Next Article in Journal
A Mixed Data-Based Deep Neural Network to Estimate Leaf Area Index in Wheat Breeding Trials
Next Article in Special Issue
Variation in Anther Extrusion and Its Impact on Fusarium Head Blight and Deoxynivalenol Content in Oat (Avena sativa L.)
Previous Article in Journal
Natural Compounds as Elicitors of Plant Resistance Against Diseases and New Biocontrol Strategies
Previous Article in Special Issue
Anther Extrusion and Its Association with Fusarium Head Blight in CIMMYT Wheat Germplasm
Open AccessArticle
Peer-Review Record

Genome-wide Association Study and Genomic Prediction for Fusarium graminearum Resistance Traits in Nordic Oat (Avena sativa L.)

Agronomy 2020, 10(2), 174; https://doi.org/10.3390/agronomy10020174
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Agronomy 2020, 10(2), 174; https://doi.org/10.3390/agronomy10020174
Received: 21 December 2019 / Revised: 16 January 2020 / Accepted: 20 January 2020 / Published: 25 January 2020

Round 1

Reviewer 1 Report

The manuscript is well written and the experiments are scientifically sound. The manuscript did not conclude with any novel finding that promotes oat breeding with respect to Fusarium head blight tolerance. However, this manuscript would be a good guide for new studies to caution for appropriate selection of germplasm.

Author Response

Thank you for your feedback. Indeed, we hope that this manuscript promotes the importance of germplasm selection, but as well we hope that more studies will be published on Fusarium resistance in oat. We fixed spelling mistakes and punctuations in the manuscript.

At this point, we as well found that our method of calculating heritability from the estimates should be altered. We have another paper accepted by Plant Breeding journal, and there a reviewer noticed this fault and wanted us to fix it. We thought that in order to be consistent in both papers, we changed the method in this paper as well. The method was changed, because our lines are inbred and the diagonal of the G-matrix is close to two. This should be taken into account when computing additive genetic variance estimates. This affects the additive variance estimates and phenotypic correlations in table 3.

With kind regards,

The authors

Reviewer 2 Report

I reviewed the manuscript entitled “Genome-wide association study and genomic prediction for Fusarium graminearum resistance traits in Nordic oat (Avena sativa L.), submitted to Agronomy (ISSN 2073-4395). The authors measured several fusarium head blight (FHB) traits and agronomic related traits in an oat set of 327 breeding lines and cultivars. The manuscript also illustrated results from GWAS and genomic selection. Even though, the authors were not successful in identifying QTL associated with resistance, I believe this manuscript enriches FHB research in oat (few previous studies on FHB in oat). FHB is a quantitative trait with high GxE interaction and that could explain the lack of success in identifying GWAS hits for FHB resistance. Genomic prediction results however look promising for faster breeding of FHB resistance in oat. The manuscript was generally well written except for few typos and punctuation mistakes.

Other comments to the authors

In the abstract, the authors should describe the origin/type of germplasm used in this study. Line 23 in abstract: “some resistance related traits” be more specific and mention them In Introduction: the authors should describe the genome of oat, ploidy level, chromosome number, how chromosomes are named. For example what does chromosomes 12/13A, 02/2D mean? (not all readers are familiar with oat genome) Can you provide subtitles for each section in Materials and Methods The authors use Bonferroni corrections and FDR (p-value <0.05) which could be a bit stringent, especially for FHB traits. I would suggest they use FDR (P-value cutoff of 0.1) and check if they can have any significant hits that could in similar position to previously published QTL in oat Line 162-163 page 4: “An unpublished genetic consensus map based on seven 163 mapping populations was used” provide reference (who did this work, is it the same group?) and some information on the consensus map (e.g. how many markers, what crosses, etc). Line 190: page 4: at what stage of plant development did you measure plant height (at maturity?) include the version of ” R “ you used for your analysis Line 296, page 296: “There was a significant…. within trial variability in DON content and qFUSG” I believe there is a missing word after significant. Line 345-366, page 9: authors already mentioned this in materials and methods Table 3: include the significance levels for the correlation values (use something like * for P <0.05, ** for P < 0.01, and *** for P < 0.001).

