Using machine learning and partial dependence to evaluate robustness of best linear unbiased prediction (BLUP) for phenotypic values.

Best linear unbiased prediction (BLUP) is widely used in plant research to address experimental variation. For phenotypic values, BLUP accuracy is largely dependent on properly controlled experimental repetition and how variable components are outlined in the model. Thus, determining BLUP robustness implies the need to evaluate contributions from each repetition. Here, we assessed the robustness of BLUP values for simulated or empirical phenotypic datasets, where the BLUP value and each experimental repetition served as dependent and independent (feature) variables, respectively. Our technique incorporated machine learning and partial dependence. First, we compared the feature importance estimated with the neural networks. Second, we compared estimated average marginal effects of individual repetitions, calculated with a partial dependence analysis. We showed that contributions of experimental repetitions are unequal in a phenotypic dataset, suggesting that the calculated BLUP value is likely to be influenced by some repetitions more than others (such as failing to detect simulated true positive associations). To resolve disproportionate sources, variable components in the BLUP model must be further outlined.

Genetic architecture of fresh-market tomato yield.

BACKGROUND: The fresh-market tomato (Solanum lycopersicum) is bred for direct consumption and is selected for a high yield of large fruits. To understand the genetic variations (distinct types of DNA sequence polymorphism) that influence the yield, we collected the phenotypic variations in the yields of total fruit, extra-large-sized fruit, small-sized fruit, or red-colored fruit from 68 core inbred contemporary U.S. fresh-market tomatoes for three consecutive years and the genomic information in 8,289,741 single nucleotide polymorphism (SNP) positions from the whole-genome resequencing of these tomatoes. RESULTS: Genome-wide association (GWA) mapping using the SNP data with or without SNP filtering steps using the regularization methods, validated with quantitative trait loci (QTL) linkage mapping, identified 18 significant association signals for traits evaluated. Among them, 10 of which were not located within genomic regions previously identified as being associated with fruit size/shape. When mapping-driven association signals [558 SNPs associated with 28 yield (component) traits] were used to calculate genomic estimated breeding values (GEBVs) of evaluated traits, the prediction accuracies of the extra-large-sized fruit and small-sized fruit yields were higher than those of the total and red-colored fruit yields, as we tested the generated breeding values in inbred tomatoes and F2 populations. Improved accuracy in GEBV calculation of evaluated traits was achieved by using 364 SNPs identified using the regularization methods. CONCLUSIONS: Together, these results provide an understanding of the genetic variations underlying the heritable phenotypic variability in yield in contemporary tomato breeding and the information necessary for improving such economically important and complex quantitative trait through breeding.

A mutant allele of the flowering promoting factor 1 gene at the tomato BRACHYTIC locus reduces plant height with high quality fruit.

Reduced plant height due to shortened stems is beneficial for improving crop yield potential, better resilience to biotic/abiotic stresses, and rapid crop producer adoption of the agronomic and management practices. Breeding tomato plants with a reduced height, however, poses a particular challenge because this trait is often associated with a significant fruit size (weight) reduction. The tomato BRACHYTIC (BR) locus controls plant height. Genetic mapping and genome assembly revealed three flowering promoting factor 1 (FPF1) genes located within the BR mapping interval, and a complete coding sequence deletion of the telomere proximal FPF1 (Solyc01g066980) was found in the br allele but not in BR. The knock-out of Solyc01g066980 in BR large-fruited fresh-market tomato reduced the height and fruit yield, but the ability to produce large size fruits was retained. However, concurrent yield evaluation of a pair of sister lines with or without the br allele revealed that artificial selection contributes to commercially acceptable yield potential in br tomatoes. A network analysis of gene-expression patterns across genotypes, tissues, and the gibberellic acid (GA) treatment revealed that member(s) of the FPF1 family may play a role in the suppression of the GA biosynthesis in roots and provided a framework for identifying the responsible molecular signaling pathways in br-mediated phenotypic changes. Lastly, mutations of br homologs also resulted in reduced height. These results shed light on the genetic and physiological mechanisms by which the br allele alters tomato architecture.

postQTL: a QTL mapping R workflow to improve the accuracy of true positive loci identification.

