Hangover regulates gene expression by limiting NSL-mediated H4K16 acetylation

Abstract

The RNA-binding protein Hangover (Hang) is essential for several stress responses in Drosophila melanogaster. Here, we discover a novel function of Hang in the regulation of gene expression. Hang binds to >2000 genes in the Drosophila genome and modulates transcription. We identify a diverse set of chromatin regulators as Hang interactors, including NSL, dMec, Sin3A, dREAM, and Ino80. Among these, the non-specific lethal complex (NSL) is the most prominent one. We show that Hang attenuates NSL-mediated H4K16 acetylation at transcriptional start sites to downregulate gene expression. Our work uncovers novel roles for Hang in epigenetic gene regulation and suggests that it coordinates the function of multiple chromatin regulators.

Graphical Abstract

Graphical Abstract.

Introduction

Cells have to respond to a variety of stimuli. These include physiological stimuli mediated by signaling molecules or cell-to-cell contacts as well as chemical and physical stresses. An adequate response includes modification of their gene expression profile and is often executed at multiple levels. Well-studied cases of this are the heat shock response as well as the response to changes in nutrient availability via the mammalian target of rapamycin pathway. These signaling cascades not only result in adaptation to the cell’s transcriptome and chromatin landscape but also adjust its mRNA translation efficiency [1–6].

The Drosophila melanogaster multi-zinc finger protein Hangover (Hang) regulates responses to oxidative stress, heat shock, and ethanol exposure. Drosophila feed on fermenting fruit and need to prevent loss of balance and to maintain postural control in high ethanol environments. Wild-type (wt) flies rapidly grow tolerant towards such continual alcohol exposure. As reflected in its name, Hang mutants fail to develop tolerance after repeated ethanol exposure, resulting in impaired movement [7]. Hang is expressed in the nuclei of motor neurons and directs the development of the neuromuscular junction. It regulates transcript levels in adult fly heads to enable an adequate stimulus response [8, 9]. Its role in these pathways has primarily been studied using behavioral readouts and survival analyses of loss-of-function animals. However, little is known about its molecular activity.

The Hang protein contains no less than 18 C₂H₂-type zinc fingers. Using two matrin-type zinc fingers, it binds and stabilizes the mRNA of the cAMP phosphodiesterase dunce [9]. This suggests that one mechanism by which Hang regulates transcript levels is through RNA stabilization. C₂H₂ zinc fingers are prevalent domains among chromatin-regulating proteins, particularly transcription factors, where they typically mediate associations with DNA and transcriptional effectors [10, 11]. Taken together with its nuclear localization and its ability to regulate mRNA levels, it is conceivable that, in addition to its RNA stabilization function, Hang might have an impact on transcription.

Here, we identify the Hang interactome in Drosophila S2 cells by combining genomic epitope-tagging via CRISPR/Cas9 and a systematic proteomics approach. We uncover a large number of chromatin complexes that associate with Hang. Chromatin immunoprecipitation (ChIP) sequencing reveals that Hang binds to chromatin and occupies 5′ regions of actively transcribed genes. We use RNA-seq to show that Hang predominantly represses mRNA expression in S2 cells. These observations position Hang as a transcriptional regulator.

To shed light on the mechanism by which Hang influences transcription, we further explore the interplay with its major interactor, the non-specific lethal (NSL) complex. We compared Hang- and NSL-dependent transcriptomes and found an antagonistic influence of both on gene expression. We investigate the NSL-catalyzed histone mark, acetylated lysine 16 in histone 4 (H4K16ac), in Hang knockout (ko) cells using ChIP sequencing. This identified transcriptional start sites that exhibit increased H4K16ac in the absence of Hang. These are associated with elevated expression and are linked to neuronal function and stimulus responses. We propose that Hang regulates the mRNA output of genes by modulating H4K16ac levels.

Our findings show that Hang not only regulates mRNA stability [9] but also impacts on chromatin modification as well as transcription in cooperation with versatile chromatin regulators.

Materials and methods

Cell culture and triptolide treatment

Drosophila melanogaster Schneider 2 cells with stable Cas9 expression (S2[Cas9] cells, clone 5-3) were kindly provided by Klaus Förstemann [12] and served as parental line for all cell lines generated in this study. Cells were cultivated in Schneider’s Drosophila Medium (2172001, Gibco) supplemented with 10% (v/v) fetal bovine serum (FBS; F7424, Sigma) and 1% (v/v) penicillin–streptomycin (15140122, Gibco). Cells were grown under standard conditions at 26°C.

To attenuate mRNA production, cells were treated with the RNA polymerase inhibitor triptolide (T3652, Sigma). Cells were seeded at a density of 5 × 10⁶ cells/ml and 3 µM triptolide was added. The appropriate amount of DMSO solvent was added to the control conditions and cells were incubated for 3 h at 26°C.

Genomic tagging and knockout using CRISPR/Cas9

Genetic modification of S2 cells was performed as previously described [12, 13]. Endogenous tagging was carried out by inserting GFP- or FLAG-tag sequences at the 3′ end of the hang or MBD-R2 gene locus, resulting in the expression of a C-terminally tagged protein. To generate knockout cell lines, a GFP sequence was inserted directly downstream of the start codon in order to disrupt subsequent transcription of the hang coding sequence.

To that end, Cas9-stable S2 cells (10⁶/ml) were transfected with double-stranded RNA (dsRNA) against key enzymes of NHEJ (Lig4) and MMEJ (mus308) (1 µg/ml each) to favor repair by homologous recombination. After three days, cells (1.5 × 10⁶/ml) were transfected with two polymerase chain reaction (PCR)-generated dsDNA constructs: the first one containing the sgRNA sequence under control of a U6 snRNA promoter and the second one providing the tag and a resistance marker flanked by 60 bp homology. Transfection was carried out using FuGENE HD (E2311, Promega). sgRNA targeting sequences were AGCCTGACTCCAAGTCAG (tagging of Hang isoform A), TAGAAATGCTACTTGCTAG (tagging of MBD-R2), and ACAAAATGTGCGACGCTG (Hang knockout). Homologous recombination repair constructs were amplified from the following plasmids: pMH3 (GFP-tag and Blasticidin resistance marker, Addgene #52 528, tagging of Hang and MBD-R2), pMH4 (2× FLAG-tag and Blasticidin resistance marker, Addgene #52529, tagging of Hang), and pSK23 (GFP sequence and Puromycin resistance marker, Addgene #72851, generation of Hang ko cells).

Four days after transfection, cells were transferred into selection medium [10 µg/ml Blasticidin (A11139, Gibco) or 5 µg/ml Puromycin (540411, Merck)]. Selection antibiotics were kept present until non-resistant control cells declined or at least for 14 days. Monoclonal cell lines were generated by serial dilution, and the insertion was verified by PCR on genomic DNA.

Primers for amplification of tagging constructs and genotyping can be found in Supplementary Table S1.

