Construction of high-density genetic map and QTL mapping of flowering traits in Perilla frutescens

Abstract

Perilla frutescens (L.) is an important plant in traditional Chinese medicine and ornamental horticulture. However, classical genetic studies in this species remain limited. To perform QTL mapping for its important traits, the whole-genome resequencing of an F2 segregating population (n = 140) was conducted to construct a high-density genetic map. The sequencing generated 2,865.46 Gb of clean data. And the resulting high-density genetic map comprised 1,321.55 cM across 20 linkage groups, consisting of 2,289 recombination bins that integrated 9,599,671 SNP markers. The map exhibited an average interval of 0.58 cM per marker. To identify candidate genes associated with flowering traits, integrative QTL analysis delineated four consensus QTL regions associated with flower bud differentiation and initial flowering time. The 122 candidate genes were identified in these regions, including PfBBX8, a member of the CCT domain-containing family. Gene Ontology (GO) enrichment analysis suggested that PfBBX8 may serve as a key regulator in photoperiod-mediated flowering pathways. Heterologous overexpression of PfBBX8 in Arabidopsis thaliana significantly promoted flowering, whereas a natural C/T mutation in PfBBX8 resulted in substantially delayed flowering relative to transgenic lines overexpressing the wild-type allele. Moreover, overexpression of PfBBX8 upregulated downstream flowering integrators such as CO and FT, indicating its role in activating flowering pathways. These findings provide valuable high-density genetic map and flowering regulation for molecular breeding in Perilla.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-08848-5.

Introduction

Perilla frutescens (L.), an annual herbaceous plant belonging to the Lamiaceae family, holds significant pharmacological and nutritional value. It is widely cultivated in Asian countries [1]. In recent years, growing research interest has focused on the bioactive compounds and nutraceutical potential of Perilla [2]. Alpha-linolenic acid (ALA), the primary constituent of Perilla seeds, accounts for 60.33–68.54% of the total seed composition [3]. With the rapid development of bioinformatics and genomics, genome sequencing and genetic map construction have become important approaches for studying genetic variation and biodiversity [3]. As an important medicinal and aromatic plant, Perilla frutescens has extensive research and application value in terms of medicinal properties, bioactive components, and stress resistance. Amid increasing attention to its genetic characteristics, the comprehensive re-sequencing of medicinal plant populations such as Perilla frutescens using high-throughput sequencing technology, together with the construction of their genetic maps, has become a research hotspot [1].Moreover, Perilla is a photoperiod-sensitive short-day plant and exhibits marked latitudinal adaptability [4]. The flowering phenotype critically determines metabolite biosynthesis and developmental regulation in Perilla [2].

In Arabidopsis thaliana, the FT protein moves from the leaves to the shoot apical meristem via the leaf phloem, inducing the transition from vegetative growth to reproductive growth. Hd1 is an ortholog of CO in Arabidopsis thaliana, encoding a zinc finger domain protein that promotes heading under short-day (SD) conditions but inhibits heading under long-day (LD) conditions. A recent study has shown that the CCAAT box is not essential for CO function, but rather enhances CO transcriptional activation. CO was the first cloned CCT gene that regulates flowering time in Arabidopsis thaliana through the photoperiod pathway, and it plays a core role in this pathway [5].

Quantitative trait loci (QTL) analysis serves as a classical genetic method for investigating the heritable components underlying polygenic traits [6]. Map-based cloning following QTL mapping has identified FLC in Arabidopsis thaliana, the Ve1 gene in maize (Zea mays), and the Ppd-D1 gene in wheat (Triticum aestivum) as key regulators of flowering time [7–9]. In Perilla, six flowering-associated QTLs were previously mapped in an F2 population; however, the functional characterization of these loci remains incomplete [10].

In this study, an F₂ segregating population was established from two parental lines with pronounced phenotypic differences. Whole-genome resequencing enabled the construction of a high-density genetic map, and quantitative trait loci (QTLs) for flowering time were identified. Key candidate genes were further characterized via heterologous expression in Arabidopsis thaliana. These findings lay a foundation for genetic studies of flowering regulation in Perilla.

