained under standard growth conditions. Each sample type consisted of three biological replicates. All collected tissues were immediately flash-frozen in liquid nitrogen and stored at −80 ℃ until subsequent processing. Genome & transcriptome sequencing, and data processing Genomic DNA was extracted using the Magnetic Plant Genomic DNA Kit (Cat. No.: 4992407; Tiangen, China). For short-read sequencing, an Illumina paired-end library with a 350 bp insert size was prepared using the TIANSeq Fast DNA Library Kit (Cat. No.: 4992261; Tiangen, China) and sequenced on the Illumina NovaSeq 6000 platform (Illumina, USA). For long-read sequencing, a ligation-based library was constructed with the ONT Ligation Sequencing Kit (SQK-LSK110; ONT, UK) and sequenced on the PromethION platform (ONT, UK). A Hi-C library was prepared following established protocols [28] using the Mbo I restriction enzyme and sequenced on the Illumina NovaSeq 6000 platform (Illumina, USA). Total RNA was isolated using the RNAprep Pure Plant Kit (Cat. No.: 4992237; Tiangen, China). RNA-seq libraries were constructed using the TIANSeq Fast RNA Library Prep Kit (Cat. No.: 4992376; Tiangen, China) and sequenced on the Illumina NovaSeq 6000 platform (Illumina, USA). Raw Illumina short reads were quality-filtered using Fastp v0.23.2 [29] to remove adapters, low-quality bases, and poly-N sequences. Raw ONT long reads were filtered with NanoFilt v2.8.0 [30], discarding reads < 2 kb or RQ < 7.0. Genome size estimation Genome size was estimated via k-mer frequency analysis of the clean Illumina short-reads using Jellyfish v2.3.0 (key parameters: jellyfish count -m 19, 21, 23; jellyfish histo -h Max_count) [31]. The k-mer profiles were analyzed with GenomeScope 2.0 [32] and GCE v1.0.2 [33], and the final genome size was determined as the mean value from both tools. Genome assembly A preliminary genome assembly was performed using clean ONT long-reads using NextDenovo v2.5.0 (key parameters: seed_depth = 999, nextgraph_options = -a 1 -q 7) [34], followed by polishing with Illumina short-reads using NextPolish v1.4.1 (key parameters: task = 1212) [35]. Clean Hi-C reads were aligned to the polished assembly using Juicer v1.6 (key parameters: -s MboI) [36]. Chromosome-level scaffolding was performed with 3D-DNA [37], followed by manual curation of scaffolds in Juicebox v1.11.08 [38]. Genome annotation Repetitive sequences were annotated via RepeatMasker v4.1.4 (http://www.repeatmasker.org), based on a combined database including the Dfam v3.7 database [39] plus a de novo custom library of P. palustris constructed via RepeatModeler v2.0.4 (key parameter: -LTRStruct) [40]. Repetitive regions were masked prior to gene annotation. Ab-initio gene annotation was conducted with AUGUSTUS v3.5.0 [41] and GeneMark-EP + v4.71 [42]. Homology-based annotation was performed using Exonerate v2.4.0 (key parameter: --model protein2genome --showtargetgff) [43] with reference proteins from A. thaliana [44], Kalanchoe fedtschenkoi [45], and Tiarella polyphylla [46]. Transcript evidence was incorporated via PASA v2.5.2 [47]. All evidence types were integrated into a consensus gene set using MAKER v3.01.03 (key parameters: softmask = 1; min_protein = 49) (https://www.yandell-lab.org/software/maker.html). Genome evaluation Completeness of the genome assembly and the gene set was assessed using BUSCO v5.2.2 (key parameters: --augustus -l eudicots_odb10) [48]. Potential contaminants were screened by aligning unplaced genomic sequences against the NCBI NT database using BLASTN 2.13.0+ (key parameter: -evalue 1e-10 -max_hsps 1 -max_target_seqs 5000) [49]. Heterozygosity was further evaluated by mapping unplaced sequences to the assembled chromosomes with minimap2 v2.17 (key parameter: -x asm5) [50]. Comparative genomics analysis Sequence similarity searches were performed using BLASTP 2.13.0+ (key parameter: -evalue 1e-5 -max_target_seqs 5) [49]. Collinearity within and across species was identified with MCScanX [51]. Gene pairs within collinear blocks were aligned at the amino acid level using MUSCLE