After hybridization, microarrays were
DNA Manipulations and Sequencing
All plasmid and DNA fragment manipulations were done according to standard protocols [95]. DNA fragment purification was done using the Geneclean Kit (MP Biomedicals, Irvine, CA, USA) according to the protocol of the manufacturer. Restriction endonucleases were from Promega (Madison, WI, USA) or Fermentas, Inc. (Hanover, MD, USA) and digestions were done according to the protocols of the manufacturers. DNA sequencing to confirm cDNA clone identities was done by the DNA Laboratory at Arizona State University using a 3730 Capillary Array Sequencer (Applied Biosystems, Foster City, CA, USA).
RNA Isolation and Blot Analyses
Total RNA was isolated from plants (typically 0.5? g of rosette tissue) using TRIzol according to the protocol of the manufacturer (Invitrogen, Carlsbad, CA, USA). RNA was separated by agaroseformaldehyde gel electrophoresis [95]. Fluorescence from ethidium bromide stained ribosomal RNA was recorded using a UVP Gel Analysis System and the LabWorks Imaging and Analysis program (UVP, Inc., Upland, CA, USA) according to the instructions of the manufacturer. RNA was then blotted to nylon membrane according to standard protocols [95]. DNA probes were labeled with Digoxigenin HighPrime Labeling and Detection Kit and used for hybridizations according to the protocols of the manufacturer (Roche Diagnostics Corporation, Indianapolis, IN, USA). For detection of AtAOX1a transcript, a labeled fragment from an AtAOX1a cDNA clone [96] was used. For other genes, labeled inserts from the following cDNA clones were used: NDB2, U12861; mtPORIN, 124F4; AtHSP70-9, U15562; mtGST, U13302;washed successively for 5 min in 26 SSC and 0.5% (w/v) SDS at 55uC, followed by 0.56 SSC and then 0.056 SSC, both at about 25uC. The microarray was dried by centrifugation at 1000 g for 3 min and scanned using a G2565BA Microarray Scanner System (Agilent, Palo Alto, CA, USA). Quantitative data were extracted from image files of microarrays and further analyzed using GenePix Pro 6.0 software (Molecular Devices, Sunneyvale, CA, USA) and the appropriate GenePix Array List file from the Oligonucleotide Microarray Facility.
Cluster Analysis
Expression of 1316 genes was statistically significantly (q#0.05) either up-regulated or down-regulated following AA treatment. By the same criteria, 606 genes were deemed MFA-responsive. These genes were used in a cluster analysis of transcript levels of the same genes during biotic and abiotic stress treatments. From public databases (Resource S5), 46 experiments were selected that used leaf or seedling treatment, and Affymetrix ATH1 or comparable arrays containing probes for loci covering nearly the entire genome. The treatment and control data were downloaded. The log2-transformed expression ratios were derived for all available replicates then averaged. From these data sets, the AA- or MFA-responsive genes that are represented in all array types were used in the respective cluster analyses. When AA-responsive genes were the query set, MFA treatment data were included in the analysis, and vice versa. The finalized data sets from these experiments were compared using Cluster [103]. Hierarchical Clustering/Average Linkage Clustering was used to generate array clusters based on Pearson correlation coefficients derived by Cluster. The clusters (Fig. 7) and heat maps (Resources S7 to S15) were visualized using Java TreeView (http://sourceforge.net/projects/ jtreeview/). The Pearson correlation coefficients derived by Cluster between experiments are listed in Resource S6.
Statistical Analysis of Microarray Data
Three biological replicate experiments (each including treatment and control) with dye-swapping (6 microarrays) and a duplicate slide of one dye condition for each of two of the biological replicates (for 8 total microarrays) were performed and used in the analyses for AA-treated versus control samples. Three biological replicate experiments (each including treatment and control) with dye-swapping (6 microarrays) were performed and used in analyses for MFA-treated versus control samples. For analyses of the AA-treated or MFA-treated versus control samples, differences in transcript levels were determined by a two-step mixed model analysis of variance (ANOVA) procedure [97], [98]. Data from GenePix software files for each array were log2transformed. In the first ANOVA step of the mixed model, these values were scaled and normalized. In the second ANOVA step, these data were used to determine statistically significant differences in gene transcript levels between treated and control samples. The ANOVA steps were performed in SAS version 9 (SAS Institute, Cary, NC, USA). A false discovery rate parameter, the q-value, was calculated using the q-value package provided by Bioconductor to correct for the multi-testing problem and was used to identify significantly differentially expressed genes [99]. A q-value cutoff of 0.05 was chosen to increase the likelihood of identifying informative changes in transcript levels but keep the risk of finding false positives low. Fold changes were obtained from the differences between the least square means obtained in the second ANOVA step of the analysis. Microarray data are compliant with the Minimum Information About a Microarray Experiment guidelines and the experimental data were deposited into the National Center for Biotechnology Information’s Gene Expression Omnibus with super-series accession number GSE29269 and subseries numbers GSE29204 and GSE29268.