Results

Population construction and variation detection

The hybrid population was created by crossing EF (early-flowering, low oil content, and red leaves) as the male parent with LF (late-flowering, high oil content, and green leaves) as the female parent (Table S1). The F1 plants were self-pollinated, resulting in a total of 140 individuals in the F₂ segregating population. Whole-genome resequencing was conducted on the two parental individuals and the F2 population lines, generating 2,865.46 Gb of data with an average alignment rate of 96.6% to the reference genome [2]. The average sequencing depth for the population was 15.55× (Table S2). A total of 1,203,258,251 single nucleotide polymorphisms (SNPs) were identified and annotated (Table S3). The distribution of SNPs was uneven across chromosomes (Fig. 1a). Specifically, Chromosomes 10 and 8 exhibited relatively high-density distributions, with 698,803 and 671,339 SNPs, respectively. In contrast, Chromosome 2 showed the lowest density, with only 425,118 SNPs (Table S3). Analysis of SNP annotations using SnpEff revealed that they were predominantly located in intragenic regions (75.33%), followed by upstream regions (10.92%) and exon regions (1.3%) (Fig. 1b, Table S4). The predominant types of base variations were G/A and C/T transitions (Fig. 1c), which were consistent with the resequencing results of natural Perilla populations [2].

Fig. 1.

Genomic Variation Profiling in the F₂ Population of Perilla. a Chromosomal distribution of SNPs. b Distribution interval type of SNPs. c The type of nucleotide variations

High-density genetic map construction

Using a sliding window approach, consecutive 100 kb genomic intervals with identical genotype patterns were consolidated into recombination bins. Potential segregation distortion was evaluated via χ² goodness-of-fit tests, and markers showing significant deviation (α = 0.05) were systematically excluded [11]. Collinear representative markers without recombination evidence were computationally merged, resulting in 2,289 high-confidence recombination bins. These bins were distributed across 20 linkage groups, with an average of 6.01 × 10⁷ SNPs per chromosome (Fig. 2; Table S5). The final integrated map spanned 1,321.55 centiMorgans (cM), a unit used to measure genetic distance based on recombination frequency, with a mean inter-marker resolution of 0.58 cM. Marker density also varied across chromosomes [12](Table S5).

Fig. 2.

High-density genetic map of Perilla frutescens F₂ population each blue line represents a bin

Flowering characterization of F₂ Perilla Population

Flowering time is a pivotal agronomic trait governing metabolite biosynthesis and developmental regulation in Perilla [13]. Three flowering phases—flower bud differentiation (FBD), initial flowering (IF), and final flowering (FF)—were quantified and recorded in the parents and F₂ Perilla population (Fig. 3a). There was significant divergence in flowering time between the parental lines.The male parents were early-flowering (EF) lines with FBD: 156 ± 1.8 d, IF: 168 ± 2.1 d, and FF: 176 ± 2.4 d,, whereas the female parent were late-flowering (LF) lines, with 14 d, 13 d and 12 d advances in FBD, IF, and FF, respectively (Fig. 3b-d).

Fig. 3.

Flowering Characterization in parental lines and F₂ Population of Perilla. a the phenotypic characterization of two parents. early-flowering (EF) and late-flowering (LF). b, c, d Statistical analysis of flower bud differentiation (FBD), initial flowering (IF) and final flowering (FF) of two parents. e Normal distribution and correlation analysis of FBD, IF, and FF traits in the F2 population. (The upper triangle shows the results of the correlation test (including significance and correlation coefficients). ***: p < 0.001; **: p < 0.01; *: p < 0.05. The diagonal shows the frequency distribution histogram of each trait, and the lower triangle shows the correlation scatter plot of each pair of phenotypes, with the straight line representing the linear regression equation.)

The F₂ population also exhibited significant segregation of flowering traits, with the ranges of FBD, IF, and FF spanning 151–176 d, 157–186 d, and 179–200 d, respectively. Quantitative trait distribution analysis indicated that the three flowering traits presented a normal distribution (Table 1; Table S6; Fig. 3e). All three traits conformed to the expected quantitative genetic inheritance patterns. Additionally, a strong positive correlation was observed between the FBD and IF stages (r = 0.87, Pearson’s correlation; P < 0.001) (Fig. 3e).

