Periodic Reporting for period 1 - ToxBiome (Characterising the microbiome and mycotoxin profiles of winter wheat varieties grown under sustainable crop protection system)
Période du rapport: 2024-01-01 au 2025-12-31
Résumé du contexte et des objectifs généraux du projet
Cereals are central to Europe’s food and feed security, but fungal infection can contaminate grain with mycotoxins that threaten human and animal health and create major economic losses when grain lots are downgraded or rejected. This is not a single toxin problem: wheat can contain mixtures of regulated and emerging mycotoxins, and co occurrence complicates monitoring and management because reducing one toxin does not automatically reduce the overall food safety burden.
Predicting and controlling mycotoxin risk remains difficult in real farming conditions. Toxin accumulation depends on weather and crop growth stage, but also on wheat genotype, since varieties differ in susceptibility, flowering traits, and how they interact with microbial communities on the wheat ear. The wider microbial community living on the ear can influence which pathogens dominate and which toxin profiles are produced. At the same time, disease control tools such as fungicides can have variable outcomes across seasons, locations, and genotypes. In practice, fungicide applications may reduce visible disease and lower some toxins, while other toxins remain unchanged or increase. These uncertainties limit the reliability of decision making based only on historical expectations or weather driven risk models and can lead to either unnecessary chemical use or missed periods of elevated risk.
Against this background, the project addressed a clear need for field relevant evidence and tools that support integrated pest management and risk based monitoring of cereal mycotoxins. The overall objective was to understand how agronomic interventions, host genotype, and microbial ecology interact to shape multi mycotoxin outcomes in wheat, and to translate that understanding into approaches that are more robust across different locations.
The specific objectives were to: (1) quantify multi mycotoxin profiles in wheat under contrasting field conditions and management regimes; (2) characterise how the wheat ear microbial community changes in response to fungicide programmes and natural variation across sites and genotypes; (3) link microbial community restructuring to shifts in toxin mixtures, moving beyond single pathogen explanations; and (4) test whether microbiome state can act as a transferable biological indicator for within season classification of mycotoxin risk, complementing or improving on approaches that rely on weather summaries alone.
The pathway to impact is improved decision support for growers, advisors, and the cereal supply chain: earlier identification of higher risk situations, more targeted sampling and intervention timing, and better awareness of trade offs where management suppresses some toxins but favours others. In the longer term, these outputs can contribute to safer grain, fewer regulatory exceedances, reduced waste, and more sustainable crop protection strategies aligned with EU priorities for food safety, environmental sustainability, and resilient agriculture.
Predicting and controlling mycotoxin risk remains difficult in real farming conditions. Toxin accumulation depends on weather and crop growth stage, but also on wheat genotype, since varieties differ in susceptibility, flowering traits, and how they interact with microbial communities on the wheat ear. The wider microbial community living on the ear can influence which pathogens dominate and which toxin profiles are produced. At the same time, disease control tools such as fungicides can have variable outcomes across seasons, locations, and genotypes. In practice, fungicide applications may reduce visible disease and lower some toxins, while other toxins remain unchanged or increase. These uncertainties limit the reliability of decision making based only on historical expectations or weather driven risk models and can lead to either unnecessary chemical use or missed periods of elevated risk.
Against this background, the project addressed a clear need for field relevant evidence and tools that support integrated pest management and risk based monitoring of cereal mycotoxins. The overall objective was to understand how agronomic interventions, host genotype, and microbial ecology interact to shape multi mycotoxin outcomes in wheat, and to translate that understanding into approaches that are more robust across different locations.
The specific objectives were to: (1) quantify multi mycotoxin profiles in wheat under contrasting field conditions and management regimes; (2) characterise how the wheat ear microbial community changes in response to fungicide programmes and natural variation across sites and genotypes; (3) link microbial community restructuring to shifts in toxin mixtures, moving beyond single pathogen explanations; and (4) test whether microbiome state can act as a transferable biological indicator for within season classification of mycotoxin risk, complementing or improving on approaches that rely on weather summaries alone.
