Elucidating Genetic Mechanisms of Summer Stress Tolerance in Chinese Cabbage through GWAS and Phenotypic Analysis

Abstract

The development of Chinese cabbage (Brassica rapa subsp. pekinensis) varieties that are resilient to climate change is becoming increasingly critical. Our study focused on developing stress-tolerant Chinese cabbage during the summer season to minimize the impacts of global climate change. We evaluated 52 Chinese cabbage accessions grown in the field during the late spring–summer season in Korea. Various phenotypic data of Chinese cabbage in adverse environments were collected from field data. In addition to field screening, we used a controlled-environment chamber to observe the direct impact of heat on young plants, particularly in the root area. A genome-wide association study was conducted using two sets of phenotypic data collected from both chamber and field studies and genotype data acquired from GBS analyses. A total of 57 SNPs distributed across all 10 B. rapa chromosomes were selected to be highly related to the target traits. Several candidate genes were annotated using the flanking sequences of these SNPs. The study revealed that most of the annotated genes seemed to be highly connected with the function of the heat stress response. Other genes were also found to be related to the environmental stress response and flowering. These candidate SNPs and genes can provide valuable tools for breeders to develop summer stress-tolerant Chinese cabbage varieties.

1. Introduction

Brassica rapa is a widely cultivated vegetable that is primarily grown in Asia and Europe. Among its ten subspecies, B. rapa subsp. pekinensis (Chinese cabbage) is consumed in east Asian countries such as China, Japan, and Korea. Among Brassica species, Chinese cabbage (Brassica rapa L. ssp. pekinensis) is used as the main ingredient of the fermented food kimchi and remains an integral part of Korean cuisine; it is served with almost every meal in Korea [1].

Globally, crop production is currently threatened by global warming. Over the past three decades, the Northern Hemisphere has experienced its warmest period in 1400 years [2]. Linear trends indicate increasing average maximum and minimum temperatures per decade from 1980 to 2011, with an average increase of 0.3 °C for Tmax and 0.2 °C for Tmin [3,4]. Moreover, climatological extremes, such as heatwaves, are expected to occur more frequently [5]. High temperatures pose a significant environmental constraint on normal crop growth, making breeding tolerant varieties more critical [3,6].

Heat stress negatively affects root elongation, leaf development, plant growth, and reproduction, leading to decreased crop yields due to shorter life cycles and accelerated senescence [7,8]. Although tremendous progress has been made in expanding genomics technologies and crop genome sequencing, the impact of genomics data on crop improvement is still far from satisfactory, largely due to a lack of effective phenotypic data [9,10].

In general, a moderate increase in air temperature accelerates plant development, but shortens the overall crop duration. Genetic and physiological mechanisms can be altered by heat stress [11]. Warm-season annuals tend to tolerate higher temperatures than cool-season annuals. Several studies have described the effects of high temperatures on cool-season annuals, such as Brassica crops, noting growth anomalies in lettuce at soil temperatures exceeding 32 °C [12] growth delays and adverse impacts on photosynthesis in Brassica juncea L. [13], and hindered seed production in B. napus due to heat stress affecting micro- and megagametophyte fertility [14].

Phenotyping is a pivotal process in precise breeding. It is also a very important step in developing useful selection markers and genomic selection systems. Despite its importance, high-throughput field phenotyping presents considerable challenges [10,15]. Improving plant tolerance is an important strategy for combating climate change; however, determining which crops are and are not tolerant in the summer season is not easy. This tolerance combines several environmental tolerances such as heat, drought, and humidity. Even if we focus on heat stress alone, the induced damage may differ among varieties. Some plants may show chlorosis, whereas others may show calcium deficiency or blight. Phenotyping in the open field is a complicated task. The phenotype can be affected by weather conditions, the soil environment, and nearby pests. This is a difficult aspect for researchers who want to distinguish the symptoms caused by heat stress from other physiological disorders.

However, since Chinese cabbage is mostly grown in open fields, it is necessary to evaluate the resources in such open fields under various environmental conditions. Accumulated phenotypic data from different environmental conditions are important for characterizing new breeding resources with the development of high-throughput genotyping technologies.

In this study, we collected phenotyping data from both open fields and controlled environments, focusing on developing selection criteria for tolerant Chinese cabbage breeding in response to climate change. The phenotype data of 52 different Chinese cabbage accessions were collected during the late spring–summer season. The root phenotype data of the plantlets under heat stress conditions were also collected. RGB image data of the roots were analyzed using the Image J program. A correlation study was performed using SNPs from genotyping-by-sequencing strategies and a genome-wide association study (GWAS). Useful candidate genes were studied after an analysis of the key SNP flanking sequencing. The obtained list of SNPs and candidate genes that might be related to the heat stress response can be used as valuable genotyping tools in summer-season suitable Chinese cabbage breeding. The characterized Chinese cabbage accessions with varying degrees of heat tolerance can also be used as tolerant breeding materials.

2. Materials and Methods

2.1. Plant Material Selection and Genomic DNA Extraction

A total of 52 accessions were sown and cultivated in a greenhouse at the National Institute of Horticultural and Herbal Science, Korea, during 2022–2023. A list of the accessions is shown in

Table 1. The list of the Chinese cabbage accessions used in this study.

