Spatially uniform establishment of chromatin accessibility in the early Drosophila embryo

As the Drosophila embryo transitions from the use of maternal RNAs to zygotic transcription, domains of open chromatin, with relatively low nucleosome density and specific histone marks, are established at promoters and enhancers involved in patterned embryonic transcription. However, it remains unclear whether open chromatin is a product of activity - transcription at promoters and patterning transcription factor binding at enhancers - or whether it is established by independent mechanisms. Recent work has implicated the ubiquitously expressed, maternal factor Zelda in this process. To assess the relative contribution of activity in the establishment of chromatin accessibility, we have probed chromatin accessibility across the anterior-posterior axis of early Drosophila melanogaster embryos by applying a transposon based assay for chromatin accessibility (ATAC-seq) to anterior and posterior halves of hand-dissected, cellular blastoderm embryos. We find that genome-wide chromatin accessibility is remarkably similar between the two halves. Promoters and enhancers that are active in exclusively one half of the embryo have open chromatin in the other half, demonstrating that chromatin accessibility is not a direct result of activity. However, there is a small skew at enhancers that drive transcription exclusively in either the anterior or posterior half of the embryo, with greater accessibility in the region of activity. Taken together these data support a model in which regions of chromatin accessibility are defined and established by ubiquitous factors, and fine-tuned subsequently by activity.


52
Spatially resolved ATAC-seq is robust and consistent with whole embryo measurements of 53 chromatin accessibility 54 To determine the extent to which chromatin accessibility is spatially patterned along the 55 anteroposterior axis in the early embryo, we manually separated anterior and posterior embryo 56 halves and performed a modified ATAC-seq [25] protocol on each half separately. Briefly, we 57 collected cellular blastoderm embryos (mitotic cycle 14, embryonic stage 5), flash froze them in 58 liquid nitrogen, and then sliced each embryo with a chilled scalpel at the anteroposterior midline, 59 separating anterior and posterior halves into separate pools (Fig. 1a). We isolated nuclei from 20 60 anterior halves (in duplicate), 20 posterior halves (in duplicate), 10 frozen unsliced embryos, and a mixed sample containing a subset of nuclei from anterior and posterior samples and applied the 93 To get a more systematic view of the relationship between transcriptional activity and 94 spatial patterns of chromatin accessibility, we used available genome annotation and functional 95 data to systematically identify A-P and D-V (as a control) patterned enhancers whose 96 transcriptional outputs are restricted to one half of the embryo [27][28][29][30][31][32][33][34][35][36][37][38] (File S1). We excluded 97 enhancers that did not overlap peaks called in any of the anterior, posterior, or whole samples 98 leaving 98 A-P and D-V patterned enhancers. 99 Although virtually all of the A-P patterned enhancers we looked at are accessible in both 100 halves, they clearly trend towards greater accessibility in the embryo half where they are active  105 We next computed a measure of differential accessibility (accessibility skew score) for 106 each enhancer by dividing the difference in accessibility in the active and inactive half by total 107 accessibility, such that positive scores denote loci that are more accessible in the active half, 108 negative scores signify loci that are more accessible in the inactive half, and loci with a score of 109 zero have no difference in accessibility (methods). We found that both anterior and posterior 110 enhancers have a significantly greater mean accessibility skew score than D-V enhancers or 111 random genomic regions with similar accessibility (p ant < 7.8 x 10 -10 and p post < 0.0016, Fig. 3c 112 and Table S1).  Promoters of A-P patterned genes are similarly accessible both when active and inactive. 120 We next examined the promoters of A-P patterned genes using expression data from 121 sections of embryos cryosliced along the A-P axis to curate lists of A-P patterned gene promoters 122 [39]. We only included promoters that overlapped accessibility peaks called on our dataset and have patterned expression confirmed by in situ hybridization assays (n= 36 anterior promoters, n = 39 posterior promoters, File S1). 125 Interestingly, accessibility in the active and inactive halves is much more similar at 126 anterior promoters (r S = 0.95) than at anterior enhancers (r S = 0.82), and is comparable to D-V 127 enhancers and promoters (Fig. 4 a-b, Dorsal promoters, r S = 0.98 ; Ventral promoters, r S = 0.96).

