Supporting Information (SI) Appendix Part 1. Impact of tissue preparation, LMD, and RNA amplification on array output. p. 2 Text S1: Detailed Experimental Design and Methods Figure S1A: Correlation analysis indicates tissue preparation has minimal impact on ATH1 array output. Figure S1B: Correlation analysis indicates RNA degradation does not significantly impact array output. Figure S1C: Correlation analysis indicates two-round amplification does not significantly impact array output. Figure S1D: Independent biological replicates of LMD samples are highly correlated. Figure S1E: Validation of LMD array expression by qPCR. Table S1: Tissue preparation-associated genes Part 2. Analysis of LMD and parallel whole leaf array data. p. 13 Table S2A: Known PM-impacted genes enriched in LMD dataset Table S2B: Dataset of LMD and whole leaf genes with PM-altered expression Table S2C: LMD PM MapMan Results and Bins Table S2D: ics1 vs. WT LMD PM MapMan Results and Bins Table S2E: Infection site-specific changes for redox and calcium categories Table S2F: cis-acting regulatory element motif analysis Part 3. Process network construction p. 149 3A. Photosynthesis 3B. Cold/dehydration response Part 4. Powdery mildew infection of WT and myb3r4 mutants p. 173 Text S4: Detailed Experimental Design and Methods (supplement to manuscript) Figure S4A. Uninfected 4 week old WT and myb3r4 plants Figure S4B. myb3r4 mutants exhibit reduced visible PM growth and reproduction Figure S4C. PM-infected WT and myb3r4 mutants do not exhibit cell death Figure S4D. Endoreduplication occurs at site of PM infection not distal to infection Figure S4E. Ploidy correlates with nuclear size. Part 1. Impact of tissue preparation, LMD, and RNA amplification on array output. Text S1: Detailed Experimental Design and Methods Figure S1A: Correlation analysis indicates tissue preparation has minimal impact on ATH1 array output. Figure S1B: Correlation analysis indicates RNA degradation does not significantly impact array output. Figure S1C: Correlation analysis indicates two-round amplification does not significantly impact array output. Figure S1D: Independent biological replicates of LMD samples are highly correlated. Figure S1E: Validation of LMD array expression by qPCR. Table S1: Tissue preparation-associated genes Text S1: Detailed Experimental Design and Methods Arabidopsis thaliana ecotype Columbia-0 (Col-0) and ics1-2 (eds16-1) mutants in the Col-0 background were grown, evenly spaced in boxes containing Metro mix 200 (Scotts Sierra Horticultural Products; Marysville, OH) in Percival AR66L growth chambers at 22ºC, 70% RH and a 12 h photoperiod with photosynthetically active radiation = 180 !mol m-2 s-1. At 4 weeks, a subset of boxes were infected with conidia from 2 half-infected leaves (10-14 dpi) per box of Golovinomyces orontii MGH isolate [1] using a settling tower and mesh screen as in [2] to maximize reproducibility. Boxes containing G. orontii-infected plants were placed in a separate Percival AR66L growth chamber under the same conditions as those used for uninfected samples. Three to 4 fully expanded mature leaves were harvested at 5 dpi from infected and uninfected plants for PM expression profiling analyses. For the parallel whole leaf samples, fresh mature leaves (~100 mg) were collected in parallel with those for LMD. Freshly harvested leaf tissue for LMD samples was dissected in 10mM Sørensen’s phosphate buffer pH 7.2, microwave fixed, paraffin embedded and sectioned as in [3]. The Leica Microsystems AS Laser Microdissection system was used for isolation of cells from 10 !m prepared tissue sections using 40x XT objective, aperture 6, intensity 35, and speed 5. Batches of 1250 cells (infected) or 2500 cells (uninfected) were collected in the cap of one PCR tube, with 2 infected cell batches pooled prior to RNA extraction. Total RNA extractions employed the PicoPure kit (Arcturus Engineering) with Qiagen RNase-free DNase columns. Two-round amplification of extracted mRNA utilized the Affymetrix two-cycle target labeling kit. Amplified RNA from 3 batches of 2500 LMD-isolated cells was combined to obtain sufficient RNA for microarray hybridization. Microarray hybridizations to Affymetrix ATH1 arrays and array scanning were performed by the Functional Genomics Laboratory (UC Berkeley). Expression values (log2) for two biological replicates per treatment type were extracted using robust multi- array analysis with perfect match correction and quantile normalization as in [2]. Genes with "2 fold change in infected vs. uninfected samples were computed using one-way ANOVA (p<0.05) with Partek Genomics suite and had a false discovery rate (q-value) ! 