Quantitative analysis of root system architecture and fresh weight biomass traits highlight phenotypic variation in radish (Raphanus sativus L.) germplasm

Abstract

Radish exhibits remarkable diversity in root morphology and architecture, which are crucial traits for nutrient and water uptake, stress adaptation and marketability. This study assessed root variability in 23 radish accessions, including a wild relative, landraces, and cultivars. Plants were grown in controlled greenhouse, and 2D images of roots acquired, using a high-resolution flatbed scanner fitted with a transparent acrylic tray (30 cm × 20 cm). Root analysis was performed using the WinRHIZO Pro software (Regent Instruments Inc., Quebec, Canada). Results of the analysis of variance revealed significant genotype variation (p < 0.01) for nearly all traits, except average length of link. Turkish accessions recorded the longest average root length and greater root branching, whereas Chinese and Korean accessions exhibited significantly larger root diameters and higher root biomass-related trait values. Landraces developed the most extensive root systems, wild relatives showed high trait variability, and cultivars were more uniform in root volume and diameter. Correlation analysis revealed strong positive associations (p < 0.01) among root length, surface area, projected area, and branching, suggesting a coordinated system for soil exploration. Principal component analysis identified five functional clusters, explaining 93.485% of total variation. This study revealed the presence of a wide range of variation in radish root traits and provides a foundation for trait selection, targeting resource-use efficiency, and market needs.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-39212-0.

Introduction

Radish (Raphanus sativus L., 2n = 2x = 18) is an ancient annual or biennial herbaceous crop belonging to the Brassicaceae family^1,1,2 and has a long history of cultivation and diversification. The genus Raphanus comprises two species, R. sativus and its wild relative R. raphanistrum, the latter encompassing several subspecies primarily distributed across the Mediterranean basin and adjacent regions³. Archaeobotanical, morphological, and genetic evidence indicate that radish originated in the broader Mediterranean–West Asian region, extending into South Asia, which is recognized as a primary center of diversity and domestication⁴. Subsequent dispersal and human selection resulted in multiple independent domestication events, giving rise to distinct European and Asian cultivated forms⁴. Cultivated radish is traditionally classified into several morphotypes, including European small-rooted, East Asian long-rooted, black, oilseed, and rat-tail types, reflecting extensive phenotypic and functional diversity^2,5. Wild relatives of radish are native not only to the Mediterranean coast but also to coastal regions of East Asia, including China, Korea, and Japan, underscoring the crop’s broad ecological adaptation and evolutionary history². Today, radish is widely cultivated across tropical, subtropical, and temperate regions for its edible roots and tender leaves, with particularly high diversification and consumption in Asia^6,7,4.

In South Korea, radish represents a major vegetable crop, accounting for approximately 10% of total vegetable cultivation area, highlighting its cultural, nutritional, and economic importance⁸. In terms of food preparation, radish plays a central role in Korean food culture through its extensive use in kimchi, Korea’s traditional fermented food⁹. Several widely consumed kimchi types—kkakdugi (made from radish roots), yeolmu kimchi (made from radish leaves), and chonggak kimchi (made from both roots and leaves) use radish as the primary ingredient and are considered representative traditional kimchi varieties in Korea^10,11. Kimchi production relies heavily on radish and cabbage, with over 200 recognized kimchi varieties, underscoring the crop’s cultural, nutritional, and economic importance across Korea¹². The morphological diversity in radish, expressed in variations of root shape, size, and color, makes it an excellent model for studies on root development and secondary metabolite accumulation². The root of radish comprises both hypocotyl-derived upper and root-derived lower regions and serves as a storage site for starch and various bioactive compounds^13,14. Additionally, the crop displays highly diverse leaf morphologies and edible siliques, with consumption habits shaped by regional preferences^15,16. Both root and shoot tissues are nutritionally valuable, containing carbohydrates, vitamins, minerals, and dietary fiber⁶; however, the leafy parts are frequently underutilized, particularly in regions where root consumption is prioritized. Substantial morphological and physiological variation exists among cultivated radish accessions worldwide, primarily shaped by consumer preferences and local agroecological conditions. These variations serve as critical selection parameters for breeding programs aiming to improve root quality, stress resilience, disease resistance, and adaptability^2,17. Characterizing and conserving radish germplasm is therefore, essential to ensure success in crop improvement initiatives. Germplasm resources, including wild relatives, landraces, and modern cultivars offer invaluable genetic diversity and underpin breeding objectives related to agronomic, nutritional, and phytochemical traits. In particular, wild radish (R. raphanistrum subsp. sativus) carries alleles linked to resistance against both biotic and abiotic stresses such as drought, salinity, and pests^18,19. The National Agrobiodiversity Center of the Rural Development Administration (RDA) in Jeonju, South Korea, maintains a diverse collection of radish germplasm, including wild relatives, and landraces². Besides, there are popular cultivars such as Asian seed twenty-day-old radish (AS20DOR) and CHERISH 1 which are characterized by rapid growth (~ 20 days post-sowing), compact root morphology, and suitability for fresh salads and urban gardening. Despite their increasing use, there is limited phenotypic characterization of these cultivars, particularly regarding root morphological traits. This gap in phenotypic data limits their optimal utilization in breeding and agronomic planning. Additionally, radish production faces persistent constraints from preharvest physiological disorders such as forking, cracking, pithiness, and internal browning²⁰. These defects, which significantly reduce both yield and market appeal, are primarily influenced by environmental factors, such as moisture stress, temperature fluctuations, soil structure, nutrient imbalances, and improper harvest timing that disrupt root tissue heirgh root traits are recognized as promising breeding targets for improved nutrient use efficiency, particularly phosphorus, potassium, and calcium^21,22, the practical assessment of root systems has lagged behind due to challenges in accessing intact roots and the labor-intensive nature of traditional methods²³. Technological advancements, such as 2D image-based root phenotyping platforms (e.g., WinRHIZO Pro), have made it possible to quantify root traits like root length, surface area, average diameter, volume, tip number, and branching^24,25, yet these tools remain underutilized in radish research. To address these gaps, the present study employed a 2D image-based phenotyping technique (WinRHIZO Pro) to systematically evaluate root morphological traits across a diverse panel of radish accessions from the RDA Genebank. Conducted under controlled greenhouse conditions, which allow for efficient environmental regulation and disease mitigation, this research introduces a robust phenotyping framework for radish root traits. The findings are expected to enhance breeding strategies targeting improved root quality, stress tolerance, and cultivation efficiency, while also promoting the sustainable utilization of genetic resources in both commercial and subsistence agriculture.