Table 1.

Statistical analysis of flowering characterization in the F₂ population of Perilla

Characteristics	EF	LF	F2
			Minimum	Maximum	Mean	Stdev	Skewness	Kurtosis	CV1(%)
FBD	156	170	151	176	162.49	5.31	-0.60	-0.02	0.03
IF	168	181	157	186	169.54	6.56	0.32	-0.52	0.04
FF	176	192	179	200	189.82	5.00	-0.54	-0.53	0.03

CV¹(%): Coefficient of variation, reflecting data dispersion and precision

QTL mapping of flowering time

Through inclusive composite interval mapping (ICIM) analysis, eight quantitative trait loci (QTLs) governing the phenology of FBD and IF were identified. The cM (centiMorgan) is a unit of genetic distance used in genetic mapping and QTL analysis.Among these, the four FBD-associated QTLs (qFBD-1 to qFBD-4) were mapped to Chromosomes 5 (31 cM), 7 (23 cM), and 17 (15/30 cM). They exhibited a mean LOD score of 5.96 ± 0.83 and collectively explained 45.6% of the phenotypic variance. qFBD-3 demonstrated peak significance (LOD = 7.13, P < 0.001), contributing 13.7% to the variance (Fig. 4a; Table 2). The four IF-associated QTLs (qIF-1 to qIF-4) were also localized to Chromosomes 5 (31 cM), 7 (24 cM), 17 (30 cM), and 18 (31 cM). As a major-effect QTL, qIF-1 showed exceptional significance (LOD = 9.42, P < 0.0001), accounting for 18.5% of the phenotypic variation (Fig. 4b; Table 2).

Fig. 4.

QTL localization of flowering traits. a QTL localization for FBD. b QTL localization for IF. c Venn diagram of overlapping candidate genes. d Distribution interval type of SNP variations in co-localized regions. e GO enrichment of common genes in co-localized regions

Table 2.

QTL localization interval in the population

	QTL	Chromosome	Position(cm)	Flanking Markers	LOD2	PVE3 (%)	Add4
FBD	qFBD-1	5	31	Chr5P48137285-Chr5P47487594	5.7962	11.1779	-2.4189
	qFBD-2	7	23	Chr7P36262804-Chr7P36096118	7.1298	13.7373	2.7586
	qFBD-3	17	15	Chr17P48306020-Chr17P48164171	6.5769	12.6195	0.6753
	qFBD-4	17	30	Chr17P35018000-Chr17P34472337	4.3312	8.0867	-1.9261
IF	qIF-1	5	31	Chr5P48137285-Chr5P47487594	7.1834	14.1103	-3.587
	qIF-2	7	24	Chr7P36096118-Chr7P35896951	9.421	18.4592	3.963
	qIF-3	17	30	Chr17P35018000-Chr17P34472337	6.0328	11.4688	-3.0525
	qIF-4	18	31	Chr18P14878547-Chr18P14339293	4.0927	7.4507	2.4911

LOD² (Logarithm of Odds): An indicator for evaluating QTL localization reliability; LOD ≥3 is generally considered significant

PVE³ (Phenotypic Variation Explained, %): The percentage of target trait variation explained by a single QTL, reflecting its effect size

ADD⁴ (Additive Effect): The average effect of allele substitution at QTL; positive for male parent allele, negative for female parent allele

Further investigation into the variat distribution revealed that the majority of FBD-related QTLs were predominantly located in intronic regions (49%), followed by upstream and downstream gene regions (28.6%), and exonic regions (22.2%) (Figure S1). The distribution of IF-related QTLs was found to be in introns (37.8%), upstream and downstream regions (38.8%), and exons (21.3%) (Figure S2). A total of 135 and 192 candidate genes were identified in the QTL regions of FBD and IF, respectively.

