Data S1. Correlations between the Agilent array results and the Affymetrix platform results. Comparison of quantile normalized log2 intensities from Affymetrix probe sets versus log2 intensities of matching Agilent probes (intensities normalized to median of the array and averaged, see main text for details) for 3 tissue types in a common background: a) msca1 Spikelets, b) Msca1 Tissue A1.0, and c) Msca1 Tissue A1.5. Experimental procedure Similar to the Ma et al. (2006) study, a separate and independent array platform, the GeneChip� (®) Maize Genome Array (Affymetrix, Santa Clara, CA, USA), was utilized to cross-validate a subset of the results obtained from the Agilent platform and detect any platform-specific biases. The GeneChip� Array is comprised of 17,555 probe sets, for approximately 14,850 transcripts representing 13,339 genes specific/maize.affx". In contrast to the Agilent arrays, each transcript is represented by fifteen 25-mer oligonucleotide probes. The probe set was compiled using the September 29, 2004 NCBI GenBank assembly and July 23, 2004 Zea mays UniGene Build 42. Twelve GeneChip� Arrays were used to look at three different anther tissue types: the spikelet stage of msca1 mutants and Msca1 fertile siblings at the A1.0 and A1.5 stages. The three RNA samples analyzed on the Agilent platform were used, along with one additonal RNA sample, for a total of four biological replicates on the GeneChip� Array. The One-Cycle Target Labeling protocol, included in the Eukaryotic Sample and Array Processing section (701025 Rev. 6) of the GeneChip� Expression Analysis: Technical Manual (701021 Rev. 5), was followed, using 2 μg of starting total RNA. Intensity data from the resulting CEL files were imported with the Affymetrix R package, followed by quantile normalization and generation of probe set expression values with the justPlier R package (an implementation of the Affymetrix PLIER algorithm). A total of 2,636 probe sets on the Affymetrix platform overlapped in the same orientation with positively hybridizing 60-mers on the Agilent platform (BLAST criteria: e-value < 1e-10, match length ≥ 45, percent identity ≥ 0.9). Affymetrix probe sets with a normalized intensity below 4 (approximately the lowest 10%) were then removed. After normalization to the array median, the Affymetrix log2 intensity values were compared to the log of the average Agilent intensity values with correlation coefficients ranging from 0.47 to 0.49 (Supplementary Figure S1). These coefficents are lower than those calculated in the Ma et al., 2006 study for probes designed to the same region of the gene transcript (~0.76) but the use of different protocols, scanner, and analysis algorithms required by the array platform likely account for much of this difference. Data S2. Differentially expressed transcripts in the 3 ms mutants at the 4 stages. Data S3. Average linkage clustering trees for both fertile and ms mutant tissues. The tree is based on correlation measures (uncentered). Distances are calculated from log-2 values of the absolute intensities normalized over the median value for 3,616 probes that displayed above-median hybridizations in all 3 fertile samples. The tree is obtained with the R package pvclust ("/) and multiscale bootstrapping resampling statistics (%; Approximately Unbiased p-values) are given at the edges. For tissue stage information see the legend to Figure 1. Data S4. Test for mis-expression of leaf genes in mutant anthers. Only differentially regulated genes (mutants compared to fertiles; up means up-regulated, and down, down-regulated) in the 3 anther stages (A1.0, A1.5, and A2.0) that are lowly expressed (relative expression value, i.e. log2 values of normalized signal intensities over median < 1) in fertile anthers were included. Then the relative expression levels of these genes were extracted from a dataset on juvenile leaves (Ma et al., 2006), and then classified as non-detectable (relative expression value is < -1), low expression (between -1 and 1), mid expression (between 1 and 3), and high expression (above 3) in leaf. The complete gene sets for msca1 and ms23 (at A1.5 stage) are provided as Supplementary Data S5. Experimental procedure In order to minimize noise that might be introduced by genes with high expression in both anther and leaf, we elected to analyze genes identified in a previous study with Version 1 arrays as having low or non-detectable expression in the A1.0 and A1.5 anther stages across three backgrounds (Ma et al., 2006). From this list, we identified genes differentially expressed by each of the mutants at the A1.0 through A2.0 anther stages in the present study. This filtering produced, in most cases, more than 30 transcripts of up or down regulated genes (relative to fertile siblings) at each developmental stage for each mutant. We then determined the levels at which these differentially expressed genes were expressed in juvenile leaves (log2 values of signals from Version 1 arrays, normalized over median, averaged across the 3 backgrounds) (Ma et al., 2006). The ~30 transcripts for each stage and each ms mutant were divided into four expression classes: non-detectable in leaf, low in leaf, mid-expression in leaf, and high expression in leaf. Data S5. Differentially expressed transcripts in the 3 ms mutants that are also highly expressed in leaf. Data S6. Lists of genes in the 4 clusters analyzed in Figure 5. Data S7. Putative maize LRR-RLK genes. (a) Unrooted phylogenetic tree of two maize putative LRR-RLKs, together with MSP1, OsBRI1, and an EXS-like gene from rice, EXS and BRI1 from Arabidopsis, and tBRI1/SR160 from L. peruvianum, was constructed similarly to Figure 7a. Expression profiles of the two maize transcripts are given in (b) ZmMSP1, and (c) ZmBRI1. Note that the big deviations in ZmMSP1 expression levels at the two ms23 stages (S and A1.5) were most likely from systematic errors. Data S8. Maize orthologs of rice anther-expressed genes. Data S9. Conservation of gene expression regulation between maize and rice anther-specific orthologs. For maize transcripts, the expression values are log2 of the absolute intensities normalized against the median value in the 3 anther stages and averaged over the 3 lines, in (A) fertile siblings and (B) ms mutants. For rice transcripts, the expression values are log2 of the averaged signal intensities from cDNA arrays normalized against the median value of the stage I anther (Endo et al., 2004). Data points not included in the linear regression analysis are indicated with open symbols. See Supplementary Data S8 for transcript identities. Data S10. Relative expression levels (Ex) of 142 meiosis-related transcripts on the array. The three parallel dashed lines represent the boundaries y = x-0.5, y = x, and y = x+0.5, respectively. (A) msca1; (B) mac1; (C) ms23. (D) A histogram of the differences in expression levels at the 4 stages in the mutant and fertile plants of msca1, i.e. Ex(msca1) – Ex(fertile).

TPJ_3074_sm_DataS1.pdf497KSupporting info item
TPJ_3074_sm_DataS2.xls414KSupporting info item
TPJ_3074_sm_DataS3.pdf3KSupporting info item
TPJ_3074_sm_DataS4.pdf6KSupporting info item
TPJ_3074_sm_DataS5.xls64KSupporting info item
TPJ_3074_sm_DataS6.pdf34KSupporting info item
TPJ_3074_sm_DataS7.pdf10KSupporting info item
TPJ_3074_sm_DataS8.xls27KSupporting info item
TPJ_3074_sm_DataS9.pdf7KSupporting info item
TPJ_3074_sm_DataS10.pdf67KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.