Number	Accession ID	Traits
HS1	Wongyo20034ho	Light yellow inner leaf, light green outer leaf Semi-opened, not firm head with low leaf number Bumpy top Clubroot race 4 resistance
HS2	Wongyo20035ho	Yellow numerous inner leaf, green outer leaf Semi-opened, firmness in middle Good round-shaped top
HS 3	Wongyo20036ho	The outer leaves are oblong oval in shape and long, the outer leaves are pale green and the inner leaves are light yellow Semi-opened Upright and tall plant stature The length of the outer leaf midrib is long, and the shape of the bulb tip is bumpy
HS 4	Wongyo20037ho	Light green outer leaf and yellow inner leaf Closed, small head type Clubroot race 4 resistance late flowering and high glucosinolate contents
HS 5	Wongyo20038ho	Light yellow inner leaf Unrigid inner head Elongated, big type Clubroot race 4
HS 6	Wongyo20039ho	Yellow inner leaf Firm, round head Small size, normal kimchi cabbage type Clubroot race 4 resistance Early maturing
HS 7	Wongyo20040ho	Deep green outer leaf and pale yellow inner leaf Opened head type Clubroot race 11 resistance
HS 8	Wongyo20041ho	Golden yellow inner leaf Fully closed head type with numerous inner leaves, small size Clubroot race 11 resistance
HS 9	Wongyo20042ho	The outer leaves are pale green and the inner leaves are yellow Small, firm, closed type head Round top, small leaf size Clubroot race 4 resistance
HS 10	Wongyo20043ho	Green outer leaf and yellow inner leaf Big closed, unrigid head type High glucosinolate contents
HS 11	Wongyo20045ho	The outer leaves are deep green and the inner leaves are light yellow Small, firm, closed type head Round top, small leaf size Clubroot race 4 Early maturing
HS 12	Wongyo20046ho	Yellow inner leaves Closed type head, low firmness Long, narrow shape Clubroot race 9
HS 13	Wongyo20047ho	Green outer leaves and deep yellow inner leaves Closed type head, low firmness Long, narrow shape High adaptability in heat stress/summer season
HS 14	Wongyo20048ho	Green outer leaves and yellow inner leaves Closed type head, low firmness Small size, erected leaf Early maturing, heat stress resistance
HS 15	Wongyo20049ho	Deep green leaves Open type Thick root, high glucosinolate contents
HS 16	Wongyo20050ho	Deep green leaves Open type Thick leaf, low leaf number High glucosinolate contents
HS 17	Wongyo20051ho	Green outer leaves, yellow inner leaves Firm, closed type, small size Cold tolerance
HS 18	Wongyo20052ho	Light green outer leaves, pale yellow inner leaves Firm, closed type with many leaves, round shape, small size Very early maturing No leaf hair Clubroot race 9 resistance
HS 41	16-FFB128-2F	Yellow inner leaves Rigid, closed type Normal kimchi cabbage type
HS 42	17-FC02-1	Pale yellow inner leaves Half-closed head Heat tolerance
HS 43	17-FC66-1	Golden yellow inner leaves Half closed, unrigid Heat tolerance
HS 44	Wongyo20053ho	Small, extremely early maturing Green outer leaves, yellow inner leaves Showing moderate resistance to various kinds of insets and extreme weather events
HS 45	17-FC69-1	Closed head, numerous yellow inner leaves, heat tolerance, summer-season adaptability
HS 48	RFC106-2	Half-closed head, light green outer leaves
HS 49	17-RFC145-3	No heading, deep-green-colored outer leaves, light yellow inner leaves
HS 50	18-BD32	White inner leaves, green outer leaves, long closed head
HS 51	18-BD109	Closed head type, yellow inner leaves, dark green outer leaves, unrigid
HS 52	19-FQ48	Closed head type, light yellow inner leaves, green outer leaves
HS 53	19-FQ49-1	Yellow inner leaf, small closed head type, unrigid, green outer leaves
HS 54	19-FQ53	Small closed head type, yellow inner leaves, green outer leaves
HS 55	19-FQ55-1	Yellow inner leaf, dark green leaf, numerous inner leaves, closed head type, adaptable in summer season
HS 56	19-FQ58	Yellow inner leaves, unrigid/closed head type, dark green outer leaves
HS 57	19-FQ108	Opened head type, green-colored leaves
HS 58	19-FQ145-1	Similar to turnip, dark green outer leaves
HS 59	19-FQ154	Yellow inner leaf, dark green outer leaf, numerous inner leaves, closed head type, adaptable in summer season
HS 60	19-FQ155	Yellow inner leaf, green outer leaves, numerous inner leaves, closed head type, adaptable in summer season Heat tolerance
HS 61	19-FQ156	Yellow inner leaf, dark green outer leaf, numerous inner leaves, closed head type Heat tolerance, summer-season adaptability
HS 62	15-CDYB2-7	Adaptable in summer season, yellow inner leaves, dark green outer leaf, small closed head type
HS 64	namdo13-2-6	Yellow inner leaves, dark green outer leaves, closed head type, early maturing, numerous inner leaves, adaptable in summer season
HS 68	18-BD85	Half-closed head type, light yellow inner leaves, green outer leaves
HS 69	15-CDYB3-1(DA180)	Adaptable in summer season, drought tolerance, half-closed head type, light green outer leaves
HS 70	16-GW06	Similar turnip, dark green outer leaves
HS 83	15-FD29	Light yellow inner leaves, light green outer leaf numerous inner leaves, adaptable in summer season
HS 85	19-FQ27-2	Closed head type, light yellow inner leaves, green outer leaves
HS 88	17-WA10-4	Half-closed head type, light green inner leaves, light green inner leaves, green outer leaves, numerous inner leaves
HS 92	18-FH115-1	Yellow inner leaves, dark green outer leaves, numerous inner leaves, adaptable in summer season, numerous inner leaves
HS 93	17-FC82-2	Yellow inner leaves, closed head type, Early–mid maturing Heat tolerance, summer-season adaptability
HS 94	18-FH119-2	Light yellow inner leaves, adaptable in summer season, dark green outer leaves, closed head type, early maturing
HS 96	17-FE101	Yellow inner leaves, unrigid closed head type, green outer leaves, unrigid
HS97	17-FE102	Opened head type, dark green outer leaves, adaptable in summer season
HS98	13-CAC14-19	Yellow inner leaves, green outer leaves, half-closed head type, elongated shape
HS99	15-CDYB7-7	Yellow inner leaves, green outer leaves, closed head type, adaptable in summer season

. Genomic DNA was extracted from young fresh leaf tissues using the cetyltrimethylammonium bromide (CTAB) method. The extracted DNA samples were diluted to a concentration of 20 ng/µL for library preparation. The DNA concentration and purity were determined using an Assay-NanoDrop (Denovix Inc., Wilmington, DE, USA) and verified through gel electrophoresis. Two sets of Chinese cabbage accessions were prepared: one set was tested at the young plant stage in a controlled-environment room, and the other set was transplanted into the field after 30 days (days after seeding).

2.2. Phenotyping of Young Chinese Cabbages Using Environmental Control Room

The goal of this experiment was to evaluate heat resistance during the early stages of Chinese cabbage cultivation. Most Chinese cabbages used for kimchi production are harvested in November, making heat stress during the transplantation period a significant issue. To simulate typical early to mid-August conditions in Korea, an artificial environmental chamber was used. The seeds were sown at intervals in a 105-cell tray. After 38 days, young Chinese cabbage plants were transferred to the chamber for 18 days.

Under heat stress conditions, the temperatures started at a minimum of 23 °C, increased from 7 AM, peaked at 40 °C between 1 PM and 3 PM, and then decreased to the minimum temperature. LED lighting conditions simulated natural sunset phenomena, while the humidity was maintained at 60% and the CO₂ concentration at 450 ppm. TH3 LED light intensity ranged from 0 pmol to 1180 pmol. For the control, the temperature was kept constant at 23 °C without diurnal fluctuations, and all other parameters were identical to those in the heat stress treatment (Supplementary Table S1). The effects on the root system under heat stress were examined using a photo box to capture images of the roots and soil from the trays. Images were taken from one side of the root plug using a smartphone, showing the densest root distribution (Samsung Galaxy S22 Ultra, manufactured in Samsung, Seoul, Republic of Korea). Four individual plants from each plant line were photographed to minimize experimental errors. The images were trimmed for easier analysis using the ImageJ software bundled with 64-bit Java 8, which employs a threshold-based pixel count measurement to calculate leaf area [16,17,18,19,20,21].

The root system shape was recognized using ImageJ, and the soil and root areas were calculated. The target root area and calibration area were kept similar to minimize errors from lens distortion. The fixed distance between the digital camera and the target plant was 1020 mm, with a blue-colored PVC screen used as the background. Misidentified small groups of background pixels were filtered before the area calculation. The proportion of the total area occupied by the roots was then computed to determine the differences in root biomass among the resources.

2.3. Phenotyping Mature Chinese Cabbage in the Open Field during the Summer Season

Several physiological traits were investigated to analyze the stress levels induced by inappropriate summer environments (Figure 1). The stress levels of the 52 Chinese cabbage accessions were analyzed based on data from each trait. Seven traits affecting Chinese cabbage quality were selected: calcium deficiency, boron deficiency, heading, early death, inner rot, insects (aphid, striolata, etc.), and disease resistance (downy mildew, bacterial soft rot). Axillary bud and flowering ratios were also screened. Calcium deficiency symptoms were assessed based on several significant symptoms described as follows: new leaf tip drying, growth point rotting, tip burn phenomena, and tissue softening. Symptoms were recorded as severe, medium, or weak, based on visual observations. When symptoms were rarely present in the plant, we screened the plant as ‘weak’. The ‘medium’ category meant 5 or fewer plants showing symptoms among 20 plants. The ‘severe’ plants showed evident symptoms in over 90% of individuals. Phenotypes were labeled as “weak (0%), medium (50%), and strong (100%)” for ease of calculation. Boron deficiency symptoms were measured by the presence of black lines in the white stem part of the outer leaves, sometimes appearing as cracks, and labeled as “present (100%) or absent (0%)”. Soft rot symptoms caused by bacterial or environmental factors were also measured. Rotting typically starts from the upper root and moves upward, and the phenotype was scored by calculating the ratio of rotten individuals to the total plants.

Axillary buds and flowering were scored by calculating the ratio of specific individuals to total plants. Pest and disease damage was assessed and scored.

Three cases of heading patterns were observed: fully closed, half-closed, and open. Phenotypes were labeled as “fully closed (0%), half closed (50%), and open (100%)” to make the data quantifiable.

Appropriate weighting of the relevant factors was provided to better explain the quantified data. A (weakness under stressed conditions) = sum of the weighted scores (average of each factor’s data). 100-A = B (production stability score during the summer season). The environmental conditions (temperature and humidity) of the open field are shown in Supplementary Figure S1.