v3.8.31 [52], and alignments were reverse-translated to nucleotides. Ks values were computed with KaKs_Calculator 3.0 [53]. Peaks in Ks distributions indicated WGD events within species and speciation events between species. Single-copy orthologs were identified using OrthoFinder v2.5.5 [54] for species tree construction. Divergence times were estimated with the MCMCtree module in PAML v4.9 [55], using calibrations from TimeTree [56]. Gene family analysis Two Pfam domains, i.e., PF03790 and PF03791, were used as queries for domain similarity search via the hmmsearch module within HMMER 3.3.2 (key parameter: -E 1e-05) [57]. Candidate KNOX proteins were validated through domain prediction with InterPro (https://www.ebi.ac.uk/interpro/search/sequence/). Conserved motifs in KNOX proteins were identified using MEME (https://meme-suite.org/meme/tools/meme). Subsequently, the identified KNOX proteins were aligned with MUSCLE v3.8.31 [52], and the alignment was trimmed with trimAl v1.4.1 (key parameter: -gt 0.50) [58]. A maximum likelihood phylogenetic tree was constructed using IQ-TREE v2.2.2 (key parameters: --seqtype AA -m MFP --alrt 1000 -B 1000) [59]. Phylogenetic trees, Pfam domains, and conserved motifs were visualized with EvolView v3 [60]. Transcriptome expression analysis Clean RNA-seq reads were aligned to the reference genome using STAR v2.4.1 [61]. Read counts per gene were quantified with featureCounts v2.4.1 [62]. Differential expression analysis was conducted using DESeq2 v1.38.2 [63], applying thresholds of |log2(fold change)| ≥ 1 and False discovery rate (FDR) < 0.001 for identifying significantly differentially expressed genes (DEGs). RT-qPCR validation Total RNA was extracted from SLB and TLB samples and reverse-transcribed into first-strand cDNA using the PrimeScript RT Master Mix Kit (TaKaRa, China). The resulting cDNA served as the template for real-time quantitative PCR (RT-qPCR), which was performed using the TB Green Premix Ex Taq II Kit (TaKaRa, China) on a CFX96 Real-Time PCR Detection System (Bio-Rad, USA). Amplification was carried out following a two-step protocol with 40 cycles. All primer sequences are provided in Table S1. Each sample was analyzed in three biological replicates, with each reaction run in technical triplicate. Relative expression levels of target genes were calculated using the 2⁻ΔΔCt method. Results Genome assembly, annotation, and evaluation of P. palustris We generated approximately 19.11 Gb of Illumina short-reads, 30.17 Gb of ONT long-reads, and 28.58 Gb of Hi-C data (Table S2). The estimated genome size was approximately 387.45 Mb (Figure S2). A contig-level assembly yielded a total contig length of 385.32 Mb, comprising 443 contigs with N50 and N90 values of 3.11 Mb and 0.62 Mb, respectively (Table 1). These contigs were further assembled into 173 scaffolds, totaling 385.35 Mb with scaffold N50 and N90 values of 54.84 Mb and 39.99 Mb (Table 1). Subsequently, 371.68 Mb of the assembly was anchored onto seven pseudochromosomes (Figure S3). The completeness of the genome assembly reached 95.61% (Table 1).Table 1Summary statistics of genome assembly and gene set of P. palustris and T. polyphyllTypesP. palustrisT. polyphyllaScaffoldsContigsScaffoldsContigsTotal number17344354207Total lengths (Mb)385.35385.32412.26412.24Total N50 lengths (Mb)54.843.1157.2312.01Total N90 lengths (Mb)39.990.6244.080.97BUSCO evaluation(eudicots_odb10, 2326)for total genome assemblyComplete: 2224 (95.61%);Fragmented: 13 (0.56%);Missing: 89 (3.83%)Complete: 2251 (96.78%);Fragmented:24 (1.03%);Missing:51 (2.19%)Total gene models22,75625,244BUSCO evaluation(eudicots_odb10, 2326)for total gene setComplete: 2243 (96.43%);Fragmented: 10 (0.43%);Missing: 73 (3.14%)Complete: 2194 (94.33%);Fragmented: 20 (0.86%);Missing: 112 (4.82%)Chromosomal (Chr) number72327160Chr lengths (Mb)371.68371.66403.10403.09Chr N50 lengths (Mb)54.843.1457.2312.81Chr N90 lengths (Mb)39.990.7944.081.22BUSCO evaluation(eudicots_odb10, 2326)for chr genome assemblyComplete: 2204 (94.75%);Fragmented: 13 (0.56%);Missing: 109 (4.69%)Complete: 