henhao
Sign in
Plant BiologybioRxiv
Preprint

The Sorghum Lipid Database (SoLD): population-scale lipidomics linking environmental and genetic variation in the Sorghum Association Panel

Nirwan TandukariD, Ruthie StokesiD, Richard Elmer BoylesiD, Zachary W. BrentoniD, Katherine LouieiD, Rubén Rellán‐ÁlvareziD

Abstract

Sorghum (Sorghum bicolor) is a climate-resilient crop whose acclimation to nutrient limitation and low temperature likely involves extensive lipidome reconfiguration. Lipids are key membrane components, carbon and energy stores, and mediators of stress signaling, yet population-scale lipidomics data for sorghum are limited. We present the Sorghum Lipid Database (SoLD), a curated lipidomics resource from the Sorghum Association Panel grown under two field regimes, (i) a nutrient- sufficient with usual planting date environment (control) and (ii) a low input treatment with reduced nitrogen and phosphorus, earlier planting, and no pesticide application (low input). Using high-resolution LC-MS, we quantified 266 lipid species and found the lipidome broadly similar across trials, with four low-input-associated signatures relative to control, (i) depletion of sulfoquinovosyldiacylglycerol, (ii) triacylglycerol enrichment, (iii) phospholipid redistribution centered on phosphatidylserine, and (iv) a shift in the lysophosphatidylcholine-to-lysophosphatidylethanolamine ratio. Genomic heritability was zero at the median under control against 0.171 under low input, and more of the lipidome could be mapped under low input. GWAS of lipid species, class sums, and class ratios recovered loci that were largely trial-specific, with candidate genes assigned by linkage disequilibrium and prioritized by LD-aware ontology enrichment and reaction mapping. Control-associated loci yielded twenty enriched gene sets, all confined to triacylglycerol traits and led by cutin biosynthesis, whereas low-input-associated loci yielded fifty-one across five lipid classes, centered on nutrient transport and signaling, cold-responsive regulators, and jasmonate and oxylipin genes, and including a chromosome 3 block carrying two LOW PHOSPHATE ROOT homologues. SoLD thus connects sorghum lipid diversity to environmental and genetic variation, and lets users move from lipid class to individual species and associated candidate loci. It is freely accessible at https://nirwan.shinyapps.io/SAP-Lipidomics-Database/.

PDFDOICodeData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
0
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (3 Oct 2026) · CC BY 4.0

SoLD, a leaf lipidomics resource (244 species) for the Sorghum Association Panel across two field trials, reveals low-input sulfoquinovosyldiacylglycerol depletion, triacylglycerol enrichment, and environment-specific GWAS loci.

Plant lipidomics resources spanning natural genetic variation are scarce compared with transcriptomic data, and large LC-MS studies suffer from signal drift, field-layout artifacts, and inconsistent lipid annotation. SoLD combines QC-normalized, spatially corrected, GNPS-annotated lipid profiles of a sorghum diversity panel with GWAS, letting users move from lipid classes or species to candidate loci under contrasting field regimes.

Key results

  1. ShownVersus control (CTL), the low-input (LIN) trial showed four lipid signatures: sulfoquinovosyldiacylglycerol (SQDG) depletion, triacylglycerol (TG) enrichment favoring longer chains, phosphatidylserine (PS)-centered phospholipid rebalancing, and altered LPC/LPE lysophospholipid balance.

    Mean %TIC: TG 2.7% (CTL) vs 5.4% (LIN); SQDG 2.9% vs 1.4%; MGDG 34.5% vs 32.1%; PC 25.6% vs 24.8%. PS was among the strongest positive class CLR shifts and rose against PC, PE and PG in TG-referenced ALR contrasts; PE weighted mean double bonds rose from about 2.68 to 3.38. LION enriched TG, DG and lipid-droplet terms in LIN. · Lipidome overview for CTL and LIN SAP; Abstract

  2. ShownCTL-LIN differences were population-wide rather than outlier-driven: all class ratios and nearly all shared species differed after Benjamini-Hochberg (BH) correction with fully stable effect direction, and high-variance lipid traits recurred across trials.