2.4. Preparation of a GBS Library and Sequencing Data Analysis

A GBS library was constructed through the following process: adaptor annealing, DNA double digestion with PstI and MsPI, adaptor ligation, sample pooling, DNA purification, and multiplexed PCR. The pooled GBS library was sequenced using the paired-end read method on an Illumina HiSeq X platform (Illumina Inc., San Diego, CA, USA). Raw sequences were demultiplexed into individual samples using barcode sequences, followed by adapter sequence removal and quality trimming. Adapter trimming was performed using cutadapt v. 1.8.3, and sequence quality trimming was performed using the DynamicTrim and LengthSort programs of SolexaQA V.1.13. The reads were aligned to the Brassica rapa reference genome (Brapa_chiifu_v41_genome20230413.fasta, http://brassicadb.cn/#/ accessed on 26 May 2024) using the Burrows–Wheeler Aligner (BWA) program v. 0.6.1-r104. Raw SNPs were detected and consensus sequences were extracted using SAMtools v.0.1.16. SNPs were classified into homozygous (SNP read depth ≥ 90%), heterozygous (40% ≤ SNP read depth ≤ 60%), and other (homozygous/heterozygous; undistinguished types) through an SNP filtering step. Further read group addition and sorting were performed using the Picard tool 1.112 package, and the genotyping and unification of the SNPs were accomplished using GATK3.1. SNPs were selected by individually comparing all sequences from the 52 individuals to avoid reference bias. The identified crude SNPs were filtered based on the following criteria: (i) short read depth > 10, (ii) homozygous and diallelic SNPs, (iii) minor allele frequency (MAF): polymorphism information content (PIC) > 0.36, (iv) 1:1 segregation ratio, and (v) distance between flanking SNPs. The PIC values of the markers were calculated with the formula PIC = 1 − ∑(Pi²) − ∑(2PiPj) (where Pi and Pj are the allele frequencies of the i and j alleles, respectively) using a Python script.

2.5. Clustering Analysis and Structure Analysis

A cluster analysis of the marker genotyping results was performed using the neighbor-joining algorithm. A phylogenetic tree was constructed using the DARWIN 6.0 software (https://darwin.cirad.fr/, accessed on 17 January 2022). Tree construction was based on a dissimilarity matrix calculated using the Manhattan index. The analysis was performed with 1000 bootstrap replicates from the generated genetic distance matrix.

The population structure of the Chinese cabbage accessions was analyzed using the STRUCTURE program ver. 2.3.4. The true number of populations (K) was determined using the ad hoc quantity (ΔK), calculated based on the second-order rate of change of the likelihood [22,23]. Five independent runs for K values ranging from 1 to 10 were performed using 100,000 burn-in lengths and 100,000 MCMC replicates in all simulations.

2.6. Genome-Wide Association Study (GWAS) and Candidate Gene Annotation

A GWAS was conducted using a fixed random model cyclic probability unification (FarmCPU) model implemented in GAPIT in R package. Principal components and a kinship matrix were used to control the influence of the population structure on the association analysis. Based on the 200,255 SNPs obtained from the GBS analysis, the multi-locus linear mixed model FarmCPU was used to identify the significant SNPs for the four pod-related traits. The threshold for significant associations was set to 1/n, where n is the number of markers and the p-value is <1/n or –log₁₀(p) ≥ 5.95. Visualization of the Manhattan plot and Q-Q plot was performed using the R package.

3. Results

3.1. Evaluation of Phenotypic Data in an Environmentally Controlled Room

In this study, we used an environmentally controlled room to test the effects of high daytime temperatures (exceeding 35 °C). We analyzed the images of the shoots and roots immediately after heat treatment. The phenotypes of the heat-treated group were compared to of those the control group. Unfortunately, it was challenging to discern significant phenotypic differences between the heat-treated and untreated shoots. Some individuals exhibited symptoms such as vein protrusion, leaf wilting, tipburn, rosette phenomena, and elongated leaves immediately after heat treatment. However, the variation in these symptoms was substantial among the heat-treated individuals. Additionally, many of these heat-stressed symptoms resembled the signs of plant senescence due to the extended growth periods in the trays (Supplementary Figure S2).

We also analyzed the root images to identify heat tolerance. Compared with the control group, most of the heat-treated groups exhibited a noticeably lower root distribution. In most experiments, the heat-treated group had a reduced calculated root area in the overall image-based measurements. In these cases, roots did not spread evenly, but were concentrated at the top or were generally sparse throughout. Some specimens that were significantly affected by heat exhibited instances of root breakage. Conversely, some resources exhibited minimal differences between the heat-treated and control groups (Figure 2 and Figure 3). According to the data, HS 4, HS 6, HS 8, HS 1, HS 69, HS 59, and HS 92 showed almost no differences between heat-stressed root density and the control. HS 50, HS 45, HS 58, HS 14, HS 43, and HS 51 showed the largest reduction in root density after heat treatment (Figure 3, Supplementary Table S2). To obtain a better result in the GWAS analysis, this reduction in root density was calculated, and the accessions were divided based on the degree of the differences, as shown in Supplementary Table S2.

3.2. Evaluation of Phenotypic Data in an Open Field during the Summer Season

If the Chinese cabbage grew and formed a normal head with fewer physiological disorders, growth delays, and inner rot, it was considered to be tolerant. The phenotypic data of the 52 Chinese cabbage accessions collected in the open field are shown in

Table 2. Phenotype data of the Chinese cabbage accessions collected in the open field from two years.