The pathway to impact is improved decision support for growers, advisors, and the cereal supply chain: earlier identification of higher risk situations, more targeted sampling and intervention timing, and better awareness of trade offs where management suppresses some toxins but favours others. In the longer term, these outputs can contribute to safer grain, fewer regulatory exceedances, reduced waste, and more sustainable crop protection strategies aligned with EU priorities for food safety, environmental sustainability, and resilient agriculture.
Travail effectué depuis le début du projet jusqu’à la fin de la période considérée dans le rapport et principaux résultats atteints jusqu’à présent
The project generated integrated field evidence linking agronomy, wheat genotype, microbial community structure, and multi mycotoxin outcomes. Work was organised around two complementary strands: first, field experiments to quantify how fungicide programmes restructure the wheat ear mycobiome and redistribute toxin mixtures; second, modelling analyses to test whether microbiome state can serve as a transferable indicator of within season mycotoxin risk across locations.
For the fungicide and genotype strand, replicated winter wheat field experiments were conducted at two trial sites on the island of Ireland, Limavady and Backweston, which differ in agronomic context and management history. A panel of 11 winter wheat genotypes spanning diverse agronomic backgrounds was grown under fungicide treated and untreated regimes in a randomised complete block design, with three independent plot replicates for each genotype and treatment at each site. Fungicide treated plots received site appropriate multiple spray programmes representative of regional practice for foliar disease and Fusarium head blight control, while untreated plots received no applications. Wheat ears were sampled during a consistent phenological window in June 2024 at BBCH growth stages 77 to 83, when mycotoxin accumulation is underway. To capture within plot heterogeneity while minimising border effects, each plot was subdivided into five sampling sectors and 40 ears per sector were collected, giving 200 ears per plot. Ears were pooled into one composite sample per plot, transported on ice, stored frozen, then freeze dried and milled. Subsamples for chemical analysis and DNA based profiling were taken from the same milled material to ensure direct comparability.
For chemical outcomes, a validated multi mycotoxin LC MS/MS workflow was applied using a SCIEX QTRAP 6500+ triple quadrupole platform in multiple reaction monitoring mode. Quantification used isotope dilution where labelled analogues were available, and matrix matched calibration for remaining analytes. Batch quality control included blanks and fortified materials, with identification criteria based on retention time agreement, ion ratio agreement, and signal to noise. Analytical performance was documented, including limits of quantification of 5 µg kg⁻¹ for DON, 2.5 µg kg⁻¹ for T-2, and 5 µg kg⁻¹ for HT-2. In parallel, fungal community composition was profiled using ITS amplicon sequencing, and interpretation was strengthened with targeted marker assays using a Luminex based approach to quantify key Fusarium species signals and the trichothecene biosynthesis gene TRI5. These targeted assays were used to distinguish proportional shifts in sequencing data from true changes in pathogen load and biosynthetic potential.
For the transferable indicator strand, field sampling was extended to three winter wheat trial sites in June 2024, Backweston, Crossnacreevy, and Limavady, with fungicide treated and untreated plots interspersed within blocks. Composite sampling used the same structured approach of 200 ears per plot, followed by freeze drying, milling, and matched subsampling for chemistry and DNA extraction. Both bacterial and fungal communities were characterised using 16S and ITS amplicon sequencing, with rigorous controls including extraction blanks, no template controls, and a fungal mock community. Sequence processing used an amplicon sequence variant workflow in R with DADA2, contaminant removal with prevalence based filtering, and robustness checks to confirm that conclusions were not sensitive to contaminant handling choices.
Predictive modelling was then used to evaluate portability across locations under strict spatial validation using leave one location out evaluation. Microbiome features were constructed in a way that supports transfer, using genus level aggregation, centred log ratio transformation, and dimension reduction fitted on training data only. Classification used penalised logistic regression with elastic net regularisation, with uncertainty quantified via bootstrap confidence intervals and above chance testing using within fold permutation tests. Drop one ablation analyses were implemented to identify genera that contribute unique, non redundant information under spatial transfer.