2215 (95.23%);Fragmented: 23 (0.99%);Missing: 88 (3.78%)Chr gene models22,23224,717BUSCO evaluation(eudicots_odb10, 2326)for chr gene setComplete: 2223 (95.57%);Fragmented: 10 (0.43%);Missing: 93 (4.00%)Complete: 2161 (92.91%);Fragmented: 20 (0.86%);Missing: 145 (6.23%) Table 1Summary statistics of genome assembly and gene set of P. palustris and T. polyphyllTypesP. palustrisT. polyphyllaScaffoldsContigsScaffoldsContigsTotal number17344354207Total lengths (Mb)385.35385.32412.26412.24Total N50 lengths (Mb)54.843.1157.2312.01Total N90 lengths (Mb)39.990.6244.080.97BUSCO evaluation(eudicots_odb10, 2326)for total genome assemblyComplete: 2224 (95.61%);Fragmented: 13 (0.56%);Missing: 89 (3.83%)Complete: 2251 (96.78%);Fragmented:24 (1.03%);Missing:51 (2.19%)Total gene models22,75625,244BUSCO evaluation(eudicots_odb10, 2326)for total gene setComplete: 2243 (96.43%);Fragmented: 10 (0.43%);Missing: 73 (3.14%)Complete: 2194 (94.33%);Fragmented: 20 (0.86%);Missing: 112 (4.82%)Chromosomal (Chr) number72327160Chr lengths (Mb)371.68371.66403.10403.09Chr N50 lengths (Mb)54.843.1457.2312.81Chr N90 lengths (Mb)39.990.7944.081.22BUSCO evaluation(eudicots_odb10, 2326)for chr genome assemblyComplete: 2204 (94.75%);Fragmented: 13 (0.56%);Missing: 109 (4.69%)Complete: 2215 (95.23%);Fragmented: 23 (0.99%);Missing: 88 (3.78%)Chr gene models22,23224,717BUSCO evaluation(eudicots_odb10, 2326)for chr gene setComplete: 2223 (95.57%);Fragmented: 10 (0.43%);Missing: 93 (4.00%)Complete: 2161 (92.91%);Fragmented: 20 (0.86%);Missing: 145 (6.23%) Summary statistics of genome assembly and gene set of P. palustris and T. polyphyll BUSCO evaluation (eudicots_odb10, 2326) for total genome assembly Complete: 2224 (95.61%); Fragmented: 13 (0.56%); Missing: 89 (3.83%) Complete: 2251 (96.78%); Fragmented:24 (1.03%); Missing:51 (2.19%) BUSCO evaluation (eudicots_odb10, 2326) for total gene set Complete: 2243 (96.43%); Fragmented: 10 (0.43%); Missing: 73 (3.14%) Complete: 2194 (94.33%); Fragmented: 20 (0.86%); Missing: 112 (4.82%) BUSCO evaluation (eudicots_odb10, 2326) for chr genome assembly Complete: 2204 (94.75%); Fragmented: 13 (0.56%); Missing: 109 (4.69%) Complete: 2215 (95.23%); Fragmented: 23 (0.99%); Missing: 88 (3.78%) BUSCO evaluation (eudicots_odb10, 2326) for chr gene set Complete: 2223 (95.57%); Fragmented: 10 (0.43%); Missing: 93 (4.00%) Complete: 2161 (92.91%); Fragmented: 20 (0.86%); Missing: 145 (6.23%) Repetitive sequences comprised 59.89% of the genome (230.77 Mb), with interspersed repeats representing the largest proportion (204.44 Mb, 53.06%), whereas tandem repeats accounted for 1.75 Mb (0.45%) (Table S3). Among interspersed repeats, long terminal repeats (LTRs) were the most abundant, constituting 47.43% of the genome, followed by DNA transposons (5.23%), long interspersed nuclear elements (LINEs, 0.29%), and short interspersed nuclear elements (SINEs, 0.0031%) (Table S3). A total of 22,756 protein-coding genes were annotated (Table 1). The gene set exhibited high completeness, achieving a BUSCO score of 96.43%. Gene density was elevated near chromosomal ends relative to pericentromeric regions (Fig. 1), consistent with typical plant genome architecture. Approximately 13.66 Mb of genomic sequences remained unanchored to chromosomes in the P. palustris genome assembly (Table 1). To characterize these unplaced sequences, we conducted further analyses. Alignment against the NCBI NT database confirmed the absence of contaminants (Table S4). Collinearity analysis revealed that a subset of the unplaced sequences aligned to small regions within the chromosomal assembly (Figure S4A), suggesting a heterozygous origin. These sequences were likely excluded during manual Hi-C polishing due to ambiguous or duplicate signals. Annotation indicated that most unplaced sequences were repetitive, as evidenced by high repeat content (Table S6). Although a small number of protein-coding genes were identified among them (Table S6), their removal resulted in only a minor decrease in BUSCO completeness (Table 1), with no substantial effect on the overall integrity