Identification of chromatin states during zebrafish gastrulation using CUT&RUN and CUT&Tag

Abstract

Background:

Cell fate decisions are governed by interactions between sequence-specific transcription factors and a dynamic chromatin landscape. Zebrafish offer a powerful system for probing the mechanisms that drive these cell fate choices, especially in the context of early embryogenesis. However, technical challenges associated with conventional methods for chromatin profiling have slowed progress toward understanding the exact relationships between chromatin changes, transcription factor binding, and cellular differentiation during zebrafish embryogenesis.

Results:

To overcome these challenges, we adapted the chromatin profiling methods CUT&RUN and CUT&Tag for use in zebrafish, and applied these methods to generate high resolution enrichment maps for H3K4me3, H3K27me3, H3K9me3, RNA polymerase II, and the histone variant H2A.Z using tissue isolated from whole, mid-gastrula stage embryos. Using this data, we identify a subset of genes that may be bivalently regulated during both zebrafish and mouse gastrulation, provide evidence for an evolving H2A.Z landscape during embryo development, and demonstrate the effectiveness of CUT&RUN for detecting H3K9me3 enrichment at repetitive sequences.

Conclusions:

Our results demonstrate the power of combining CUT&RUN and CUT&Tag methods with the strengths of the zebrafish system to define emerging chromatin landscapes in the context of vertebrate embryogenesis.

Introduction

Interaction between proteins and DNA is central to regulation of transcription and generation of the complex gene expression patterns required for normal development. Transcription factor binding helps to control the rate of gene expression in a given tissue, while modification of the histones that package DNA creates localized chromatin environments that promote or repress transcription factor binding^1,2. Modification of histone tails can be highly dynamic, driving dramatic reshaping of chromatin, especially during early animal embryogenesis. For example, in zebrafish, large scale de novo establishment of histone lysine 4 trimethylation (H3K4me3), histone lysine 27 trimethylation (H3K27me3) and histone lysine 9 trimethylation (H3K9me3) is first noted during blastula stages, with genome-wide accumulation of these modifications loosely coinciding with zygotic genome activation^3,4. Similar dynamic changes in histone modifications have been noted during early embryogenesis in mouse, Drosophila, and C. elegans^5–7.

For roughly three decades, Chromatin Immunoprecipitation (ChIP) has been the primary method used to detect protein-DNA interactions⁸. In its standard form, this method requires crosslinking of proteins to DNA, physical fragmentation of chromatin, and antibody mediated selection of chromatin fragments of interest. ChIP has proven especially powerful when coupled with high throughput sequencing (ChIP-seq) and has been widely implemented to profile genome-wide chromatin states and transcription factor binding^9–11. However, the need for crosslinking, relatively low signal to noise ratios, and input requirements of ~1 million cells present challenges to effective implementation of ChIP, especially when applied to early-stage embryos and embryonic tissues. Modifications of conventional protocols that allow ChIP to be performed in the absence of crosslinking and with low cell numbers have been described^12,13. However, these methods can require deep sequencing, and generally work best for abundant proteins.

Cleavage Under Targets and Release Using Nuclease (CUT&RUN) offers a recently described alternative to ChIP that may be especially appealing for studies requiring early-stage embryos and embryonic tissues^14,15. Based on Chromatin Immunocleavage (ChIC) technology¹⁶, CUT&RUN employs a targeted nuclease strategy to isolate DNA fragments associated with proteins of interest. In CUT&RUN, a fusion of micrococcal nuclease and protein A/G (pAG-MNase) selectively cleaves antibody-bound chromatin in unfixed permeabilized cells. Released fragments are then isolated from solution and used for sequencing. The strategy bypasses immunoprecipitation steps, thereby shortening time required for purification. By only releasing the relevant fraction of the genome, CUT&RUN promotes high signal-to-noise ratios that allow significantly lower sequencing read depths (1/10th that of ChIP) and use of far fewer cells for analysis (1/10 to 1/1000th of ChIP)^14,15. Compared to physical fragmentation, endonuclease digestion also generates shorter fragments that reflect true protein footprints, allowing for base pair resolution mapping of protein binding in the absence of crosslinking.