OBJECTIVE: The determination of the location of quantitative trait loci (QTL) (i.e., QTL mapping) is essential for identifying new genes. Various statistical methods are being incorporated into different QTL mapping functions. However, statistical errors and limitations may often occur in a QTL mapping, implying the risk of false positive errors and/or failing to detect a true positive QTL effect. We simulated the power to detect four simulated QTL in tomato using cim() and stepwiseqtl(), widely adopted QTL mapping functions, and QTL.gCIMapping(), a derivative of the composite interval mapping method. While there is general agreement that those three functions identified simulated QTL, missing or false positive QTL were observed, which were prevalent when more realistic data (such as smaller population size) were provided. RESULTS: To address this issue, we developed postQTL, a QTL mapping R workflow that incorporates (i) both cim() and stepwiseqtl(), (ii) widely used R packages developed for model selection, and (iii) automation to increase the accuracy, efficiency, and accessibility of QTL mapping. QTL mapping experiments on tomato F2 populations in which QTL effects were simulated or calculated showed advantages of postQTL in QTL detection.

Breeding for Resistance to Fusarium Wilt of Tomato: A Review.

For over a century, breeders have worked to develop tomato (Solanum lycopersicum) cultivars with resistance to Fusarium wilt (Fol) caused by the soilborne fungus Fusarium oxysporum f. sp. lycopersici. Host resistance is the most effective strategy for the management of this disease. For each of the three Fol races, resistance has been introgressed from wild tomato species, predominately in the form of R genes. The I, I-2, I-3, and I-7 R genes have each been identified, as well as the corresponding Avr effectors in the fungus with the exception of Avr7. The mechanisms by which the R gene protein products recognize these effectors, however, has not been elucidated. Extensive genetic mapping, gene cloning, and genome sequencing efforts support the development of tightly-linked molecular markers, which greatly expedite tomato breeding and the development of elite, Fol resistant cultivars. These resources also provide important tools for pyramiding resistance genes and should support the durability of host resistance.

Identification and characterization of GLOBE, a major gene controlling fruit shape and impacting fruit size and marketability in tomato.

Within large-fruited germplasm, fruit size is influenced by flat and globe shapes. Whereas flat fruits are smaller and retain better marketability, globe fruits are larger and more prone to cuticle disorders. Commercial hybrids are often developed from crosses between flat and globe shaped parents because flat shape is thought to be dominant and fruit size intermediate. The objectives of this study were to determine the genetic basis of flat/globe fruit shape in large-fruited fresh-market tomato germplasm and to characterize its effects on several fruit traits. Twenty-three advanced single plant selections from the Fla. 8000 × Fla. 8111B cross were selectively genotyped using a genome-wide SNP array, and inclusive composite interval mapping identified a single locus on the upper arm of chromosome 12 associated with shape, which we termed globe. A 238-plant F2 population and 69 recombinant inbred lines for this region from the same parents delimited globe to approximately 392-kilobases. A germplasm survey representing materials from multiple breeding programs demonstrated that the locus explains the flat/globe shape broadly. A single base insertion in an exon of Solyc12g006860, a gene annotated as a brassinosteroid hydroxylase, segregated completely with shape in all populations tested. CRISPR/Cas9 knock out plants confirmed this gene as underlying the globe locus. In silico analysis of the mutant allele of GLOBE among 595 wild and domesticated accessions suggested that the allele arose very late in the domestication process. Fruit measurements in three genetic backgrounds evidenced that globe impacts fruit size and several fruit shape attributes, pedicel length/width, and susceptibility of fruit to weather check. The mutant allele of GLOBE appears mostly recessive for all traits except fruit size where it acts additively.

Characterization and elimination of linkage-drag associated with Fusarium wilt race 3 resistance genes.