of the chromosome-level assembly. For comparative purposes, equivalent analyses were performed on unplaced sequences from the T. polyphylla genome. Results showed a similar profile: no contamination was detected (Table S5), a subset exhibited collinearity to chromosomal regions (Figure S4B), sequences were predominantly repetitive (Table S7), and their exclusion led to a comparable slight reduction in BUSCO score (Table 1). These consistent findings demonstrate that the quality and characteristics of the P. palustris genome assembly are comparable to those of the published T. polyphylla genome. Whole-genome duplication events in P. palustris To investigate whether P. palustris underwent WGD events, we performed intra- and interspecific synteny analyses using T. polyphylla (Saxifragales), which has experienced only the γ-WGT event [46], and Aristolochia contorta (Piperales), which has not experienced the γ-WGT event [64], as reference species. The Ks distribution analysis of intraspecific collinear gene pairs indicated that P. palustris exhibited a single peak at Ks ≈ 1.58 (Peak 2; Fig. 2A). Similarly, T. polyphylla displayed a single peak at Ks ≈ 1.22 (Peak 1; Fig. 2A), whereas A. contorta exhibited no discernible peaks (Fig. 2A). These findings are consistent with those previously reported [46, 64].Fig. 2Comparative genomic analysis and divergence time estimation among P. palustris, T. polyphylla, and A. contorta. A Density distribution of synonymous substitution rates (Ks) for intraspecific collinear gene pairs within each species. B Ks density curves for interspecific comparisons. C Species tree and divergence time estimates inferred from single-copy orthologs Comparative genomic analysis and divergence time estimation among P. palustris, T. polyphylla, and A. contorta. A Density distribution of synonymous substitution rates (Ks) for intraspecific collinear gene pairs within each species. B Ks density curves for interspecific comparisons. C Species tree and divergence time estimates inferred from single-copy orthologs Interspecific synteny analysis between P. palustris and T. polyphylla revealed a major peak at Ks ≈ 1.05 (Peak 3; Fig. 2B), while the P. palustris vs. A. contorta and T. polyphylla vs. A. contorta comparisons showed peaks at Ks ≈ 1.80 (Peak 5) and Ks ≈ 1.59 (Peak 4), respectively (Fig. 2B). Phylogenetic analysis inferred that the common ancestor of P. palustris and T. polyphylla underwent the γ-WGT event at approximately 129 million years ago (MYA) (Ks ≈ 1.40; Fig. 2C), consistent with previously reported estimates [46, 64, 65]. Divergence between P. palustris and T. polyphylla occurred later, around 85 MYA (Ks ≈ 1.05; Fig. 2C). In contrast, the divergence between their common ancestor and A. contorta occurred earlier, approximately 163 MYA (Ks ≈ 1.70). Intraspecific synteny analysis revealed that A. contorta exhibited limited collinearity among chromosomes, with few detectable collinear blocks (Figure S5C, S5I). Only 2.85% of its gene pairs were classified as WGD/segmental replication-type, and among these, 86.42% exhibited a replication depth of 1× (Fig. 3), further supporting the absence of the γ-WGT event. In contrast, both P. palustris and T. polyphylla displayed numerous syntenic blocks (Figure S5A, S5B, S5G, S5H). Although these blocks appeared fragmented and disordered without a clear 1:2 syntenic ratio, the proportion of WGD/segmental replication-type genes was substantially higher in P. palustris (19.89%) and T. polyphylla (18.72%) compared to A. contorta. Moreover, the majority of genes in these two species showed a replication depth of 2× (i.e., 1:2 replication ratio; Fig. 3). Collectively, these findings suggest that both P. palustris and T. polyphylla probably underwent only the γ-WGT event, with no subsequent WGD events.Fig. 3Gene replication types and replication depths within intraspecific collinear regions of P. palustris, T. polyphylla, and A. contorta Gene replication types and replication depths within intraspecific collinear regions of P. palustris, T. polyphylla, and A. contorta To