Materials and methods

Plant growth and experimental setup

Seeds of 23 radish accessions of different geographic origins (nine countries, Table S1) were obtained from the Rural Development Administration (RDA) Genebank, National Agro-biodiversity Center, Jeonju, Republic of Korea. The classification of each accession as a cultivar, landrace, or wild relative was based on the passport and characterization data provided by the RDA Genebank, which documented the biological status, origin, and improvement history of each accession at the time of collection. All accessions belonged to the Cherry Belle radish market class, characterized by rapid growth and early maturity (20 to 30 days), and are well adapted for vegetable garden cultivation. The experiment was arranged in a completely randomized design (CRD) with three replications. Each replication consisted of 10 individual plants per accession, each plant grown in a separate polyvinyl chloride (PVC) pipe, resulting in a total of 30 plants per genotype. Prior to sowing, seeds were surface-sterilized in 70% ethanol (Sigma-Aldrich, MO, USA) for 1 min and rinsed three times with sterile distilled water to minimize microbial contamination. Two seeds were sown in PVC pipes measuring 6 cm diameter x 40 cm height, containing commercial horticultural substrate (Tobirang, Baekkwang Fertility, Andong, Korea). Seedlings were thinned to one healthy plant per pipe after emergence. The experiment was conducted in a controlled Venlo-type greenhouse under day/night temperatures of 25 ± 1 °C (maximum 32 ± 3 °C) and 18 ± 1 °C, respectively, with relative humidity maintained at 60–70%. Root phenotyping was conducted at 20 days after sowing, corresponding to the early harvestable stage for Cherry Belle radish.

Plant harvesting and sampling

At 20 days after germination, corresponding to the early root bulking stage, plants of each genotype were carefully harvested by loosening the soil in each pipe to preserve root integrity. Residual soil was removed by rinsing roots under low-pressure tap water. Shoots were separated from roots by excision at the root–shoot junction using sterilized scissors. Immediately after separation, shoot fresh weight (SFW) and root fresh weight (RFW) were recorded using an analytical balance with 0.001 g precision. For imaging, roots were gently blotted with absorbent paper to remove surface moisture and arranged on a transparent acrylic tray (30 cm × 20 cm) containing a shallow layer of clean water to minimize overlap and glare. Root systems were carefully spread using tweezers to ensure flatness and clear visibility before image acquisition for morphometric analysis. Before imaging, the root shape of each accession was classified according to the International Union for the Protection of New Varieties of Plants (UPOV) guidelines (1999; https://www.upov.int/documents/d/upov/tg-documents-en-tg064_06.pdf ).