Chromatin immunoprecipitation

10⁸ S2[Cas9] cells were fixed with 1% (v/v) formaldehyde at room temperature (RT) for 10 min. The reaction was stopped with 240 mM glycine for 10 min at RT. Cells were washed with phosphate buffered saline (PBS) and then lysed in 1 ml ChIP lysis buffer [50 mM Tris/HCl, pH 8.0, 10 mM ethylenediaminetetraacetic acid (EDTA), 1% (w/v) SDS, 1 mM DTT]. After 10 min on ice, chromatin was fragmented using a Bioruptor UCD-200TM-EX (Diagenode) supplied with ice water for 30 min (30 s sonication at high power followed by 30 s cool down). Lysate was cleared by centrifugation at 21 100 × g for 30 min at 4°C and stored at −80°C. Fragment size was evaluated on agarose/TAE gels by decrosslinking 50 µl of lysate overnight at 65°C including RNase A (400 ng/µl; A3832, Applichem) and Proteinase K (400 ng/µl; 7528.1, Roth). One hundred forty microliters of lysate was diluted 1:10 with ChIP IP buffer [16.7 mM Tris/HCl, pH 8.0, 1.2 mM EDTA, 167 mM NaCl, 1.1% (w/v) Triton X-100, 0.01% (w/v) SDS, 1 mM DTT] and pre-cleared for 1 h using 40 µl of Protein A Sepharose (nProtein A Sepharose 4 Fast Flow, 17–5280, GE Healthcare) blocked with 2 mg/ml BSA and 2% (w/v) fish skin gelatin. For ChIP of histone modifications, 14 µl (1%) of HEK cell chromatin was added as a spike-in control.

Antibodies were added to pre-cleared chromatin and incubated for 3 h at 4°C with rotation. Immune complexes were precipitated by adding 45 µl blocked Protein A Sepharose and incubation overnight at 4°C with rotation. GFP-tagged proteins were immunoprecipitated using 25 µl of blocked GFP-Trap Agarose (GTA, Proteintech).

Protein-coupled resin was washed thrice with ChIP low salt buffer [20 mM Tris/HCl, pH 8.0, 2 mM EDTA, 150 mM NaCl, 1% (w/v) Triton X-100, 0.1% (w/v) SDS, 1 mM DTT], thrice with ChIP high salt buffer [20 mM Tris/HCl pH 8.0, 2 mM EDTA, 500 mM NaCl, 1% (w/v) Triton X-100, 0.1% (w/v) SDS, 1 mM DTT], once with LiCl buffer [10 mM Tris/HCl, pH 8.0, 1 mM EDTA, 250 mM LiCl, 0.1% (w/v) NP-40, 1 mM DTT], and twice with TE buffer [10 mM Tris/HCl pH 8.0, 1 mM EDTA]. Washing took place at 4°C for 5 min with rotation, followed by centrifugation (4 min, 400 × g, 4°C).

Precipitated material was eluted in 250 µl ChIP elution buffer [200 mM NaHCO₃, 2% (w/v) SDS] for 45 min at RT with rotation, followed by incubation at 95°C for 15 min. The resin was precipitated by centrifugation for 3 min at 21 100 × g, and the supernatant was diluted 1:2 with nuclease-free water. Fourteen microliters of pre-cleared chromatin was added to 250 µl ChIP elution buffer and 250 µl nuclease-free water as an “input” sample. To separate protein–DNA crosslinks, samples were treated with 40 µM NaCl overnight at 65°C with agitation. Upon addition of 40 mM Tris/HCl (pH 6.8) and 1 mM EDTA, samples were treated with 40 ng/μl Proteinase K (7528.1, Roth) for 1 h at 45°C, and DNA was purified using the QIAquick PCR purification kit (28 106, Qiagen).

The following antibodies were used for precipitation: H3 (ab1791, Abcam), H4K16ac (07-329, Millipore), and mouse IgG (12-371, Millipore).

Precipitated DNA was analyzed by high-throughput sequencing and qPCR.

ChIP sequencing and data analysis

Precipitated DNA from six ChIP reactions was concentrated (Concentrator 5301, Eppendorf) and quantified using the Qubit dsDNA High-Sensitivity Assay Kit (Q32851, Thermo Fisher Scientific). Sequencing libraries were generated from 0.5–2 ng of DNA using the MicroPlex Library Preparation Kit v2 (C05010012, Diagenode) according to the manufacturer’s instructions. Quality of sequencing libraries was controlled on a Bioanalyzer 2100 using the Agilent High Sensitivity DNA Kit (Agilent). Pooled sequencing libraries were sequenced on the NextSeq 550 platform (Illumina) with 50 base single reads.

Single-end reads were aligned to the Drosophila melanogaster genome (dm6) using Bowtie2 (2.5.3, [14]) with default settings. Reads resulting from ChIP experiments using human spike-in chromatin were aligned to a composite genome (dm6 and hg38). Aligned reads were deduplicated using Samtools markdup (1.15.1, [15]) and, in the case of spike-in controlled experiments, separated according to genome origin using Split BAM (Samtools, 2.5.2). Scaling factors were calculated based on the number of human reads in ChIP from control and Hang knockout cells. RPKM-normalized coverage vectors were generated using bamCoverage and bamCompare of Deeptools (3.5.4, [16]) with a bin size of 5. Here, reads were extended to an average fragment size of 500 bp (Hang and H4K16ac ChIP) or 200 bp (Rcd1 and MBD-R2 ChIP). Duplicates were ignored, and spike-in scaling factors were included accordingly. Peaks were called against input control or H3 ChIP with MACS2 callpeak (2.2.9.1, [17]), and medium fragment size was estimated as 500 bp (Hang, H3 and H4K16ac ChIP) or 200 bp (Rcd1 and MBD-R2 ChIP). Peaks that lay less than 500 bp apart were fused using mergeBED.

To define common peaks between two BED files, peak sets were intersected with a default of 1 bp overlap. For Hang ‘high-confidence’ binding sites, intersecting peaks of replicate #1 with replicate #2 were calculated (-wa) and vice versa. Both intersections were combined and merged with a maximum distance of 500 bp on either strand. This way, for a given ‘high-confidence’ peak, the bigger region will be counted. Bedtools version 2.31.1 was used for all manipulations of BED files [18]. Heatmaps were generated using Deeptools (3.5.4) and sorted according to average occupancy [defined as RPKM (ChIP)/RPKM (input)]. Read counts were retrieved with MultiCovBed within Bedtools.

Metagene plots were generated over unique transcriptional start sites (TSSs) retrieved from the UCSC table browser using bamCompare within Deeptools (ChIP over input, ratio of read number, bin size: 5). Peak distribution over genomic elements was determined using CEAS (1.0.0, [19]) within the Cistrome platform with default settings. To that end, Hang peaks were mapped to dm3 using the UCSC liftOver tool. Genomic elements were summarized as follows: promoter [promoter (≤250 bp) + promoter (250–500 bp) + promoter (500–1000 bp)] and intergenic [distal intergenic + distal (≤250 bp) + distal (250–500 bp) + distal (500–1000 bp)]. Remaining definitions were adopted from CEAS.

To identify the top 100 protein-coding genes that display the highest Hang occupancy, genomic locations of all genes were retrieved from flybase.org (FB2024_2), and their position was expanded +500 bp upstream of their TSSs. Reads in these extended gene regions were counted using MultiCovBed within Bedtools, and occupancy was calculated for both ChIP-seq replicates [RPKM(ChIP)/RPKM(input)]. Genes were sorted according to average occupancy, and the top 100 protein-coding genes were compiled into a list for gene set enrichment analysis (GSEA).

Paired ATAC-seq reads were aligned as mentioned above. Mitochondrial reads, duplicates, and pairs with a mapping quality <30 were removed (Filter BAM 2.5.2; MarkDuplicates 3.1.1.0). Bedgraph files were generated using MACS2 (2.2.9.1; set extension size: 200; set shift size: −100) and converted to bigWig files using wigtobigwig (447, [20]).