further substantiate this conclusion, we performed interspecific collinearity analysis and quantitatively assessed syntenic mapping types across the three species. When using A. contorta as a reference, the most frequent mapping type observed in T. polyphylla was 1:3 (43.43%; Fig. 4A), indicating that one gene in A. contorta corresponded to three homologs in T. polyphylla, a pattern consistent with the γ-WGT event (i.e., triplication). Had an additional WGD occurred, a predominant 1:6 mapping type would be expected. Similarly, the predominant mapping type from A. contorta to P. palustris was also 1:3 (21.59%; Fig. 4C). Using T. polyphylla as the reference, the most mapping type in P. palustris was 1:4 (23.91%), slightly exceeding that of the 1:3 type (23.49%; Fig. 4B). Although the syntenic blocks appeared fragmented and disorganized without a clear 1:3 ratio (Figure S5D-S5F, S5J-S5L), these results still align with the conclusion that P. palustris probably underwent only the γ-WGT event, as a recent WGD would have resulted in prevalent 1:6 mapping types.Fig. 4Distribution of syntenic mapping types among P. palustris, T. polyphylla, and A. contorta. A Syntenic mapping types in T. polyphylla using A. contorta as reference. B Syntenic mapping types in P. palustris using T. polyphylla as reference. C Syntenic mapping types in P. palustris using A. contorta as reference Distribution of syntenic mapping types among P. palustris, T. polyphylla, and A. contorta. A Syntenic mapping types in T. polyphylla using A. contorta as reference. B Syntenic mapping types in P. palustris using T. polyphylla as reference. C Syntenic mapping types in P. palustris using A. contorta as reference Candidate KNOX genes associated with heterophylly in P. palustris We identified nine KNOX genes in the P. palustris genome, distributed across six chromosomes (Figure S6). Phylogenetic analysis of 30 KNOX proteins, including nine from P. palustris, nine from A. thaliana, ten from T. polyphylla, and two functionally validated KNOX proteins, classified them into three clades: KNOX I, KNOX II, and KNOX M (Fig. 5). KNOX I proteins were characterized by the presence of four Pfam domains, i.e., PF03790 (KNOX1), PF03791 (KNOX2), PF05920 (Homeobox_KN), and PF03789 (ELK). KNOX II proteins comprised two distinct subgroups: one retaining all four domains, and another lacking PF03789 (Fig. 5). KNOX M proteins were characterized by the presence of either PF03789 or PF03791 only (Fig. 5). Conserved motif analysis further supported clade-specific differentiation: KNOX I proteins possessed seven conserved motifs, including clade-specific motifs 6 and 8; KNOX II proteins also contained seven motifs, with motifs 4, 5, and 9 being unique; KNOX M proteins exhibited only two conserved motifs (Fig. 5). Fig. 5Phylogenetic analysis and conserved domain/motif architecture of KNOX proteins. A phylogenetic tree was constructed using 30 KNOX proteins, including nine from P. palustris, nine from A. thaliana, ten from T. polyphylla, and two functionally validated homologs (i.e., HdSTM [20] and RaSTM [21]). The proteins were categorized into three clades: KNOX I (16 members), KNOX II (11 members), and KNOX M (3 members). Four Pfam domains were identified: PF03790 (KNOX1), PF03791 (KNOX2), PF05920 (Homeobox_KN), and PF03789 (ELK). Among these, PF03790 and PF03791 belong to the MEINOX domain, while PF05920 represents the homeodomain. Additionally, ten conserved motifs were detected across all proteins. Bootstrap values are indicated at the nodes of the phylogenetic tree Phylogenetic analysis and conserved domain/motif architecture of KNOX proteins. A phylogenetic tree was constructed using 30 KNOX proteins, including nine from P. palustris, nine from A. thaliana, ten from T. polyphylla, and two functionally validated homologs (i.e., HdSTM [20] and RaSTM [21]). The proteins were categorized into three clades: KNOX I (16 members), KNOX II (11 members), and KNOX M (3 members). Four Pfam domains were identified: PF03790 (KNOX1), PF03791 (KNOX2), PF05920 (Homeobox_KN), and PF03789 (ELK). Among