Root trait imaging and analysis

Cleaned root samples were transported to the laboratory for two-dimensional (2D) image acquisition, using a high-resolution flatbed scanner (Expression 12000XL, Epson, Japan) equipped with a transparent acrylic tray (30 cm × 20 cm). Roots were submerged in a thin layer of water during scanning to maximize overlap and structural distortion. Image were captured at 600 dpi resolution and saved in PNG format to maintain a high-resolution for quantitative analysis. Image processing and root trait extraction were performed using WinRHIZO Pro software (Regent Instruments Inc., Quebec, Canada), based on the WinRHIZO description (https://regent.qc.ca/assets/winrhizo_software.html), as described in earlier studies²⁶. Images were captured at a fixed resolution of 600 dpi, which was used by WinRHIZO Pro to convert pixel measurements into metric units. Scanner calibration was verified prior to analysis to ensure measurement accuracy. The software automatically segmented the root system and quantified morphological parameters. All genotypes were scanned and analyzed under identical conditions. Individual roots were gently separated using tweezers to avoid overlap, and images with visible root crossing were excluded from analysis. Only well-dispersed root samples were scanned and analyzed using WinRHIZO Pro, which reliably extracts root traits when roots are non-overlapping. The workflow for root image analysis is presented in Fig. 1, and the measured traits are measured in Table 1, including fresh weight biomass used to assess biomass allocation among radish accessions.

Fig. 1.

Procedure for radish fresh weight biomass and root system architecture.

Table 1.

Description of the 16 quantitative traits studied in the 23 radish germplasm.

Trait	Abbreviation	Category	Description
Root length	RL	Morphology	Total length of roots traced within the 2D scanned image (cm).
Projected area	PA	Morphology	2D area occupied by roots (cm²)
Surface area	SA	Morphology	Estimated total surface area of roots (cm²)
Root volume	RV	Morphology	Total estimated root volume (cm³)
Average diameter	AD	Morphology	Mean diameter of all root segments (cm)
Average projected area of link	APAL	Architecture	Mean projected area per segment (cm²)
Average surface area of link	ASAL	Architecture	Surface area per root segment (cm²)
Average diameter of link	ADL	Architecture	Mean diameter per segment (cm)
Number of root tips	NRT	Architecture	Total number of terminal root ends (count)
Forks	-	Architecture	Branching points where roots split (count)
Number of crossings	NOC	Architecture	Points where roots overlap in 2D (count)
Average length of link	ALOL	Architecture	Mean length between forks/tips (cm)
Average branching angle of link	ABAL	Architecture	Mean angle at which lateral roots emerge (°)
Root fresh weight	RFW	Fresh biomass	Total mean weight of the harvested root (g)
Shoot fresh weight	SFW	Fresh biomass	Total mean weight of the harvested shoot (g)
Root- Shoot fresh weight	RSFW	Fresh biomass	Root fresh weight to shoot fresh weight (ratio)

Statistical analysis

Measurements from the 10 plants within each replication were averaged prior to statistical analysis. Data were expressed as means ± standard deviations (SD) based on three biological replicates per genotype. Descriptive statistics, including minimum, maximum, mean, standard deviation, and coefficient of variation (CV) were calculated to assess variability among the measured traits. A two-way analysis of variance (ANOVA) was conducted with genotype treated as a fixed effect and replication treated as a random effect. Mean comparisons among genotypes were performed using Tukey’s Honest Significant Difference (HSD) test at p < 0.05. To identify major sources of phenotypic variation and classify genotypes based on root traits, multivariate analyses were performed. Principal component analysis (PCA) was used to summarize trait variation and identify key contributing traits, while hierarchical cluster analysis (HCA) grouped genotypes according to phenotypic similarity. Pearson’s correlation coefficients (two-tailed, p < 0.05) were calculated to examine interrelationships among root traits. All statistical analyses and heatmaps were generated in RStudio software (v4.5.0), and PCA visualizations were created using SIMCA-P software (v13.0, Umetrics, Umeå, Sweden).

Results

Phenotypic variation in root system architecture and biomass in radish

The radish root, the main edible organ, showed substantial genotypic variation in both root and shoot traits (Table S2&S3). Sixteen traits, including 14 root attributes, shoot fresh weight, and root-to-shoot ratio, were evaluated under greenhouse conditions. ANOVA revealed highly significant genotypic effects (p < 0.01) for nearly all traits (Table 2). Wide variability was observed, with forks ranging from 64 to 11,204, root tips from 173 to 7602, crossings from 1 to 2564, and root length from 23.10 to 1435.11 mm (Table 2). Chinese accessions such as Hong yingtao luobo and Ying tiao shui luobo produced exceptionally long roots (> 900 mm), while Hong bai 20 ri had very short roots (< 50 mm). The largest diameters were found in Hong bai 20 ri (2.22 cm), Kruglaya chernaya (2.01 cm), and Dunganskiy (1.66 cm). Root surface area and volume varied widely, with Hong yingtao luobo, PI140433, and Ying tiao shui luobo showing the highest projected areas (> 28 cm²). Biomass differed significantly, with Negrityanka showing the highest shoot weight (20.43 g) and UZB-GJG-2009-10/3–13 the highest root weight (13.23 g). Full genotype-specific data are presented in Supplementary Table S2.

Table 2.

Summary and phenotypic variations of 16 root system architecture and fresh weight biomass traits in radish.