To date, CUT&RUN has been used primarily to profile global transcription factor binding and histone modification enrichment in mammalian cells^{14,15,17–27}. Success has also been reported adapting CUT&RUN for profiling genome wide chromatin states in S. cerevisiae¹⁴, Drosophila^28,29, Arabidopsis³⁰, mouse germ cells^31–33 and mouse and human embryos^34,35. Recently, CUT&RUN has been used to directly test interactions between the transcription factor Pu.1 and known binding sites, as well as to identify Hox13 binding sites in dissected zebrafish tissues^36,37.

CUT&Tag represents a recent adaptation of CUT&RUN that similarly takes advantage of immunotethering to identify regions of antibody-bound chromatin³⁸. However, in this method, a fusion of protein A/G and Tn5 transposase is used to catalyze simultaneous cleavage and sequencing adapter ligation at antibody-bound chromatin, thus eliminating the need for downstream library preparation. Similar to CUT&RUN, this streamlined approach yields improved signal-to-noise ratios, especially when mapping stable protein-DNA interactions such as those involving histones³⁸. Advantages of CUT&Tag include increased amenability to ultra-low sample inputs and high-throughput approaches³⁸. However, compared to CUT&RUN, current protocols may have more limited effectiveness at detecting transient protein-DNA interactions involving transcription factors, and in some cases, the use of Tn5 transposase could potentially bias data toward open chromatin^38,39. Thus far, CUT&Tag has been applied to assay enrichment of modified histones in plant and mammalian tissues^38,40–45.

Here we describe successful implementation of CUT&RUN and CUT&Tag methods for profiling protein-DNA interactions in zebrafish embryos. Zebrafish have long been recognized as a powerful system for the study of early vertebrate development, as external fertilization facilitates molecular analysis at the earliest stages of embryogenesis. However, classical ChIP experiments in zebrafish can still require hundreds of embryos, resolution is limited, and developmentally emerging sites of enrichment can be difficult to identify above background. We apply modified CUT&RUN and CUT&Tag methods to generate high resolution maps of enrichment for H3K4me3, H3K27me3, H3K9me3, RNA polymerase II (pol II) and the histone variant H2A.Z during zebrafish gastrulation. Using this data, we identify a conserved subset of developmental genes that are enriched in H3K4me3 and H3K27me3 in bulk chromatin isolated from gastrulating embryos, provide evidence for a changing H2A.Z landscape during embryogenesis, and demonstrate the effectiveness of CUT&RUN for detecting protein enrichment at repetitive sequences with reduced mappability. Our work demonstrates the power of combining CUT&RUN and CUT&Tag with the strengths of the zebrafish system to better understand the changing embryonic chromatin landscape and its roles in shaping development.

Results

CUT&RUN effectively detects H3K4me3 and H3K27me3 enrichment near gene transcriptional start sites (TSS).

To test the feasibility of using CUT&RUN in zebrafish, we initially profiled two well-characterized histone modifications, H3K4me3 and H3K27me3, in mid-gastrula stage embryos (shield stage, ~6 hours post fertilization (hpf)). H3K4me3 is typically associated with the TSS of active genes, while H3K27me3 is associated with repressed genes. The combined presence of H3K4me3 and H3K27me3 is also observed at a subset of gene promoters during embryogenesis, with these bivalently marked genes existing in a poised state that enables rapid activation at appropriate developmental time points⁴⁶. Previous studies have identified gene regulatory regions marked by both H3K27me3 and H3K4me3 in blastula stage zebrafish embryos^47–49. However, the capacity for these dually marked domains to persist during zebrafish gastrulation or for new domains of dual enrichment to emerge during this period has not been addressed.

For each histone modification, we performed two replicate CUT&RUN experiments using pools of 25 embryos per replicate. CUT&RUN was performed using an adapted version of the protocol first described by Skene et al for use in yeast and mammalian cells⁵⁰. A detailed version of this adapted protocol is included in supplementary materials. In total, we identified 29,562 sites of H3K4me3 enrichment and 9,948 sites of H3K27me3 enrichment. Consistent with previous studies, we observed strong enrichment of H3K4me3 within the 500 base pairs (bp) surrounding the TSS of known genes (Figure 1A–C). Enrichment of H3K27me3 was found at the TSS of a smaller subset of genes, with broad H3K27me3 peaks noted in additional regions including Hox gene clusters (Figure 1A–D). Consistent with known associations, published RNA-seq data from shield stage embryos⁵¹ demonstrated the presence of RNA transcripts derived from 96% of genes exclusively marked by H3K4me3, while only 6% of those marked exclusively by H3K27me3 were associated with RNA transcripts at shield stage. Taken together, these data demonstrate that CUT&RUN can be used for robust detection of modified histones in the embryonic zebrafish genome.