these, PF03790 and PF03791 belong to the MEINOX domain, while PF05920 represents the homeodomain. Additionally, ten conserved motifs were detected across all proteins. Bootstrap values are indicated at the nodes of the phylogenetic tree To identify KNOX genes associated with heterophylly regulation, we conducted RNA-seq sequencing on SLB and TLB. Differential expression analysis revealed 3,559 DEGs, including 1,951 upregulated and 1,608 downregulated genes in TLB compared to SLB (Fig. 6A). Among the nine KNOX genes identified in P. palustris, three exhibited significant differential expression: PPA02C1640 and PPA03C1159, and PPA01C1293 (Fig. 6B). Transcriptome expression analysis revealed that PPA02C1640 and PPA03C1159 (both KNOX I) were highly expressed in SLB but repressed in TLB (Table S8, Fig. 6B), a pattern consistent with two functionally validated homologs (i.e., HdSTM [20] and RaSTM [21]) within the same KNOX I sub-clade (Fig. 5). In contrast, PPA01C1293 (KNOX II) exhibited an opposing expression profile (Table S8, Fig. 6B). Subsequent RT‑qPCR validation confirmed that the expression patterns of the KNOX II gene (PPA01C1293) and the KNOX I genes (PPA02C1640 and PPA03C1159) were mutually antagonistic between SLB and TLB (Fig. 6C). These opposing expression dynamics between KNOX I and KNOX II genes suggest a potential regulatory mechanism for heterophylly in P. palustris.Fig. 6Analysis of differentially expressed genes (DEGs) and KNOX gene expression in P. palustris. SLB, Submerged leaves and buds; TLB, Terrestrial leaves and buds. A Volcano plot of DEGs between SLB and TLB. B Expression patterns of KNOX genes in SLB vs. TLB. C RT‑qPCR validation of three differentially expressed KNOX genes (PPA01C1293, PPA02C1640 and PPA03C1159) across SLB and TLB Analysis of differentially expressed genes (DEGs) and KNOX gene expression in P. palustris. SLB, Submerged leaves and buds; TLB, Terrestrial leaves and buds. A Volcano plot of DEGs between SLB and TLB. B Expression patterns of KNOX genes in SLB vs. TLB. C RT‑qPCR validation of three differentially expressed KNOX genes (PPA01C1293, PPA02C1640 and PPA03C1159) across SLB and TLB Discussion Recent advances in genome sequencing technology Current genome sequencing strategies often integrate multiple technologies to leverage their complementary strengths. Short-read sequencing is widely used for genome surveys, including estimations of genome size and heterozygosity, due to its high base accuracy, exceptional throughput, and cost efficiency [66]. However, its limited read length often impedes de novo assembly, particularly in complex genomic regions. In contrast, long-read sequencing technologies, primarily represented by Pacific Biosciences (PacBio) and Oxford Nanopore Technologies (ONT), are essential for generating highly contiguous assemblies [66–68]. PacBio systems produce continuous long reads (CLR) and high-fidelity circular consensus sequencing (CCS) reads [67]– [68], while ONT delivers long and ultra-long reads that excel in resolving repetitive regions and improving assembly continuity [67]. The effectiveness of ONT has been demonstrated in the assembly of various plant genomes, including T. polyphylla, A. contorta, and Saururus chinensis [46, 64, 69]. Additionally, Hi-C sequencing captures three-dimensional chromatin architecture and has become a cornerstone method for anchoring contigs into chromosome-scale scaffolds [68]. Re-diploidization following γ-WGT in P. palustris The assembled genome of P. palustris exhibits a haploid chromosome number of 1n = 7, consistent with previous cytogenetic studies [70]– [71]. This chromosome number is shared by most species in the Haloragaceae family, where base numbers are typically 1n = 7 or multiples thereof (Figure S7). Although P. palustris underwent the ancient γ-WGT common to core eudicots, it does not display clear hexaploid characteristics, and its chromosome number is substantially reduced from the ancestral state of 1n = 21 [72]– [73]. We propose that, following the γ-WGT event, P. palustris underwent extensive structural