Differential H4K16 acetylation was determined by comparing levels in Hang wt and ko cells. Global differences in the H4K16ac signal were evaluated by normalizing total H4K16ac and H3 read numbers to the respective numbers of human spike-in reads. Acetylation was then defined as H4K16ac/H3 signal and normalized to Hang wt. TSSs were recognized as H4K16 acetylated if an H4K16ac/H3 peak was found within a 1-kb window. H4K16ac and H3 reads in wt and Hang ko conditions were counted in these 1 kb intervals using MultiCovBed. H4K16ac counts were normalized to H3 counts, and for each TSSs, a difference was calculated between wt and ko (ΔH4K16ac/H3). A frequency distribution of ΔH4K16ac/H3 was generated in Prism (10.0.3) with a bin width of 0.01. Mean (µ = 0.00048) and standard deviation (μ = 0.12631) of ΔH4K16ac/H3 were determined, and only those TSSs were considered as differentially acetylated that lay outside of a µ±2μ interval. Gene Ontology term analysis was performed using Metascape with “Express Analysis” settings (metascape.org; [21]).

The following publicly available datasets were used: H3K4me3 (GSM3106567-GSM3106568, GSM3106577-GSM3106578, [22]), H3K27ac and H3K27me3 (GSM2175510-GSM2175513, GSM2175500-GSM2175503, [23]), ATAC-seq (GSM3381127, [24]), and Rcd1 and MBD-R2 (E-MTAB-1085, [25]).

RNA interference and RNA isolation

Sequences of dsRNA and corresponding primers were designed using E-RNAi (e-rnai.dkfz.de; [26]). dsRNA was generated using the MEGAscript T7 kit according to the manufacturer’s instructions (AMB1334, Invitrogen). Templates for in vitro transcription were synthesized by PCR from S2[Cas9] cDNA using primers containing minimal T7 promoter sequences (Supplementary Table S1). For RNAi S2[Cas9], cells were seeded at 0.1–0.33 × 10⁶ cells/ml and 5 µg/ml dsRNA was added. After four days cells were harvested for RNA or whole cell extract preparation.

RNA was isolated from S2[Cas9] cells using the peqGOLD Total RNA Kit (13–6834-02, VWR Life Science) with the peqGOLD DNase I Digest Kit (13–1091-01, VWR Life Science) according to the manufacturer’s instructions.

Protein extract preparation

For preparation of whole-cell extracts, cells were washed in PBS and subsequently lysed in RIPA buffer [50 mM Tris/HCl pH 8.0, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% (w/v) NP-40, 0.5% (w/v) sodium deoxycholate, 0.1% (w/v) SDS, 10% (v/v) glycerol, 1 mM dithiothreitol (DTT)]. Suspension was incubated for 20 min at 4°C with rotation, followed by freeze/thaw lysis in liquid nitrogen. Cell debris was removed by centrifugation at 21 100 g at 4°C for 30 min.

Nuclear proteins were isolated by a two-step lysis. After washing cells with PBS, cells were lysed in hypotonic buffer B (10 mM Hepes/KOH, pH 7.6, 10 mM KCl, 1.5 mM MgCl₂, 1 mM DTT) for 20 min at 4°C with rotation. Nuclei were isolated by centrifugation at 1900 × g for 20 min at 4°C. The cytoplasmic fraction was discarded. Nuclei were resuspended in hypertonic buffer C [20 mM Hepes/KOH, pH 7.6, 420 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 20% (v/v) glycerol, 1 mM DTT], and proteins were extracted for 30 min at 4°C with rotation. Lysate was cleared by centrifugation at 21 100 × g for 30 min at 4°C.

Protein concentrations were estimated against a BSA standard using a Bradford-based method (Protein Assay, 5 000 006, BioRad) according to the manufacturer’s instructions.

Subcellular fractionation

Fractionation of proteins according to their subcellular localization was performed as previously described [27] with minor changes. 25 × 10⁶ cells were washed in PBS and resuspended in 300 µl buffer 1 [10 mM Hepes/KOH, pH 7.6, 10 mM KCl, 1.5 mM MgCl₂, 340 mM sucrose, 10% (v/v) glycerol, 1 mM DTT]. Five microliters of 10% (w/v) Triton X-100 was added to achieve a final concentration of 0.1% (w/v), and the suspension was incubated for 5 min at 4°C with rotation. After centrifugation (1300 × g, 15 min, 4°C), the supernatant (“cytosolic fraction”) was removed and cleared by centrifugation (21 000 × g, 15 min, 4°C). The remaining pellet was washed twice in buffer 1, subsequently resuspended in 150 µl buffer 2 (3 mM EDTA, 0.2 mM EGTA), and the suspension was incubated for 30 min at 4°C with rotation. After centrifugation (1700 × g, 4 min, 4°C), the supernatant (“nucleoplasmic fraction”) was removed and cleared by centrifugation (21 000 × g, 15 min, 4°C). The remaining pellet was washed twice in buffer 2, subsequently resuspended in 150 µl RIPA buffer, and 100 units of turbonuclease (T4330, Sigma) were added. The suspension was incubated for 20 min at 4°C with rotation and cleared by centrifugation (21 000 × g, 15 min, 4°C). Supernatant was removed (“chromatin fraction”). Protein concentrations of fractions were estimated against a BSA standard using a Bradford-based method (Protein Assay, 5 000 006, BioRad) according to the manufacturer’s instructions. For adequate comparison of protein abundance on western blot, the following amounts of each fraction were loaded: 60 µg whole-cell extract, 30 µg cytosolic fraction, 15 µg nucleoplasmic fraction, and 15 µg chromatin fraction.

Size-exclusion chromatography

Nuclear extracts were diluted to a concentration of 4 mg/ml, 0.15 U/µl Benzonase (70 664, Millipore) was added, and the suspension was incubated at 4°C overnight with rotation. Potential precipitates were removed by centrifugation at 21 000 × g for 20 min. Extracts were loaded onto a Superose^TM 6 HR 10/30 GL column (17-0537-01, GE Healthcare) using an ÄKTA pure chromatography system (29 018 226, GE Healthcare) with a 200-µl injection loop. The column was operated in buffer EX300 [10 mM Hepes/KOH, pH 7.6, 300 mM KCl, 1.5 mM MgCl₂, 0.5 mM EGTA, 10% (v/v) glycerol, 1 mM DTT]. Five hundred microliters of fractions were collected using a Fraction collector F9-R (29 011 362, GE Healthcare). Molecular weight across different fractions was calibrated using the Gel Filtration Cal Kit High Molecular Weight (28-4038-42, GE Healthcare).

To precipitate proteins in each fraction, 5 µl of StrataClean resin (400 714, Agilent Technologies) was added, and proteins were bound to the resin for 30 min at 4°C with rotation. The resin was precipitated by centrifugation at 3500 × g for 5 min at 4°C, and proteins were eluted in 2× SDS loading buffer [100 mM Tris/HCl, pH 6.8, 4% (w/v) SDS, 20% (v/v) glycerol, 0.2% (w/v) bromophenol blue, 200 mM DTT] for 5 min at 95°C.