Figure 1: CUT&RUN detects H3K4me3 and H3K27me3 at the transcriptional start sites of annotated genes in shield stage embryos.

(A) Heat maps of replicate data for H3K4me3 (green) and H3K27me3 (purple) enrichment as detected by CUT&RUN. (B) Profile plot of mean enrichment of H3K4me3 and H3K27me3 at the TSS of annotated genes. (C) Genome browser view of H3K4me3 and H3K27me3 enrichment at select loci. H3K4me3 is detected at the TSS of three genes with associated RNA transcripts (alkbh6, capns1a, and zgc153990) and H3K27me3 is detected at the TSS of the inactive dlb gene. (D) Genome browser view of broad H3K27me3 enrichment over the hoxA gene cluster on chromosome 3. In C and D replicate data are shown as individual tracks. Rep= Replicate.

Next, we investigated the co-occurrence of H3K4me3 and H3K27me3. In total, analysis of CUT&RUN data revealed 6,570 regions that showed enrichment signals for both H3K4me3 and H3K27me3 in bulk chromatin isolated from shield stage embryos (Figure 2A–B). Genes showing enrichment for both H3K4me3 and H3K27me3 at their TSS in CUT&RUN data were heavily biased toward control of developmental processes, suggesting that these genes might be subject to bivalent regulation (Figure 2C). Alternatively, because bulk tissue was used for analysis, detection of dual marks could simply reflect sample heterogeneity, with some cells marked by H3K4me3 and others marked by H3K27me3. To discriminate between these possibilities, we examined RNA levels at dually marked genes in shield stage embryos, hypothesizing that if dual marks arose from mixed cell populations, transcripts from active H3K4me3 marked genes would be detectable in the homogenized tissue. We found 42% of genes harboring both modifications in CUT&RUN data had associated transcripts as would be predicted for dual marks resulting from cellular heterogeneity. The remaining 58% lacked detectable transcripts, as would be predicted for bivalently regulated genes held in a poised state.

Figure 2. Regions marked with both H3K4me3 and H3K27me3 are detected in chromatin isolated from whole shield stage zebrafish embryos.

(A) Venn diagram indicating the number of regions enriched in H3K4me3, H3K27me3 or both modified histones in CUT&RUN data from whole shield stage embryos. Regions where one peak of H3K27me3 overlapped more than one peak of H3K4me3, or vice versa, were counted as one overlapping peak. (B) Genome browser view illustrating the co-occurrence of H3K4me3 and H3K27me3 enrichment peaks at the TSS of the igfbp2a gene, which lacks detectable transcripts at shield stage. At the same stage, the TSS of the nearby tra2b gene is exclusively marked by H3K4me3 and is associated with RNA transcripts. Replicate data are shown as individual tracks. (C) GO analysis reveals genes marked by both H3K4me3 and H3K27me3 at shield stage are associated with developmental processes including nervous system development (3831 query terms). Rep= Replicate.

Genes gaining strong bivalency during gastrulation have recently been identified in mouse⁵². Intriguingly, we find that 40% of genes acquiring strong bivalency during mouse gastrulation are also dually marked in our CUT&RUN data, with the majority of this gene set (84%) lacking detectable RNA transcripts during zebrafish shield stage (Supplementary Table 1). This overlap suggests the potential for a conserved bivalent chromatin program involved in guiding cell fate decisions during germ layer formation.

CUT&RUN effectively detects C-terminal domain serine 5 phosphorylated RNA pol II at the 5’ end of transcribed genes.

To assess the potential for CUT&RUN to detect interactions between DNA and non-histone proteins, we performed CUT&RUN on shield stage embryos using an antibody that recognizes the serine 5 phosphorylated C-terminal domain of RNA polymerase II (Ser5P-pol II). This modified form of poI II is highly enriched near the TSS of genes undergoing active transcription^53,54. Consistent with known localization patterns, CUT&RUN revealed Ser5-pol II enrichment near the TSS of zebrafish genes, with 71% of enrichment peaks found at genes producing transcripts that were detectable by RNA-seq at shield stage (Figure 3A–C).