rearrangements (Figure S5A, S5G), which disrupted the ancestral hexaploid architecture, erased traces of hexaploidization, and resulted in a re-diploidized genome. A similar evolutionary trajectory has been observed in other Saxifragales species. For example, T. polyphylla (Saxifragaceae), which experienced only the γ-WGT and subsequent structural rearrangements (Figure S5B, S5H), retains 1n = 7 chromosomes [46]; Paeonia ludlowii (Paeoniaceae), also having undergone solely the γ-WGT, currently has 1n = 5 [73]. These cases highlight the central role of structural rearrangements in reshaping karyotypes. Roles of KNOX genes in regulating leaf morphology KNOX I and KNOX II genes play antagonistic roles in regulating leaf morphological complexity. They exhibit distinct spatiotemporal expression patterns: KNOX I genes are highly expressed in the SAM but downregulated in mature tissues, whereas KNOX II genes show the opposite trend, with elevated expression in mature organs and suppression in the SAM [24–26, 74]. This functional antagonism is evident in leaf development. In A. thaliana, overexpression of KNOX I genes enhances leaf serration and lobing, while KNOX II overexpression suppresses marginal dissection [22, 27]. This regulatory opposition is conserved in compound-leaved species, such as S. lycopersicum, C. hirsuta, and M. truncatula, where KNOX I activity increases leaf complexity and KNOX II activity reduces it [15, 17]– [18]. KNOX genes also contribute to heterophylly in amphibious plants. KNOX I genes are linked to the development of complex leaf morphologies under submerged conditions. In M. aquaticum, MaKN1 is predominantly expressed in submerged leaves; similarly, in H. difformis and R. aquatica, HdSTM and RaSTM are upregulated in dissected submerged leaves compared to simple terrestrial leaves [18–21]. These findings suggest that KNOX I genes promote intricate leaf architectures by sustaining meristematic potential and modulating hormone signaling. In contrast, the role of KNOX II genes in heterophylly remains poorly characterized. In this study, we identified three KNOX genes in P. palustris that were differentially expressed between SLB and TLB. The KNOX I genes, i.e., PPA02C1640 and PPA03C1159, were upregulated in SLB, while the KNOX II gene, i.e., PPA01C1293, was significantly upregulated in TLB. This antagonistic expression mirrors patterns in established models such as A. thaliana [22, 27], supporting the hypothesis that heterophylly involves not only KNOX I activation but also competitive repression by KNOX II. Although these three KNOX genes are promising candidates for regulating heterophylly in P. palustris, their precise biological functions and regulatory mechanisms require further experimental validation. Conclusion In this study, we report the first chromosome-level genome assembly for P. palustris, which also represents the inaugural high-quality reference genome within the genus Proserpinaca (Haloragaceae). This assembled genome spans 385.32 Mb, achieves a BUSCO completeness of 95.61%, and comprises 59.89% repetitive sequences along with 22,756 protein-coding genes. Comparative genomic analyses indicate that P. palustris underwent the ancestral γ-WGT event without subsequent WGD. Although the species experienced ancient hexaploidization, its extant genome no longer exhibits typical hexaploid architecture. Our results suggest that extensive structural rearrangements have substantially reconfigured the genome, resulting in a re-diploidized karyotype that has largely obscured structural traces of ancient hexaploidy. Furthermore, we identified nine KNOX gene family members in the P. palustris genome and analyzed their expression patterns in SLB and TLB. Among these, three genes, i.e., PPA02C1640, PPA03C1159, and PPA01C1293, displayed significant differential expression, suggesting their potential role as key regulators underlying heterophyllous phenotypic plasticity in P. palustris. Supplementary Information Supplementary Material 1 Supplementary Material 1 Supplementary Material 2 Supplementary Material 2