Accession ID	Year	Plant Death	Internal Rot	Calcium Deficiency	Boron Deficiency	Soft Rot Disease	Insect Invasion	Head Formation	Stress Weakness	Production Stability Score
Weighted		0.2	0.1	0.2	0.1	0.1	0.1	0.2
HS1	2022	-	0%	79%	0%	0%	-	100%	47.9	52.1
	2023	0%	-	67%	80%	0%	93%	100%
	Average	0%	0%	73%	40%	0%	93%	100%
	Sum	0.0	0.0	14.6	4.0	0.0	9.3	20.0
HS2	2022	-	100%	100%	11%	0%	-	100%	53.6	46.4
	2023	27%	-	91%	9%	27%	18%	50%
	Average	27%	100%	95%	10%	14%	18%	75%
	Sum	5.4	10.0	19.0	1.0	1.4	1.8	15.0
HS3	2022		0%	53%	47%	0%	-	100%	48.3	51.7
	2023	20%	-	33%	100%	17%	75%	100%
	Average	20%	0%	43%	74%	8%	75%	100%
	Sum	4.0	0.0	8.6	7.4	0.8	7.5	20.0
HS4	2022	-	100%	92%	0%	0%	-	0%	47.8	52.2
	2023	33%	-	70%	60%	40%	50%	50%
	Average	33%	100%	81%	30%	20%	50%	25%
	Sum	6.6	10.0	16.2	3.0	2.0	5.0	5.0
HS5	2022	-	100%	94%	17%	6%	-	100%	54.7	45.3
	2023	7%	-	100%	31%	8%	8%	100%
	Average	7%	100%	97%	24%	7%	8%	100%
	Sum	1.4	10.0	19.4	2.4	0.7	0.8	20.0
HS6	2022	-	0%	11%	84%	5%	-	0%	15.4	84.6
	2023	13%	-	23%	46%	8%	23%	0%
	Average	13%	0%	17%	65%	6%	23%	0%
	Sum	2.6	0.0	3.4	6.5	0.6	2.3	0.0
HS7	2022	-	100%	100%	100%	0%	-	100%	63.3	36.7
	2023	0%	-	100%	87%	0%	40%	100%
	Average	0%	100%	100%	93%	0%	40%	100%
	Sum	0.0	10.0	20.0	9.3	0.0	4.0	20.0
HS8	2022	-	50%	58%	16%	5%	-	100%	44.5	55.5
	2023	0%	-	67%	53%	0%	33%	100%
	Average	0%	50%	62%	35%	3%	33%	100%
	Sum	0.0	5.0	12.4	3.5	0.3	3.3	20.0
HS9	2022	-	0%	20%	100%	0%	-	0%	25.2	74.8
	2023	7%	-	86%	29%	7%	14%	50%
	Average	7%	0%	53%	64%	4%	14%	25%
	Sum	1.4	0.0	10.6	6.4	0.4	1.4	5.0
HS10	2022	-	100%	100%	73%	0%	-	100%	52.7	47.3
	2023	7%	-	71%	50%	0%	29%	50%
	Average	7%	100%	86%	62%	0%	29%	75%
	Sum	1.4	10.0	17.2	6.2	0.0	2.9	15.0
HS11	2022	-	0%	0%	20%	0%	-	0%	8.8	91.2
	2023	0%	-	53%	33%	0%	7%	0%
	Average	0%	0%	27%	27%	0%	7%	0%
	Sum	0.0	0.0	5.4	2.7	0.0	0.7	0.0
HS12	2022	-	100%	100%	100%	11%	-	100%	54.5	45.5
	2023	0%	-	60%	20%	0%	20%	100%
	Average	0%	100%	80%	60%	5%	20%	100%
	Sum	0.0	10.0	16.0	6.0	0.5	2.0	20.0
HS13	2022	-	100%	100%	100%	0%	-	100%	60.0	40.0
	2023	0%	-	100%	73%	0%	13%	100%
	Average	0%	100%	100%	87%	0%	13%	100%
	Sum	0.0	10.0	20.0	8.7	0.0	1.3	20.0
HS14	2022	-	50%	94%	100%	6%	-	100%	52.0	48.0
	2023	0%	-	100%	7%	0%	20%	100%
	Average	0%	50%	97%	53%	3%	20%	100%
	Sum	0.0	5.0	19.4	5.3	0.3	2.0	20.0
HS15	2022	-	0%	5%	100%	0%	-	100%	37.9	62.1
	2023	0%	-	0%	100%	0%	73%	100%
	Average	0%	0%	3%	100%	0%	73%	100%
	Sum	0.0	0.0	0.6	10.0	0.0	7.3	20.0
HS16	2022	-	50%	95%	95%	0%	-	100%	49.8	50.2
	2023	0%	-	60%	53%	0%	20%	100%
	Average	0%	50%	77%	74%	0%	20%	100%
	Sum	0.0	5.0	15.4	7.4	0.0	2.0	20.0
HS17	2022	-	0%	39%	17%	0%	-	0%	22.4	77.6
	2023	7%	36%	71%	7%	0%	7%	0%
	Average	7%	18%	55%	12%	0%	7%	0%
	Sum	1.4	1.8	11.0	1.2	0.0	7.0	0.0
HS18	2022	-	50%	26%	100%	11%	-	0%	18.8	81.2
	2023	7%	-	14%	7%	7%	21%	0%
	Average	7%	50%	20%	54%	9%	21%	0%
	Sum	1.4	5.0	4.0	5.4	0.9	2.1	0.0
HS19	2022	-	0%	0%	100%	11%	-	0%	17.8	82.2
	2023	13%	0%	15%	85%	0%	38%	0%
	Average	13%	0%	8%	92%	6%	38%	0%
	Sum	2.6	0.0	1.6	9.2	0.6	3.8	0.0
HS20	2022	-	100%	68%	100%	16%	-	0%	46.7	53.3
	2023	33%	-	70%	100%	10%	50%	0%
	Average	33%	100%	69%	100%	13%	50%	0%
	Sum	6.6	10.0	13.8	10.0	1.3	5.0	0.0
HS21	2022	100%	-	-	-	100%	-	0%	34.5	65.5
	2023	40%	-	22%	22%	22%	78%	0%
	Average	70%	-	22%	22%	61%	78%	0%
	Sum	14.0	-	4.4	2.2	6.1	7.8	0.0
HS22	2022	-	0%	11%	100%	11%	-	0%	19.8	80.2
	2023	7%	14%	36%	57%	7%	43%	0%
	Average	7%	7%	23%	79%	9%	43%	0%
	Sum	1.4	0.7	4.6	7.9	0.9	4.3	0.0
HS23	2022	-	0%	5%	100%	5%	-	0%	12.9
	2023	7%	0%	7%	43%	0%	29%	0%
	Average	7%	0%	6%	71%	3%	29%	0%
	Sum	1.4	0.0	1.2	7.1	0.3	2.9	0.0		87.1
HS24	2022	-	0%	0%	0%	0%	-	100%	25.7
	2023	0%	-	13%	20%	0%	33%	100%
	Average	0%	0%	7%	10%	0%	33%	100%
	Sum	0.0	0.0	1.4	1.0	0.0	3.3	20.0		74.3
HS25	2022	-	100%	100%	53%	0%	-	100%	46.6	53.4
	2023	0%	-	100%	0%	0%	40%	0%
	Average	0%	100%	100%	26%	0%	40%	50%
	Sum	0.0	10.0	20.0	2.6	0.0	4.0	10.0
HS26	2022	-	100%	63%	100%	5%	-	0%	27.7	72.3
	2023	0%	-	40%	13%	0%	13%	0%
	Average	0%	100%	52%	57%	3%	13%	0%
	Sum	0.0	10.0	10.4	5.7	0.3	1.3	0.0
HS27	2022	-	100%	89%	100%	0%	-	100%	77.1	22.9
	2023	33%	-	100%	100%	50%	90%	100%
	Average	33%	100%	95%	100%	25%	90%	100%
	Sum	6.6	10.0	19.0	10.0	2.5	9.0	20.0
HS28	2022	-	50%	16%	100%	0%	-	0%	26.7	73.3
	2023	20%	17%	75%	92%	0%	8%	0%
	Average	20%	33%	45%	96%	0%	8%	0%
	Sum	4.0	3.3	9.0	9.6	0.0	0.8	0.0
HS29	2022	-	0%	0%	100%	0%	-	0%	18.1	81.9
	2023	0%	13%	53%	100%	0%	20%	0%
	Average	0%	7%	27%	100%	0%	20%	0%
	Sum	0.0	0.7	5.4	10.0	0.0	2.0	0.0
HS30	2022	-	100%	53%	100%	0%	-	100%	55.9	44.1
	2023	0%	-	100%	60%	0%	27%	100%
	Average	0%	100%	76%	80%	0%	27%	100%
	Sum	0.0	10.0	15.2	8.0	0.0	2.7	20.0
HS31	2022	-	0%	0%	100%	0%	-	0%	19.9	80.1
	2023	7%	43%	100%	0%	0%	14%	0%
	Average	7%	21%	50%	50%	0%	14%	0%
	Sum	1.4	2.1	10.0	5.0	0.0	1.4	0.0
HS32	2022	-	100%	68%	100%	11%	-	100%	46.4	53.6
	2023	0%	-	53%	60%	0%	7%	50%
	Average	0%	100%	61%	80%	5%	7%	75%
	Sum	0.0	10.0	12.2	8.0	0.5	0.7	15.0
HS33	2022	-	0%	61%	0%	0%	-	100%	36.5	63.5
	2023	7%	-	43%	0%	7%	43%	100%
	Average	7%	0%	52%	0%	4%	43%	100%
	Sum	1.4	0.0	10.4	0.0	0.4	4.3	20.0
HS34	2022	-	0%	0%	0%	0%	-	100%	25.3	74.7
	2023	0%	-	0%	0%	0%	53%	100%
	Average	0%	0%	0%	0%	0%	53%	100%
	Sum	0.0	0.0	0.0	0.0	0.0	5.3	20.0
HS35	2022	-	0%	11%	100%	6%	-	0%	18.2	81.8
	2023	7%	0%	14%	100%	7%	36%	0%
	Average	7%	0%	13%	100%	6%	36%	0%
	Sum	1.4	0.0	2.6	10.0	0.6	3.6	0.0
HS36	2022	-	0%	0%	100%	7%	-	0%	16.7	83.3
	2023	7%	0%	0%	100%	0%	50%	0%
	Average	7%	0%	0%	100%	3%	50%	0%
	Sum	1.4	0.0	0.0	10.0	0.3	5.0	0.0
HS37	2022	-	0%	0%	100%	0%	-	0%	12.1	87.9
	2023	7%	0%	0%	86%	0%	14%	0%
	Average	7%	0%	0%	93%	0%	14%	0%
	Sum	1.4	0.0	0.0	9.3	0.0	1.4	0.0
HS38	2022	-	0%	0%	37%	0%	-	0%	11.5	88.5
	2023	8%	8%	25%	8%	8%	42%	0%
	Average	8%	4%	13%	23%	4%	42%	0%
	Sum	1.6	0.4	2.6	2.3	0.4	4.2	0.0
HS39	2022	-	50%	0%	100%	0%	-	100%	27.1	72.9
	2023	7%	0%	36%	43%	7%	21%	0%
	Average	7%	25%	18%	71%	4%	21%	50%
	Sum	1.4	2.5	3.6	7.1	0.4	2.1	10.0
HS40	2022	-	-	-	-	-	-	-	52.5	47.5
	2023	13%	-	100%	46%	15%	38%	100%
	Average	13%	-	100%	46%	15%	38%	100%
	Sum	2.6	-	20.0	4.6	1.5	3.8	20.0
HS41	2022	-	100%	81%	100%	10%	-	0%	44.6	55.4
	2023	33%	-	60%	100%	50%	10%	0%
	Average	33%	100%	70%	100%	30%	10%	0%
	Sum	6.6	10.0	14.0	10.0	3.0	1.0	0.0
HS42	2022	-	-	-	-	-	-	-	36.0	64.0
	2023	0%	-	20%	33%	0%	87%	100%
	Average	0%	-	20%	33%	0%	87%	100%
	Sum	0.0	-	4.0	3.3	0.0	8.7	20.0
HS43	2022	-	-	-	-	-	-	-	20.3	79.7
	2023	13%	0%	23%	85%	15%	31%	0%
	Average	13%	0%	23%	85%	15%	31%	0%
	Sum	2.6	0.0	4.6	8.5	1.5	3.1	0.0
HS44	2022	-	-	-	-	-	-	-	2.7	97.3
	2023	0%	0%	0%	0%	0%	27%	0%
	Average	0%	0%	0%	0%	0%	27%	0%
	Sum	0.0	0.0	0.0	0.0	0.0	2.7	0.0
HS45	2022	-	-	-	-	-	-	-	37.9	62.1
	2023	0%	-	73%	0%	0%	33%	100%
	Average	0%	-	73%	0%	0%	33%	100%
	Sum	0.0	-	14.6	0.0	0.0	3.3	20.0
HS46	2022	-	-	-	-	-	-	-	8.0	92.0
	2023	0%	0%	13%	47%	0%	7%	0%
	Average	0%	0%	13%	47%	0%	7%	0%
	Sum	0.0	0.0	2.6	4.7	0.0	0.7	0.0
HS47	2022	-	-	-	-	-	-	-	10.0	90.0
	2023	0%	0%	40%	0%	0%	20%	0%
	Average	0%	0%	40%	0%	0%	20%	0%
	Sum	0.0	0.0	8.0	0.0	0.0	2.0	0.0
HS48	2022	-	-	-	-	-	-	-	12.5	87.5
	2023	0%	0%	31%	50%	0%	13%	0%
	Average	0%	0%	31%	50%	0%	13%	0%
	Sum	0.0	0.0	6.2	5.0	0.0	1.3	0.0
HS49	2022	-	-	-	-	-	-	-	39.9	60.1
	2023	27%	-	100%	0%	36%	9%	50%

