Targeted DamID detects cell-type-specific histone modifications in intact tissues or organisms

We fused Dam to proteins that bind with high affinity and specificity to posttranslational modifications, enabling identification of genome-wide distribution of these marks. We took advantage of several previously characterized mintbodies and protein-binding domains (chromatin readers) that recognize histone modifications on native chromatin (Fig 1A). The PHD domain of TAF3 recognizes H3K4me3 [17], the chromodomain of CBX7 binds H3K27me3 [18], and the single chain monoclonal antibody fragments 19E5 [19] and 15E11 [20] recognize H3K9ac and H4K20me1, respectively (Fig 1B). All chromatin readers were previously validated by ChIP-seq or peptide arrays, and the H3K9ac and H4K20me1 mintbodies were optimized to detect histone modifications upon intracellular expression [19,20]) and proven compatible with DamID [14].

We fused these chromatin readers to the E. coli Dam methylase, and expressed the Dam-fusions specifically in the NSCs of Drosophila third-instar larvae using TaDa [12]. Spatiotemporal control of TaDa expression was achieved with the GAL4 system [21] and GAL80^ts [22]. All Dam-fusion proteins generated distinct and reproducible methylation profiles (Figs 1C, 1D and S1A). As expected, H3K4me3 was enriched in promoter regions, whereas H3K9ac, H4K20me1 and H3K27me3 covered both promoter and coding sequences (Figs 1C, 1D and S1B). Focusing on genes actively transcribed in NSCs at this stage [13], H3K4me3 was present at both promoters and coding sequences, while H4K20me1was mainly located in coding sequences. H3K9ac was not highly prevalent but, as expected from its role in a switch from transcriptional initiation to elongation [15], was biased towards the 5′ end of the coding sequence.H3K27me3, which is associated with repressed regions, was absent from expressed genes (Fig 1E and 1F).

We then compared the NSC-specific normalized- and averaged-binding profiles of the chromatin readers and mintbody with published TaDa of PolII, Brahma, Pc, H1 and HP1a [13] to validate the distribution of the profiled histone marks. PolII and H3K4me3 correlated with one another well, as expected, whereas H3K9ac correlated better with Brahma, usually linked to enhancer regions marked by H3K27ac. This suggests that in Drosophila NSCs H3K9ac might also be associated with enhancers rather that the promoters of expressed genes (S1C, S1E Fig). In terms of repressive marks, H3K27me3 normalized signal correlated highly with the Pc and H1 TaDa, indicating that the chromatin reader is correctly recognizing the presence of H3K27me3 (S1F Fig). In fact, this can be clearly observed at loci known to be targeted by Polycomb Group of proteins, such as the Antennapedia complex or vnd locus, where H3K27me3 and Pc-binding patterns overlap (Figs 1E and S1D). Overall, we concluded that our strategy can accurately profile histone marks in vivo.

Radial glial cells (RGCs) that line the ventricular zone of the developing mouse cortex can be transiently transfected through injection of plasmid DNA into the telencephalic ventricles of a mouse embryo in utero, followed by transcranial electroporation (Fig 2A–2C). We fused the chromatin readers for H3K4me3, H3K9ac, H4K20me1 to intron-Dam for transient transfection [23] and expressed them downstream of pHes5, an established RGC-specific promoter [24,25]. We found that Dam fusions generated distinct methylation profiles when transfected in utero (Figs 2D, 2E and S2A), as shown by H3K4me3, H3K9ac and H4K20me1 signal present at Foxg1 locus, which is expressed in the early mouse telencephalon [26]. In contrast, the active marks were lacking at HoxB locus, which is normally only expressed in the hindbrain, rather than the anterior brain (Fig 2E) [27].

Fig 2. TaDa in mouse RGCs.