5%. We performed parallel experiments to separately assess the impact of tissue preparation, RNA amplification, and laser microdissection on gene expression (array output). Samples were collected in parallel from the same experiment. Independent replicates were collected and processed for all sample types. In total these samples included: 1. Laser microdissected samples (prepared tissue, laser microdissected, amplified RNA). 2. Whole leaf amplified sample (fresh tissue, amplified RNA). These samples are referred to as “whole leaf” in the text as they are the most accurate whole leaf comparison. A 10 ng aliquot of RNA isolated from each of the whole leaf samples below is subject to two- round amplification and microarray hybridization. 3. Whole leaf sample (fresh tissue). Fresh mature whole leaves (~100 mg sample) are harvested and immediately frozen in liquid Nitrogen. RNA isolation, without 2-round amplification, and microarray hybridization. 4. Whole leaf scrape sample (prepared tissue, amplified RNA). Whole leaf tissue was prepared as for LMD. The prepared, sectioned, deparaffinized whole leaf samples were then scraped from the PEN slides prior to laser microdissection. RNA isolation, two- round amplification and microarray hybridization was performed. References 1. Plotnikova, J.M., T.L. Reuber, and F.M. Ausubel, Powdery mildew pathogenesis of Arabidopsis thaliana. Mycologia, 1998. 90: p. 1009-1016. 2. Chandran, D., et al., Temporal Global Expression Data Reveals Known and Novel Salicylate- Impacted Processes and Regulators Mediating Powdery Mildew Growth and Reproduction on Arabidopsis. Plant Physiology, 2009. 149: p. 1435-51. 3. Inada, N. and M.C. Wildermuth, Novel tissue preparation method and cell-specific marker for laser microdissection of Arabidopsis mature leaf. Planta, 2005. 221(1): p. 9-16. Figure S1A. Correlation analysis indicates tissue preparation has minimal impact (<0.5%) on ATH1 array expression data. Log2 expression correlation plots for (i) whole leaf scrape vs. whole leaf amplified, infected samples, (ii) LMD vs. whole leaf amplified, infected samples and (iii) LMD vs. whole leaf amplified, uninfected samples. r-value denotes Pearson’s correlation. A small subset of 107 genes (shown in red) were not correlated between prepared tissue samples (whole leaf scrape and LMD amplified) and fresh tissue samples (whole leaf amplified). These genes exhibited a "10- fold change in expression in whole leaf scrape vs. whole leaf-amplified samples. These same 107 genes also exhibited a "10-fold change in expression when laser microdissected samples were compared with whole leaf amplified samples as would be expected if their expression difference was due to the tissue preparation method. This was true when uninfected samples were compared or when infected samples were compared, again indicating that the altered expression of these genes was not associated with infection but with the preparation method. Tissue preparation-associated genes are listed in SI Table S1. A SI Figure S1B. Correlation analysis indicates RNA degradation does not significantly impact array output. ATH1 array probesets were designed to be 3’ biased (within 600 bp of the 3’ end of the transcript) to minimize the impact of processive degradation of mRNA from the 5’ end. Standard RNA degradation plots generated using AffyRNAdeg in Bioconductor plot expression values versus probes presented in order from 5’ to 3’ for each transcript. However, this plot does not take into account the relative position of the probes in the transcript and their physical distance from the 5’ end. For example, for shorter transcripts, individual probes might reside near the 5’ end. In addition, updated annotations (e.g. TAIR9) indicate some of the ATH1 probesets are not within 600 bp of the 3’ end. Therefore, to globally assess whether RNA degradation impacted our array output, we selected probesets for which all probes are “far” from the 5’ end (i.e. within 0.70 fractional distance of the 5’ end) using current annotation. Shown below is the distribution of probesets based on fractional distance from the 5’end for all genes on array (i), and (ii) for 3’ biased probesets where the distance between all probes and the 5’end of the transcript was at least 70% of the coding sequence length. log2 expression correlation plots to assess the impact of tissue preparation, RNA amplification, and LMD were then obtained for all genes and 3’biased probesets as shown for whole leaf amplified (fresh tissue, amplified) versus whole leaf scrape (prepared tissue, amplified) infected samples for all genes (iii) or for 3’ biased probesets (iv). r-value denotes Pearson’s correlation. The r values for all probesets versus 3’ biased probesets was not significantly different, indicating that RNA degradation associated with tissue preparation does not negatively impact ATH1 expression