Interestingly, QTLs for FBD and IF shared common genomic regions on Chromosomes 5 and 17, with 122 loci putatively involved in the regulation of both traits (Fig. 4c). The common variats were primarily concentrated in the intron variant (37.8%) (Fig. 4d). Gene ontology (GO) enrichment analysis was performed to summarize the genes in QTL regions into their GO classes, and the genes related to both traits were mainly categorized into the term of long-day photoperiodism in flowering (GO:0048586) for subsequent candidate gene selection. (Fig. 4e, Tables S7 and S8). There are more genes associated with flowering traits, including RING/FYVE/PHD-type zinc finger proteins, U-box proteins, and AAA-type ATPase family proteins. These have been reported as homologous transcription factors that influence flowering by regulating hormone signaling and environmental responses in Arabidopsis and other plants [5](Table S9).

Variation analysis of candidate gene

Variation analysis of GO enrichment genes revealed that two core regulators may exert a major effect on the flowering time of Perilla. The CCT family member gene (Chr07_36110108_36113567) contained exon-localized missense mutations, while the zinc finger gene (Chr05_48096409_48099620) exhibited synonymous coding changes that potentially influence mRNA splicing efficiency [14] (Tables S9-S10).

Given the important roles of CCT family genes in plant flowering, further evolutionary and variation analyses were conducted. A total of 57 proteins containing the CCT domain were identified in the Perilla genome (Figure S3). A phylogenetic tree was constructed with Arabidopsis CCT proteins, which were classified into three subfamilies: the CONSTANS-like subfamily (COL), Pseudo Response Regulator (PRR), and CCT Motif (CMF) (Fig. 5a). The gene chr07_36110108_36113567 belongs to the COL subfamily, which consists of two B-box domains and one CCT domain. Due to its high homology with AtBBX8 (AT5G48250.1), the gene was renamed PfBBX8 (Figure S4).

Fig. 5.

Phylogenetic analysis of the CCT Family and genotyping variation of PfBBX8. a Phylogenetic tree of the CCT family in Perilla and Arabidopsis. b Identification of genetic variations in PfBBX8. c, d Genotyping analysis of PfBBX8 variants associated with flower bud differentiation (FBD) and initial flowering (IF) phenotypes. e, f Protein structure prediction of C-T mutation in PfBBX8

Further variation analysis revealed a non-synonymous C/T variation in the second exon and a G/A variation in the intron of PfBBX8 (Fig. 5b). Focusing on the non-synonymous variation in the exon, the homozygous C allele was present in the early-flowering (EF) parents, while the homozygous T allele was found in the late-flowering (LF) parents. In the F₂ population, the genotype frequencies at this locus were CC: C/T: TT = 35:71:34. (Table S11). Genotyping analysis showed that the average flower bud differentiation (FBD) was 161 days, 163 days, and 167 days for the CC, C/T, and TT genotypes, respectively. The genotyping of initial flowering (IF) also exhibited a similar pattern (Table S12). The C / T variations significantly delayed flowering time in Perilla (Fig. 5c, d). Further analysis revealed that the SNP variation resulted in an amino acid change from proline to leucine, leading to alterations in the local helix and tertiary structure of the PfBBX8 protein (Figs. 5e, f; Table S12).

Functional validation of PfBBX8

To verify the gene function and the effect of locus variation, PfBBX8 was cloned and the C/ T mutation was constructed. Both over-expression vectors were transformed into Arabidopsis thaliana, respectively. Transgenic lines were successfully obtained via resistance screening and PCR verification (Figure S5). Under similar cultivation and photoperiod conditions, transgenic PfBBX8 plants significantly promoted flowering. The average FBD time and IF time were 5–6 days earlier compared to wild-type lines (Col-0) (Fig. 6a). The flowering time of transgenic PfBBX8 with the mutation was later than that of transgenic PfBBX8 plants but earlier than that of Col-0, suggesting that the mutation reduced the genetic effect. This trend was consistent with the mutation detection results in the Perilla population (Fig. 6b, c).

Fig. 6.