Figure 3. CUT&RUN detects Ser-5 CTD phosphorylated RNA polymerase 2 (Ser5P-pol II) at the TSS of transcribed genes in shield stage embryos.

(A) Heat maps of replicate data for Ser5P-pol II enrichment at the TSS of annotated genes (B) Profile plot showing mean enrichment of Ser5-pol II at the TSS of annotated genes. (C) Genome browser view showing enrichment of Ser5-pol II at the start of setdb1a, which produces RNA transcripts at shield stage. Similar S5-pol II enrichment is not observed at the silent si:ch211-81a5.5 gene. Replicate data are shown as individual tracks. Rep= Replicate.

CUT&RUN detects H3K9me3 enrichment at repetitive sequences, including in regions with low mappability.

To assess CUT&RUN effectiveness in interrogating the repetitive fraction of the zebrafish genome, we examined enrichment of the heterochromatic histone modification H3K9me3 in shield stage embryos (Figure 4A). H3K9me3 is canonically associated with transcriptional silencing at repetitive sequences including many transposable elements^55,56. As expected, only a small fraction (2.6%) of H3K9me3 enrichment peaks shared between replicates overlapped with transcriptional start sites of known genes, while 93% percent were associated with annotated repetitive sequences (Figure 4B). Among sites of enrichment, the largest number of peaks were associated with DNA transposons, followed by long terminal repeat (LTR) transposons and simple repeats (Figure 4B).

Figure 4. CUT&RUN detects H3K9me3 enrichment at repeated sequences in shield stage embryos.

(A) Heat map of H3K9me3 enrichment around peak centers in CUT&RUN replicates (B) Pie chart depicting H3K9me3 enrichment at different repeat sequence classes as detected by CUT&RUN. (C) Venn diagram showing shared and unique peaks between shield stage CUT&RUN and ChIP data. (D) Genome browser view comparing enrichment and called peaks in CUT&RUN (dark blue) and ChIP (light blue) data from shield stage embryos. (E) Genome browser view of a select genomic loci containing a ChIP-only peak. (F) Genome browser view of a select genomic loci containing multiple CUT&RUN-only peaks on the long arm of chromosome 4. (G) Bar graph of CUT&RUN-only, shared and ChIP-only peaks detected in regions with mappability scores below 0.5 (H) Bar graph comparing CUT&RUN-only and ChIP-only peaks on the long vs short arms of chromosome 4. In D-F replicate data are shown as individual tracks. Rep= Replicate.

We previously analyzed H3K9me3 in shield stage embryos by ChIP using the same antibody used for CUT&RUN analysis⁵⁷. Comparing H3K9me3 CUT&RUN data to published ChIP data revealed 10,229 sites of enrichment that were shared between ChIP and CUT&RUN datasets. We also noted 20,271 enrichment peaks that were unique to CUT&RUN data and 7,564 peaks were unique to ChIP (Figure 4C). Browser assessment of select loci suggested that CUT&RUN yielded improved signal to noise ratios over ChIP (Figure 4D–E), and consistent with this initial impression, we found that the globally calculated fraction of reads in peaks (FRiP score)⁵⁸ was roughly 6-fold higher for CUT&RUN data compared to ChIP (10.4% vs. 1.6%). This finding raises the possibility that some ChIP only peaks may arise from spurious peak calling in a high background environment.

Further browser inspection of CUT&RUN-only peaks suggested that this subset of peaks was common in repetitive, low-mappability regions where it can be difficult to map reads accurately (Figure 4D, F)⁵⁹. Consistent with this initial impression, we calculated that genome wide, there were roughly 10 times more CUT&RUN-only peaks in regions with mappability scores of less than 0.5 compared to ChIP-only or shared peaks (Figure 4G). This distinction was particularly noticeable on the long arm of chromosome 4 (4q), which is very highly enriched in repetitive elements and exists in a densely heterochromatic state^60,61 (Figure 4F, H). Here, the number of CUT&RUN-only peaks was 4 times more frequent than on the euchromatic short arm (4p) of the same chromosome. In contrast, the number of ChIP-only peaks on 4p and 4q were similar. Collectively, these data suggest that CUT&RUN may be especially well-suited for detecting protein-DNA interactions in repetitive regions.