(A–C) Schematic of in utero TaDa in mouse RGCs. (D) TaDa profiles at FoxG1 in mouse RGCs. Signal files in bigwig format for (D–E) are available as supplementary at GSE278272, with the following filename prefixes: H4K20me1: GSE278272_iue115_5_phes5_15f11nls_5-vs-iue60_i2phes5_dam_S6, H3K4me3: GSE278272_iue62_phes5_taf3-vs-iue53_i2phes5_dam_S2, H3K9ac: GSE278272_iue92_i2phes5_19e5-vs-iue61_i2phes5_dam_S1. (E) TaDa profiles at the HoxB locus in mouse RGCs. (F) Schematic of TaDa during mouse cortical neurogenesis. (G) H3K4me3 TaDa profiles at the Nestin locus throughout neurogenesis. Signal files in bigwig format for (G–K) are available as supplementary at GSE278272, with the following filename prefixes: RGCs: GSE278272_iue62_phes5_taf3-vs-iue53_i2phes5_dam_S2, IPCs: GSE278272_taf3_pta1_iue67_S5-vs-dam_pta1_iue67_S4 Neurons: GSE278272_taf3_flox_iue101-vs-dam_flox_iue72. (H, I) H3K4me3 TaDa profiles at the NeuroG2 locus throughout neurogenesis and the mRNA expression pattern of NeuroG2 at E14.5. (J, K) H3K4me3 TaDa profiles at the Dcx locus throughout neurogenesis and the mRNA expression pattern of Dcx at E14.5. In situ images in (I) and (K) were obtained from the Eurexpress database [59]. All sequencing files are available at GSE278272.

https://doi.org/10.1371/journal.pbio.3002944.g002

The genomic distribution of the profiled modifications mimicked the known distribution of these marks, with H3K4me3 enriched at promoters and H3K9ac and H4K20me1 at gene bodies (S2B Fig). Histone mark peaks in mouse NSCs also reflected peak distribution previously identified by ChIP-seq [28] (S2D Fig), validating our approach. H4K20me1 was not studied as part of the ENCODE project (ENCODE Project Consortium., 2012), but H4K20me1 TaDa signal was enriched in the gene bodies of genes highly expressed in mouse RGCs (S2C and S2D Fig). The promoters of genes that are highly expressed in mouse RGCs were correlated with the presence of activating H3K4me3 and H3K9ac histone modifications (S2C and S2D Fig).

During neurogenesis, RGCs in the developing mouse cortex differentiate into intermediate progenitors (IPCs), and postmitotic neurons that migrate to the cortical plate (Fig 2F). To profile histone modifications during neurogenesis, we used a Hes5 promoter-fragment (pHes5 [24,25]) to restrict H3K4me3 TaDa to RGCs and the promoter of the Tuba1a gene (pTα1 [25,29,30]) to restrict H3K4me3 TaDa to IPCs (Fig 2F). To target postmitotic neurons, we expressed Cre recombinase from a NeuroD1 promoter (pND1 [31]) with a floxDam construct [32] (Fig 2F). The combination gave continuous expression of H3K4me3 TaDa after recombination in young postmitotic neurons (Fig 2G). The histone modification profiles were cell-type-specific, as illustrated by identification of H3K4 trimethylation at the promoter of the progenitor-specific gene Neurog2 in RGCs and IPCs, but not neurons (Fig 2H and 2I). Similarly, the promoter of the neuronal gene Dcx was marked by H3K4me3 in IPCs and neurons, but not in RGCs (Fig 2J and 2K). TaDa generated accurate genome-wide profiles of histone modifications from small numbers of cells, including transient progenitor populations like IPCs, without dissociating them from the complex cellular environment of the developing cerebral cortex.