KEY MESSAGE: Reducing the size of the I-3 introgression resulted in eliminating linkage-drag contributing to increased sensitivity to bacterial spot and reduced fruit size. The I-7 gene was determined to have no effect on bacterial spot or fruit size, and germplasm is now available with both the reduced I-3 introgression and I-7. Tomato (Solanum lycopersicum) production is increasingly threatened by Fusarium wilt race 3 (Fol3) caused by the soilborne fungus, Fusarium oxysporum f. sp. lycopersici. Although host resistance based on the I-3 gene is the most effective management strategy, I-3 is associated with detrimental traits including reduced fruit size and increased bacterial spot sensitivity. Previous research demonstrated the association with bacterial spot is not due to the I-3 gene, itself, and we hypothesize that reducing the size of the I-3 introgression will remedy this association. Cultivars with I-7, an additional Fol3 resistance gene, are available but are not widely used commercially, and it is unclear whether I-7 also has negative horticultural associations. To characterize the effect of I-3 on fruit size, segregating populations were developed and evaluated, revealing that the large I-3 introgression decreased fruit size by approximately 21%. We reduced the I-3 introgression from 5 to 140 kb through successive recombinant screening and crossing efforts. The reduced I-3 introgression and I-7 were then separately backcrossed into elite Florida breeding lines and evaluated for effects on bacterial spot sensitivity and fruit size across multiple seasons. The reduced I-3 introgression resulted in significantly less bacterial spot and larger fruit size than the large introgression, and it had no effect on these horticultural characteristics compared with Fol3 susceptibility. I-7 was also found to have no effect on these traits compared to Fol3 susceptibility. Together, these efforts support the development of superior Fol3-resistant cultivars and more durable resistance against this pathogen.

Aldoxime Metabolism Is Linked to Phenylpropanoid Production in Camelina sativa.

Plants produce diverse secondary metabolites. Although each metabolite is made through its respective biosynthetic pathway, plants coordinate multiple biosynthetic pathways simultaneously. One example is an interaction between glucosinolate and phenylpropanoid pathways in Arabidopsis thaliana. Glucosinolates are defense compounds made primarily from methionine and tryptophan, while phenylpropanoids are made from phenylalanine. Recent studies have shown that the accumulation of glucosinolate intermediate such as indole-3-acetaldoxime (IAOx) or its derivatives represses phenylpropanoid production via the degradation of phenylalanine ammonia lyase (PAL) functioning at the entry point of the phenylpropanoid pathway. Given that IAOx is a precursor of other bioactive compounds other than glucosinolates and that the phenylpropanoid pathway is present in most plants, we hypothesized that this interaction is relevant to other species. Camelina sativa is an oil crop and produces camalexin from IAOx. We enhanced IAOx production in Camelina by overexpressing Arabidopsis CYP79B2 which encodes an IAOx-producing enzyme. The overexpression of AtCYP79B2 results in increased auxin content and its associated morphological phenotypes in Camelina but indole glucosinolates were not detected in Camelina wild type as well as the overexpression lines. However, phenylpropanoid contents were reduced in AtCYP79B2 overexpression lines suggesting a link between aldoxime metabolism and phenylpropanoid production. Interestingly, the expression of PALs was not affected in the overexpression lines although PAL activity was reduced. To test if PAL degradation is involved in the crosstalk, we identified F-box genes functioning in PAL degradation through a phylogenetic study. A total of 459 transcript models encoding kelch-motifs were identified from the Camelina sativa database. Among them, the expression of CsKFBs involved in PAL degradation is up-regulated in the transgenic lines. Our results suggest a link between aldoxime metabolism and phenylpropanoid production in Camelina and that the molecular mechanism behind the crosstalk is conserved in Arabidopsis and Camelina.

Integration of lodging resistance QTL in soybean.

Poor lodging resistance could limit increases in soybean yield. Previously, a considerable number of observations of quantitative trait loci (QTL) for lodging resistance have been reported by independent studies. The integration of these QTL into a consensus map will provide further evidence of their usefulness in soybean improvement. To improve informative QTL in soybean, a mapping population from a cross between the Harosoy and Clark cultivars, which inherit major U.S. soybean genetic backgrounds, was used along with previous mapping populations to identify QTL for lodging resistance. Together with 78 QTL for lodging collected from eighteen independent studies, a total of 88 QTL were projected onto the soybean consensus map. A total of 16 significant QTL clusters were observed; fourteen of them were confirmed in either two or more mapping populations or a single population subjected to different environmental conditions. Four QTL (one on chromosome 7 and three on 10) were newly identified in the present study. Further, meta-analysis was used to integrate QTL across different studies, resulting in two significant meta-QTL each on chromosomes 6 and 19. Our results provide deeper knowledge of valuable lodging resistance QTL in soybean, and these QTL could be used to increase lodging resistance.