Author Response

Thank you for the positive feedback. We appreciate it. Here are the detailed answers to the specific questions:

In the abstract, the authors should describe the origin/type of germplasm used in this study.

The information was included in the abstract. Please see the revised abstract.

Line 23 in abstract: “some resistance related traits” be more specific and mention them

The studied traits were added. Please see the revised abstract.

In Introduction: the authors should describe the genome of oat, ploidy level, chromosome number, how chromosomes are named. For example what does chromosomes 12/13A, 02/2D mean? (not all readers are familiar with oat genome)

This information was added to the introduction. Please see the revised introduction starting from row 125 onwards.

Can you provide subtitles for each section in Materials and Methods

Subtitles were added.

The authors use Bonferroni corrections and FDR (p-value <0.05) which could be a bit stringent, especially for FHB traits. I would suggest they use FDR (P-value cutoff of 0.1) and check if they can have any significant hits that could in similar position to previously published QTL in oat

The table 4 values were changed to correspond the significance level of 0.1. Please see the altered values in the table. Unfortunately, no new significant associations for K or QK models were found.

Line 162-163 page 4: “An unpublished genetic consensus map based on seven 163 mapping populations was used” provide reference (who did this work, is it the same group?) and some information on the consensus map (e.g. how many markers, what crosses, etc).

We refer to the text which was written to the other reviewer during the last round: “The map used in the study is a joint cooperation between many companies (who wish to be unnamed) and, for now, no information on it can be included due to plausible publication coming. We could have used an old consensus map (Chaffin et al. 2016), but the number of SNPs that had position information on it was very low. Therefore, we decided to use the unpublished map information.”

Line 190: page 4: at what stage of plant development did you measure plant height (at maturity?)

The information was included (please see row 208).

include the version of ” R “ you used for your analysis

R version was included (please see row 183).

Line 296, page 296: “There was a significant…. within trial variability in DON content and qFUSG” I believe there is a missing word after significant.

Actually, the sentence was changed, because it was poorly constructed. Please see the revised sentence.

Line 345-366, page 9: authors already mentioned this in materials and methods

The paragraph was revised. Please see the revised version in rows 384-386.

Table 3: include the significance levels for the correlation values (use something like * for P <0.05, ** for P < 0.01, and *** for P < 0.001). 

Overall, the computation method of phenotypic correlation was not included in the m&m. The method was added as well as specifications for the two-trait model. Due to calculation method of the phenotypic correlation, computing p-values or even confidence intervals are very difficult tasks. For the genetic correlation, confidence intervals were calculated through the given standard error estimates and the values were included in the table 3. For the phenotypic correlation, we used estimated values computed in the variance component estimation and, in the correlation, both additive and residual effects were included. Therefore, the estimation of the standard error is troublesome, while additive and residual are both present in the phenotypic correlation.

At this point, we as well found that our method of calculating heritability from the estimates should be altered. We have another paper accepted by Plant Breeding journal, and there a reviewer noticed this fault and wanted us to fix it. We thought that in order to be consistent in both papers, we changed the method in this paper as well. The method was changed, because our lines are inbred and the diagonal of the G-matrix is close to two. This should be taken into account when computing additive genetic variance estimates. This affects to the additive variance estimates and phenotypic correlations in table 3.

With kind regards,

The authors

 

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Title:  Genome-wide association study and genomic prediction for Fusarium graminearum resistance traits in Nordic oat (Avena sativa L.)