Early development of the Xenopus embryo is a powerful system to study cellular differentiation and reprogramming [33] and the role of epigenetic modifications [34]. ChIP experiments from early-stage embryos require large amounts of starting material and have proven difficult due to the yolk and protein-rich cytoplasm. We generated constructs for in vitro transcription of H3K4me3 TaDa mRNA and injected these into the cytoplasm of fertilized eggs. DamID-seq was performed on stage 11 gastrulae, and on dissected dorsal (ecto- and mesoderm) or ventral (endoderm) halves of embryos at neurula stage 21 (Fig 3A). Despite having only a few embryos as starting material, the H3K4me3 profiles were highly sensitive and specific as demonstrated by selective Dam-methylation on the promoters of genes known to be differentially expressed in mesoderm (Meox2; Fig 3B and 3C) or endoderm (Sox17 locus; Fig 3D). In comparison with ChIP-seq, requiring 50 gastrulae or the tissue from 75 stage 21 embryos [34]), TaDa enabled us to profile H3K4me3 genome-wide in a with high sensitivity in very few embryos.

Fig 3. Cell-type-specific histone mark TaDa in Xenopus embryos.

(A) Schematic of H3K4me3 TaDa during Xenopus embryogenesis after injection of in vitro transcribed mRNA in 1-cell stage embryos. (B–D) H3K4me3 TaDa profiles at the indicated developmental stages on the Meox2 and Sox17 loci. Signal files in bigwig format are available as supplementary at GSE278272, with the following filename prefixes: Stage 11: GSE278272_TAF3_S11_2-vs-Dam_St11_2, Stage 21 ventral: GSE278272_TAF3_S21bot-vs-Dam_S21bot_25, Stage 22 dorsal: GSE278272_TAF3rn_S21top_22-vs-Damrn_S21top_23. All reads files available at GSE278272. All sequencing files are available at GSE278272.

https://doi.org/10.1371/journal.pbio.3002944.g003

pHes5, a 764 bp fragment of the mouse Hes5-promoter including the 5’UTR [24], pTa1, a 1097 bp fragment of the mouse Tuba1a promoter [25,29,30], pNeuroD1-Cre and pCAG-Venus were described before [23].

All mouse husbandry and experiments were carried out in a Home Office-designated facility, according to the UK Home Office guidelines upon approval by the local ethics committee, Animal Welfare and Ethical Review Body (AWERB) (Project Licence PPL70/8727). Experiments were done in wild-type MF1 mice. Timed natural matings were used, where noon of the day of plug-identification was E0.5. IUE was performed as previously described [49,50] at E13.5 with 50 ms, 40 V unipolar pulses (BTX ECM830) using CUY650P5 electrodes (Sonidel). DamID plasmids were injected at 1 µg/µl together with pCAG-Venus at 0.25 µg/µl. Embryos were harvested after 24 h (E14.5) for TaDa with pHes5 and pTa1 or after 72 h (E16.5) for TaDa with pND1-Cre.

Mature Xenopus laevis males and females were obtained from Nasco. Work with X. laevis is covered under the Home Office Project Licence PPL70/8591 after approval by AWERB. Frog husbandry and all experiments were performed according to the relevant regulatory standards, essentially as previously described [34]. For mRNA injection, eggs were in vitro fertilized, dejellied using 2% Cystein solution in 0.1× MMR, pH 7.8, washed three times with 0.1× MMR and transferred into 0.5× MMR for injections. 2–3 pg of mRNA was used per injection. Embryos were cultured at 23 °C and collected at gastrula stage 11 or neurula stage 21. Genomic DNA extraction for DamID library preparation was performed using a Qiagen DNAeasy blood and tissue kit, as per manufacturer instructions.

UAS-mCherry-i2Dam was injected into y,sc,v,nos-phiC31;attP40;+ (Bl#25709) and y,sc,v,nos-phiC31;+;attP2 (Bl#25710) embryos. All other Drosophila constructs were only injected into y,sc,v,nos-phiC31;attP40;+ (Bl#25709) embryos. Male transgenic TaDa flies were crossed to w¹¹¹⁸;Worniu-GAL4 (from [51]);Tub-Gal80^ts (active in NSCs) virgins. Flies were reared in cages at 25 °C, embryos were collected on food plates for 3 hr and transferred to 29 °C for 24 h before dissection at third instar.