Impact of Rhg1 copy number, type, and interaction with Rhg4 on resistance to Heterodera glycines in soybean.

KEY MESSAGE: Evaluations of soybean populations showed that both Rhg1 copy number and type were important in determining soybean cyst nematode resistance with higher copy number within Rhg1 type conferring greater resistance. Rhg1 and Rhg4 are important loci conferring resistance to soybean cyst nematode (SCN; Heterodera glycines). Alleles at Rhg1 have been shown to vary for copy number and type and the importance of this variation in conferring resistance is not well defined. The repeat number ranges from one to 10 and there are three variant repeat sequence types [plant introduction (PI) 88788-'Fayette' type (F), 'Peking' type (P) and Williams 82 type (W)] across diverse soybean germplasm. We developed populations segregating for Rhg1 copy number and type and Rhg4 allele type to investigate the effect of these factors and their interaction on SCN resistance. F2 plants from each cross were evaluated for the segregation of Rhg1 and Rhg4 alleles and for SCN reproduction after infesting plants with HG type 2.5.7 and HG type 7 populations. Within repeat types, an increase in repeat number was associated with greater resistance. The P type Rhg1 showed an advantage over F + W type for SCN population HG type 2.5.7 but this was not observed for SCN HG type 7. While plants with P type Rhg1 required Rhg4 to achieve full resistance, Rhg4 did not increase resistance in the background of F + W type Rhg1 repeat. This study demonstrates the importance of both Rhg1 copy number and type in determining resistance and can assist soybean breeders in determining what alleles would best fit their breeding goals.

An efficient method for measuring copy number variation applied to improvement of nematode resistance in soybean.

Copy number variation (CNV) is implicated in important traits in multiple crop plants, but can be challenging to genotype using conventional methods. The Rhg1 locus of soybean, which confers resistance to soybean cyst nematode (SCN), is a CNV of multiple 31.2-kb genomic units each containing four genes. Reliable, high-throughput methods to quantify Rhg1 and other CNVs for selective breeding were developed. The CNV genotyping assay described here uses a homeologous gene copy within the paleopolyploid soybean genome to provide the internal control for a single-tube TaqMan copy number assay. Using this assay, CNV in breeding populations can be tracked with high precision. We also show that extensive CNV exists within Fayette, a released, inbred SCN-resistant soybean cultivar with a high copy number at Rhg1 derived from a single donor parent. Copy number at Rhg1 is therefore unstable within a released variety over a relatively small number of generations. Using this assay to select for individuals with altered copy number, plants were obtained with both increased copy number and increased SCN resistance relative to control plants. Thus, CNV genotyping technologies can be used as a new type of marker-assisted selection to select for desirable traits in breeding populations, and to control for undesirable variation within cultivars.

Identification of genome-specific transcripts in wheat-rye translocation lines.

Studying gene expression in wheat-rye translocation lines is complicated due to the presence of homeologs in hexaploid wheat and high levels of synteny between wheat and rye genomes (Naranjo and Fernandez-Rueda, 1991 [1]; Devos et al., 1995 [2]; Lee et al., 2010 [3]; Lee et al., 2013 [4]). To overcome limitations of current gene expression studies on wheat-rye translocation lines and identify genome-specific transcripts, we developed a custom Roche NimbleGen Gene Expression microarray that contains probes derived from the sequence of hexaploid wheat, diploid rye and diploid progenitors of hexaploid wheat genome (Lee et al., 2014). Using the array developed, we identified genome-specific transcripts in a wheat-rye translocation line (Lee et al., 2014). Expression data are deposited in the NCBI Gene Expression Omnibus (GEO) under accession number GSE58678. Here we report the details of the methods used in the array workflow and data analysis.

Evolution and selection of Rhg1, a copy-number variant nematode-resistance locus.