Co-immunoprecipitation of epitope-tagged proteins

Nuclear extracts were diluted to a final concentration of 1 mg/ml protein, 100 mM NaCl, and 0.1% (w/v) NP-40 with buffers C-0 [20 mM Hepes/KOH, pH 7.6, 1.5 mM MgCl₂, 0.2 mM EDTA, 20% (v/v) glycerol, 0.131% (w/v) NP-40, 1 mM DTT] and C-100 [20 mM Hepes/KOH, pH 7.6, 100 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA, 20% (v/v) glycerol, 0.1% (w/v) NP-40, 1 mM DTT]. 0.03 U/µl Benzonase (70664, Millipore) was added, diluted extracts were incubated at 4°C for 1 h with rotation, and contingent precipitates were removed by centrifugation (20 min, 5000 × g, 4°C). GFP-Trap Agarose (GTA, Proteintech) or ANTI-FLAG M2 Affinity Gel (A2220, Sigma) was washed in buffer C-100 and blocked with 1 mg/ml BSA and 1% (v/v) fish skin gelatin for 1 h at 4°C with rotation. 1–2 mg of diluted extract was added to 25 µl of resin and incubated overnight at 4°C with rotation. After four washes with 1 ml IP150 buffer [25 mM Hepes/KOH, pH 7.6, 150 mM NaCl, 12.5 mM MgCl₂, 0.1 mM EDTA, 10% (v/v) glycerol, 0.1% (w/v) NP-40, 1 mM DTT], proteins were eluted in SDS loading buffer [50 mM Tris/HCl, pH 6.8, 2% (w/v) SDS, 10% (v/v) glycerol, 0.1% (w/v) bromophenol blue, 100 mM DTT] for 5 min at 95°C. Purified proteins were analyzed by SDS–PAGE and western blot.

Mass spectrometry

Immunoprecipitation was performed from 10 mg of nuclear extract. Diluted extracts were added to 200 µl of non-blocked GFP-Trap Agarose (GTA, Proteintech) or ANTI-FLAG M2 Affinity Gel (A2220, Sigma). Washing was performed on a Poly-Prep Chromatography Column (731–1550, BioRad) with 4 × 10 ml IP150 buffer and twice with 100 mM NH₄HCO₃ buffer pH 7.5. Resin was removed from the column, and bound proteins were analyzed by SDS–PAGE and silver staining or digested with trypsin for LC-MS analysis.

For protein quantification via LC-MS, beads were treated with 10 ng/l of trypsin in 1 M urea and 50 mM NH₄HCO₃ for 30 min, then rinsed with 50 mM NH₄HCO₃. Supernatants were digested overnight with 1 mM DTT. Before LC-MS analysis, peptides were alkylated and desalted. Peptides were then injected into a 25-cm analytical column (75 µm ID, 1.6 µm C18, Aurora-IonOpticks) with a 50-minute gradient from 2 to 35% acetonitrile in 0.1% formic acid. The effluent from the HPLC was electrosprayed directly into an Orbitrap Exploris 480 instrument operating in data-dependent mode to automatically transition between full-scan mass spectrometry (MS) and MS/MS acquisition. Typical mass spectrometric conditions were spray voltage, 1.5 kV; no sheath and auxiliary gas flow; heated capillary temperature, 275°C; ion selection threshold, 5 × 10³ counts; and dynamic exclusion, 20 s.

Specifically, Exploris 480 was set to survey full scan MS spectra (m/z 350–1400) acquired with resolution 60 000 at m/z 400 (max IT 40 ms, AGC target of 3 × 10⁶). The 15 most intense peptide ions with charge states between 2 and 6 were sequentially isolated to a target value of 2 × 10⁵, fragmented at 30% normalized collision energy, and acquired with resolution 15 000 at m/z 400 (max IT 40 ms).

MaxQuant 2.0.1.0 was used to identify and quantify proteins using iBAQ with the following parameters: DB: uniprot_AUP000000803_Dmelanogaster_20 210325; MS tol: 10 ppm; MS/MS tol: 20 ppm Da; Peptide FDR: 0.1; Protein FDR: 0.01 min. peptide length: 7; Variable modifications: Oxidation (M); Fixed modifications: Carbamidomethyl (C); Peptides for protein quantitation: razor and unique; Min. peptides: 1; Min. ratio count: 2.

The quantified proteins (MaxQuant iBAQ Z-score normalized values) were compared using the adjusted t-test function from the volcano plot option from Perseus (missing values from the normal distribution replaced), width: 0.3 and downshift: 4, the false discovery rate (FDR): 0.05, and the S0 value: 0.1.

Proteins with a Δlog₂ iBAQ (log₂ iBAQ IP in control cells—log₂ iBAQ IP in cells expressing GFP—or FLAG-tagged Hang) >2 and a P-value <.05 were considered significantly enriched. Area-proportional Venn diagrams were generated using BioVenn [28]. The STRING database (string-db.org, version 12.0) was used to generate a physical subnetwork based on experiments and databases. Proteins were grouped using MCL clustering (inflation parameter: 3). Gene Ontology term analysis was performed using Metascape with “Express Analysis” settings (metascape.org; [21]).

SDS–PAGE, western blot, and silver staining

SDS–polyacrylamide gel electrophoresis was used to separate proteins for subsequent western blotting. Here, proteins were transferred to polyvinylidene difluride (PVDF) membranes (T830.1, Roth) in Pierce Western Blot Transfer Buffer (35040, Thermo Fisher Scientific). Blocking was conducted for 1 h at room temperature in blocking buffer [PBS, 0.1% (w/v) Tween-20, 5% (w/v) non-fat dry milk], and primary antibodies were applied in blocking buffer overnight at 4°C. Membranes were washed four times in washing buffer (PBS, 0.1% Tween-20), and HRP-coupled secondary antibodies were applied in blocking buffer for 2 h at room temperature [anti-mouse IgG (NA931, GE Healthcare), anti-rabbit IgG (NA934, GE Healthcare), and anti-rat IgG (31470, Thermo Fisher Scientific)]. Again, membranes were washed four times in washing buffer, and chemiluminescence was detected using Immobilon Western Blot Chemiluminescence HRP substrate (WBKLS0500, Millipore) on a ChemiDoc^TM Touch Imaging System from Bio-Rad.

The following antibodies and antisera were used for detection: Hang (1:5 000; [7]), GFP (1:5 000; clone [3H9] from Chromotek), FLAG (1:8 000; clone M2 from Sigma), FLAG (1:8 000; F7425 from Sigma), Tubulin beta (1:8 000; clone KMX-1 from Merck Millipore), Lamin Dm0 (1:5 000; clone ADL67.10 from DHSB), Actin (1:1 000; clone JLA20 from DHSB), H3 (1:10 000; ab1791 from Abcam), dMi-2 (1:8 000; [29]), dRPD3 (1:10 000; [29]), Nsl1 (1:5 000; [30]), MBD-R2 (1:5 000; [30]), Mof (1:12 000; [30]), Mle (1:12 000; [30]), and Rcd5 (1:5 000; [31]).

For silver staining of purified proteins, samples were supplemented with NuPAGE^TM LDS sample buffer (NP0007, Invitrogen), loaded on NuPAGE^TM 4-12% Bis-Tris gradient gels (NP0321, Invitrogen), and electrophoresed in NuPAGE^TM MOPS SDS running buffer (NP0001, Invitrogen). Gradient gels were then stained with the SilverQuest Silver Staining Kit (LC6070, Invitrogen) according to the manufacturer’s instructions.

qPCR

One microgram of total RNA was utilized for cDNA preparation using the SensiFAST cDNA Synthesis Kit (BIO-65054, Bioline), and the reaction was subsequently diluted 1:10. Eluted DNA from ChIP experiments was diluted 1:4. Diluted DNA samples were measured by qPCR using the SensiFast SYBR Lo-ROX Kit (BIO-94050, Bioline) according to the manufacturer’s instructions. Each sample was measured in triplicate on a Stratagene Mx3000P thermocycler (Agilent Technologies). Sequences of primers used for amplification and their experimentally determined efficiencies can be found in Supplementary Table S1.

