Relationships Among Arsenic-Related Traits, Including Rice Grain Arsenic Concentration and Straighthead Resistance, as Revealed by Genome-Wide Association

Overview

This genome-wide association study (GWAS) used the USDA Rice Minicore Collection (RMC, approximately 166 to 168 accessions) to identify quantitative trait loci (QTL) for grain arsenic concentration, hull silica concentration, straighthead disease (StHD) resistance, and related agronomic traits (days to heading, plant height) in rice. The study also measured grain concentrations of phosphorus, sulfur, calcium, and copper. A machine-learning Bayesian network approach was used alongside conventional GWA to model causal relationships among these arsenic-related traits. The study identified 195 GWA-QTL across nine traits, including 15 QTL for grain arsenic and 23 for StHD. Grain arsenic varied 3- to 100-fold among the diverse rice genotypes, and StHD and grain arsenic were positively correlated (r = 0.23, p = 0.011 in the full RMC).

This is a genetic/breeding study reporting grain-level arsenic concentrations across a diverse rice germplasm collection grown under field conditions (including MSMA-treated plots to induce arsenic stress). It provides context for understanding genetic variation in arsenic accumulation rather than finished-food survey data.

Key numbers

All values as reported in the paper’s own units (ppm, equivalent to mg/kg, for grain concentrations).

Grain arsenic (whole RMC, n = 168):

• Median approximately 0.5 ppm (from Figure 1I box plot)
• Range: approximately 0.1 to 2.2 ppm across all accessions
• 3- to 100-fold differences in grain arsenic observed among diverse rice genotypes

Grain arsenic by subpopulation:

• Aromatic (ARO) and tropical japonica (TRJ) subpopulations had the highest median grain-As
• Temperate japonica (TEJ) had notably lower grain-As, and was also more resistant to StHD, earlier maturing, and shorter

Straighthead disease (StHD) ratings (MSMA-treated plots, RMC):

• Average StHD 2-year LSmean: 6.7 +/- 1.3 (scale 1 to 9)
• Range: 3.4 to 9
• None of the accessions were as StHD resistant as Zhe733 (resistant check)
• Majority of RMC (101 of 156 accessions) had LSmeans greater than 6.0

Hull silica (ppm):

• Beaumont TX average: 188 ppm
• Stuttgart AR average: 215 ppm

Correlation between StHD and grain-As: r = 0.23 (p = 0.011) for the entire RMC; r = 0.36 (p = 0.003) among indica accessions; not significant among japonica accessions (r = 0.50, p = 0.10).

QTL identified:

• 33 QTL for hull silica concentration (Si)
• 23 QTL for StHD
• 15 QTL for grain arsenic
• Remaining QTL for P (11), S (18), Ca (19), Cu (9), DHD (39), PHT (28)
• Nine of 15 grain-As QTL coincided with previously reported arsenic (As) QTL
• Thirty-two percent (41/128) of StHD and elemental QTL were validated by other studies

Bayesian network findings: The DAG model indicated that StHD, grain-As, and grain-Cu are all likely caused by DHD (days to heading), either directly or indirectly through PHT (plant height). The model found no direct relationships between Si, P, S, Ca, and Cu with either StHD or grain-As.

Evidence fitness

EF-2: This study provides grain-level arsenic concentration data for 168 accessions of the USDA Rice Minicore Collection grown across two US field sites (Stuttgart AR and Beaumont TX) over two years. Grain arsenic was measured by ICP-MS from field-harvested grain. The GWAS design with high-density SNP mapping (3.2 million markers) is methodologically rigorous. The data are from US rice production environments, which limits direct applicability to other regions but provides a well-characterized genetic framework for grain arsenic variation.

Methods (brief)

USDA Rice Minicore Collection (RMC): 166 to 168 accessions from 14 global rice-growing regions. Genetic data from 156 PCR-based markers and 3,200,320 SNP markers. Field trials at Stuttgart, AR and Beaumont, TX. Randomized complete block design with plots of StHD-susceptible “Cocodrie” and resistant “Zhe733” checks. StHD evaluated in MSMA-treated soil plots at Stuttgart AR over 2 years. Grain arsenic, P, S, Ca, and Cu measured in grains from 2 replications at Beaumont TX by ICP-MS. Hull Si measured from rice harvested at both locations. GWAS conducted using GAPIT (R package) with MLMM and BLINK models. Bayesian network analysis using the bnlearn R package with iterative learning and 10-fold cross-validation.

Implications

This study is relevant as context for understanding the genetic architecture of arsenic accumulation in rice grain. The key findings for the wiki are: (1) grain arsenic varies 3- to 100-fold among diverse rice genotypes, confirming that cultivar selection is a meaningful lever for reducing grain arsenic; (2) grain arsenic and straighthead disease susceptibility are positively correlated and share underlying QTL, suggesting that breeding for StHD resistance may co-select for lower grain arsenic; (3) the temperate japonica subpopulation was notably lower in grain arsenic, more StHD resistant, and shorter, which may inform cultivar recommendations for arsenic-prone environments; (4) the Bayesian network suggests that days to heading (maturity) may be a root cause driving both grain arsenic and StHD, rather than arsenic directly causing StHD.

Provenance notes

• Open access under CC BY 4.0.
• Received 1 October 2021, accepted 23 December 2021, published 14 March 2022.
• Field trials at USDA-ARS Dale Bumpers National Rice Research Center, Stuttgart AR, and USDA-ARS Plant Science Research Unit, Beaumont TX.
• The Rice Minicore Collection is a public USDA-ARS resource.
• Grain-As measured from Beaumont TX harvests only (2 replications, 2 years).

Related evidence

• Rice
• Arsenic

Update history

The five most recent substantive edits to this page, classified major (evidence or structure moved), correction (a published value or statement was wrong and has been fixed), or minor (narrative rewritten without changing the underlying evidence). Each description is derived from what the edit did to this page; the linked commit is the authoritative record, routine regeneration passes are excluded, and the full version history lives in git.