ATAC-seq analysis

Adapter sequences were removed from raw sequence reads using NGMerge (Gaspar, 2018). ATAC-seq reads were aligned to the Drosophila melanogaster (dm6) genome using bowtie2 with the following parameters: --very-sensitive, --no-mixed, --no-discordant, -X 5000, -k 2. Reads with a mapping quality score < 30 were discarded, as were reads aligning to scaffolds or the mitochondrial genome. Analysis was restricted to fragments < 100 bp, which, as described previously, are most likely to originate from nucleosome-free regions (Buenrostro et al., 2013). To maximize the sensitivity of peak calling, reads from all replicates of GAF^deGradFP and control embryos were combined. Peak calling was performed on combined reads using MACS2 with parameters -f BAMPE --keep-dup all -g 1.2e8 --call-summits. This identified 64,133 accessible regions. To facilitate comparison to published datasets with different sequencing depths, peaks were filtered to include only those with a fold-enrichment over background > 2.5 (as determined by MACS2). This resulted in 36,571 peaks that were considered for downstream analysis. Because greater sequencing depth can result in the detection of a large number of peaks with a lower level of enrichment over background, this filtering step ensured that peak sets were comparable across all datasets. 201 bp peak regions (100 bp on either side of the peak summit) were used for downstream analysis. Reads aligning within accessible regions were quantified using featureCounts, and differential accessibility analysis was performed using DESeq2 with an adjusted p-value<0.05 and a fold change > 2 as thresholds for differential accessibility. Previously published ATAC-seq datasets (Hannon et al., 2017), {"type":"entrez-geo","attrs":{"text":"GSE86966","term_id":"86966"}}GSE86966; (Blythe and Wieschaus, 2016), {"type":"entrez-geo","attrs":{"text":"GSE83851","term_id":"83851"}}GSE83851; (Nevil et al., 2020), {"type":"entrez-geo","attrs":{"text":"GSE137075","term_id":"137075"}}GSE137075. were re-analyzed in parallel with ATAC-seq data from this study to ensure identical processing of data sets. Heatmaps and metaplots of z score-normalized read depth were generated with DeepTools. MEME-suite was used to for de novo motif discovery for differentially accessible ATAC peaks and to match discovered motifs to previously known motifs from databases. Genome browser tracks were generated using bigWigs with the Gviz package in R (Hahne and Ivanek, 2016). Numbers used for all Fisher’s exact tests are included in Supplementary file 4.