Authors: Hanna Haikka , Outi Manninen , Juho Hautsalo , Leena Pietilä , Marja Jalli , Merja Veteläinen

A brief summary:

The Authors investigated the genetics of resistance to FHB in an oat population from Finland.  A genome-wide association mapping was performed in an oat panel of 328 lines for frequency of infected kernels, DON accumulation, days to heading, plant height, germination capacity, and maturity class. Four GWAS methods were used (naiive, Q, K, and Q+K). For FHB and DON, only the naiive model that detected QTL, where other models were unable to identify significant hits. Genomic prediction was evaluated in the population for all traits. The accuracy of prediction varied among traits and was high for maturity class, heading date, and plant height.

Specific Comments

(Page, Lines) Comments

3, 140: Fhb1 is not a single gene. Fhb1 is a major quantitative trait locus. Please re-write this sentence.

3, 146: Remove “genomic” before “estimates of the GEBV”.

4, 151: What are the filial generations of these inbreds?.

4, 163-175: There is no mention of using resistant or susceptible checks in the materials and methods. Please clarify including checks with a list of their names and if the same checks were used across all trials?.

4, 156-162:  Linkage disequilibrium (LD) was not estimated using the 2,785 markers. It is important to present LD estimates as an average adjacent marker LD or an average sliding window of adjacent markers. The level of LD in the population will determine if marker coverage is sufficient to establish marker-trait association.

4, 184: “FIK was determined by plating 100 seeds on a selective media”, Were 100 seeds plated for each line and this was done for all replications. Please clarify this?.

5, 197: A clear methodology is needed for correction of spatial field variability. It is not clear how the row-column correction was performed in the materials and methods.

6, 247: I see that the genomic heritability “heritability estimated from marker data” was used in the study. Broad-sense heritability estimates for all traits based on phenotypic evaluation should be included. The experiments are unbalanced, but still heritability based on phenotypic evaluation could be estimated.

6, 247-257: There is no mentioning for analysis of variance in the materials and methods. These estimates are essential to know how much genetic and environmental variability based on the phenotypic evaluation?.

8, 297: The three check varieties Niklas, Belinda, and Mirella were not included in the materials and methods section. They should be included with the number of times they were replicated in all trials?.

10, 356: The Authors reported the correlation between two markers =0.998. It is better to report this value as LD estimated as r2.

10, 357: Using the sentence “same linkage group” is not the best because the Authors did generate linkage map in this panel. It is more accurate to say that these markers belong to a same haplotype block. This is another reason for LD to be estimated in a GWAS panel.

11, 398: Why within trial variation is not shown. Within and among trial variation should be included in the main manuscript or as supplemental information?. It is important to see that there is a presence or absence of significant variation among lines for all trials?.

11, 402-403: “Experimental designs improve the records, but most likely fail in removing the entire field induced variation”. This statement is general and not very accurate. Experimental design were used in numerous studies and showed the effectiveness in eliminating confounding factors and reducing error. Please remove this sentence?.

Figures S5 and S6: Clearly there is a peak specially for FIK on chromosomes 1 and 9 that are undetected when using Q, K, Q+K GWAS models. May be both Bonferroni and FDR corrections were too stringent to detected QTL in this population. In such situations, it was suggested to declare SNP marker p-values in the bottom 0.1 percentile of the distribution (DOI: 10.1534/genetics.109.108522; http://www.biomedcentral.com/1471-2229/12/16).

Reviewer 2 Report

Dear Authors,

Please find my comments in the attachment.

Kind regards

Agronomy-639319: Genome-wide association study and genomic prediction for Fusarium graminearum resistance traits in Nordic oat (Avena sativa L.)

Due to the larger geographical distribution of Fusarium pathogens, the FHB disease causes, nowadays, serious problems for the oat production in Nordic countries. Hence, the subject of presented manuscript is of high agricultural relevance. Among the common disease management strategies, including crop rotation and pesticide application, breeding for resistant cultivars is still the most promising strategy for a sustainable disease management. Against this background, the authors have carried out a GWAS and GS on FHB resistancerelated and selected agronomic traits using a germplasm set of 328 breeding lines adapted to Nordic climatic and photoperiodic conditions.