Variable	Minimum	Maximum	Mean	SD	CV	Skewness	Kurtosis	G	R	G x R
RL	23.10	1435.11	354.84	370.355	104.37	1.04	0.01	***	NS	NS
PA	3.01	39.78	13.98	9.07	64.85	0.96	-0.12	***	NS	NS
SA	9.47	124.96	43.93	28.49	64.85	0.96	-0.12	***	NS	NS
AD	0.26	2.77	0.84	0.65	77.65	1.28	0.81	***	*	NS
RV	0.14	1.83	0.64	0.29	45.81	1.52	3.43	***	*	*
NRT	173.00	7602.00	1866.33	1573.56	84.31	1.14	1.26	***	**	**
Forks	64.00	11204.00	2303.12	2696.28	117.07	1.20	0.70	***	**	**
NOC	1.00	2564.00	511.97	635.94	124.21	1.21	0.54	***	NS	NS
ALOL	0.07	0.18	0.09	0.02	19.69	1.56	5.53	***	NS	*
APAL	0.00	0.03	0.01	0.01	113.46	0.98	0.87	***	NS	NS
ASAL	0.01	0.10	0.02	0.02	76.05	1.86	4.81	***	NS	NS
ADL	0.18	0.74	0.33	0.09	27.02	1.84	6.11	NS	NS	NS
ABAL	0.00	58.12	52.84	7.31	13.83	-5.75	40.94	*	NS	**
SFW	1.40	25.20	13.86	3.63	26.20	0.17	2.01	***	*	NS
RFW	0.60	15.30	7.08	3.73	52.69	0.00	-0.48	***	NS	NS
RSFW	0.05	1.00	0.51	0.23	44.64	-0.53	-0.51	***	NS	*

RL: Root length, PA: Projected area, SA: Surface area, RV: Root volume, AD: Average diameter, APAL: Average projected area of link, ASAL: Average surface area of link, ADL: Average diameter of link, NRT: Number of root tips, NOC: Number of crossings, ALOL: Average length of link, ABAL: Average branching angle of link, SFW: Shoot fresh weight, RFW: Root fresh weight, RSFW: Root- Shoot fresh weight. G: Genotype, R: Replication.

Variation in root system architecture and biomass by country of origin

Substantial variation was observed in root architectural and biomass traits among radish accessions from nine countries (Table 2; Table S4). Japanese and Turkish radishes exhibited the broadest trait ranges, indicating high within-country diversity. Japanese accessions showed particularly wide variation in root length and surface area, reflecting both slender and thick-rooted morphotypes. Turkish accessions displayed extensive branching, with the highest mean number of root forks (> 7000) and crossings (> 700), as well as the longest average roots (~ 1050 mm). Chinese accessions were characterized by thicker and more voluminous roots, showing the largest mean diameter (~ 1.5 cm) and root volume (> 0.8 cm³), consistent with their adaptation to heavier soils and high-yield breeding lines. In contrast, American and Russian radishes exhibited shorter and finer roots with narrower diameters, reflecting more compact root systems. Korean accessions showed relatively high root–shoot ratios and larger link surface areas, indicating greater allocation of biomass to belowground organs.

Variation in root system architecture and biomass by genotype

Significant genotypic variation (p < 0.05) was detected for all root morphological and architectural traits (Table 3). Wild relatives showed the widest variability in projected area (3.01–5.37 cm²), surface area (9.47–16.88 cm²), root tips (173–648), link length (0.08–0.13 mm), link dimensions, and branching angle (44.87–51.37°), with the greatest spread in surface area, tips, and link diameter. Landraces exhibited the highest variability in forks (2561–4048), crossings (616–836), shoot fresh weight (14.67–19.13 g), root fresh weight (5.23–8.70 g), and root–shoot ratio (0.33–0.52). Cultivars varied most in root length (260–325 mm), average diameter (0.83–1.11 cm), and root volume (0.59–0.79 cm³). Wild relatives had the highest means for diameter (1.16 cm), link length (0.11 mm), root fresh weight (7.27 g), and root–shoot ratio (0.58), while landraces showed the highest values for root length (497.59 mm), surface area (54.52 cm²), tips (2633), forks (3351), and shoot weight (16.84 g). Cultivars had the highest root volume (0.67 cm³). Wild relatives recorded the lowest averages for most traits, landraces for diameter-related traits, and cultivars only for root fresh weight (6.96 g).

Table 3.

Comparison of root system architecture and fresh weight biomass traits across radish wild relative, landraces and cultivated varieties.