For expression analysis using RT-qPCR, samples were first normalized to the control condition and then scaled according to a normalization factor composed of the geometric mean of three reference genes (Act5C, βTub56D, and RpL32). For that, the following calculations were applied:

1. Relative quantity of treatment condition (t) compared to control (c), with consideration of the amplification efficiency (E):

2. Normalization factor (NF) calculated as the geometric mean of Act5C, βTub56D, and RpL32 values:

3. Relative mRNA expression level exp:

Relative mRNA levels upon triptolide treatment were not normalized to reference genes since their expression is likely also affected upon RNA polymerase inhibition.

Error bars represent the standard deviation of five independent experiments.

For protein occupancy analysis by ChIP-qPCR, ChIP and input DNA (1%) were quantified using site-specific primers (Supplementary Table S1). Enrichment was calculated as “% input”:

Error bars represent the standard deviation of five independent experiments.

RNA sequencing and data analysis

To assess changes to gene expression upon Hang depletion, total RNA upon Hang RNAi was isolated from three independent transfections of each construct. RNA quality was assessed using the Experion RNA StdSens Analysis Kit (BioRad). RNASeq libraries were prepared from total RNA using the TruSeq Stranded mRNA kit (Illumina) according to the manufacturer’s instructions. Quality of sequencing libraries was controlled on a Bioanalyzer 2100 using the Agilent High Sensitivity DNA Kit (Agilent). Pooled sequencing libraries were quantified and sequenced on the NextSeq 550 platform (Illumina) with 50 base single reads.

Read quality was assessed with FastQC (0.74), and adaptor sequences and low-quality bases were removed using Trimmomatic (0.39, [32]). Single-end or paired-end reads were mapped to the D. melanogaster genome (dm6) using RNA Star with 2-pass mapping (2.7.11a, [33]). Reads were counted with featureCounts (default settings, 2.0.3, [34]), using the dmel_r6.56 GTF annotation file from flybase.org. Fold change values and normalized counts were determined using DESeq2 (2.11.40.8, [35]). A gene was defined as “regulated” when its adjusted p-value was lower than 0.05.

To intersect regulated genes with genomic binding sites, gene locations were retrieved from flybase.org (FB2024_2). To include promoter regions, locations were expanded by adding 500 bp upstream of the TSSs. These genomic regions were intersected with Hang binding sites using bedtools (default settings, 2.31.1, [18]), and hypergeometric testing was performed using R (4.5.1). Heatmaps were generated on fold change values that were scaled by row using heatmap2 (3.1.3.1). Gene set enrichment analysis was performed using GSEA (4.3.3, [36, 37]) with custom gene sets.

The following publicly available datasets were used: Nsl1 RNAi (GSM3344618–GSM3344623, [38]), Nsl1 and Nsl3 RNAi (GSM4029652–GSM4029654, GSM4029657–GSM4029659, GSM4029662–GSM4029664, [31]).

Software, analysis platforms, and data availability

The following software was used for analysis and representation of data: Image Lab (version 6.1.0 build 7, Bio-Rad), GraphPad Prism (version 10.0.3), Perseus (version 1.6.15.0, [39]), and R (version 4.5.1). Genome-wide data were analyzed on the following Galaxy web platforms [40]: Galaxy Europe (usegalaxy.eu), Galaxy (usegalaxy.org), and Galaxy/Cistrome (cistrome.org/ap/root).

Results

Hangover interacts with chromatin-regulating proteins

We identified Hang-interacting proteins in order to shed light on its molecular function. To this end, we used a CRISPR/Cas9-based approach to introduce GFP- or FLAG-tag sequences at the 3′ end of the hang gene in S2 cells. This ensures the expression of a C-terminally tagged protein from its endogenous locus. Insertion of the epitope tag sequences was verified by PCR on genomic DNA and western blot (Supplementary Fig. S1A and B). Subcellular fractionation showed that Hang is enriched in a fraction that contains prominent chromatin regulators like dMi-2, dRPD3, and dCTCF (Supplementary Fig. S1C). We performed ⍺-GFP and ⍺-FLAG immunoprecipitation on nuclear extracts from control cells and cells expressing Hang-GFP or Hang-FLAG. SDS–PAGE and silver staining of the precipitated material revealed the coprecipitation of several proteins (Fig. 1A). Upon size-exclusion chromatography, we detected Hang in fractions of high molecular weight, suggesting that it is part of larger protein complexes (Supplementary Fig. S1D). The Hang elution behavior was not altered upon Benzonase digestion, which indicates that interactions with nucleic acids are not critical for Hang to be incorporated into high molecular weight complexes.

Figure 1.

The Hangover interactome. (A) Silver-stained polyacrylamide gel of ⍺-GFP (lanes 1 and 2) and ⍺-FLAG (lanes 3 and 4) immunoprecipitates (IPs) from nuclear extracts of control S2 cells (Hang non-tagged lanes 1 and 3) and cells expressing Hang-GFP (lane 2) and Hang-FLAG (lane 4). Molecular weight marker bands are indicated on the left. Arrowhead points to Hang, and dots to polypeptides that are specifically enriched in Hang-GFP and Hang-FLAG IPs, respectively. Asterisks indicate heavy and light chains of the FLAG antibody. IP-MS/MS results of ⍺-GFP (B) and ⍺-FLAG IP (C) represented as volcano plots. Proteins with a Δlog₂ iBAQ enrichment >2 and P < 0.05 are marked in green (B) and red (C), respectively. The number of these proteins is indicated above. (D) Venn diagram showing the overlap of Hang-GFP and Hang-FLAG interacting proteins. (E) STRING interaction network of Hang high-confidence interactors enriched in both GFP- and FLAG-IP after clustering. Lines between proteins indicate physical interactions. Line density represents the interaction confidence. Prominent chromatin-associated complexes are marked in gray.

We identified Hang-interacting proteins by analyzing Hang-GFP and Hang-FLAG coprecipitates with mass spectrometry (LC-MS/MS). Compared to control IPs from extracts containing non-tagged Hang, 552 proteins were enriched in Hang-GFP precipitates and 153 proteins were enriched in Hang-FLAG precipitates (Fig. 1B and C). In total, 141 proteins were enriched in both conditions, which we refer to as “high-confidence interactors” (Fig. 1D and Supplementary Table S2). These results support the notion that Hang does not operate as a monomer but cooperates with other proteins. In order to assign biological functions to Hang interactors, we employed Gene Ontology (GO) analysis. The most significantly enriched GO terms were associated with chromatin and transcription (“chromatin organization,” “ATP-dependent chromatin remodeling,” “negative regulation of transcription,” and “positive regulation of transcription”) (Supplementary Fig. S1E). In fact, 83 out of the 141 high-confidence interactors (58.9%) were assigned to one or more transcription-related GO terms. This is unexpected since Hang has so far not been implicated in the regulation of transcription or chromatin. Using STRING interaction analysis, we sorted Hang high-confidence interactors into their respective protein complexes (Fig. 1E and Supplementary Fig. S1F and G). This identified several multi-subunit chromatin regulators: Hang copurified with chromatin remodelling complexes [inositol-requiring mutant 80 (Ino80) and Drosophila MEP-1 complex (dMec)] and histone-modifying complexes [NSL and Swi-independent 3 (Sin3)]. It contacts activators (TFIID) as well as repressors [Drosophila Rbf, E2F, and Myb-interacting proteins (dREAM)] of transcription. Moreover, it interacts with a cluster of insulator-binding proteins.