Phenotypic Traits and Gene Expression in Transgenic PfBBX8 and PfBBX8-mutant Arabidopsis. a Phenotypic Traits differences; (b) and (c) The flower bud times and flowering times in Transgenic PfBBX8 and PfBBX8-mutant Arabidopsis; d. Gene expression levels in transgenic PfBBX8 plants. e Mode of PfBBX8 promoting early flowering mechanism. Up-regulated expression genes are indicated in red, while down-regulated expression genes are indicated in green. Solid black arrows represented activation, bars represented inhibition, and dashed lines indicatde speculative functionality. AtFT (AT1G65480.2), AtCO (AT5G15840.1), AtAP1 (AT1G69120.1), AtSOC1 (AT2G45660.1), AtFUL (AT5G60910.1), AtLFY (AT5G61850.2), AtFLC (AT5G10140.1), AtSVP (AT2G22540.1), and AtPRR9 (AT2G46790.1)

Furthermore, four lines of transgenic plants overexpressing PfBBX8 were selected to analyze the transcript levels of key flowering-related regulatory genes (Fig. 6d; Table S13). Notably, the core flowering regulator AtCO was significantly upregulated in transgenic PfBBX8 plants. Additionally, overexpression of PfBBX8 resulted in the upregulated expression of AtFUL, AtSOC1, and AtLFY, which are important flowering promoter factors. In contrast, the expression levels of flowering repressors, including AtFLC, AtSVP, and AtPRR9, were suppressed in these transgenic PfBBX8 plants. This regulatory pathway involving PfBBX8 was established to elucidate its mechanism of promoting flowering in Arabidopsis (Fig. 6e).

Discussion

High-density genetic maps are essential for QTL localization and genetic breeding. Recent advancements have been made in the development of high-density genetic maps for medicinal plants [15–17]. For example, using SRAP and SSR marker methods, Dendrobium officinale genetic map, constructed from 140 individuals with 157 markers, comprises 27 linkage groups with a total map distance of 1,580.4 cM and an average interval of 11.89 cM, using SRAP and SSR marker methods [18]. Similarly, Aquilegia viridiflora comprising 141 markers, forms 7 linkage groups with a total map distance of 1,073.4 cM and an average interval of 8.3 cM using AFLP and SSR methods [19]. The hybrid population of Perilla is challenging to establish due to its strict self-pollination and small flower size. Fortunately, we successfully established an F2 hybrid population in Perilla. Based on whole-genome resequencing, a high-density genetic map of Perilla was constructed, with a total distance of 1,321.55 cM and an average interval of 0.58 cM across 20 linkage groups. The genetic map incorporates a total of 9,599,671 SNPs, with an average of approximately 479,983 SNP markers per chromosome. This high-quality genetic map of Perilla, derived from an F₂ population, will mainly provide a foundation for our subsequent genetic localization research using other trait data from the same experiment. Flowering in plants is regulated by multiple factors and precise regulatory networks [20]. Understanding the genetic diversity that influences flowering time and photoperiod is crucial for determining growth periods, yields, and metabolite accumulation. Here, eight QTL loci for FBD and IF were successfully identified in our population. The phenotypic contribution rates of the detected minor-effect QTLs of FF in this study were all below the threshold, so no major-effect QTLs were identified. PfBBX8, a CCT family gene, was identified based on GO enrichment and variation annotation analysis.

Genes containing the CCT domain are major genetic determinants that influence flowering time, morphological traits, and biotic/abiotic stresses [5, 21]. Both CO and BBX proteins belong to the COL subfamily, which are well-known flowering time regulators in various crops, including rice [14], wheat [22] and maize [20]. Previous research has demonstrated that AtBBX8, a B-box zinc-finger transcription factor, can interact with CO transcription factors to control flowering time and respond to photoperiod in Arabidopsis [23–25]. CmBBX8 in chrysanthemum acts as a floral transition regulator; its overexpression can influence circadian rhythms and accelerate flowering [25].

Regarding the C/T mutation in PfBBX8, functional validation (Fig. 6) showed the mutant form still promotes flowering vs. Col-0 but less effectively than the non-mutant form, indicating partial loss of function rather than complete loss. Protein alignment with orthologs from other species confirmed the Proline-to-Leucine mutation occurs in a conserved domain, highlighting this residue’s critical role in PfBBX8 activity (Fig. 5b, Figure S4).