Across these activities, the main achievements were the delivery of a quality controlled multi layer dataset linking wheat genotype, fungicide management, fungal community restructuring, targeted pathogen and pathway markers, and quantified mycotoxin outcomes, plus a validated modelling framework for assessing whether microbiome state provides transferable information on within season toxin risk beyond weather summaries alone.
For the fungicide and genotype strand, replicated winter wheat field experiments were conducted at two trial sites on the island of Ireland, Limavady and Backweston, which differ in agronomic context and management history. A panel of 11 winter wheat genotypes spanning diverse agronomic backgrounds was grown under fungicide treated and untreated regimes in a randomised complete block design, with three independent plot replicates for each genotype and treatment at each site. Fungicide treated plots received site appropriate multiple spray programmes representative of regional practice for foliar disease and Fusarium head blight control, while untreated plots received no applications. Wheat ears were sampled during a consistent phenological window in June 2024 at BBCH growth stages 77 to 83, when mycotoxin accumulation is underway. To capture within plot heterogeneity while minimising border effects, each plot was subdivided into five sampling sectors and 40 ears per sector were collected, giving 200 ears per plot. Ears were pooled into one composite sample per plot, transported on ice, stored frozen, then freeze dried and milled. Subsamples for chemical analysis and DNA based profiling were taken from the same milled material to ensure direct comparability.
For chemical outcomes, a validated multi mycotoxin LC MS/MS workflow was applied using a SCIEX QTRAP 6500+ triple quadrupole platform in multiple reaction monitoring mode. Quantification used isotope dilution where labelled analogues were available, and matrix matched calibration for remaining analytes. Batch quality control included blanks and fortified materials, with identification criteria based on retention time agreement, ion ratio agreement, and signal to noise. Analytical performance was documented, including limits of quantification of 5 µg kg⁻¹ for DON, 2.5 µg kg⁻¹ for T-2, and 5 µg kg⁻¹ for HT-2. In parallel, fungal community composition was profiled using ITS amplicon sequencing, and interpretation was strengthened with targeted marker assays using a Luminex based approach to quantify key Fusarium species signals and the trichothecene biosynthesis gene TRI5. These targeted assays were used to distinguish proportional shifts in sequencing data from true changes in pathogen load and biosynthetic potential.
For the transferable indicator strand, field sampling was extended to three winter wheat trial sites in June 2024, Backweston, Crossnacreevy, and Limavady, with fungicide treated and untreated plots interspersed within blocks. Composite sampling used the same structured approach of 200 ears per plot, followed by freeze drying, milling, and matched subsampling for chemistry and DNA extraction. Both bacterial and fungal communities were characterised using 16S and ITS amplicon sequencing, with rigorous controls including extraction blanks, no template controls, and a fungal mock community. Sequence processing used an amplicon sequence variant workflow in R with DADA2, contaminant removal with prevalence based filtering, and robustness checks to confirm that conclusions were not sensitive to contaminant handling choices.
Predictive modelling was then used to evaluate portability across locations under strict spatial validation using leave one location out evaluation. Microbiome features were constructed in a way that supports transfer, using genus level aggregation, centred log ratio transformation, and dimension reduction fitted on training data only. Classification used penalised logistic regression with elastic net regularisation, with uncertainty quantified via bootstrap confidence intervals and above chance testing using within fold permutation tests. Drop one ablation analyses were implemented to identify genera that contribute unique, non redundant information under spatial transfer.
Across these activities, the main achievements were the delivery of a quality controlled multi layer dataset linking wheat genotype, fungicide management, fungal community restructuring, targeted pathogen and pathway markers, and quantified mycotoxin outcomes, plus a validated modelling framework for assessing whether microbiome state provides transferable information on within season toxin risk beyond weather summaries alone.
Progrès au-delà de l’état des connaissances et impact potentiel prévu (y compris l’impact socio-économique et les conséquences sociétales plus larges du projet jusqu’à présent)
The project produced two advances that go beyond common practice in cereal mycotoxin research and risk management.