We conclude that Hang might associate with several chromatin-regulating protein complexes. Given the predominance of transcriptional regulators among its high-confidence interactors, we hypothesize that Hang itself binds and regulates chromatin.

Hangover binds to actively transcribed genes

In order to determine whether Hang associates with chromatin, we performed ChIP of Hang in Hang-GFP-expressing cells. We analyzed two replicates by high-throughput sequencing and identified >3000 binding sites in the genome of S2 cells that were shared between both replicates (Fig. 2A). We determined the genomic distribution of Hang binding sites. Hang signal was most strongly enriched at the 5′ end of genes, with its highest binding intensity around 150 base pairs downstream of transcriptional start sites (TSS) (Fig. 2B and C). To determine if Hang associates with actively transcribed and/or repressed genes, we investigated the chromatin environment at Hang-bound regions. Hang colocalizes with histone modifications associated with active transcription (H3K4me3 and H3K27ac). By contrast, its binding sites are mostly devoid of H3K27me3, a mark of polycomb-mediated gene repression. Hang-bound regions also show high ATAC-seq signal, indicating increased chromatin accessibility (Fig. 2D). To quantify their transcriptional status, expression levels of Hang-occupied genes were determined. Genes underlying Hang binding sites exhibited significantly higher mRNA levels in comparison to all genes expressed in S2 cells (Fig. 2E). To test whether gene activity and the magnitude of Hang ChIP-seq signal are correlated, we divided all genes into quartiles based on their expression level (Supplementary Fig. S2A). We then examined Hang occupancy over the TSS of genes in each quartile for each replicate (Supplementary Fig. S2B). Indeed, we find that genes with higher expression also display higher average Hang binding. Collectively, these observations demonstrate that Hang binds to actively transcribed genes.

Figure 2.

Hangover is a chromatin-binding protein. (A) Heatmap of Hang-GFP and input ChIP-seq signals within a 10-kb interval of Hang-GFP binding sites. Results of two Hang-GFP ChIP-seq replicates in S2 cells are shown. (B) Metagene analysis of Hang-GFP ChIP-seq signal at TSSs. Normalized Hang signal was evaluated within a 4-kb interval surrounding all genomic TSSs. The two graphs represent two Hang-GFP ChIP-seq replicates (light blue: replicate #1, dark blue: replicate #2). (C) Distribution of Hang-GFP binding sites over genomic regions compared to the distribution of these regions in the genome. The enrichment fraction of each genomic region is indicated. The most highly enriched region (5′ UTR) is marked in blue. (D) Heatmap of Hang-GFP ChIP-seq, histone modification ChIP-seq, and ATAC-seq signals within a 10-kb interval of Hang binding sites. (E) Average transcript levels (average RPKM) of all genes expressed in S2 cells (gray) and genes bound by Hang (blue). Their distribution is shown as violin plots. Dashed lines represent the median and quartiles. Distributions were compared using two-tailed Mann–Whitney statistics.

Given that Hang binds to actively transcribed genes, Hang binding peaks downstream of the TSSs, and Hang has previously been shown to bind mRNA in vitro [9], it is tempting to speculate that an interaction with nascent RNA contributes to Hang chromatin association. To test this hypothesis, we suppressed nascent RNA synthesis by treating Hang-GFP expressing cells with the RNA polymerase inhibitor triptolide followed by Hang ChIP at genes that were identified as Hang targets in our ChIP-seq analysis. Triptolide treatment did not significantly alter Hang-GFP levels (Supplementary Fig. S2C) but reduced mRNA expression at five of the six genes tested (Supplementary Fig. S2D). Hang occupancy in Hang-GFP-expressing cells could be verified at the six Hang target genes as well as at an intergenic Hang binding site (Supplementary Fig. S2E). By contrast no significant ChIP signal could be detected at these loci in cells not expressing Hang-GFP. Upon triptolide treatment, ChIP signals were not decreased in Hang-GFP expressing cells. This result does not support the hypothesis that RNA binding makes a significant contribution to Hang chromatin association.

Hangover represses transcription

Its association with actively transcribed genes and its interaction with transcriptional regulators strongly suggest that Hang affects gene expression. In order to identify Hang-dependent transcripts, we performed RNA sequencing after depletion of Hang from S2 cells using RNAi. We used two different dsRNA sequences and confirmed effective depletion by western blot (Supplementary Fig. S3A). RNA from cells transfected with control dsRNA (dsEGFP) and from cells transfected with dsRNA targeting Hang (dsHang#1 and dsHang#2) was analyzed by high-throughput sequencing. Depletion of Hang with the first construct resulted in the upregulation of 356 genes, whereas 149 showed lower expression (Supplementary Fig. S3B). Depletion with the second construct led to significant derepression of 442 genes and decreased the levels of 243 transcripts (Supplementary Fig. S3C). Both data sets showed a high degree of correlation of the significantly differentially expressed genes (Fig. 3A, R² = 0.9414). A total of 260 genes responded with an increased expression upon Hang depletion, whereas 65 genes were downregulated in both experiments (Fig. 3A and Supplementary Table S3). We examined Hang occupancy at Hang-dependent genes. 41.9% of genes (109 out of 260) that showed increased expression levels and 32.3% of genes (21 out of 65) that showed lower expression levels upon Hang depletion also contained a Hang binding site (Fig. 3B). We sought to test whether genes with more pronounced Hang binding are also more likely to be regulated by Hang. To this end, we compiled a list of the top 100 protein-coding genes exhibiting the highest Hang ChIP-seq signals. We performed a gene set enrichment analysis (GSEA) of these genes utilizing our RNA-seq data (Fig. 3C). We find a strong correlation of Hang binding and gene repression.

Figure 3.

Hangover regulates transcription. (A) Correlation analysis of fold gene expression changes upon depletion with two different dsRNA constructs used to target Hang. Colors represent genes whose fold change (FC) is associated with a P-value <0.05 in both (blue), only one (light blue), or neither condition (gray). FC values are calculated across three independent transfections (n = 3). Numbers of genes with P < 0.05 in both conditions are indicated in the first and third quadrants. Linear regression of genes with P < 0.05 in both conditions is indicated in the upper left. (B) Pie chart depicting Hang binding to Hang-regulated genes. Charts are separated by genes whose expression is repressed by (left) or dependent on (right) Hang. Blue sections represent genes that contain one or more Hang binding sites. P-values of hypergeometric testing are indicated. (C) GSEA of the top 100 protein-coding genes with the highest Hang occupancy at their extended locus. GSEA was performed on datasets derived from two different RNAi constructs. Black lines represent genes from the respective gene sets and their position within the list of genes ranked according to differential expression. Normalized enrichment scores (NES) and P-values are indicated.

Collectively, these results suggest that Hang binds and directly regulates a significant fraction of differentially expressed genes. Moreover, they reinforce the view that Hang acts predominantly as a repressor. Furthermore, we propose that Hang employs the activity of its interacting chromatin regulators to modulate transcription.

