Steven K. St. Martin

Global Transcription Profiling Reveals Multiple Sugar Signal Transduction Mechanisms in Arabidopsis

Complex and interconnected signaling networks allow organisms to control cell division, growth, differentiation, or programmed cell death in response to metabolic and environmental cues. In plants, it is known that sugar and nitrogen are critical nutrient signals; however, our understanding of the molecular mechanisms underlying nutrient signal transduction is very limited. To begin unraveling complex sugar signaling networks in plants, DNA microarray analysis was used to determine the effects of glucose and inorganic nitrogen source on gene expression on a global scale in Arabidopsis thaliana. In whole seedling tissue, glucose is a more potent signal in regulating transcription than inorganic nitrogen. In fact, other than genes associated with nitrate assimilation, glucose had a greater effect in regulating nitrogen metabolic genes than nitrogen itself. Glucose also regulated a broader range of genes, including genes associated with carbohydrate metabolism, signal transduction, and metabolite transport. In addition, a large number of stress responsive genes were also induced by glucose, indicating a role of sugar in environmental responses. Cluster analysis revealed significant interaction between glucose and nitrogen in regulating gene expression because glucose can modulate the effects of nitrogen and vise versa. Intriguingly, cycloheximide treatment appeared to disrupt glucose induction more than glucose repression, suggesting that de novo protein synthesis is an intermediary event required before most glucose induction can occur. Cross talk between sugar and ethylene signaling may take place on the transcriptional level because several ethylene biosynthetic and signal transduction genes are repressed by glucose, and the repression is largely unaffected by cycloheximide. Collectively, our global expression data strongly support the idea that glucose and inorganic nitrogen act as both metabolites and signaling molecules.

Analysis of genes underlying soybean quantitative trait loci conferring partial resistance to Phytophthora sojae.

Few quantitative trait loci (QTL) have been mapped for the expression of partial resistance to Phytophthora sojae in soybean and very little is known about the molecular mechanisms that contribute to this trait. Therefore, the objectives of this study were to identify additional QTL conferring resistance to P. sojae and to identify candidate genes that may contribute to this form of defense. QTL on chromosomes 12, 13, 14, 17, and 19, each explaining 4 to 7% of the phenotypic variation, were identified using 186 RILs from a cross of the partially resistant cultivar ‘Conrad’ and susceptible cultivar ‘Sloan’ through composite interval mapping. Microarray analysis identified genes with significant differences in transcript abundances between Conrad and Sloan, both constitutively and following inoculation. Of these genes, 55 mapped to the five QTL regions. Ten genes encoded proteins with unknown functions, while the others encode proteins related to defense or physiological traits. Seventeen genes within the genomic region that encompass the QTL were selected and their transcript abundance was confirmed by quantitative reverse transcription polymerase chain reaction (qRT‐PCR). These results suggest a complex QTL‐mediated resistance network. This study will contribute to soybean resistance breeding by providing additional QTL for marker‐assisted selection as well as a list of candidate genes which may be manipulated to confer resistance.

Infection and genotype remodel the entire soybean transcriptome

BackgroundHigh throughput methods, such as high density oligonucleotide microarray measurements of mRNA levels, are popular and critical to genome scale analysis and systems biology. However understanding the results of these analyses and in particular understanding the very wide range of levels of transcriptional changes observed is still a significant challenge. Many researchers still use an arbitrary cut off such as two-fold in order to identify changes that may be biologically significant. We have used a very large-scale microarray experiment involving 72 biological replicates to analyze the response of soybean plants to infection by the pathogen Phytophthora sojae and to analyze transcriptional modulation as a result of genotypic variation.ResultsWith the unprecedented level of statistical sensitivity provided by the high degree of replication, we show unambiguously that almost the entire plant genome (97 to 99% of all detectable genes) undergoes transcriptional modulation in response to infection and genetic variation. The majority of the transcriptional differences are less than two-fold in magnitude. We show that low amplitude modulation of gene expression (less than two-fold changes) is highly statistically significant and consistent across biological replicates, even for modulations of less than 20%. Our results are consistent through two different normalization methods and two different statistical analysis procedures.ConclusionOur findings demonstrate that the entire plant genome undergoes transcriptional modulation in response to infection and genetic variation. The pervasive low-magnitude remodeling of the transcriptome may be an integral component of physiological adaptation in soybean, and in all eukaryotes.

Dissection of two soybean QTL conferring partial resistance to Phytophthora sojae through sequence and gene expression analysis

BackgroundPhytophthora sojae is the primary pathogen of soybeans that are grown on poorly drained soils. Race-specific resistance to P. sojae in soybean is gene-for-gene, although in many areas of the US and worldwide there are populations that have adapted to the most commonly deployed resistance to P. sojae ( Rps) genes. Hence, this system has received increased attention towards identifying mechanisms and molecular markers associated with partial resistance to this pathogen. Several quantitative trait loci (QTL) have been identified in the soybean cultivar ‘Conrad’ that contributes to the expression of partial resistance to multiple P. sojae isolates.ResultsIn this study, two of the Conrad QTL on chromosome 19 were dissected through sequence and expression analysis of genes in both resistant (Conrad) and susceptible (‘Sloan’) genotypes. There were 1025 single nucleotide polymorphisms (SNPs) in 87 of 153 genes sequenced from Conrad and Sloan. There were 304 SNPs in 54 genes sequenced from Conrad compared to those from both Sloan and Williams 82, of which 11 genes had SNPs unique to Conrad. Eleven of 19 genes in these regions analyzed with qRT-PCR had significant differences in fold change of transcript abundance in response to infection with P. sojae in lines with QTL haplotype from the resistant parent compared to those with the susceptible parent haplotype. From these, 8 of the 11 genes had SNPs in the upstream, untranslated region, exon, intron, and/or downstream region. These 11 candidate genes encode proteins potentially involved in signal transduction, hormone-mediated pathways, plant cell structural modification, ubiquitination, and basal resistance.ConclusionsThese findings may indicate a complex defense network with multiple mechanisms underlying these two soybean QTL conferring resistance to P. sojae. SNP markers derived from these candidate genes can contribute to fine mapping of QTL and marker assisted breeding for resistance to P. sojae.