. The aboveground growth of Chinese cabbage resources (HS41, HS61, HS60, HS55, and HS45) with production stability scores ranging from 80.1 to 88 were observed by drone. Figure 4a shows a photograph of these accessions taken 55 days after planting. The outermost leaves showed varying degrees of chlorosis, depending on the source. The head part of the plants formed stably, even in summer conditions, and physiological disorders, such as calcium deficiency, were barely observed in this group.

The stability scores of ten Chinese cabbage accessions that scored above 70 points were investigated for their cross-sectional phenotypes. Figure 4b shows an image of the vertically cut Chinese cabbages. The stability scores of the accessions in the picture ranged from 73.3 to 92 (Figure 4b). The overall heading process of these plants was stable, and physiological disorders such as calcium deficiency were either absent or minimal. Symptoms of internal rot, which can be caused by calcium deficiency, were mildly observed in HS45, HS52, HS61, and HS92, whereas no symptoms were found in HS41, HS44, HS59, and HS60, and very slight symptoms were observed in HS55 and HS93. Despite the relatively high stability score of HS92, internal rot symptoms were suspected in the root area. Additionally, some Chinese cabbages exhibited signs of bolting under high-temperature and long-day conditions. Drone images of the Chinese cabbages with stability scores ranging from 22.9 to 50 were also acquired (Figure 4c). Due to high temperatures, most plants in this group experienced delayed heading, calcium deficiency, severe internal rot, or decay. Harvesting and producing vertical-cut images of these plants were not possible (Figure 4c).

3.3. Searching Useful SNPs by Genotyping-by-Sequencing

The GBS library was constructed from 52 Chinese cabbage accessions. The samples were sequenced using the Illumina HiSeq X platform(SEEDERS Inc. (Daejeon, Republic of Korea). A summary of the sequencing results is presented in

Table 3. Total raw data of genotyping by sequencing (GBS).

Data	No. of Barcode	No. of Sample	No. of Reads	Total Length (bp)	GC (%) *1	Q30 (%) *2	No. of Demultiplexed Reads (%)
Total	96	52	856,098,684	129,270,901,284	46	93	650,426,476 (75.98%)

*¹: GC (%): GC content. *²: Q30 (%): Ratio of bases that have phred quality score of over 30.

. The total length of the generated reads was 129,270,901,284 bp and the read number was 856 million. The GC content was 46% and Q30 was 93%. Pooled reads were demultiplexed prior to SNP calling. The number of demultiplexed reads was 650,426,476 and the mapping rate was 75.98% (Table 3). After demultiplexing, adapter sequences, primers, and other unwanted sequences from the high-throughput sequencing reads were removed. Raw data for each of the 52 accessions are listed in Supplementary Table S3. The range of the reads was between 5 million reads and 24 million reads, and the length range of the reads ranged from 776 mb (Plant ID: HS96) to 3659 mb (Plant ID: HS50).

The average sum of the raw reads of the 52 Chinese cabbages was 24,544,395, and the average total length of the raw reads was 3,706,203,693.4 bp. After the five SNP filtering steps ((i) short read depth > 10, (ii) homozygous and diallelic SNPs, (iii) minor allele frequency (MAF): polymorphism information content (PIC) > 0.36, (iv) 1:1 segregation ratio, and (v) distance between flanking SNPs), 200,255 SNPs were selected for use in the association study. These SNPs were distributed throughout all ten chromosomes. The number of SNPs ranged from 13,490 to 29,555 on the chromosomes. The range of SNP densities was between 0.55 and 0.84 SNPs/Mb, with an average SNP density of 0.64 SNPs/Mb. Chromosome A10 showed the highest SNP density, whereas chromosome A08 showed the lowest density (

Table 4. Selected SNP numbers and densities of each chromosome.

	Chr 1	Chr 2	Chr 3	Chr 4	Chr 5	Chr 6	Chr 7	Chr 8	Chr 9	Chr 10	Total	Average
SNP numbers	20,638	17,907	27,626	13,490	21,408	21,487	18,280	14,753	29,555	15,111	200,255	20,026
Chromosome size (Mb)	32,850	29,510	35,350	22,750	35,020	38,650	28,860	26,880	52,890	17,990
SNP density	0.63	0.61	0.78	0.59	0.61	0.56	0.63	0.55	0.56	0.84		0.64

).

3.4. Structure and Clustering Analysis of Chinese Cabbage Accessions

A STRUCTURE analysis was performed using the GBS genotype data (Table 4). The mean log likelihood curve reached a maximum value around K = 4, indicating that the best K value for this SNP dataset was 4 (Figure 5). At K = 4, four distinct sue cabbage accessions could be divided into four groups (Figure 6). Group bpopulations were observed. Phylogenetic analysis revealed that the 52 ChinesA included eight plants: HS3, HS50, HS68, HS16, HS97, HS11, HS18, and HS98. Two phenotypes were observed in this group: five pak-choi-shaped accessions and three cabbage-shaped accessions. Group B included 11 plants: HS1, HS57, HS43, HS49, HS5, HS7, HS48, HS58, HS70, HS15, and HS2. Most of the accessions belonging to this group possessed open- or half-open-type heads. Three radish-like accessions were included in this group.