The soybean cyst nematode (SCN) resistance locus Rhg1 is a tandem repeat of a 31.2 kb unit of the soybean genome. Each 31.2-kb unit contains four genes. One allele of Rhg1, Rhg1-b, is responsible for protecting most US soybean production from SCN. Whole-genome sequencing was performed, and PCR assays were developed to investigate allelic variation in sequence and copy number of the Rhg1 locus across a population of soybean germplasm accessions. Four distinct sequences of the 31.2-kb repeat unit were identified, and some Rhg1 alleles carry up to three different types of repeat unit. The total number of copies of the repeat varies from 1 to 10 per haploid genome. Both copy number and sequence of the repeat correlate with the resistance phenotype, and the Rhg1 locus shows strong signatures of selection. Significant linkage disequilibrium in the genome outside the boundaries of the repeat allowed the Rhg1 genotype to be inferred using high-density single nucleotide polymorphism genotyping of 15 996 accessions. Over 860 germplasm accessions were found likely to possess Rhg1 alleles. The regions surrounding the repeat show indications of non-neutral evolution and high genetic variability in populations from different geographic locations, but without evidence of fixation of the resistant genotype. A compelling explanation of these results is that balancing selection is in operation at Rhg1.

Expression analysis of individual homoeologous wheat genome- and rye genome-specific transcripts in a 2BS.2RL wheat-rye translocation.

Wheat-rye translocations are widely used in wheat breeding to confer resistance against abiotic and biotic stress. Studying gene expression in wheat-rye translocations is complicated due to the presence of homoeologous genes in hexaploid wheat and high levels of synteny between wheat and rye chromatin. To distinguish transcripts expressed from each of the three wheat genomes and those from rye chromatin, genomic probes generated from diploid progenitors of wheat and rye were synthesized on a custom array. A total of 407 transcripts showed homoeologous genome ('A', 'B' or 'D' genome)- or rye genome ('R')-specific differential expression, based on unequal values of probe hybridization. In a 2BS.2RL wheat-rye translocation, thirteen of the 407 transcripts showed preferential expressions from rye chromatin. As well as quantifying variation in homoeologous transcript in wheat-rye translocations, this study also provides a potential aid to examine the contribution of the subgenomes to complex allohexapolyploids.

Copy number variation of multiple genes at Rhg1 mediates nematode resistance in soybean.

The rhg1-b allele of soybean is widely used for resistance against soybean cyst nematode (SCN), the most economically damaging pathogen of soybeans in the United States. Gene silencing showed that genes in a 31-kilobase segment at rhg1-b, encoding an amino acid transporter, an α-SNAP protein, and a WI12 (wound-inducible domain) protein, each contribute to resistance. There is one copy of the 31-kilobase segment per haploid genome in susceptible varieties, but 10 tandem copies are present in an rhg1-b haplotype. Overexpression of the individual genes in roots was ineffective, but overexpression of the genes together conferred enhanced SCN resistance. Hence, SCN resistance mediated by the soybean quantitative trait locus Rhg1 is conferred by copy number variation that increases the expression of a set of dissimilar genes in a repeated multigene segment.

Comparative physical mapping between wheat chromosome arm 2BL and rice chromosome 4.

Physical maps of chromosomes provide a framework for organizing and integrating diverse genetic information. DNA microarrays are a valuable technique for physical mapping and can also be used to facilitate the discovery of single feature polymorphisms (SFPs). Wheat chromosome arm 2BL was physically mapped using a Wheat Genome Array onto near-isogenic lines (NILs) with the aid of wheat-rice synteny and mapped wheat EST information. Using high variance probe set (HVP) analysis, 314 HVPs constituting genes present on 2BL were identified. The 314 HVPs were grouped into 3 categories: HVPs that match only rice chromosome 4 (298 HVPs), those that match only wheat ESTs mapped on 2BL (1), and those that match both rice chromosome 4 and wheat ESTs mapped on 2BL (15). All HVPs were converted into gene sets, which represented either unique rice gene models or mapped wheat ESTs that matched identified HVPs. Comparative physical maps were constructed for 16 wheat gene sets and 271 rice gene sets. Of the 271 rice gene sets, 257 were mapped to the 18-35 Mb regions on rice chromosome 4. Based on HVP analysis and sequence similarity between the gene models in the rice chromosomes and mapped wheat ESTs, the outermost rice gene model that limits the translocation breakpoint to orthologous regions was identified.

Functional characterization of pectin methylesterase inhibitor (PMEI) in wheat.