For Drosophila DamID, between 20 and 50 larval brains were dissected. For DamID on mouse cortex after IUE, embryos were cooled on ice, and the electroporated cortex was identified with a fluorescent binocular microscope (Fig 2B). Meninges were removed and the electroporated region microdissected. For Xenopus DamID, five injected embryos were pooled for genomic DNA extraction, and one fifth of the genomic DNA was processed for DamID. All samples were processed for DamID as described previously [11]. DamID fragments were prepared for Illumina sequencing according to a modified TruSeq protocol [11]. All sequencing was performed as single end 50 bp reads generated by the Gurdon Institute NGS Core using an Illumina HiSeq 1500.

Analysis of fastq-files from DamID-experiments was performed with the damidseq pipeline script [52] that maps reads to an indexed bowtie2 genome, bins into GATC-fragments according to GATC-sites and normalizes reads against a Dam-only control. Preprocessed fastq-files from mouse, Drosophila and Xenopus were mapped respectively to the mm10 (GRCm38.p6), dm6 (BDGP6) or Xla.v9.1 genome assemblies. Reads were binned into fragments delineated by 5′-GATC-3′ motifs (GATC-bins). Individual replicates (see S1 Fig for n-numbers) for the Dam-fusion constructs were normalized against separate Dam-only replicates with a modified version of the damidseq_pipeline [52] (RPM normalization, 300 bp binsize) and all resulting-binding profiles for one Dam-fusion construct were quantile normalized to each other [52]. The resulting logarithmic profiles in bedgraph format were averaged for all GATC-bins across the genome and subsequently backtransformed (“unlog”). Files were converted to the bigwig file format with bedGraphToBigWig (v4) for visualization with the Integrative Genomics Viewer (IGV) (v2.4.19).

Macs2 (v2.1.2) [53] was used to call broad peaks for every dam-fusion/dam-only pair on the set of *.bam-files generated by the damidseq_pipeline, using Dam-only as control. Peaks were filtered stringently for FDR −2 (mouse) or FDR −5 (Drosophila and Xenopus) and were only considered if present in all pairwise comparisons for a particular experimental condition. Peaks were merged when they were within 1 kb from each other with the merge-function from bedtools (v2.26.0). A similar approach was used for the published ChIP-seq datasets, without control. Associating peaks with the defined genomic features was performed using ChIPseeker (v3.10), with plotAnnoBar function and promoter defined as 2 kb upstream of the transcription start site.

Drosophila NSC list of expressed genes came from [54]. Expressed genes were called from the Pol II DamID-seq bedgraph output using the polii.gene.call R script with default parameters. Mouse RGC gene expression came from [55]. The top or bottom 1,500 genes from the aRG bulk RNA seq dataset (average of 4 replicates) were used for further analysis. Mouse gene coordinates of protein-coding transcripts defined by ensembl were downloaded from biomaRt (v2.38.0). Drosophila gene coordinates of “canonical genes” were obtained from the UCSC Table Browser [56].

Genome browser views were generated using the Integrative Genomics Viewer IGV (v2.4.19) [57] with the midline for TaDa ratio tracks set at 1 and for ChIP-seq set at 0. Peak or gene coordinates were saved in.bed format and supplied as features to Seqplots (v1.12.1) [58]; for S2C Fig quantile normalized, averaged and backtransformed TaDa profiles were provided in bigwig format; for Fig 1F quantile normalized and averaged files in bigwig format were provided.. plotAverage and plotHeatmap were used to visualize and average the binding intensities across all supplied coordinates with bin-size of 50 bp. Figures were assembled in Adobe Illustrator.

In situ hybridization images were from the Eurexpress database [59]. Histone mark ChIP-seq datasets from mouse E14.5 forebrain (ENCFF002AME for H3K4me3; ENCFF002ANR for H3K9ac; ENCFF002AFY for H3K27me3) was downloaded respectively from the ENCODE [28] portal as fastq-files and remapped to the mm10 genome assemblies using the just_align function of the damidseq pipeline [52]. Drosophila NSC RNA PolII-binding data came from GSE77860 [54]. Mouse RGC gene expression came from GSE65000 [55].