A panel of spring wheat lines was evaluated to determine the feasibility of screening single plants for both seminal root angle and root biomass. The panel included two International Maize and Wheat Improvement Center (CIMMYT) varieties (Kingbird and Borlaug100), two Australian commercial varieties (Suntop and Mace) and six accessions from a diversity panel studied by Voss-Fels et al. . Borlaug100 was selected as the recipient background to introgress key root traits. It is a high-yielding wheat which was developed at CIMMYT and first imported into Australia in 2015 via the CIMMYT-Australia-ICARDA Germplasm Evaluation project (CAIGE). The six accessions from the diversity panel have known root biomass phenotypes and haplotypes for the root biomass QTL on chromosome 5B . Three of the six accessions were selected as donor parents for root trait introgression: SW107 and SW388 for high root biomass (both positive for the 5B QTL), and SW309 for low root biomass (negative for the 5B QTL).
Testing the ability to integrate seminal root angle and root biomass screening protocols
Two experiments were conducted to determine the feasibility of screening individual wheat plants for both seminal root angle and root biomass. The goal was to develop a non-destructive method suitable for SPS, which integrated two established protocols: 1) the ‘clear-pot’ method , which enables phenotyping for seminal root angle through image analysis, and 2) a hydroponic sand-based system , which allows efficient root washing to phenotype root dry biomass. However, the root dry biomass phenotyping method typically involves root and shoot dissection and drying, resulting in plant destruction. To integrate these methods, firstly, seminal root angle screening using the clear pot method was performed and selected plants were transplanted into sand-filled pots to grow-on for subsequent root biomass assessment. Next, visual assessment of biomass was recorded for washed intact root systems, and selected plants were transplanted and grown-on to enable generation advance or crossing using selected individual plants directly. Prior to applying this method to segregating wheat populations, it was established through experiments that root biomass phenotypes were not compromised during the transplanting process, and root biomass screening could be faithfully conducted visually rather than destructively.
Comparison of root biomass phenotypes: direct sowing versus transplants
Two genotypes were selected to test whether wheat seedlings could be transplanted from clear pots into sand pots for root biomass assessment. SW300 was included as the low root biomass standard (lacks the 5B QTL for high biomass), whereas SW411 was included as the high root biomass standard (carries the 5B QTL for high root biomass). Seeds were initially sown into clear pots for seminal root angle assessment, as described by Richard et al. . A total of 24 seeds per line were sown across two 4 L clear pots (ANOVApot®, 200 mm top diameter, 190 mm height, http://www.anovapot.com/php/anovapot.php) adopting a randomized complete block design (RCBD). Clear pots were filled with a pine bark potting media (70% composted pine bark 0-5 mm, 30% coconut peat, pH 6.35, EC = 650 ppm, nitrate = 0, ammonium < 6 ppm and phosphorus = 50 ppm), and seeds were sown using tweezers by carefully placing the seed vertically, at a depth of 2 cm and every 2.5 cm with embryo downwards and facing the pot wall to facilitate root growth along the transparent wall. After sowing, the clear pots were placed inside 4 L black pots (ANOVApot®, 200 mm diameter, 190 mm height) to block light from reaching the developing roots. Plants were grown in the glasshouse at a constant temperature (17 ±2 °C) over 24 hours with diurnal (12 hour) natural light. Imaging of seminal root angle was performed at five days after sowing, using a digital camera (Canon PowerShot SX600 HS 16MP Ultra-Zoom) and root angle was measured manually using ImageJ software (http://imagej.nih.gov/ij/). Individual plants representing the population tails or extreme root angle phenotypes were selected, including both narrow and wide root angles.
After the imaging step, seedlings were transplanted into pots filled with sand for subsequent root biomass phenotyping, as described by Voss-Fels et al. . Each 1.4 L ANOVA pot (ANOVApot®, 137 mm diameter, 140 mm height) was filled with approximately 1.7 kg of coarse washed sand (particle size ranging 0.075-4.75 mm) to facilitate root washing. With gentle water flow, roots could be easily and cleanly separated from the sand, which minimised damage to roots. The experimental design consisted of 16 pots with two plants of the same genotype in each pot, placed inside a clear plastic storage container (65 cm length and 35 cm width with a capacity of 36 L), allowing eight replications per genotype. Containers were fitted with capillary mats to ensure water and nutrient uptake and hand-watered daily using a commercial hydroponic solution with complete nutrients (Cultiplex Extra-Nutrex Grown). The hydroponic solution was diluted to adjust the nutrient concentrations so that the growth requirements of the plants were met as they developed (days 1-10: 1.50 mL/L, days 11-17: 2 mL/L, days 18-21: 2.50 mL/L).