Group C comprised 14 plants: HS69, HS85, HS17, HS9, HS12, HS14, HS42, HS51, HS13, HS56, HS54, HS52, HS88, and HS10. In this group, most Chinese cabbages showed a small closed-head type, except for HS 9, which had a relatively larger head than the rest of the lines. HS69 had an open-type head, and HS14, HS42, HS88, and HS10 had half-open-type heads. Group D contained 19 plants, HS92, HS45, HS59, HS44, HS94, HS61, HS83, HS55, HS60, HS41, HS6, HS53, HS96, HS8, HS62, HS93, HS4, HS99, and HS64. Group D was the largest group. Most Chinese cabbages in Group A shared the traits of small, long-shaped heads or no heads. Group B showed incomplete heads, with some plants exhibiting radish-like traits. Group C included variously shaped plants without specific traits. The most common Chinese cabbages used for kimchi production seemed to be gathered in Group D.

3.5. Genome-Wide Association Study of Six Horticulturally Important Traits

The data of seven phenotype traits from 52 accessions were used for the GWAS study (

Table 5. The phenotype data used for the GWAS study.

Place	Field		Environmental Control Room		Field	Field	Field	Field	Field
	Tolerance against Summer Season		Heat Tolerance of Root		Bolting Rate	Axillary Bud Rate	Calcium Deficiency Rate	Boron Deficiency Rate	Heading Rate
Plant ID	Data #1	Data #2	Data #1	Data #2
HS 01	52.1	2	−0.02	1.00	100.00	100.00	66.67	80.00	100
HS 02	46.4	2	0.01	2.00	100.00	109.09	90.91	9.09	47
HS 03	51.7	2	0.08	4.00	0.00	116.67	33.33	116.67	100
HS 04	52.2	2	−0.12	1.00	0.00	150.00	70.00	60.00	26
HS 05	45.3	1	0.04	3.00	0.00	115.38	100.00	30.77	47
HS 06	84.6	5	−0.06	1.00	0.00	23.08	23.08	46.15	24
HS 07	36.7	1	0.10	5.00	0.00	100.00	100.00	86.67	47
HS 08	55.5	3	−0.02	1.00	0.00	80.00	66.67	53.33	14
HS 09	74.8	4	0.12	5.00	0.00	107.14	85.71	28.57	40
HS 10	47.3	2	0.02	2.00	0.00	35.71	71.43	50.00	40
HS 11	91.2	6	0.03	3.00	100.00	0.00	53.33	33.33	40
HS 12	45.5	1	0.10	5.00	0.00	73.33	60.00	20.00	61
HS 13	40	1	0.07	4.00	100.00	0.00	100.00	73.33	61
HS 14	48	2	0.18	6.00	0.00	86.67	100.00	6.67	100
HS 15	62.1	3	0.04	3.00	0.00	100.00	0.00	100.00	100
HS 16	50.2	2	0.06	3.00	0.00	13.33	60.00	53.33	100
HS 17	77.6	5	0.02	2.00	0.00	28.57	71.43	7.14	33
HS 18	81.2	5	0.03	2.00	0.00	100.00	14.29	7.14	27
HS 41	82.2	5	0.09	4.00	0.00	100.00	15.38	84.62	33
HS 42	53.3	2	0.10	5.00	0.00	120.00	70.00	150.00	33
HS 43	65.5	3	0.19	6.00	0.00	0.00	22.22	22.22	27
HS 44	80.2	5	0.09	4.00	0.00	107.14	35.71	57.14	27
HS 45	87.1	6	0.16	6.00	0.00	64.29	7.14	42.86	27
HS 48	74.3	4	0.07	4.00	0.00	100.00	13.33	20.00	100
HS 49	53.4	2	0.03	3.00	0.00	93.33	100.00	0.00	40
HS 50	72.3	4	0.13	6.00	0.00	0.00	40.00	13.33	33
HS 51	22.9	1	0.20	6.00	50.00	20.00	130.00	150.00	40
HS 52	73.3	4	0.09	5.00	0.00	125.00	75.00	91.67	40
HS 53	81.9	5	0.10	5.00	0.00	100.00	53.33	100.00	33
HS 54	44.1	1	0.06	3.00	46.67	40.00	100.00	60.00	61
HS 55	80.1	5	0.09	4.00	0.00	107.14	42.86	107.14	27
HS 56	53.6	2	0.00	2.00	0.00	40.00	53.33	60.00	40
HS 57	63.5	3	0.10	5.00	0.00	92.86	42.86	0.00	61
HS 58	74.7	4	0.16	6.00	0.00	100.00	0.00	0.00	100
HS 59	81.8	5	0.00	1.00	0.00	85.71	14.29	107.14	27
HS 60	83.3	5	0.05	3.00	0.00	92.86	0.00	100.00	27
HS 61	87.9	6	0.04	3.00	0.00	107.14	0.00	85.71	27
HS 62	88.5	6	0.12	5.00	0.00	0.00	25.00	8.33	40
HS 64	72.9	4	0.09	5.00	0.00	85.71	35.71	42.86	40
HS 68	47.5	2	0.03	2.00	0.00	100.00	100.00	46.15	40
HS 69	55.4	3	0.00	1.00	0.00	150.00	60.00	150.00	100
HS 70	64	3	0.12	5.00	0.00	46.67	20.00	33.33	100
HS 83	79.7	5	0.02	2.00	0.00	115.38	23.08	84.62	27
HS 85	97.3	6	0.08	4.00	0.00	0.00	0.00	0.00	40
HS 88	62.1	3	0.06	3.00	40.00	0.00	73.33	0.00	100
HS 92	92	6	0.00	1.00	0.00	33.33	13.33	46.67	27
HS 93	90	6	0.05	3.00	0.00	0.00	40.00	0.00	40
HS 94	87.5	6	0.11	5.00	0.00	93.75	31.25	50.00	100
HS 96	60.1	3	0.06	4.00	0.00	0.00	100.00	0.00	40
HS 97	70	4	0.01	2.00	0.00	13.33	40.00	20.00	100
HS 98	87.7	6	0.05	3.00	16.67	91.67	25.00	0.00	27
HS 99	49.7	2	0.12	5.00	0.00	8.70	100.00	8.70	100

). Six phenotypic traits were acquired from the field screening data and one from the environmentally controlled room. The six traits included tolerance against the summer season (stability score), bolting rate, axillary bud rate, calcium deficiency rate, boron deficiency rate, and heading rate. The phenotype data of the roots were collected from the heat stress experiment. For the GWAS analysis, a total of 200,255 filtered SNPs with a minor allele frequency greater than 5% and missing data lower than 30% were used. Seven models (BLINK, CMLM, FarmCPU, MLMM, GLM, MLM, and SUPER) were applied to perform the GWAS analysis with the default parameters in the GAPIT R package. Seven specific traits related to heat tolerance were analyzed. Manhattan plots showed numerous SNPs associated with the Bonferroni threshold (

Table 6. The candidate SNP results of genome-wide association study.

Trait	SNP	Chromosome	Position	p-Value	Maf	Nobs	Effect	Phenotype_Variance_Explained (%)
Bolting	SA02_32779781	2	32,779,781	6.32 × 10−13	0.14	50	−6.73902	1.567524925
	SA02_36125892	2	36,125,892	8.35 × 10−14	0.04	50	−39.5783	55.43468979
	SA05_1527161	5	1,527,161	1.12 × 10−08	0.07	36	NA	5.23 × 10−09
	SA05_1928956	5	1,928,956	1.33 × 10−09	0.13	50	−17.9368	11.01177681
	SA05_40364503	5	40,364,503	1.64 × 10−11	0.29	50	−5.40308	1.789144987
	SA05_40519151	5	40,519,151	7.83 × 10−08	0.02	36		0.014316435
	SA06_38680523	6	38,680,523	2.14 × 10−11	0.02	50	58.74881	30.46097055
	SA06_46190907	6	46,190,907	7.83 × 10−08	0.02	36	NA	0.000205104
	SA09_8067736	9	8,067,736	3.13 × 10−23	0.17	50	−15.5337	10.01210093
	SA09_8404885	9	8,404,885	9.86 × 10−08	0.06	36		5.40 × 10−05
	SA09_56761182	9	56,761,182	7.83 × 10−08	0.02	36		1.87 × 10−05
	SA09_58874214	9	58,874,214	7.05 × 10−16	0.39	50	−6.19219	2.085168086
Heat	SA09_55392231	9	55,392,231	1.39 × 10−29	0.39	50	−11.6034	34.43961
	SA09_55392192	9	55,392,192	1.47 × 10−09	0.43	36		0
	SA10_28010298	10	28,010,298	2.18 × 10−22	0.49	50	−5.2023	6.185935
	SA10_29194037	10	29,194,037	2.71 × 10−13	0.29	50	0.42079	5.04724
	SA09_7986935	9	7,986,935	5.57 × 10−09	0.15	50	0.374991	5.990016
	SA09_8108971	9	8,108,971	4.51 × 10−08	0.31	36		1.70 × 10−11
Boron	SA08_28311462	8	28,311,462	2.17 × 10−07	0.16	36		0.115844
	SA08_15162711	8	15,162,711	3.88 × 10−24	0.17	50	35.2342	13.48117

).