Pectin, one of the main components of plant cell wall, is deesterified by the pectin methylesterase (PME). PME activity is regulated by inhibitor proteins known as the pectin methylesterase inhibitor (PMEI), which plays a key role in wounding, osmotic stress, senescence and seed development. However, the role of PMEI in many plant species still remains to be elucidated, especially in wheat. To facilitate the expression analysis of the TaPMEI gene, RT-PCR was performed using leaf, stem and root tissues that were treated with exogeneous application of phytohormones and abiotic stresses. High transcription was detected in salicylic acid (SA) and hydrogen peroxide treatments. To elucidate the subcellular localization of the TaPMEI protein, the TaPMEI:GFP fusion construct was transformed into onion epidermal cells by particle bombardment. The fluorescence signal was exclusively detected in the cell wall. Using an enzyme assay, we confirmed that PME was completely inhibited by TaPMEI. These results indicated that TaPMEI was involved in inhibition of pectin methylesterification and may play a role in the plant defense mechanism via cell wall fortification.

Development and functional assessment of EST-derived 2RL-specific markers for 2BS.2RL translocations.

ESTs-derived markers are useful for comparative genomic analysis and can also serve as phenotype-linked functional markers. Here, we report the development of EST-derived 2RL-specific markers and the evaluation of the possibility of functional assessment of markers tagging 2RL, which carries Hessian fly resistance genes (loci). To identify transcripts specific to 2RL, unigene sequences in combination with wheat progenitor genomes were used. Total 275 contigs mapped to the long arms of homoeologous group 2 chromosomes were downloaded. To obtain a cluster corresponding to each of the wheat 275 contigs, unigene sequences of wheat, rice, barley, and rye were pooled for cross-species clusters. Out of 275 clusters examined, it was possible to design 112 cross-species primer pairs for genome-specific amplifications. Out of 112 cross-species primer pairs, 45 primer pairs (40%) produced amplicons from at least one species (three wheat progenitors or rye). Among the 45 contigs, 73% were associated with one of known functions and 82% of the contigs associated with known functions were also associated with one of the GO categories. On the basis of the oligonucleotide sequence alignment of each of 45 genome-specific amplifications, 21 amplifications (47%) were suitable for designing RR genome-specific primers, which are specific to translocated rye chromatin 2RL. Six primer pairs (13%) successfully produced amplicons in the 2BS.2RL translocation lines and not in the non-2RLs. Functional assessment of one of the 2RL-specific markers, NSFT03P2_Contig4445, was performed on Hessian fly infested NILs. Under Hessian fly infestation, significantly high expression of a gene tagged by a 2RL-specific marker (NSFT03P2_Contig4445) was observed 1 day after infestation. EST-derived 2RL-specific marker development from this study provides a basis for the development of ESTs-derived markers for detecting wheat-rye translocations. In addition, these markers could be employed in elucidating functional analysis of genes on 2RL.

Evolution of non-specific lipid transfer protein (nsLTP) genes in the Poaceae family: their duplication and diversity.

Previously, the genes encoding non-specific lipid transfer proteins (nsLTPs) of the Poaceae family appear to evidence different genomic distribution and somewhat different shares of EST clones, which is suggestive of independent duplication(s) followed by functional diversity. To further evaluate the evolutionary fate of the Poaceae nsLTP genes, we have identified Ka/Ks values, conserved, mutated or lost cis-regulatory elements, responses to several elicitors, genome-wide expression profiles, and nsLTP gene-coexpression networks of both (or either) wheat and rice. The Ka/Ks values within each group and between groups appeared to be similar, but not identical, in both species. The conserved cis-regulatory elements, e.g. the RY repeat (CATGCA) element related to ABA regulation in group A, might be reflected in some degree of long-term conservation in transcriptional regulation post-dating speciation. In group A, wheat nsLTP genes, with the exception of TaLTP4, evidenced responses similar to those of plant elicitors; however, the rice nsLTP genes evidenced differences in expression profiles, even though the genes of both species have undergone purifying selection, thereby suggesting their independent functional diversity. The expression profiles of rice nsLTP genes with a microarray dataset of 155 gene expression omnibus sample (GSM) plates suggest that subfunctionalization is not the sole mechanism inherent to the evolutionary history of nsLTP genes but may, rather, function in concert with other mechanism(s). As inferred by the nsLTP gene-coexpression networks, the functional diversity of nsLTP genes appears not to be randomized, but rather to be specialized in the direction of specific biological processes over evolutionary time.