    All 33 class ratios were BH-significant (padj<0.05) with leave-one-out sign stability 1.0 (median absolute log10 effect 0.4743). Of 244 species, 153 were shared; 151 of these were significant and all 153 had jackknife stability 1.0 (median absolute CLR effect 0.7084). MGDG(18:3/18:3) had top genotype-level variance in both trials; seven of ten top-variance lipids were shared. · Lipid variation across genotypes; Statistical Tests

  3. SuggestedUnder CTL, recurrence-filtered GWAS prioritized a small cross-trait locus set tied to lipid oxidation, neutral-lipid turnover and isoprenoid supply (LOX6-like, OBL1-like lipase, FPS-like genes), chloroplast and carbon metabolism, root and reproductive development, and cell-wall organization.

    At p ≤ 1 × 10^−7, individual-lipid GWAS gave 3,478 genes; ≥13 phenotypes kept 17 (6 lead SNPs, 5 regions). Sum/ratio GWAS gave 4,385 loci; ≥72 phenotypes kept 20. Strongest displayed associations: LOX6-like SORBI_3006G248300 p = 1.37 × 10^−10; DRO1/DRL1-like SORBI_3002G373700 p = 7.77 × 10^−13; OsPG1-homolog SORBI_3003G141800 p = 1.30 × 10^−10. · CTL recurrent GWAS signals highlight basal…; CTL candidates highlight lipid oxidation…; Figure legends

  4. SuggestedUnder LIN, recurrent loci pointed to nutrient-status signaling and remobilization (tryptophan synthase pair, PHR1-like regulator, ATG8-interacting protein), callose and expansin wall remodeling, defense receptors, and cold-linked wax (CER1/WDA1-like) and cullin proteostasis genes.

    At p ≤ 1 × 10^−7, ≥7 phenotypes yielded 48 individual-lipid candidates (30 annotated; top recurrence 11 phenotypes) and ≥50 phenotypes yielded 31 sum/ratio candidates. Strongest displayed associations: SLO1-like SORBI_3001G256400 p = 2.08 × 10^−16; CER1/WDA1-like SORBI_3001G222700 p = 1.96 × 10^−13; PHR1-like SORBI_3001G384300 p = 5.72 × 10^−9. · LIN recurrent GWAS signals highlight stress and…; Nutrient sensing and recycling candidates…; Cuticle barrier formation and ubiquitin-mediated…; Figure legends

  5. SuggestedRice-referenced LINEX2 reaction branches showed LIN shifts toward TG-side product balances in diacylglycerol (DG)-TG and phospholipid acyl-transfer branches (opposite for TG hydrolysis), overlapping GWAS hits for DGAT1, an SDP1-like PNPLA lipase, and an LCAT-like gene.

    Paired LIN-CTL reaction-balance effects: PNPLA3 +0.327, LCAT* +0.238, LRO1 +0.232, PNPLA1 −0.268 (BH p < 10^−56). GWAS overlap: DGAT1 SORBI_3010G170000 (22 LIN traits, 7 reaction-linked), LCAT-like SORBI_3001G103800 (19, 14 linked), PNPLA-domain SORBI_3001G041900 (8, 8 linked); best p 3.16 × 10^−15 to 1.13 × 10^−9. · LINEX-based reaction mapping, reaction-balance…; Reaction-supported lipid-remodeling candidates…; Lipid Metabolic Network Analysis (LINEX2)

Methods, data and reproducibility

Leaf lipids of Sorghum Association Panel accessions from control (CTL) and low-input (LIN) trials were LC-MS/MS profiled, GNPS-annotated, SERRF- and SpATS-corrected, and compared as %TIC, centered log-ratios (CLR) and class log-ratios with Wilcoxon tests; GEMMA mixed-model GWAS per species, class sum and ratio fed recurrence filtering and LINEX2 mapping.