CUT&Tag effectively identifies sites enriched for the histone variant H2A.Z

Finally, we investigated whether CUT&Tag could also be applied for effective profiling of DNA-histone interactions in zebrafish. For this analysis, we examined the localization of the histone variant H2A.Z in whole shield stage embryos. In vertebrates, H2A.Z-containing nucleosomes are known to reside at gene promoters, and this histone variant has been implicated in protecting these sites from DNA methylation^62–64. Previous studies have described H2A.Z localization in zebrafish sperm and blastula stage embryos by ChIP⁶². However, profiles have yet to be assessed during gastrulation or later stages of embryogenesis. CUT&Tag was performed using an adapted version of the protocol first described by Kaya-Okur, et al. for use in mammalian cells³⁸. A detailed version of this adapted protocol is included in supplementary materials. Using this protocol, we detected robust signals for H2A.Z among replicates in shield stage embryos, and in keeping with previous findings, we found that H2A.Z was enriched at gene promoters (Figure 5A–C). Consistent with previous reports that have suggested a complex relationship between H2A.Z enrichment and transcription⁶⁵, approximately 74% of H2A.Z marked genes are expressed at shield stage (TPM > 0.5). This result demonstrates that CUT&Tag can be used effectively to detect H2A.Z enriched sites during zebrafish embryogenesis.

Figure 5. CUT&Tag detects H2A.Z in shield stage zebrafish embryos.

(A) Heatmaps of replicate data for H2A.Z enrichment at annotated genes as detected by CUT&Tag in shield stage embryos. (B) Profile plot of H2A.Z mean enrichment at annotated genes. (C) Genome browser views of H2A.Z and RNA transcripts at selected loci. Replicate data for H2A.Z CUT&Tag are shown as individual track. Rep= Replicate.

To further characterize potential changes in H2A.Z patterning during developmental transitions, we also performed CUT&Tag for H2A.Z using 24 hpf embryos. Correlation analysis showed that replicates were clustered according to developmental stages, indicating that H2A.Z enrichment patterns changed during developmental progression (Figure 6A). Further comparison of H2A.Z profiles between shield stage and 24 hpf embryos revealed that H2A.Z normalized cumulative enrichment in aligned fragments generally increased from shield stage to 24 hpf and that there were more sites that were unique to 24 hpf embryos than to shield stage embryos (Figure 6B–F). We further performed gene ontology analysis of those unique H2A.Z peaks to shield stage and 24 hpf. Genes related to developmental differentiation occurred in proximity to sites where H2A.Z increased and decreased from 6 hpf to 24 hpf (Figure 6E). Taken together these data suggest that H2A.Z profiles are dynamic and evolve over the course of embryogenesis.

Figure 6. CUT&Tag identifies changes of H2A.Z patterning from shield stage to 24 hpf zebrafish embryos.

(A) Correlation heatmap demonstrating high correlation among CUT&Tag replicates and separation based on developmental stages. (B) Heatmaps of H2A.Z enrichment at H2A.Z-marked regions as detected by CUT&Tag in shield stage and 24 hpf embryos. Normalized H2A.Z signals were generated using merged replicate data. (C) Profile plot of H2A.Z enrichment at H2A.Z-marked regions in shield stage and 24 hpf embryos. (D) Venn diagram showing shared and unique H2A.Z peaks between H2A.Z CUT&Tag data in shield stage and 24 hpf embryos. (E) Gene ontology analysis of shield stage and 24 hpf unique H2A.Z peaks. Only top 6 non-redundant GO terms were shown. (F) Genome browser views of H2A.Z enrichment in shield stage and 24 hpf embryos at selected loci.

Discussion

Our results demonstrate the feasibility of using CUT&RUN and CUT&Tag approaches to profile chromatin during zebrafish embryogenesis, and provide the first high-resolution enrichment maps of H3K4me3, H3K27me3 and H3K9me3, Ser5P-pol II, and H2A.Z during zebrafish gastrulation. The reduced embryo requirements of CUT&RUN and CUT&Tag eliminate a significant barrier to current chromatin profiling during zebrafish embryogenesis, while increased signal-to-noise ratios and more precise foot printing afforded by these methods facilitate accurate identification of enrichment sites in these data sets.