Hangover and NSL have opposite effects on gene expression

Hang binds to chromatin, it interacts with transcriptional regulators and impacts gene expression. Based on these observations, we hypothesized that Hang contributes to shaping the chromatin environment at actively transcribed genes. To elucidate how Hang modulates transcription, we focused on the NSL complex since NSL subunits were among the most highly enriched proteins in our Hang interactome analysis (Fig. 4A and B and Supplementary Table S2). We confirmed the association of Hang with NSL by coimmunoprecipitation followed by western blot. The presence of the NSL subunits Nsl1, MBD-R2, Rcd5, and Mof was observed in both Hang-GFP and Hang-FLAG precipitates (Fig. 4C). Next, we generated an S2 line expressing a GFP-tagged version of the NSL complex subunit MBD-R2 (Supplementary Fig. S4A and B). Anti-GFP immunoprecipitation from nuclear extracts of this cell line not only co-precipitated Nsl1 but also robust amounts of Hang (Fig. 4D). Additionally, the bulk of the Hang protein coeluted with NSL complex components in size-exclusion chromatography (Supplementary Fig. S4C). These findings confirm that NSL is the most prominent Hang interactor in S2 cells.

Figure 4.

Hangover and NSL have opposite effects on gene expression. (A, B) ⍺-GFP (A) and ⍺-FLAG (B) IP-MS/MS results represented as volcano plots. Proteins that surpassed Δlog₂ iBAQ and p-value cutoffs (Δlog₂ iBAQ > 2, P < 0.05; represented by dotted lines) are considered significantly enriched. The number of these proteins is indicated above. Hang protein isoforms are marked in blue, NSL complex subunits in green, and male-specific lethal (MSL) subunits in purple. (C) Coimmunoprecipitation of NSL complex subunits with Hang-GFP (left panels) and Hang-FLAG (right panels). ⍺-GFP and ⍺-FLAG IP were performed from nuclear extracts of control cells (Hang non-tagged) and cells expressing Hang-GFP or Hang-FLAG, respectively. Immunoprecipitates were verified using GFP, FLAG, and Hang antibodies and probed with antibodies against NSL and MSL subunits as indicated on the right. Molecular weight markers are indicated on the left. (D) Coimmunoprecipitation of Hang with MBD-R2-GFP. ⍺-GFP IP was performed from nuclear extracts of control cells (MBD-R2 non-tagged) and cells expressing MBD-R2-GFP. Immunoprecipitates were verified using GFP and MBD-R2 antibodies and probed with antibodies against Nsl1, Mle, and Hang. Molecular weight markers are indicated on the left. (E) Heatmap of ChIP-seq signals at Hang and NSL binding sites. Signals of both Hang-GFP ChIP-seq replicates (blue) and the NSL complex subunits Rcd1 and MBD-R2 (green) are plotted in a 10-kb window surrounding peak centers. Heatmap is divided according to co-occupied (top), Hang only (middle), and NSL only (bottom) regions. (F) Heatmap depicting fold changes of Hang-regulated genes upon depletion of Hang (lanes 1 and 2) and the NSL complex subunits Nsl1 and Rcd1 (lanes 3–5). Top panel depicts 260 genes with increased expression; bottom panel depicts 65 genes with decreased expression upon Hang depletion. Values are scaled by row and sorted according to FC upon Hang RNAi.

NSL is a ubiquitous chromatin regulator that contains the histone acetyltransferase Mof. By acetylating histone 4 at lysine 16 (H4K16), it activates gene transcription [41]. The catalytic subunit of NSL, Mof, is also found in a second protein assembly, the male-specific lethal (MSL) complex, which is involved in dosage compensation [42]. To test whether Hang might also interact with MSL, we probed Hang-GFP and Hang-FLAG precipitates for the occurrence of the MSL subunit Mle. We did not detect Mle as an interactor of Hang (Fig. 4C). Additionally, the MSL subunits Msl-1, Msl-2, and Msl-3 were not part of Hang high-confidence interactors (Fig. 4A and B). This indicates that Hang specifically contacts NSL and not MSL.

Given that Hang and NSL interact in nuclear extracts, we asked whether they also colocalize on chromatin. We compared Hang binding sites to published ChIP sequencing data of the NSL subunits Rcd1 and MBD-R2 [25]. 46.4% of Hang binding sites (1432 out of 3085) colocalized with both Rcd1 and MBD-R2, whereas 53.6% (1653 out of 3085) did not show substantial enrichment of NSL (Fig. 4E). Moreover, there was a large number of additional NSL binding sites (7084, 83.2% of all NSL-occupied regions) lacking significant Hang signal. This is not surprising since NSL is an abundant complex that regulates the expression of thousands of genes (Supplementary Fig. S4D-S4F). We conclude that NSL and Hang not only interact in the soluble nuclear fraction but also share binding sites on chromatin.

Our RNA sequencing analysis suggests that Hang predominantly represses gene expression (Fig. 3). However, the NSL complex is an activator of transcription. To investigate the apparently opposing effects on transcription, we elucidated the influence of NSL on the expression of Hang-regulated genes. To this end we analyzed RNA sequencing data upon depletion of Nsl1 and Rcd1 (Supplementary Fig. S4D–F; [31, 38]). Gene expression changes were investigated over the set of genes that responded to Hang depletion. Intriguingly, genes that were repressed by Hang required NSL for their activation. On the contrary, genes downregulated upon Hang depletion tended to respond with higher expression to the loss of NSL (Fig. 4C). This opposite influence on genes was most strikingly observed at Hang-repressed genes upon depletion of Nsl1 (Supplementary Fig. S4G). We conclude that Hang and NSL indeed have opposite effects on gene expression.

Loss of Hangover causes local changes in H4K16ac

We hypothesized that Hang might antagonize NSL function. To test this hypothesis, we used CRISPR/Cas9 to generate Hang knock-out cells (Hang ko) by inserting a GFP expression cassette in the hang locus of Hang-FLAG expressing cells (Hang wt). The GFP cassette was inserted into codon 2 of the hang gene, thereby disrupting its open reading frame. This setup allowed us to monitor remaining Hang expression by western blot against the FLAG epitope. Insertion was confirmed by PCR on genomic DNA (Supplementary Fig. S5A). Hang ko cells showed no remaining Hang-FLAG signal in western blot but were positive for GFP expression (Supplementary Fig. S5B) confirming absence of Hang in these cells.

To explore the effects of Hang on Mof-catalyzed histone acetylation, we examined the genomic distribution of H4K16ac using ChIP sequencing in Hang wt and ko cells. We quantified the spike-in normalized genome-wide H4K16ac ChIP-seq signal (Supplementary Fig. S5C). We did not observe global changes in H4K16ac upon Hang knock-out. This is not unexpected since, in male S2 cells, the majority of H4K16ac is deposited by the MSL complex and resides on the X chromosome [42]. To investigate locus-specific effects we focused our analysis on the TSSs of genes. We identified 192 TSSs that gained H4K16 acetylation upon Hang ko and 237 TSSs that showed decreased H4K16ac (Fig. 5A and B and Supplementary Fig. S5D). These changes were especially pronounced on TSSs that showed low H4K16ac levels in wild-type cells. It is noteworthy that loss of Hang changes H4K16 acetylation on a narrow subset of TSSs. This corresponds well to the limited number of genes that are deregulated upon depletion of Hang compared to the thousands of genes that are controlled by NSL (Fig. 3 and Supplementary Fig. S4D–F). Importantly, TSSs on the X chromosome were affected to a lower extent than TSSs on other chromosomes, further supporting the notion that altered H4K16ac levels in Hang ko cells are likely attributed to NSL rather than to MSL function (Supplementary Fig. S5E).