  • Sorghum Association Panel trials, Pee Dee Research and Education Center (SC): control (CTL, 2019) and low-input (LIN, 2022; reduced N and P, earlier planting, no pesticides) · Sorghum bicolor · n = 400 genotypes
  • Leaf samples (V6 to V8) profiled by positive- and negative-mode LC-MS/MS at JGI/LBNL; 244 annotated lipid species (153 shared, 49 CTL-only, 42 LIN-only) · Sorghum bicolor · n = 244 lipid species
  • SAP SNP genotypes filtered at MAF 0.05 and heterozygosity 0.8, used for centered kinship and GEMMA GWAS · Sorghum bicolor
  • Sorghum bicolor reference genome v3.1 (BTx623); genes within ±25 kb of SNPs annotated via Phytozome, SorghumBase and TAIR orthologs · Sorghum bicolor
  • LINEX2 rice (OSA) reference lipid-reaction network with RHEA annotations, substituting for an unavailable sorghum reference · Oryza sativa (rice)
Methods & tools
Thermo Exploris 120 Orbitrap with Agilent 1290 UHPLC (LC-MS/MS), MZmine 2, MS-FLO, GNPS spectral library matching, SERRF (random forest QC normalization), SpATS (two-dimensional P-spline spatial correction, R), GEMMA v2.3 mixed linear model, LINEX2 (Lipid Network Explorer) with RHEA reactions, LION/web lipid ontology enrichment, LIPID MAPS classification, Phytozome, SorghumBase, TAIR, Shiny (SoLD web application)
Code
https://github.com/rr-lab/SAP-Lipidomics-Database-Database
Data & accessions
https://nirwan.shinyapps.io/SAP-Lipidomics-Database/
Software & environment
R 4.3.3; GEMMA v2.3; MZmine 2; SERRF online server; SpATS (smoothing settings stated to be in GitHub); Sorghum bicolor reference v3.1 (BTx623); Thermo Exploris 120 Orbitrap with Agilent 1290 UHPLC
Reproducibility gaps
No raw LC-MS or genotype data accession is given; the Shiny URL appears in two forms (one with a repeated -Database suffix) and the GitHub URL also ends in a doubled -Database; extraction and LC-MS settings are deferred to cited protocols, and SpATS parameters to GitHub.
Limitations (6)
  • CTL (2019) and LIN (2022) were grown in different years, so contrasts are between-trial differences that confound reduced N and P, earlier planting meant to mimic cold, absent pest management, and year effects; the authors call for same-year common-garden trials. · stated by the authors
  • LINEX2 lacks a sorghum reference, so the rice (Oryza sativa) network was used; reaction branches are annotation-guided hypotheses, and reaction-balance scores summarize lipid-pool ratios, not enzyme activity or in vivo flux. · stated by the authors
  • Recurrence filtering near the top 0.5% favors pleiotropic or upstream regulators and under-represents trait-specific candidates; most gene functions are inferred from rice, maize or Arabidopsis homologs and still need validation in sorghum. · stated by the authors
  • Lipid identities are putative GNPS matches (levels 2/3) that may be isomers or near-mass analogs; features lacking library matches were dropped, and adduct or in-source fragment degeneracy may inflate feature counts. · stated by the authors
  • The main text gives no per-trial count of genotypes retained after excluding samples with ≥70% zero features, nor SNP marker counts, so GWAS power and the size of the paired-genotype tests cannot be assessed without supplements. · noted in this summary
  • Genes were assigned via fixed ±25 kb windows around significant SNPs without LD-based intervals, conditional analysis or heritability estimates; the 17 CTL recurrent genes collapse to 5 regions, so many may be co-located passengers. · noted in this summary

Who should read:Read this if you study plant lipid remodeling under low nutrients or cold, map metabolite traits by GWAS in sorghum, or want a queryable lipid-to-locus resource for choosing candidate genes or extreme lines.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISNigam '26Liu '26Bandara '19Maina '21Raju '23Boatwright '22
81% semantic matchBioinformatics
Deepti Nigam, Sarah Metwally, Fang Chen+1 more·bioRxiv·Mar 16, 2026