Trait	Replication	Wild relative	Landrace	Cultivar	Trait	Replication	Wild relative	Landrace	Cultivar
RL	Min	23.10	485.94	260.42	ALOL	Min	0.08	0.08	0.09
	Max	48.96	510.10	325.21		Max	0.13	0.09	0.10
	Mean	34.89	497.59	302.22		Mean	0.11	0.08	0.10
	SD	13.08	12.10	36.26		SD	0.02	0.01	0.00
	CV	37.49	2.43	12.00		CV	22.51	7.85	3.30
PA	Min	3.01	16.67	11.25	APAL	Min	0.01	0.00	0.01
	Max	5.37	17.97	13.48		Max	0.02	0.00	0.01
	Mean	3.96	17.35	12.59		Mean	0.01	0.00	0.01
	SD	1.25	0.65	1.18		SD	0.00	0.00	0.00
	CV	31.45	3.77	9.39		CV	44.31	17.78	24.32
SA	Min	9.47	52.37	35.34	ASAL	Min	0.02	0.01	0.02
	Max	16.88	56.47	42.36		Max	0.05	0.01	0.03
	Mean	12.44	54.52	39.55		Mean	0.03	0.01	0.03
	SD	3.91	2.05	3.72		SD	0.01	0.00	0.01
	CV	31.45	3.77	9.39		CV	44.46	18.43	24.38
AD	Min	1.07	0.44	0.83	ADL	Min	0.21	0.25	0.33
	Max	1.31	0.64	1.11		Max	0.32	0.35	0.35
	Mean	1.16	0.51	0.94		Mean	0.28	0.31	0.34
	SD	0.13	0.11	0.15		SD	0.06	0.05	0.01
	CV	11.06	22.46	15.67		CV	22.90	17.27	2.97
RV	Min	0.29	0.51	0.59	ABAL	Min	44.87	53.62	48.91
	Max	0.46	0.56	0.79		Max	51.37	56.24	53.74
	Mean	0.36	0.54	0.67		Mean	48.28	54.84	52.02
	SD	0.09	0.03	0.11		SD	3.26	1.32	2.69
	CV	26.33	5.92	16.20		CV	6.76	2.40	5.18
NRT	Min	173.00	2567.33	1451.92	SFW	Min	11.80	14.67	12.20
	Max	648.00	2674.00	1815.62		Max	13.10	19.13	14.55
	Mean	343.67	2633.44	1666.49		Mean	12.47	16.84	13.71
	SD	264.21	57.74	190.47		SD	0.65	2.24	1.31
	CV	76.88	2.19	11.43		CV	5.22	13.27	9.57
Forks	Min	64.00	2561.33	1409.69	RFW	Min	7.10	5.23	6.50
	Max	170.00	4048.33	2058.00		Max	7.40	8.70	7.33
	Mean	123.33	3350.78	1813.08		Mean	7.27	7.11	6.96
	SD	54.12	747.75	352.01		SD	0.15	1.75	0.42
	CV	43.88	22.32	19.41		CV	2.10	24.63	6.06
NOC	Min	1.00	616.33	304.92	RSFW	Min	0.54	0.33	0.47
	Max	24.00	836.33	470.46		Max	0.62	0.52	0.55
	Mean	16.33	742.00	407.00		Mean	0.58	0.40	0.50
	SD	13.28	113.30	89.27		SD	0.04	0.10	0.04
	CV	81.30	15.27	21.93		CV	6.66	25.58	8.07

RL: Root length, PA: Projected area, SA: Surface area, RV: Root volume, AD: Average diameter, APAL: Average projected area of link, ASAL: Average surface area of link, ADL: Average diameter of link, NRT: Number of root tips, NOC: Number of crossings, ALOL: Average length of link, ABAL: Average branching angle of link, SFW: Shoot fresh weight, RFW: Root fresh weight, RSFW: Root- Shoot fresh weight.

Variations in root system architecture and biomass by root shape

Root shape significantly influenced variation in radish root morphological and architectural traits (Table 2 & Table S5). Triangular and transverse triangular roots showed the widest variability, with triangular roots exhibiting the largest ranges in projected area (21.12–26.76 cm²), surface area (66.35–84.08 cm²), root tips (3469–4178), forks (4728–6730), crossings (1076–1537), and branching angle (46.50–57.01°). Transverse elliptic roots showed the greatest variation in average diameter (1.96–2.63 cm), root volume (0.83–1.34 cm³), link length (0.08–0.12 cm), link projected area (0.01–0.03 cm²), link surface area (0.04–0.10 cm²), and shoot weight (10.30–14.40 g). Cylindric shapes varied most in link diameter (0.26–0.33 cm) and root–shoot ratio (0.51–0.68), while elliptic roots showed the broadest range in root length (393.78–581.62 mm). Spheric and inverse triangle shapes displayed minimal variability. Generally, traits with the widest ranges also had the highest means, whereas triangular and transverse elliptic shapes, despite high variability, showed low mean values for most traits.