It is important to appreciate that just like any antibody-based method, the specific efficacy of a given antibody-antigen interaction will be a critical determinant in whether CUT&RUN and CUT&Tag will be successful in profiling zebrafish chromatin. Given that these methods are typically performed on unfixed chromatin, in some cases, antibodies that perform well in other methods that lack fixation such as immunofluorescence may outperform those validated for ChIP. Although related, CUT&RUN and CUT&Tag methods have distinct subsets of strengths and weaknesses, making direct comparisons between these methods highly context dependent. For example, the streamlined workflow and ultra-low inputs associated with CUT&Tag may make this method appealing for hard to obtain tissues or applications involving many samples, while increased capacity to detect more transient interactions and elimination of potential Tn5 biases may make CUT&RUN more appealing in other contexts.

Of particular note, our data demonstrate improved power of CUT&RUN for analysis of protein-DNA interactions that impact repetitive sequences. While we were initially surprised by the substantial number of non-overlapping peaks detected between H3K9me3 CUT&RUN and ChIP data, our analysis suggests that this discrepancy may be due to improved enrichment peak calling at repetitive sequences, especially in regions of low mappability. Reduced background in CUT&RUN allows for effective global peak calling at lower sequencing depths than can be used for ChIP¹⁴. This feature likely contributes to CUT&RUN’s increased effectiveness in low mappability regions, as more sequencing reads in these regions will typically be discarded due to low mapping quality.

Our analysis of chromatin in shield stage zebrafish embryos provides an important reference for understanding how histone modifications and variants may help shape early cell fate decisions during development. For example, our analysis of shield stage CUT&RUN data suggests the presence of bivalently regulated genes during zebrafish gastrulation. Although sequential methods will be required to fully rule out the possibility that the dually marked domains detected in CUT&RUN data result from mixed cell populations, limited transcription at shield stage, and associated roles in developmental processes suggest many of these genes may be sites of bivalent regulation. Overlap between dually marked domains identified in our data and those reported to gain strong bivalency during mouse gastrulation raises the possibility of a conserved bivalent chromatin program for directing early cell fate decisions.

Our data also reveal changes in H2A.Z deposition as embryos develop from shield stage to 24 hpf. A large number of enriched H2A.Z sites were called as peaks both at shield stage and at 24 hpf (88% of shield stage peaks), providing partial validation of our CUT&Tag methods, and suggesting that there is general maintenance of embryonic H2A.Z localization over this developmental period. In addition to these 30,869 maintained sites, de novo H2A.Z accumulated at 15,172 genomic locations in 24 hpf embryos, indicating that broad expansion of H2A.Z localization occurs as embryos progress through segmentation stages of development. This shifting H2A.Z landscape suggests genomic H2A.Z reorganization may help drive differentiation over this developmental window. Consistent with this model, prior studies indicate that disruption of H2A.Z over similar developmental periods in zebrafish and frog embryos leads to defects in neural crest derived tissues^66,67.

Taken together, our data validate the use of CUT&RUN and CUT&Tag in zebrafish. The high-resolution chromatin maps of zebrafish embryos undergoing gastrulation generated in this study provide an important resource for probing the relationship between chromatin states and early cell fate decisions. At the same time, the detailed protocols for CUT&RUN and CUT&Tag accompanying this study will facilitate efficient adoption of these methods by the zebrafish field, allowing for large scale, high resolution, time course experiments to broadly define chromatin dynamics across the entirety of early zebrafish embryogenesis.

Experimental Procedures

Zebrafish

Zebrafish husbandry and care were conducted in full accordance with animal care and use guidelines with ethical approval by the Institutional Animal Care and Use Committees at the University of Georgia and the University Committee on Animal Resources at the University of Rochester Medical Center. Zebrafish were raised and maintained under standard conditions in compliance with relevant protocols and ethical regulations. Fertilized eggs were obtained from the zebrafish AB strain. Embryos were reared in system water at 28.5 °C and staged according to morphology.

CUT&RUN and CUT&Tag

CUT&RUN and CUT&Tag were performed according to published methods, with minor adaptations^38,50. Detailed protocols describing the exact CUT&RUN and CUT&Tag methods used in this study, and preparation of CUT&RUN libraries are provided in supplementary materials. Antibodies used in this study are provided in Table 1.

Table 1.