Major issue
The main objective was to identify resistance-related traits, genes, alleles or haplotypes with the potential to provide novel breeding strategies, - and the main problem of the presented study is unquestionable that this objective was not achieved. Except for germination capacity, which, based on genomic prediction, has been identified as a potential trait for selecting FHB resistance. Indeed, the correlation between mycotoxin contamination and reduced germination capacity has previously been reported for FHB diseased oats, particularly for Nordic countries. In this context, information on the scoring method for germination capacity, given in Materials and Methods, indicates that solely seed emergence has been scored but not seedling growth. However, for example, the DON toxin in Fusarium-damaged seed may not affect the initiation of germination but the seedling growth, - which has not been scored.
One important issue of the presented study is applied germplasm collection. First, the lack of genetic variation since primarily inbred lines and cultivars adapted to Nordic climatic conditions have been used. I must admit that I am not an expert in the field of oat production and breeding, especially regarding Nordic conditions. However, I wonder whether the integration of more varieties or even landraces from other geographical regions with similar climate conditions would have improved the genetic diversity within the germplasm collection. Second, the narrow range of available disease phenotypes, - in combination with the lack of genotypes representing stable (high) resistance and susceptibility. The low variation in disease phenotypes might result from the fact that oat has generally a good resistance against disease spread throughout panicles. This is, as stated by the authors, a probable reason for the observed unreliable DON and fungal biomass scoring as well as the high within trial variability observed for both traits. Here, the use of a disease index, combining different disease severity traits (see Rutkoski et al. 2012), could have been a possible solution which, however, has not been tested.
Final remark. I must give the authors high marks for mentioning clearly the limitations and problems of their study. However, the value of scientific novelty is undoubtedly very low, especially regarding a practical application in breeding programs.
I know that the main problem does not lie on the methodological side, meaning the applied approaches for disease assessment, association mapping or genomic prediction, but rather on oat specific features. Oat is generally considered more resistant to FHB if compared to wheat and barley. Other problematic aspects are: Genetic mechanisms underlying FHB resistance are largely unknown; a putative lack of genetic variance regarding FHB and DON response; high complexity of oat genome; and limited availability of molecular tools.
At the end, the design of germplasm collection seems to be a major issue. For instance, in a previous GWA study spring oat lines from North America and Scandinavia have been used, which enabled to gain a much higher value of scientific information (He et al. 2017). A diverse germplasm collection is a prerequisite for allelic heterogeneity, the number of functional alleles of the same gene and their association with different phenotypes.
Another factor that may have limited resolution of association mapping and heritability in the presented study could be the level of marker density, and the level to which the available SNPs can explain the phenotypic variation. Although GWAS requires a genomic map in which the marker density is higher than the extent of LD, the marker density is not well presented and discussed in the manuscript.

Minor issues
(1) A table showing the composition of used germplasm collection is missing. (2) The section Materials and Methods lacks information on fungal inoculum generation and plant infection (inoculum concentration, application procedure, number and timepoints of plant infection). This aspect can significantly affect the FHB field testing and thus, needs to be described in detail. (3) An unpublished genetic consensus map based on seven mapping populations (4) L281-281: “Belinda with moderate resistance to moderate susceptibility”, I suggest the designation “intermediate resistance” for this phenotype. Generally, comparative data on FHB response of check varieties are missing. (5) Figure 1: To ensure better traceability, the respective FHB response of check varieties should be given in the legend. In this context, I suggest using the term marker instead of check. Moreover, the figure legend contains information, which should be mentioned in the manuscript text (L300-304). (6) Fig. 1 and Tab. 2. Here, I suggest arranging the trials according to location instead of location. This would allow better overview on the respective annual variations per location. (7) The readability may be improved if the analysed traits are written out in the text (except the trait fungal biomass), and abbreviation are restricted to tables and figures.
(8) The Discussion section would require some revisions regarding English spelling and grammar. (9) The Discussion section is missing an informative comparison with previous QTL and genome-wide association studies.

Comments for author File: Comments.pdf

Back to TopTop