Cluster analysis of root system architecture and fresh weight biomass traits

Hierarchical cluster analysis separated the radish accessions into two major groups based on root morphological and architectural traits (Fig. 2). Cluster 1 contained eight accessions, mostly of Russian origin, including Puthan Red (G3), Negrityanka (G21), PI140433 (G1), HA17 (G18), Kruglaya chernaya (G11), UZB-GJG-2009-10/3–13 (G9), Ranniy Krasniy (G5), and Chempion (G8). These accessions generally exhibited high values for root surface area, projected area, root length, crossings, forks, tips, and branching angle, except Negrityanka, which showed a lower branching angle. Cluster 2 contained the majority of accessions, characterized by lower values for these traits but higher average surface area of link, projected area of link, average diameter, root–shoot fresh weight ratio, root and shoot fresh weights, and link length. Kvarta (G19) and CHERISH-1 (G22) displayed distinct profiles. Overall, clustering patterns did not align with geographic origin, indicating that root architectural variation is driven more by functional differences than by regional grouping.

Fig. 2.

Heatmap showing the phenotypic relationship between accessions and root system architecture and fresh weight biomass traits in 23 radish accessions. As displayed in the color bond scale, the blue and red colors in the heatmap denotes higher and lower relative values. ABAL: Average branching angle of link, NRT: Number of root tips, NOC: Number of crossings, RL: Root length, PA: Projected area, SA: Surface area, ALOL: Average length of link, ADL: Average diameter of link, SFW: Shoot fresh weight, RFW: Root fresh weight, RSFW: Root- Shoot fresh weight, RV: Root volume, AD: Average diameter, APAL: Average projected area of link, ASAL: Average surface area of link. G1 - G23 represents the 23 radish germplasm studied (Table S2) (RFW) were positioned far from the axis. Average branching angle of link (ABAL) had a weak contribution, showing a positive loading on PC1 and a negative loading on PC2, with an isolated placement, indicating minimal influence on the major variance. The PCA score plot (Fig. 3B) clearly distinguished genotypic groups based on trait associations, reinforcing the clustering patterns and uncovering both central and outlier behaviors among the radish accessions.

Principal components analysis

Principal component analysis of the 16 traits showed that the first five components explained 93.49% of total variation (Fig. 3; Table S6). PC1 accounted for 60.40% and was dominated by strong positive loadings for root length, tips, forks, crossings, projected area, surface area, and average link projected area (> 0.90), while average diameter, link surface area, root fresh weight, and root–shoot ratio loaded negatively (Fig. 3; Table S6). PC2 (12.51%) was driven mainly by root volume and average link diameter, with moderate contributions from projected and surface area. PC3 (9.91%) reflected branching angle and shoot and root fresh weights, with negative loading for link length. PC4 (5.48%) emphasized shoot weight and link length, and PC5 (5.19%) was defined by negative loadings for link diameter and positive contributions from branching angle. Loading patterns showed strong clustering of major architectural traits along PC1 and PC2 (Fig. 4A). Genotypes such as Puthan Red, PI140433, HA17, and Negrityanka aligned positively on both axes, reflecting higher values for multiple root-related traits and overall strong root performance (Fig. 4B).

Fig. 3.

Heatmap illustrating the contribution (loadings) of individual traits to the first five principal components. Color intensity represents the relative magnitude and direction of contribution of each trait, thus, highlighting traits that strongly influence each principal component.

Fig. 4.

Principal component analysis loading plot of 16 quantitative traits (A) and score plot of the 23 radish germplasm (B) along the first two principal components. ABAL: Average branching angle of link, NRT: Number of root tips, NOC: Number of crossings, RL: Root length, PA: Projected area, SA: Surface area, ALOL: Average length of link, ADL: Average diameter of link, SFW: Shoot fresh weight, RFW: Root fresh weight, RSFW: Root- Shoot fresh weight, RV: Root volume, AD: Average diameter, APAL: Average projected area of link, ASAL: Average surface area of link. G1 to G23 represents the 23 radish germplasm studied (Table S1 and S2).

Analysis of correlation coefficients

Pearson correlation analysis revealed biologically meaningful associations among independently measured radish root traits (Fig. 5; Table S7). Strong positive correlations were observed among core architectural traits, particularly between root length (RL), number of forks, number of crossings (NOC), and number of root tips (NRT) (r = 0.96–0.99; p < 0.001), indicating coordinated development of root elongation and branching. Root fresh weight (RFW) showed a positive correlation with shoot fresh weight (SFW) (r = 0.64; p < 0.001) and a strong association with the root–shoot fresh weight ratio (RSFW) (r = 0.88; p < 0.001)), reflecting linked biomass accumulation patterns. In contrast, average diameter (AD) and branching angle–related traits exhibited weak or negative correlations with root length and branching parameters, suggesting partial independence between root thickness, angular traits, and overall architectural expansion.

Fig. 5.

Pearson’s correlation analysis of 16 traits in radish. The traits (abbreviations) are the same as those described in Fig. 2. The bigger the circle the stronger and higher the significance. Red and blue circles indicate positive and negative correlations, respectively.