PfBBX8, which consists of two B-box domains and a CCT domain, belongs to the COL subfamily and exhibits 65% homology with AtBBX8 and AtCO. The heterologous expression data in Arabidopsis suggest that PfBBX8 may regulate flowering time by modulating the expression of the CO/FT module Key flowering factors were significantly activated in PfBBX8-transgenic Arabidopsis, including AtFT, AtCO, AtAP1, AtFUL, AtSOC1, and AtLFY. AtCO mediates photoperiodic signals by directly activating the transcription of the AtFT gene in Arabidopsis [26]. AtAP1 is a transcription factor involved in flower development, while AtFUL regulates inflorescence development [27]. In contrast, the expression levels of flowering repressors were significantly downregulated. AtFLC and AtSVP negatively regulate AtFT and AtSOC1, delaying flowering [28]. Additionally, AtPRR9 is a clock gene that regulates photoperiod response [29]. These results suggest that the expression of PfBBX8 can influence downstream flowering genes to promote flowering.

The F₂ segregating population used in this study consisted of 140 individuals, which is a reasonable size for initial QTL mapping studies, yet it may impose certain constraints on effective recombination resolution, thereby potentially affecting QTL mapping precision and candidate gene selection efficiency [30]. In genetic mapping, the frequency of recombination events (including rare ones) is associated with population size, and a moderate population size may lead to relatively fewer recombination breakpoints in specific genomic regions.

To further improve the precision of QTL mapping and the accuracy of candidate gene selection in future research, we plan to expand the population scale (e.g., constructing F₃ populations or recombinant inbred line (RIL) populations with larger sample sizes) [31]. This will help increase the number of recombination events, improve recombination resolution, narrow QTL intervals, and provide more robust evidence for elucidating the genetic mechanism underlying flowering traits in Perilla. Heterologous expression in Arabidopsis suggests that PfBBX8 may regulate flowering time by modulating the expression of the CO/FT module. However, the direct regulatory relationship between PfBBX8 and its CO/FT orthologs in Perilla remains to be verified through in planta experiments, such as yeast two hybrid (Y2H), chromatin immunoprecipitation (ChIP), and dual luciferase reporter assays in Perilla itself.

Experimental procedures

Population construction and flowering investigation

Using EF (early-flowering, low oil content, and red leaves) as the male parent and LF (late-flowering, high oil content, and green leaves) as the female parent, two cultivars of Perilla were crossed using the emasculation-hybridization method to establish a hybrid population. The F₁ generation was self-crossed to obtain the F₂ generation. Parental plants and individual F₂ plants were cultivated in an experimental field in Guangzhou (113°35′E, 22°95′N). Specifically, first bud differentiation (FBD) refers to the initial stage when the apical meristem of the plant transforms into a floral meristem; initial flowering (IF) is defined as the first flower on a single plant starts to open; final flowering (FF) representing the end of the flowering period.

Resequencing and SNP calling

Leaf samples from parental plants and F₂ population lines were collected, and total DNA was extracted using the CTAB method. After quality control, DNA was sequenced on the DNBSEQ-T5 platform for paired-end sequencing. Sequencing data were filtered and aligned to the Perilla reference genome, followed by variant detection [32]. Variant calling followed the GATK (v4.0.3.0) best practices pipeline: (1) Raw reads were quality-filtered using Trimmomatic (v0.27); (2) Filtered reads were aligned to the Perilla reference genome [2] using BWA-MEM (v0.7.12) with default parameters; (3) Aligned BAM files were sorted and deduplicated; (4) HaplotypeCaller was used to generate GVCF files for each sample; (5) GVCF files from all 140 samples were merged for joint genotyping to identify SNPs and InDels. The identified SNPs and InDels were further filtered using the following criteria. For SNPs, variants with QD < 2.0, FS > 60.0, MQ < 40.0, SOR > 3.0, MQRankSum < − 12.5, or ReadPosRankSum < − 8.0 were removed. For InDels, variants with QD < 2.0, FS > 200.0, SOR > 10.0, MQRankSum < − 12.5, or ReadPosRankSum < − 8.0 were excluded. In addition, genotype calls with a sequencing depth < 2 or > 50, variants with more than two alleles, and variants with a missing rate > 10% or a minor allele frequency (MAF) < 0.05 were also removed. Only uniquely mapped reads were retained for SNP calling and genotyping [18]. Parental filtering retained variants that were homozygous and polymorphic between parents. Variants were annotated using SnpEff (version 4.3r) for genomic positions (intergenic regions, exons, introns, splice sites, UTRs, and 1-kb flanking regions) and functional effects (nonsynonymous mutations, start/stop codon changes, and splicing variants).