Five traits were available for identifying useful associated SNPs using the GWAS. The traits were bolting, axillary bud, calcium deficiency, boron deficiency, and the stability score (Figure 7).

Significant SNPs related to bolting traits were also analyzed. A total of 267 significant associations were identified between the SNP markers and bolting traits. The three models showed significant marker–trait associations. Three significant marker–trait associations from chromosomes 2, 5, and 6 were detected by the BLINK model, and eight from chromosomes 2, 5, 6, 8, and 9 were detected by the MLMM method. Using the SUPER model, numerous significant associations (256 in total) were identified, mainly clustered on chromosomes 5 and 9. Seven SNP markers exhibited significant associations identified by several models. One marker from chromosome 2, located between 32,779,781 and 36,125,892, showed bolting associations in the three models (BLINK, MLMM, and SUPER). Two SNP markers on chromosome 5 (1,527,161–1,928,956 and 40,364,503–40,519,151) were identified using two models (MLMM and SUPER). Two other SNP markers from chromosome 6 (38,680,523 and 46,190,907–46,789,754) were also identified using two models (MLMM and SUPER). SNP markers located on chromosomes 9 (8,067,736–8,404,885) and 9 (5,887,421–56,761,182) were also revealed by the same two models.

Significant SNPs related to axillary bud traits were also analyzed. The two models showed marker–trait associations. Five significant marker–trait associations on chromosomes 1, 4, and 5 were detected using the MLMM model. Three SNP markers were mapped on chromosome 5. Ten significant marker–trait associations on chromosomes 3, 4, 7, 8, and 10 were detected using the SUPER method. Three SNP markers were mapped to chromosomes 8 and 10. Two SNP markers were found on chromosome 7. No co-identified regions were found for this trait among the different models.

Significant SNPs related to the stability score were analyzed. Two versions of labeled data from the stability score were prepared for GWAS analysis. These two phenotypes showed different results in the GWAS analysis. The MLMM and SUPER models showed significant marker–trait associations in both phenotype datasets. From phenotype data #1, seven significant marker–trait associations from chromosomes 2, 3, 5, 6, 9, and 10 were detected by the MLMM model, and one significant marker–trait association on chromosome 1 was found by the SUPER model. For phenotype data #2, seven significant marker–trait associations from chromosomes 3, 5, 8, 9, and 10 were detected using the MLMM model. A total of 138 significant associations were identified between SNP markers and traits using the SUPER model. SNP regions co-identified by SUPER and MLMM were also detected. Three SNP markers mapped to chromosomes 9 (53,922,931–55,392,192 and 7,986,935–8,108,971) and 10 (28,010,298–29,194,037) were significantly associated with phenotype data #2.

Seven significant marker–calcium-deficiency trait associations on chromosomes 1, 2, 5, 6, and 8 were detected using the MLMM model. Two SNP markers were mapped on chromosomes 5 and 6 each. A total of 76 SNP markers were identified using the SUPER model, with 34 markers clustered on chromosome 1 and 12 markers clustered on chromosome 9. No co-identified regions were found among the different models for these traits.

Eight significant markers related to boron deficiency traits were identified on chromosomes 2, 3, 8, and 9 using the MLMM model. Two SNP markers were mapped onto chromosomes 3 and 8, and three SNP markers were mapped onto chromosome 9. Ten significant marker–trait associations on chromosomes 1, 4, 8, and 10 were detected using the SUPER model. Five SNP markers were mapped on chromosome 1, and three SNP markers were mapped on chromosome 4. No co-identified regions were found among the different models for these traits.

3.6. Assigning Significant SNPs to Potential Candidate Genes

Among all the associated SNPs related to heat tolerance, bolting, and boron deficiency tolerance, 10 SNPs were selected for candidate gene research. The selected SNPs showed the highest log₁₀(p) values over 25. To assess the putative candidate genes associated with the significant SNPs for each trait in Chinese cabbage, we retrieved all possible genes in the 100 kb window (LD region) around each peak SNP (

Table 7. The annotation of candidate genes.

Trait	Chr	Position	Searching Range (kb)	Candidate Gene	Reference Referring Gene Function
Summer tolerance	Chr2	33,894,746	200	3-ketoacyl-CoA synthase 21	[24]
	Ch9	7,986,935	200	F-box/kelch-repeat protein like	[25]
			200	tyrosine decarboxylase 1	[26]
			200	phospholipase A2-beta	[27,28]
	Ch10	28,010,298	200	SWI/SNF-related matrix-associated actin-dependent regulator	[29]
			200	laccase-12	[30]
			200	expansin-A2	[31]
			200	transcription factor MYBC1-like	[32]
			200	DEAD-box ATP-dependent RNA helicase 18	[33,34]
	Ch10	29,194,037	200	dnaJ homolog subfamily B	[35]
			200	R3H domain-containing protein 2	[36]
			200	transcription factor bHLH143	[37,38]
			200	E3 ubiquitin-protein ligase ARI16	[39]
			200	VQ motif-containing protein 31	[40]
Bolting	A02	36,125,892	200	zinc transporter 12	[41,42]
			200	ethylene-responsive transcription factor ERF114	[43]
	A05	1,928,956	200	transcription factor JUNGBRUNNEN 1	[44]
			200	pentatricopeptide repeat-containing protein	[45,46]
			200	defensin-like protein 2	[47]
		40,364,503	200	B3 domain-containing protein REM7-like	[48]
			200	LOB domain-containing protein 22	[49]
			200	AAA-ATPase	[50,51]
	A09	8,067,736	200	tyrosine decarboxylase 1	[52]
			200	ABC transporter C family member 5	[53]
		58,874,214	200	F-box protein	[54]
			200	glycosyltransferase family 92 protein	[55,56]
			200	serine/arginine-rich splicing factor RSZ21-like	[57,58]
Boron deficiency	A03	5,577,165	200	Brassica rapa GATA transcription factor 21	[59,60]
			200	auxin efflux carrier component 2	[61,62,63]
			200	transcription factor bHLH35-like	[37,38]
			200	cytochrome P450	[64]
			200	VQ motif-containing protein 10	[40]

).

The three traits (summer tolerance, bolting, and boron deficiency tolerance) showed explainable results in the gene prediction around their SNPs. The number of genes around the four peak SNPs ranged from one to five for heat stress tolerance. The total number of predicted genes was 14. Three genes (F-box/kelch-repeat protein-like, tyrosine decarboxylase 1, and phospholipase A2-beta) were found on the upper part of chromosome 9, and ten genes (SWI/SNF-related matrix-associated actin-dependent regulator, etc.) were found on chromosome 10. One gene (3-ketoacyl-CoA synthase 21) was found on the lower part of chromosome 2.

The number of genes around the five peak SNPs ranged from one to three for the bolting traits. Zinc transporter 12 and ethylene-responsive transcription factor ERF114 were found on the lower part of chromosome 2. On the upper part of chromosome 5, three genes (transcription factor JUNGBRUNNEN 1, pentatricopeptide repeat-containing protein, and defensin-like protein 2) were designated as candidate genes. On the lower part of chromosome 5, three genes, including B3 domain-containing protein REM7-like, LOB domain-containing protein 22, and AAA-ATPase, were detected. On the upper part of chromosome 9, tyrosine decarboxylase 1 and ABC transporter C family member were detected, and on the lower part, F-box protein, glycosyltransferase family 92 protein, and serine/arginine-rich splicing factor RSZ21-like genes were detected. Thirteen genes were found on three chromosomes.