Antibodies used in this study

Antibody	Catalog number	Dilution
H3K4me3	ab8580 (Abcam)	1/100
H3K27me3	07-449 (Millipore)	1/50
H3K9me3	ab8898 (Abcam)	1/100
RNA pol-II CTD (phospho S5)	ab5408 (Abcam)	1/75
H2A.Z	39113 (Active motif)	1/50
IgG	ab46540 (Abcam)	1/100

Sequencing data

CUT&RUN libraries were pooled and sequenced on a NextSeq500 instrument at the Georgia Genomics Facility. Sequencing details are provided in Supplementary Table 2, and raw data generated in this study can be found at NCBI GEO (accession GSE178343). CUT&Tag libraries from 24 hpf embryos were pooled and sequenced with services from NovoGene Co. on a HiSeq2500 instrument, and libraries from shield stage embryos were sequenced with Genewiz Co. on a NextSeq550 instrument. Sequencing details are provided in Supplementary Table 2, and raw data generated in this study can be found at NCBI GEO (accession GSE178559). Publicly available RNA data used in this study can be found at the EBI European Nucleotide Archive (accession # PRJEB12982). Publicly available H3K4me3 & H3K27me3 ChIP data used in this study can be found at NCBI GEO (accession # GSE110663). Publicly available H3K9me3 ChIP data used in this study can be found at NCBI GEO (accession # GSE113086).

CUT&RUN data analysis

Short reads (<20 bp) and adaptor sequences were removed using TrimGalore (version 0.6.5), cutadapt version 2.8, and Python 3.7.4, with fastqc command (version 0.11.9) (https://github.com/FelixKrueger/TrimGalore). Trimmed Illumina reads were aligned to the current zebrafish genome assembly (GRCz.11, Ensembl release 103) using Bowtie2 (version 2.4.1) with the “very-sensitive-local” parameters which assigns multi-mapped reads to their best alignment based on MAPQ scores⁶⁸. Reads were also aligned to the spike-in genomes using the same parameters as above. For the histone modifications, the spike-in was S. cerevisiae (R64-1-1, Ensembl release 48) and for the pol-II, the spike-in was E. coli (assembly GCF_000005845.2_ASM584v2). Alignments were filtered using SAMtools (version 1.10) for a MAPQ score of 20^69,70. Using a modified version of the script from Skene et al 2017, CUT&RUN data was normalized for spike-in and library-size with Bedtools “genomecov” (version 2.29.2)⁷¹. This outputs a bedgraph file which was adapted to a bed file for peak-calling and converted to a bigwig using ucsc “bedGraphToBigWig” (version 359) for data visualization in the Integrated Genome Viewer. The Hypergeometric Optimization of Motif EnRichment (HOMER) software package (version 4.11) was used to identify peaks over input⁷². For pol-II CUT&RUN, the parameters used were -style factor -gsize 1.5e9. For histone modification CUT&RUN, the parameters used were -style histone -minDist 1000 -F 6 -gsize 1.5e9 -fdr 0.0001. Bedtools “intersect” was used to compare peak locations between samples⁷¹. Peak annotation was performed using HOMER “annotatePeaks.pl” with the masked reference annotation⁷². Only peaks within 1000 bps of a TSS were considered to be associated with that gene. For the H3K9me3 CUT&RUN, peaks that were further than 1000 bps from a genic TSS were then re-annotated using an un-masked reference annotation to determine which enrichment domains were associated with repetitive elements. deepTools (version 3.3.1) was used to construct heatmaps (plotHeatmap) and metaplots (plotProfile)⁷³. Mappability was calculated with GenMap using a kmer of 76 bp and a mismatch allowance of 2⁵⁹. To identify H3K9me3 peaks associated with regions of low mappability, Bedtools “intersect” was used with -f 0.3, requiring at least 30% of a peak be covered with mappability scores <50%. Proportional venn diagrams were made with the help of BioVenn⁷⁴. Similarity between CUT&RUN replicates was confirmed by multiBigwigSummary (deepTools version 3.3.1) and visualized using plotCorrelation and plotPCA (Figure 7A–B). Gene Ontology Analysis was performed using gProfiler with the g:SCS multiple testing correction and significance threshold of 0.05⁷⁵. Scripts used for data processing can be found at https://github.com/klduval/CutNRun_2021.

Figure 7. Similar enrichment profiles between CUT&RUN replicates.