First, it showed that fungicide management can reshape mycotoxin risk as a redistribution problem rather than a simple reduction problem. In field conditions, fungicide treated plots did not uniformly suppress all toxins. Instead, reductions in some major Fusarium associated toxins were accompanied by increases in other toxin groups, including the sum of T-2 and HT-2 and several emerging metabolites such as enniatins and beauvericin. This matters because real world safety outcomes are determined by the full toxin mixture and by regulatory thresholds that focus on specific compounds or sums. The work linked these shifts to changes in the wheat ear fungal community, indicating that intervention can drive turnover in dominant taxa and therefore change the functional toxin profile that accumulates in grain. Targeted molecular measurements of Fusarium markers and trichothecene pathway potential supported the interpretation that the observed toxin patterns were consistent with changes in the underlying pathogen assemblage and biosynthetic potential, not only with analytical variation.
Second, the project provided evidence that microbiome state can act as a transferable biological indicator of within season mycotoxin risk across locations. Under strict spatial validation using leave one location out evaluation, microbiome informed models classified higher versus lower contamination for key toxins with substantially better performance than models based on station level seasonal weather summaries. Importantly, combining weather with microbiome information did not consistently outperform microbiome alone, suggesting that community state captures integrated, local biological information that is not easily represented by coarse weather aggregates. This is a meaningful step beyond many existing predictive approaches because it prioritises portability across locations, not only good fit within a single site.
Together, these findings open a practical route toward earlier, risk based decisions in cereal production. They support the idea that disease control and monitoring should explicitly account for trade offs across toxin classes, and that microbial indicators could complement conventional agronomic and meteorological information to improve within season targeting of sampling and interventions.
Key needs for further uptake include expansion to multi year datasets to capture broader climate variability, harmonisation of sampling and analytical workflows across laboratories to support comparability, and demonstration studies that embed the approach within advisory decision support and supply chain monitoring. This would enable evaluation of operational value, including whether microbiome guided risk stratification can reduce the frequency of exceedances, improve the efficiency of testing, and support more sustainable disease control strategies.
First, it showed that fungicide management can reshape mycotoxin risk as a redistribution problem rather than a simple reduction problem. In field conditions, fungicide treated plots did not uniformly suppress all toxins. Instead, reductions in some major Fusarium associated toxins were accompanied by increases in other toxin groups, including the sum of T-2 and HT-2 and several emerging metabolites such as enniatins and beauvericin. This matters because real world safety outcomes are determined by the full toxin mixture and by regulatory thresholds that focus on specific compounds or sums. The work linked these shifts to changes in the wheat ear fungal community, indicating that intervention can drive turnover in dominant taxa and therefore change the functional toxin profile that accumulates in grain. Targeted molecular measurements of Fusarium markers and trichothecene pathway potential supported the interpretation that the observed toxin patterns were consistent with changes in the underlying pathogen assemblage and biosynthetic potential, not only with analytical variation.
Second, the project provided evidence that microbiome state can act as a transferable biological indicator of within season mycotoxin risk across locations. Under strict spatial validation using leave one location out evaluation, microbiome informed models classified higher versus lower contamination for key toxins with substantially better performance than models based on station level seasonal weather summaries. Importantly, combining weather with microbiome information did not consistently outperform microbiome alone, suggesting that community state captures integrated, local biological information that is not easily represented by coarse weather aggregates. This is a meaningful step beyond many existing predictive approaches because it prioritises portability across locations, not only good fit within a single site.
Together, these findings open a practical route toward earlier, risk based decisions in cereal production. They support the idea that disease control and monitoring should explicitly account for trade offs across toxin classes, and that microbial indicators could complement conventional agronomic and meteorological information to improve within season targeting of sampling and interventions.
Key needs for further uptake include expansion to multi year datasets to capture broader climate variability, harmonisation of sampling and analytical workflows across laboratories to support comparability, and demonstration studies that embed the approach within advisory decision support and supply chain monitoring. This would enable evaluation of operational value, including whether microbiome guided risk stratification can reduce the frequency of exceedances, improve the efficiency of testing, and support more sustainable disease control strategies.