For the boron deficiency trait, five genes were found on the upper part of chromosome 3. The names of the genes were Brassica rapa GATA transcription factor 21, auxin efflux carrier component 2, transcription factor bHLH35-like, cytochrome P450, and VQ motif-containing protein 10. The candidate genes were mainly involved in plant responses to several environmental stresses such as drought, salt, and heat. Some of the discovered genes were also found to be involved in flowering.

4. Discussion

Open-field environments have become increasingly unpredictable owing to climate change. Even when the same Chinese cabbage cultivars are grown and assessed during the same period each year, phenotypic variations are common, and the types of prevalent diseases or pests fluctuate annually. Consequently, resources demonstrating heat tolerance in controlled growth chambers often do not exhibit a similar resilience in open fields due to substantial environmental influences. This inconsistency emphasizes the complexity of breeding heat-tolerant crops and the need for a multi-faceted approach in evaluating tolerance traits.

In Korea, most Chinese cabbages utilized for kimchi production are predominantly sown during the hot summer season and harvested in autumn. In recent years, certain crops have experienced damage from high temperatures during planting periods, while others have been affected by unusually elevated temperatures during harvesting periods. Therefore, both seedling-stage and mature-stage heat tolerance are vital factors for consistent Chinese cabbage production. In this study, we developed selection criteria for tolerant Chinese cabbage breeding in response to climate change using phenotyping data from open fields and controlled environments. The numerical stability score shown in this study was in the selection criteria for the open-field Chinese cabbage selection. According to our research, the stability score should be over 70 to be selected as a tolerant breeding material. While stability scores may be useful for assessing the viability of Chinese cabbage in open-field cultivation, they were less effective for evaluating detailed quality such as inner rot symptoms, which is also a fatal trait in summer harvesting. Cross-sectional survey data could be incorporated into future studies to improve this assessment.

Heat stress impacts both the shoots and roots of seedlings significantly. This study focused on analyzing the response of seedling roots to high-temperature stress. Root growth was quantified indirectly through image processing. White root parts and brown and black soil parts were mixed into the acquired images. We used in-house codes for the Image J processing to remove the soil parts from the pictures. In this way, the roots of the plants were protected from washing and the selected young plants were vernalized for flowering safely. Further research is necessary to investigate the differences in recovery patterns between genotypes. This understanding could provide insights into the mechanisms of heat stress resilience and help to improve breeding strategies for heat-tolerant varieties.

To make a useful genotyping tool for the efficient selection in the future, useful SNPs and candidate genes were studied. We identified ten candidate SNPs for heat tolerance on chromosome 10. In this chromosome, we identified candidate genes encoding laccase-12, expansin-A2, and the transcription factor MYBC1-like from one identified SNP. Previous studies have indicated a relationship between these genes and heat stress. Laccases are crucial enzymes in plants that play a significant role in the biosynthesis of lignin, a complex aromatic polymer that provides structural integrity and protection against environmental stresses, including high and low temperatures [30,65]. Lignin helps plants to adapt to their environment by acting protectively, sustainably, or disruptively depending on the specific stress. Expansins contribute to heat stress tolerance by modifying cell wall properties, integrating with other stress-responsive pathways, and enhancing protective mechanisms such as reactive oxygen species (ROS) scavenging [31,66]. These roles are known to help plants to maintain their growth and survive under high-temperature conditions. MYB transcription factors enhance plant tolerance to heat stress by regulating the expression of heat-responsive genes, interacting with stress signaling pathways, and contributing to various protective mechanisms [32]. Additionally, we identified candidate genes on chromosome 10 encoding dnaJ homolog subfamily B, transcription factor bHLH143, E3 ubiquitin-protein ligase ARI16, and VQ motif-containing protein 31. DNAJ proteins help to manage heat-induced protein aggregates by interacting with Hsp101, aiding in thermotolerance [35]. This interaction is crucial for solubilizing protein aggregates and for enhancing thermotolerance. bHLH transcription factors contribute to the heat stress response through transcriptional regulation, post-translational modifications, interaction with phytohormones, the integration of stress signals, and cross-talk with other stress pathways [37,38]. E3-ubiquitin ligases regulate protein stability, modulate stress signals, ensure proper development, and activate protective responses, thereby enhancing the plant’s ability to survive and thrive under high temperatures [39,67]. VQ motif-containing proteins in Chinese cabbage likely play a multifaceted role in enhancing heat stress tolerance by regulating stress-responsive genes, modulating hormone signaling, and activating protective mechanisms at both the cellular and tissue levels [40].

Plant flowering has been extensively studied, many of the genes involved have been identified, and their functions are largely understood, providing a wealth of data for our research. In B. rapa, long-day conditions, together with vernalization, promote reproductive growth over vegetative growth. Plants bolting before reaching the harvesting stage is a serious problem in Chinese cabbage [68]. Late-bolting traits are more desirable than early flowering for the cultivation of spring-sown Chinese cabbage [69]. Among the candidate genes found for bolting (flowering) in this study, chromosome 5 had several key factors related to flowering. Genes such as pentatricopeptide repeat-containing protein, defensin-like protein 2, B3 domain-containing protein REM7-like, and AAA-ATPase were found around two SNP regions of chromosome 5. Pentatricopeptide repeat (PPR) proteins play a crucial role in plant flowering, particularly in the context of cytoplasmic male sterility (CMS) and fertility restoration. PPR proteins manage mitochondrial gene expression, particularly under CMS conditions, where they restore male fertility by ensuring the proper development of pollen [45,46]. DEF2 plays a vital role in the regulation and development of floral organs, particularly in processes such as pollen formation and viability [47]. On chromosome 9, genes encoding tyrosine decarboxylase 1, ABC transporter C family member 5, and F-box proteins were involved in flowering. Tyrosine decarboxylases (TyDCs), known as a stress-induced protein in Arabidopsis thaliana, catalyze the conversion of L-tyrosine to tyramine(a key precursor in the biosynthesis of Amaryllidaceae alkaloids) in Lycoris radiata. Tyramine and its derivatives may act as signaling molecules or precursors in pathways that regulate flowering [26,52]. The ABC (ATP-Binding Cassette) transporters play crucial roles in various physiological processes in flowering by regulating hormone levels and other signaling compounds that influence flower development and stress responses [53]. F-box proteins play an essential role in plant flowering by regulating the stability of key proteins involved in the photoperiod and hormonal signaling pathways. Their ability to target specific proteins for degradation allows plants to adapt their flowering time to environmental cues and internal developmental signals [54,70].

In the case of boron deficiency, several annotated genes have been identified, but their direct roles in the stress response are unclear. However, their described functions suggest involvement in various stress responses, aiding plant survival under inappropriate conditions. Brassica rapa GATA transcription factor 21, auxin efflux carrier component 2, and cytochrome P450 were identified. There have been few reports referring the relationship between GATA transcription factor and abiotic stress response [59,60]. Among these genes, Auxin efflux carrier components, particularly the PIN-FORMED (PIN) proteins, play crucial roles in regulating the distribution and transport of auxin within plant tissues [63]. Auxin is a vital plant hormone that influences various aspects of plant growth and development, including flowering [61,62]. Cytochrome P450 enzymes (CYPs) are integral to the plant abiotic stress response. They participate in a variety of biochemical processes, including the detoxification of harmful substances and the biosynthesis of vital compounds, such as secondary metabolites, antioxidants, and phytohormones. CYPs help plants to manage oxidative stress by regulating the levels of reactive oxygen species (ROS) through antioxidant biosynthesis. Additionally, these enzymes are involved in the synthesis of hormones like jasmonic acid, salicylic acid, and abscisic acid, which play critical roles in plant responses to environmental stresses such as drought, salinity, and extreme temperatures [64].

The candidate SNPs and genes studied in this paper can provide valuable tools for breeders in developing stress-resistant Chinese cabbage. The results also provide a more comprehensive understanding of plant stress management, which can lead to better breeding practices. However, further studies are required to collect phenotype data under different environmental conditions. In addition, data from extreme weather conditions may also be necessary. There are several indications that the rate of global warming is increasing, making this study even more critical for future agricultural resilience.