Discussion

Root system development is a key quantitative trait influencing plant adaptability across environments, and understanding root behavior is critical for improving yield stability, breeding resilient cultivars, and conserving genetic diversity²⁷. The substantial variation observed among radish accessions confirms that RSA is strongly genotype-dependent, while some traits remain environmentally responsive. This combination of genetic control and phenotypic plasticity highlights RSA as an important breeding target for enhancing radish performance under variable and climate-sensitive conditions. Although most root traits were stable across replications, several traits, including root volume, number of root tips, number of forks, average link length, average branching angle, and root–shoot fresh weight ratio, exhibited significant genotype × replication interactions. This indicates that these traits are partially modulated by environmental conditions in addition to genetic effects. Similar genotype-dependent variation in RSA traits has been reported in Brassica species under nutrient stress, reinforcing the adaptive significance of root plasticity^28,29. Among the evaluated traits, root length, branching intensity (tips and forks), and root surface area emerged as the most discriminative features among accessions, reflecting strong genetic differentiation in soil exploration strategies. Deep and highly branched root systems, observed in accessions such as Hong yingtao luobo and Ying tiao shui luobo, are likely advantageous for water and nutrient acquisition under drought-prone or low-input conditions. In contrast, shallow and weakly branched phenotypes, including Hong bai 20 ri and HA17, may be more vulnerable to moisture and nutrient limitations, suggesting environment-specific suitability rather than broad adaptation. Variation in root diameter and volume further suggests contrasting adaptive strategies. Thick-rooted accessions may favor biomass accumulation, mechanical penetration, and carbohydrate storage, consistent with the functional roles of root diameter in soil–root interaction and internal transport^30,31. Conversely, fine-rooted accessions may support more efficient soil foraging networks, albeit with lower overall root mass. These trade-offs are important considerations for breeding programs targeting either root yield or stress resilience. The observed diversity in RSA traits was efficiently captured using image-based phenotyping with WinRHIZO, supporting its utility in root phenomics for crop improvement^24,32. Overall, the extensive RSA diversity identified in this study provides valuable opportunities for selecting radish genotypes with improved resource-use efficiency and adaptability, particularly in the context of increasing climate variability.

Our findings reinforce the value of RSA traits as effective selection criteria in radish breeding programs. Current breeding priorities, early maturity, root uniformity, abiotic stress tolerance, and bolting resistance are closely linked to root development and function⁴. Under increasing climate variability, optimizing RSA is becoming essential, as water scarcity is a major constraint on global crop productivity³³. Radish is particularly sensitive to drought stress, which reduces biomass accumulation and impairs photosynthetic performance^34,35. Traits such as root length and shoot fresh weight are known to decline under water-limited conditions^33,36, making them critical indicators of drought susceptibility and resilience. The wide variability observed in key RSA traits, including root length, surface area, and branching intensity, provides opportunities to identify genotypes with enhanced drought resilience and nutrient-use efficiency. Accessions exhibiting deep and highly branched root systems, such as Hong yingtao luobo, Ying tiao shui luobo, and PI140433, represent promising ideotypes for stress-prone environments, where efficient water capture and soil exploration are essential³⁷. The high variability of root length and shoot fresh weight—traits sensitive to drought stress^33,36 further supports their utility as selection targets for developing robust radish cultivars. Root systems are central to plant adaptation and productivity, with RSA traits such as root length, lateral spread, and branching exhibiting high phenotypic plasticity in response to environmental conditions¹⁷. This plasticity enables accessions from diverse geographic origins to develop distinct adaptive strategies, emphasizing the importance of evaluating broad germplasm collections for breeding purposes³⁸. The observed variation in root morphological and architectural traits among radish accessions reflects both underlying genetic diversity and ecological adaptation³⁹. Geographic patterns in RSA traits further highlight region-specific adaptive strategies. Japanese accessions displayed broad variation in root length, branching traits, and shoot biomass, indicating genetically diverse material suitable for improving drought tolerance and yield stability⁴⁰. Turkish accessions showed high mean values for root length, forks, tips, and crossings, suggesting vigorous and spatially extensive root systems advantageous under drought-prone conditions^42,4. Chinese accessions were characterized by greater average diameter and root volume, traits associated with storage root development and marketable yield^25,41. American accessions exhibited notable variation in link length, root fresh weight, and root–shoot ratio, reflecting adaptive biomass allocation under water-limited environments^42,43. Russian genotypes varied strongly in branching angle and link diameter, traits linked to mechanical stability and horizontal soil exploration^30,31,44, while Korean accessions showed higher surface area and root–shoot ratios, indicating efficient resource acquisition and biomass partitioning^39,43. RSA traits such as root length, diameter, surface area, and branching directly influence water and nutrient uptake, stress tolerance, and yield performance, making them key targets for radish improvement across diverse and changing environments^25,41,44.