At the time of transplanting, a second root biomass experiment was initiated, where seeds were sown directly into sand pots. The transplant and direct-seeded experiments were performed simultaneously, adopting a similar design and layout. The plants in both assays were grown in the same temperature-controlled glasshouse set to 22/17 °C (day/night) under natural (12 hours) photoperiod. The experiments were designed to assess the impact of transplanting root biomass and on the ability to accurately differentiate high and low root biomass phenotypes. At 21 days after transplanting and direct sowing, plants were extracted with minimum root disruption by placing the pot in a bucket of water and washing off the sand with clean water. Roots and shoots were separated using scissors, and roots were placed in a dehydrator at 65 °C for 72 hours. A Tukey’s test was performed to determine differences in root dry biomass within and between the experiments using the corrected multiple comparison method with a confidence interval of 95% and an error rate of 5%, using the R package ‘agricolae´ (software Version 4.0.2, R Core team 2020).
Evaluating the ability to perform non-destructive visual assessment of root biomass
The protocol for phenotyping root biomass reported by Voss-Fels et al.  is destructive, as roots must be dissected from the shoot before weighing, making it impossible to use plants with desirable phenotypes for further crossing. Therefore, non-destructive visual scoring of the size of the root system was assessed, which could serve as a surrogate for root biomass weight. Seeds of 17 genotypes were sown directly into sand-filled pots (as described above), using 6 replicate pots per genotype, and four plants per pot. A total of 102 pots were arranged according to a RCBD design across six containers (i.e. 17 pots/container). Plants were watered daily using the hydroponic solution as described above.
Roots were washed 21 days after sowing and were arranged on a clear flat surface to facilitate visual scoring. The size of the root system for each plant was visually assessed using a scale of 1–6, where 1 = very fine root system with short root length and very few surface roots, 2 = fine root system, short root length and few surface roots, 3 = fine root system, short root length and some surface roots, 4 = intermediate root system, long root length and intermediate surface roots, 5 = strong root system, long root length and strong surface roots, and 6 = strong root system, long root length and strong surface roots with nodal roots clearly visible (Fig. 1a). To minimise error and variability, visual scoring was performed by the same person. Prior to scoring, an assessment for a full range of phenotypes was performed and used as an ‘eye-adjustment’. After root scoring, roots and shoots were separated 26 days after sowing from the stem tissue above the crown and both sections were placed in an air-forced dehydrator at 65°C for 72 hours. Dry weight of root and shoot biomass was recorded using a scale (AND, HR-200 scales) with 0.0001 g accuracy. The reliability of visual scoring for root biomass was examined through correlation with the actual root dry biomass (Fig. 1b). The relationship between root dry biomass and shoot dry biomass was also explored to determine if selection targeting root biomass would result in indirect selection for shoot biomass (Fig. 1c). Furthermore, to investigate the potential genetic variation in root-shoot biomass configurations, root:shoot ratio (R:S) was calculated for each of the 17 genotypes. Following ANOVA, a Fisher's least significant difference (LSD) test was conducted to compare the means to detect differences between genotypes with a 95% family-wise confidence interval with the function LSD.test using agricolae in R (software Version 4.0.2, R Core team 2020).
Overview of the single plant selection approach for root trait introgression
A visual summary of the six key steps involved in the selection pipeline is provided in Fig. 2a. This process integrates non-destructive phenotypic screening for seminal root angle and root biomass, MAS for a major root biomass QTL, and backcrossing under speed breeding to accelerate plant development. This approach was used to rapidly generate elite introgression lines using Borlaug100 as the recipient background. Selection aimed to create introgression lines with four different root trait configurations (Fig. 2b): wide-high root biomass, wide-low root biomass, narrow-high root biomass and narrow-low root biomass. A summary of each step is provided below and a detailed list of materials used for SPS with corresponding descriptions are exemplified in the Supplementary info.
Step 1 – Seminal root angle screening
The SPS approach started with assessment of a large segregating population for seminal root angle using the clear pot method, as per Richard et al. . Five days after sowing the seminal roots were imaged (Canon PowerShot SX600 HS 16MP Ultra-Zoom) and seminal root angle measured using ImageJ software (http://imagej.nih.gov/ij/). Individual plants representing the population tails or extreme root angle phenotypes were selected, including both narrow and wide root angles.
Step 2 – Transplanting into sand (semi-hydroponic sand-based system)
The selected plants were carefully extracted from clear pots and transplanted into sand-filled pots (two plants per pot) for root biomass assessment. Pots were placed into containers fitted with capillary mats where 15 pots were placed in each container in an RCBD design. Plants were watered daily using hydroponic solution (1.50 mL of Cultiplex Extra-Nutrex Grown per litre of water), where concentration slowly increased according to plant growth: days 1-10: 1.50 mL/L, days 11-17: 2 mL/L, days 18-21: 2.50 mL/L.