128
On the other hand, posterior promoters and enhancers show similar accessibility correlations 129 between the anterior and posterior halves (r S = 0.86), though as a group their mean accessibility 130 skew score is not significantly different than D-V patterned promoters or random regions (Fig. 131 4c and Table S1). Additionally, there is no distinct skew of accessibility in the direction of 132 activity seen at A-P promoters, in contrast to what we observed for A-P enhancers.

134
Anterior accessibility is associated with Bicoid binding while similarly accessible regions 135 are enriched for Zelda binding 136 We then used published ChIP-seq data of A-P patterning factors from stage 5 Drosophila 137 embryos to examine binding patterns at similarly and differentially accessible A-P enhancers 138 [40]. We analyzed Bicoid,Caudal,Knirps,Giant,Hunchback,Kruppel,and Zelda binding data,139 normalized by the mean signal for each factor. A-P enhancers that are more accessible in the 140 anterior (shades of orange) generally seem to be dominated by Bicoid binding with strikingly 141 little binding from other transcription factors, though there are some exceptions (Fig. 5, Fig. S4).

142
Enhancers more accessible in the posterior (shades of blue) generally have high Caudal, Knirps, 143 Giant, and Kruppel binding with strikingly more diversity in factors bound than at anteriorly 144 accessible enhancers. Interestingly, enhancers with similar accessibility in both halves (shades of 145 white) generally have a high diversity of factors binding -including Zelda (Fig. 5b). These 146 patterns reveal that while transcription factor binding clearly does not completely explain 147 differential chromatin accessibility, there are clear differences in factor composition and density 148 between differentially and similarly accessible enhancers.

155
This finding adds to a growing body of evidence that the chromatin landscape of the early 156 embryo is distinct from that of differentiated tissues found in later stage embryos. A recent study 157 assaying chromatin accessibility in single nuclei of developing Drosophila melanogaster 158 embryos found that 2 to 4 hour old embryos cluster distinctly from both 6 to 8 and 10 to 12 hour 159 embryos, with far less differentiation among the 2 to 4 hour nuclei than is observed for the older   [22,[51][52][53] and is likely associated with changes in the nucleo-cytoplasmic ratio [50,54]. 185 It is intriguing that, while A-P enhancers were accessible in both embryo halves 186 regardless of where they are active, the magnitude of their accessibility was modestly but 187 significantly skewed in the direction of activity. There are two obvious explanations for this. independent system with a spatial bias. We believe available data support the latter of these two 192 possibilities.

193
The strongest accessibility skew is in enhancers with an anterior bias, which as a group  Slicing frozen embryos NC14 embryos were placed in a custom freezing buffer consisting of ATAC-seq lysis 215 buffer [25] without detergent, 5% glycerol, and 1ul of bromoblue dye. Embryos were then taken 216 out of the freezing buffer and placed onto a glass slide which was then put on dry ice for 2-5 217 minutes. Once embryos were completely frozen, the glass slide was removed and embryos were 218 sliced with a chilled razorblade. Sliced embryo halves were moved to tubes containing ATAC-219 seq lysis buffer with 0.15mM spermine added to help stabilize chromatin.  Replicates were merged and peaks were called on the merged bed file using MACS2 with 251 the following parameters: --nomodel --nolambda --keep-dup all --call-summits.
where X active is the wig signal in the half where the region is activating gene expression and 277 X inactive is wig signal in the half where the region is not supposed to activate gene expression.

278
Accessibility skew score measures whether a region is differentially accessible in the expected 279 direction. This score is useful when comparing differential accessibility regardless of which half 287 Where X anterior is the wig signal in the anterior sample and X posterior is the wig signal in the 288 posterior sample. Significance for each region was determined by computationally matching 289 each region to a random region that has the same total normalized wig score (Fig. S5).

290
Accessibility skew score was calculated for each random region (termed RandSkewScore).

291
These scores were distributed normally and allowed for determining a Z-score for each region of       Drosophila embryos were flash frozen over dry ice in a buffer containing 5% glycerol and manually sliced in half with a scalpel. Twenty anterior and posterior halves were collected, homogenized, and the nuclei were isolated. ATAC-seq was then performed as described in (Buenrostro et al. 2013) with three times Tn5 transposase. (B) Scatterplot of normalized ATACseq signal over 1kb adjacent windows that tile the Drosophila genome in posterior (x) and anterior (y) samples shows high degree of correlation between the anterior and posterior halves. The Spearman correlation coefficient (denoted by rS) is 0.94. X and Y are log transformed. Light blue circles denote point density.