Significant genotypic differences in root morphological and architectural traits were observed among wild relatives, landraces, and cultivars, reflecting distinct evolutionary histories and selection pressures. Wild relatives exhibited the greatest variability in fine-scale architectural traits, including projected area, surface area, number of root tips, average link length, and branching angle. These traits enhance root complexity, soil exploration, and nutrient uptake, conferring adaptive advantages under marginal and stress-prone environments^{24,30,45–47}. Their high variation in surface area and root tips highlights wild germplasm as a critical resource for improving absorptive capacity and root–soil contact under water- and nutrient-limited conditions. Landraces displayed greater variability in branching and biomass-related traits, such as forks, crossings, shoot and root fresh weight, and root–shoot ratio, reflecting intermediate RSA patterns shaped by long-term farmer selection^48,49. Higher mean values for root length, projected area, surface area, forks, and crossings indicate denser and deeper root systems, traits associated with improved soil penetration and enhanced nutrient acquisition, particularly phosphorus, under resource-limited environments^30,40. These characteristics position landraces as versatile genetic materials bridging wild adaptability and cultivated performance. Cultivars showed substantial variation in root length, volume, fork number, and average diameter, traits closely linked to marketable yield and consumer preferences^24,50. However, their comparatively lower root fresh weight reflects breeding emphasis on uniformity, shape, and quality rather than root biomass allocation⁴. In contrast, wild relatives exhibited the highest average diameter, link length, root fresh weight, and root–shoot ratio, underscoring their potential value for drought tolerance and low-input agricultural systems^46,51,52. Elevated root–shoot ratios indicate preferential biomass allocation belowground, enhancing water and nutrient uptake, while variation in root diameter and branching contributes to hydraulic conductivity and stress resilience^30,47. Root shape further modulates RSA expression⁵³. Triangular roots displayed the greatest architectural variability, whereas cylindrical and elliptic roots were more stable, with cylindrical roots exhibiting higher root–shoot ratios and elliptic roots showing greater average length. These patterns suggest functional specialization, where root shape influences resource acquisition and storage dynamics. This agrees with earlier reports that ancient radish varieties were predominantly long and tapered, facilitating deep soil penetration⁴, and that root growth trajectories differ by shape class⁵⁴.

Multivariate and correlation analyses reinforced these interpretations. Accessions 1 and 18, which loaded positively on both PC1 and PC2, combined greater root length, surface area, and branching intensity, identifying them as strong candidates for breeding programs targeting enhanced RSA and overall performance. Pearson correlation analysis revealed strong positive associations among root length, number of tips, forks, and crossings, indicating coordinated development between primary and lateral roots. Such integration enhances soil exploration and resource acquisition and has been documented in maize and Brassica under both optimal and stress conditions^28,55–57. The positive correlation between root and shoot fresh weight reflects coordinated biomass accumulation, while the absence of correlation between shoot fresh weight and root–shoot ratio suggests genotype-specific allocation strategies. Negative or weak associations between root diameter traits and root length or branching parameters indicate that thicker roots do not necessarily confer greater architectural complexity, supporting the existence of alternative RSA strategies. The independence of root length from root volume and of average link diameter from shoot fresh weight further highlights the multidimensional regulation of root traits, consistent with reports in soybean and other crops²³. The contrasting RSA patterns observed among wild relatives, landraces, and cultivars emphasize the importance of integrating diverse germplasm and multidimensional root traits into radish breeding programs aimed at improving stress resilience, resource-use efficiency, and yield stability⁵⁸.

Conclusion

This study revealed substantial phenotypic variation in root morphological and architectural traits among 23 diverse radish accessions, encompassing wild relatives, landraces, and cultivars from nine countries. Significant differences were observed among the genotypes in key traits such as root length, number of forks, crossings, root tips, and biomass allocation, which revealed the rich genetic diversity within the radish germplasm. Correlation analysis identified tightly linked trait clusters related to root architecture, while PCA and cluster analysis effectively differentiated accessions based on trait expression, identifying both integrative and divergent phenotypic patterns.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

K.O. : Conceptualization, Methodology, Data analysis, writing – original draft, Writing–review and editing. D.-W.K.: Methodology, Writing –review and editing. S.M .: Methodology, Writing –review and editing. M. N.B. : Methodology, Writing –review and editing. S-H.K.: Conceptualization, Methodology, investigation, Writing – review and editing, supervision, Funding acquisition.

Funding

This research was carried out with the support of the “Research Program for Agricultural Science and Technology Development (Project NO. PJ01425501/RS-2019-RD007776), National Institute of Agricultural Sciences, Rural Development Administration (RDA), Republic of Korea.

Data availability

The data sets supporting the results of this article are included within the article and its supplementary files.

Declarations

Competing interests

The authors declare no competing interests.