Step 3 – KASP marker screening
Leaf tissues were sampled from wheat plants at the seedling stage to ensure quality DNA was extracted. Four pieces of 3 cm long leaf tissue were placed in 1.2 mL cluster tube (96-tube racks) and freeze-dried for 48 hours prior to dispatchment to collaborators at the Department of Plant Breeding, Justus Liebig University, Giessen, Germany. Samples were then genotyped using the high-quality extracted DNA and genotypic data were obtained to assist in selecting individuals for crossing. Selection for root biomass was based on KASP markers developed for the major QTL reported on chromosome 5B . Three robust KASP assays (HapB3-2, HapB6-1 and HapA2-2) for the 5B locus were developed by Makhoul et al.  to distinguish the haplotype combination associated with high root biomass from other haplotype combinations associated with low root biomass . The high biomass trait is associated with ‘T’ allele for marker BS00029852_51 and ‘T’ allele for marker Tdurum_contig48959_1172 in haploblock b and with ‘T’ allele for marker Excalibur_c25522_755 in haploblock a.
Step 4 – Root biomass scoring and selection
Twenty-one days after transplanting, the plants were extracted with minimum disruption to the roots and the sand was washed off by placing the pot in a bucket of clean water. Following root washing, all the plants within the same category of root angle were lined up based on the numbering of the pots on a clear surface. Root biomass for each plant was scored visually using the 1-6 scale. Individual plants representing the population tails or extreme root biomass phenotypes were selected, including both high and low root biomass. Prior to scoring, an assessment for a full range of phenotypes was performed and used as an ‘eye-adjustment’ where recurrent parent and donor lines were scored first, followed by the progenies. Further selection within the tails was applied based on the KASP marker results. This ensured that individuals selected for high root biomass displayed both superior phenotypes and carried the desirable allele for the 5B QTL region. To capture other loci that could be important for trait expression, individuals which lacked the QTL but displayed high root biomass were also retained. Hence, to increase the confidence and accuracy of selection, a combination of both MAS and phenotypic selection was employed.
Step 5 and step 6 – Growing-on selected plants for backcrossing or line development
Finally, the selected plants carrying the desired combinations of root traits were transplanted into potting mix and grown-on under speed breeding conditions  to accelerate plant development and enable rapid backcrossing or selfing for line development.
Development of introgression lines with different root configurations
The three donor lines (SW107, SW388 and SW309) for high and low root biomass were crossed to Borlaug100 to create F1 seeds. An overview of the crossing scheme is provided in Supplementary Fig. 1. The F1 plants were backcrossed to the recurrent parent, and the BC1F1 plants were self-pollinated to produce large segregating (BC1F2) populations. To accelerate population development, all generations were grown under speed breeding conditions at UQ glasshouse facilities . The SPS approach outlined in Fig. 2, was applied to the BC1F2 populations to select individual plants representative of all four root system configurations. The phenotypic screening process was repeated for every consecutive plant generation to increase the homozygosity for root traits in the selected lines. A Tukey’s test was performed to determine differences between the mean of each density population (narrow and wide tails) and the recurrent parent.
Investigating the effectiveness of single plant selection in segregating populations
The progenies from BC2F2 plants were selected for ‘narrow’ and ‘wide’ seminal root angles by growing the BC2F2:F3 populations and screening for seminal root angle. The population distributions for both ‘wide’ and ‘narrow’ groups were compared with the recurrent parent.
A Pearson’s correlation coefficient was calculated for the BC2F2:F3 populations for seminal root angle and root biomass derived by visual score, to further explore the relationship between the two root traits. A total of 120 BC2F3:F4 lines, along with the recurrent parent and donor lines, were evaluated for above-ground traits under field conditions in 2020 at The University of Queensland Gatton Research Farm, Gatton, Queensland, Australia (27°33’4’’ S, 152°16’32’’ E). The lines were sown in single 6 m long rows and key agronomic traits were recorded. Based on plant height and flowering time data collected, a total of 20 introgression lines were selected for yield evaluations in 2021 (Supplementary Table 2). This strong selection for flowering time and plant height, ensured that the introgression lines displayed a high degree of similarity to their recurrent parent for these above-ground traits. The 20 selected introgression lines were re-genotyped with KASP markers to confirm the QTL status and the lines were also phenotyped for root traits under controlled conditions. Root phenotypes displayed by introgression lines were compared to the recurrent parent Borlaug100 using a Fisher-LSD test to determine significant differences.