Construction of linkage maps

SNP genotypes were coded in QTL IciMapping 4.2 as A (maternal genotype), B (paternal genotype), H (heterozygous), and X (missing data). SNP markers with > 20% missing data were removed using the BIN function. Linkage maps were constructed using the TwoOpt algorithm [33].

QTL mapping analysis

The constructed linkage maps were analyzed using inclusive composite interval mapping (ICIM) in QTL IciMapping v4.2 to identify QTLs associated with initial and full flowering periods. A LOD threshold of 3.5 was established, controlling the global alpha error at 0.05 through 1,000 permutation tests [34]. ICIM analysis employed a 0.1 cM mapping step, with identified QTLs receiving standardized nomenclature [35]. The support intervals for these QTLs were calculated using a 1.5-LOD drop interval.

Identification of candidate genes

Candidate genes within major QTL confidence intervals were annotated using Blast2GO 4.0 and the Phytozome database for Gene Ontology (GO) classification. GO enrichment analysis of candidate genes was visualized using OmicShare tools (https://www.omicshare.com/tools).

Gene evolution and typing analysis

Protein domain analysis was performed by retrieving Hidden Markov Model profiles from the Pfam database. Candidate gene families were characterized using SMART and CDD databases for sequence alignment and domain identification. Multiple sequence alignments were conducted with MUSCLE in MEGA 7.0, followed by phylogenetic tree construction using the neighbor-joining method with 1,000 bootstrap replicates to assess branch support. Phylogenetic trees were visualized using iTOL (https://itol.embl.de/) and Evolview (https://www.evolgenius.info/evolview/). Gene structure variations were analyzed with IBS 2.0 (https://ibs.renlab.org/), while protein tertiary structures were predicted and compared using SWISS-MODEL (https://swissmodel.expasy.org/).

Genetic transformation and functional validation of candidate genes

The full-length cDNA sequences of candidate genes were cloned into the pCAMBIA1301 vector downstream of the CaMV 35 S promoter. Following PCR verification, the recombinant vectors were transformed into Agrobacterium tumefaciens strain GV3101 and introduced into Arabidopsis thaliana via floral dip transformation. Transgenic Arabidopsis plants were selected on hygromycin-containing medium, with positive transformants confirmed by PCR screening. Plants were grown under controlled conditions: the fully expanded true leaf was sampled from 4-week-old transgenic Arabidopsis seedlings grown under long-day conditions (16 h light/8 h dark, 22 °C). At this developmental stage, the seedlings enter the floral transition period [6].​.

Flowering time phenotypes of transgenic Arabidopsis lines were recorded to assess the functional effects of candidate genes. Gene expression analysis was performed by quantitative real-time PCR (qRT-PCR). Total RNA was extracted using the Sigma RT-50 kit, followed by cDNA synthesis with AMV reverse transcriptase (Promega A3500). qRT-PCR amplifications were conducted using SYBR Green chemistry.​.

For qRT-PCR quantification, the Arabidopsis actin gene served as the endogenous reference gene. Relative gene expression levels were determined using the comparative ΔΔCt method (2^−ΔΔCt), calculating fold-change values normalized to the actin reference and relative to control samples [35].

Authors’ contributions

Q.S., T.W. and G.-Y.Z. designed the experiments; X.-Z. Z. and W.-H. Z performed the experiments. M.-T.L., G.-W.X., D.W.analysed the data. X.-Z. Z wrote the manuscript. All the authors have reviewed and approved the manuscript.

Funding

This research was supported by Natural Science Foundation of China (Grant No. 32570417, 32400298, U22A20446); The Outstanding Young Talent Project of the Guangdong Special Support Program, Health Commission of Guangdong Province (No. 0820250223); Postdoctoral fellowships (363416).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.