Load-balanced isosurfacing on multi-GPU clusters

Isosurface extraction is a common technique applied in scientific visualization. Increasing sizes of volumes over which isosurfacing is to be applied combined with increasingly hierarchical parallel architectures present challenges for efficiently distributing isosurfacing work loads. We propose a technique that, with a modest amount of preprocessing, efficiently distributes isosurfacing load to GPU compute resources within a cluster. Load uniformity is maximized over a set of user-defined isovalues, enabling improved scalability over naive, non-data-centric, work distribution approaches.

Functional Genomics and Bioinformatics of the Phytophthora sojae Soybean Interaction

Oomycete plant pathogens such as Phytophthora species and downy mildews cause destructive diseases in an enormous variety of crop plant species as well as forests and native ecosystems. These pathogens are most closely related to algae in the kingdom Stramenopiles, and hence have evolved plant pathogenicity independently of other plant pathogens such as fungi. We have used bioinformatic analysis of genome sequences and EST collections, together with functional genomics to identify plant and pathogen genes that may be key players in the interaction between the soybean pathogen Phytophthora sojae and its host. In P. sojae, we have identified many rapidly diversifying gene families that encode potential pathogenicity factors including protein toxins, and a class of proteins (avirulence or effector proteins) that appear to have the ability to penetrate plant cells. Transcriptomic analysis of quantitative or multigenic resistance against P. sojae in soybean has revealed that there are widespread adjustments in host gene expression in response to infection, and that some responses are unique to particular resistant cultivars. These observations lay the foundation for dissecting the interplay between pathogen and host genes during infection at a whole-genome level.

Transformations for volumetric range distribution queries

Volumetric datasets continue to grow in size, and there is continued demand for interactive analysis on these datasets. Because storage device throughputs are not increasing as quickly, interactive analysis workflows are becoming working set-constrained. In an ideal workflow, the working set complexity of the interactive analysis portion of the workflow should depend primarily on the size of the analysis result being produced, rather than on the size of the data being analyzed. Past works in online analytical processing and visualization have addressed this problem within application-specific contexts, but have not generalized their solutions to a wider variety of visualization applications. We propose a general framework for reducing the working set complexity of the interactive portion of visualization workflows that can be built on top of distribution range queries, as well as a technique within this framework able to support multiple visualization applications. Transformations are applied in the preprocessing phase of the workflow to enable fast, approximate volumetric distribution range queries with low working set complexity. Interactive application algorithms are then adapted to make use of these distribution range queries, enabling efficient interactive workflows on large-scale data. We show that the proposed technique enables these applications to be scaled primarily in terms of the application result dataset size, rather than the input data size, enabling increased interactivity and scalability.

Histogram spectra for multivariate time-varying volume LOD selection

Level of detail techniques are widely applied to minimize sampling error subject to working set size constraints. Typical large data sets being produced today have many variables sampled across time-varying volumes. Visualization of these multivariate volumes is commonly phrased in terms of conditional expressions such as “show variable A where variable B is between B1 and B2.” The bounds, B1 and B2, tend to be specified during the interactive portion of the workflow. Thus, to maximize quality over the salient interval, level of detail selection should also be interactive. We introduce the concept of histogram spectra to quickly and compactly quantify the statistical sensitivity of volumes to sampling. Salient interval volumes of one or more variables are used to select which parts of the histogram spectra are important. The level of detail selection problem, over a time-varying, multivariate, multiresolution volume, is then posed as an integer programming problem using the histogram spectra. We propose an efficient solution enabling interactive LOD selection on large, out-of-core volumes and show its efficacy on two real data sets from different problem domains.

Interactive transfer function design on large multiresolution volumes

Interactive transfer function design techniques seek to leverage user knowledge to facilitate the discovery of data salience. In this process, interactive volume rendering is typically a necessity. Interactive volume rendering of large-scale data on workstations is often accomplished through the use of level of detail techniques, prioritizing information deemed to be salient over information deemed to be unimportant. If salience is not known a priori, and interactive transfer function design techniques that depend on volume rendering are to be applied to large-scale data using level of detail selection, then there is a cyclic dependency. Techniques must be applied that can support simultaneous development of salience both for the transfer function design technique and the level of detail selection technique. Building on recent work in LOD selection, we propose an interactive transfer function design technique that enables incremental salience discovery to support simultaneous construction of transfer functions and LOD selections on large-scale data.

Efficient Rendering of Extrudable Curvilinear Volumes

We present a technique for memory-efficient and time-efficient volume rendering of curvilinear adaptive mesh refinement data defined within extrudable computational spaces. One of the main challenges in the ray casting of curvilinear volumes is that a linear viewing ray in physical space will typically correspond to a curved ray in computational space. The proposed method utilizes a specialized representation of curvilinear space that provides for the compact representation of parameters for transformations between computational space and physical space, without requiring extensive preprocessing. By simplifying the representation of computational space positions using an extrusion of a profile surface, the requisite transformations can be greatly simplified. Our implementation achieves interactive rates with minimal load time and memory overhead using commodity graphics hardware with real-world data.