(A) Principal component analysis (PCA) plot of CUT&RUN sample. (B) Heatmap depicting Spearman correlation of enrichment between CUT&RUN samples, correlation coefficients shown for each pairwise comparison. Rep= Replicate.

ChIP data analysis

Short reads (<20 bp) and adaptor sequences were removed using TrimGalore (version 0.6.5), cutadapt version 2.8, and Python 3.7.4, with fastqc command (version 0.11.9) (https://github.com/FelixKrueger/TrimGalore). Trimmed Illumina reads were aligned to the current zebrafish genome assembly (GRCz.11, Ensembl release 103) using Bowtie2 (version 2.4.1) with the “very-sensitive-local” parameters which assigns multi-mapped reads to their best alignment based on MAPQ scores. Zhu 2019 ChIP paired reads had become disordered so were repaired using BBMap “repair.sh” (version 38.83) before alignment. Alignments were filtered using SAMtools (version 1.10) for a MAPQ score of 20^69,70. The Hypergeometric Optimization of Motif EnRichment (HOMER) software package (version 4.11) was used to identify peaks over input with the following parameters: -style histone -size 1000 -gsize 1.5e9. Bedtools “intersect” was used to compare peak locations between samples⁷¹. To plot the relative distribution of mapped ChIP reads, read counts were determined for each 10 bp window across the genome using deepTools “bamCoverage” (version 3.3.1) and data were displayed using the Integrated Genome Viewer⁷³.

CUT&Tag data analysis:

For H2A.Z CUT&Tag paired-end sequencing reads, adaptor sequences were removed using cutadapt (version 2.7). Trimmed reads were aligned to zebrafish genome assembly (GRCz.11, Ensembl release 103) using Bowtie2 (version 2.2.5) with default parameters⁶⁸. Unmapped reads were filtered using samtools (version 1.9), and PCR duplicates were removed using picard MarkDuplicates (version 2.5.0)^69,70. H2A.Z replicate data were merged using picard MergeSamFiles, and genome browser tracks were generated using deepTools bamCoverage (version 3.5.1) with the following setting: --normalizeUsing RPKM --binSize 10⁷³. Peak calling was performed using macs2 callpeak (2.2.6) with the following parameters: -f BAMPE -g 1.5e9 --nomodel --broad. Heatmaps and profile plots were generated using plotHeatmap and plotProfile from deepTools (version 3.5.1), and fonts and labels were adjusted in Affinity Designer (version 1.9.1)⁷³. Overlapping peak analysis was performed using bedtools intersect (version 2.30.0), and venn diagram was generated using R package eulerr (version 6.1.0 and R version 4.0.3)⁷⁶. To accommodate potential peak calling issues due to differences in sequencing depths, samples H2A.Z from 24 hpf were down-sampled to match shield stage sequence depth using Picard (version 2.5.0). Gene Ontology Analysis was performed using gProfiler with the g:SCS multiple testing correction and significance threshold of 0.05.

RNA-seq data analysis

Short reads (<20 bp) and adaptor sequences were removed using TrimGalore (version 0.6.5), cutadapt version 2.8, and Python 3.7.4, with fastqc command (version 0.11.9). Trimmed Illumina reads were aligned to the current zebrafish genome assembly (GRCz.11, Ensembl release 103) using STAR (version 2.7.3a) with option “--outSAMmultNmax 1” to keep multimapping reads but only report their best alignment⁷⁷. TPMCalculator (version 0.0.4) was used to determine which genes are actively expressed (TPM >0.5) at the shield stage⁷⁸. For pol-II CUT&RUN, Bedtools “intersect” was used to compare peaks with expressed genes. For H3K4me3 and H3K27me3 CUT&RUN, Bedtools “intersect” was also used to compare peaks within 1000bps of a TSS with expressed exons to ascertain which domains are associated with active transcription.

Key findings:

• CUT&RUN and CUT&Tag methods allow genome wide identification of protein-DNA interactions during zebrafish embryogenesis using far fewer embryos than traditional methods

• Low background in CUT&RUN and CUT&Tag helps promote effective detection of protein-DNA interactions in developing zebrafish embryos, especially at repetitive sequences

• CUT&RUN identifies a conserved set of genes that may be bivalently marked by H3K4me3 and H3K27me3 during both zebrafish and mouse gastrulation

• CUT&Tag reveals developmentally regulated changes in histone variant H2A.Z deposition during zebrafish embryogenesis