Figure 5.

Hangover modulates H4K16 acetylation. (A) Heatmap depicting H4K16ac ChIP-seq signal in a 6 kb interval surrounding TSSs in wt (left panel) and Hang ko cells (middle panel). Difference between wt and ko signals is represented in the right panels; red denotes increased and blue decreased H4K16ac ChIP signal. Most highly affected TSSs are shown on top (192 TSSs with increased H4K16ac in Hang ko) and bottom (237 TSSs with decreased H4K16ac in Hang ko). (B) Genome browser snapshots of exemplary regions containing differentially H4K16 acetylated TSSs upon Hang ko. H4K16ac signal is shown in Hang wt (top) and ko cells (middle), as well as difference between both cell lines (bottom, red: increase, blue: decrease). (C) GSEA of RNA-seq data from Hang depleted cells (top panel: RNAi with construct #1, bottom panel: RNAi with construct #2). Tested gene sets were comprised of genes with increased (left) and decreased (right) H4K16ac at their TSSs in Hang ko cells. Black lines represent genes from the respective gene sets and their position within the list of genes ranked according to differential expression. NES and P-values are indicated. (D) Gene Ontology term analysis of genes with decreased H4K16ac signal at their TSSs in Hang ko cells. GO terms associated with stimulus response and intracellular signaling are highlighted in yellow. GO terms associated with neural development and function are highlighted in orange. (E) RTqPCR analysis of selected genes with increased H4K16ac at their TSSs in Hang ko cells. Expression was evaluated following RNAi in wt and Hang ko cells upon control transfection (dsLuc) or depletion of NSL complex subunits (dsNsl1, dsMBD-R2, dsRcd5). Values are expressed as FC over control transfection in wt cells and are normalized to expression of Act5C, βTub56D, and RpL32. Error bars represent standard deviation from independent transfections (n = 5), which are depicted as dots. Distributions were compared using unpaired, Welch-corrected T statistics, and P-values are indicated above.

In order to examine whether the acetylation change upon Hang loss is predictive of an expression change of the underlying gene, we used GSEA on RNA sequencing data upon Hang RNAi. We compiled sets of genes that exhibited higher or lower H4K16ac levels at their TSSs in Hang ko cells. Genes with an increase in H4K16ac showed significantly higher expression levels upon Hang depletion (Fig. 5C). This suggests that Hang-mediated repression is, in part, realized by restricting acetylation of H4K16. The expression levels of genes that exhibited decreased H4K16ac at their TSSs were not significantly affected upon Hang RNAi, indicating that here lower H4K16ac does not cause detectable transcriptional changes.

We used Gene Ontology term analysis to gain insight into the function of those genes that gained H4K16ac in Hang ko cells. Remarkably, we recovered GO terms associated with neural development and function (“locomotory behavior,” “neuron differentiation,” and “nervous system process”), as well as gene groups linked to stimulus responses and intracellular signaling (“response to external stimulus,” “detection of light stimulus,” “signaling by receptor tyrosine kinases,” “cellular response to chemical stimulus,” and “intracellular signal transduction”) (Fig. 5D). This indicates that Hang particularly limits acetylation of gene promoters linked to response to environmental signals and neuronal function.

The observations above evoke a scenario in which Hang restricts the activity of NSL and, in this way, prevents transcriptional activation. We chose four examples of genes with increased H4K16ac in Hang ko cells to test this model: CG10654, Scgβ, Galt, and Gabat (Supplementary Fig. S5D). Three of these genes are highly expressed in adult brains (FlyAtlas2; [43, 44]) and the central nervous system of larvae (modENCODE): the zinc finger transcription factor CG10654, the sarcoglycan complex component Scgβ, and the GABA transaminase Gabat. Using RNAi, we depleted the NSL complex subunits Nsl1, MBD-R2, and Rcd5 in Hang wt and ko cells, followed by RTqPCR (Fig. 5E and Supplementary Fig. S5F). All four genes are induced upon loss of Hang. In wild-type cells depletion of NSL components had little or no effect on their expression. In Hang ko cells, however, the increased mRNA levels caused by Hang loss were significantly impaired upon NSL RNAi, resulting in decreased expression (Fig. 5E). This demonstrates that activation of transcription upon Hang depletion depends on the NSL complex.

We conclude that Hang represses transcription by limiting NSL-mediated H4K16 acetylation at a number of TSSs associated with neural development and stimulus response.

Discussion

Hangover is a transcriptional regulator

We have shown that the multi-zinc finger protein Hang binds to chromatin and is predominantly localized at the 5′ end of genes. This region harbors regulatory sequences crucial for gene transcription, such as promoters and the TSSs, as well as RNA polymerase II pausing sites. More than 300 genes are deregulated upon Hang depletion, and a substantial number of their gene loci are occupied by Hang. We therefore propose that Hang regulates gene expression, in part, by directly influencing transcription. This is further supported by the fact that Hang knockout results in local changes to H4K16ac levels, a histone mark linked to transcriptional activation.

We have shown that Hang occupies active genes; however, it predominantly represses transcription. Moreover, it binds to both activating and repressive chromatin-regulating proteins. Therefore, we consider it likely that Hang is required to modulate the mRNA output of active genes rather than establishing silent heterochromatin.

Hang directly binds and stabilizes mRNA in neuronal tissue [9]. The novel chromatin functions of Hang identified here represent a second tier of regulating mRNA levels. Thus, Hang adds to the growing number of transcription factors that are capable of binding to both chromatin and RNA [45].

Hangover binds to multiple chromatin regulators

The majority of Hang interactors are regulators of chromatin and transcription. Our list of interactors also includes the Polycomb group factor Sfmbt, which has previously been shown to bind to Hang [46]. Prominent Hang-associated complexes include dMec and Ino80, which contain ATPases responsible for nucleosome remodelling [47, 48]. The Sin3 complex contains the histone deacetylase RPD3 and is linked to transcriptional repression [49]. The NSL complex, on the other hand, activates gene expression by acetylating histones with its catalytic subunit Mof [41]. We recovered almost all subunits of the general transcription factor TFIID, suggesting a broader role for Hang in transcription initiation.

Its chromatin association and its interaction with effectors of gene expression evoke a scenario in which Hang recruits these factors to their sites of action. The simultaneous interaction with multiple regulators might allow Hang to direct combinations of their activities to certain gene loci. In doing so, Hang might serve as a hub or scaffold onto which chromatin-regulating complexes can assemble [50, 51].

Additionally, Hang interacts with insulator-binding proteins (e.g. CTCF, Cp190, and Ibf1/2) indicating that it might partake in chromatin insulation or even spatial genome organization [52, 53].

Hangover counteracts NSL-mediated H4K16 acetylation

The most prominent binding partner of Hang in S2 cells is the NSL complex. We identified a group of genes that are repressed by Hang and activated by NSL. We therefore consider a recruitment model in which Hang tethers NSL to their shared binding sites unlikely. Rather, Hang antagonizes NSL function. Here, we can envision two mechanisms: First, Hang might counteract NSL binding to a subset of NSL-bound TSSs; second, Hang might modulate the catalytic activity of NSL.

The NSL complex is classically known to maintain expression of housekeeping genes [25]. However, it has also been linked to stress-induced transcription [54]. Since Hang is required for appropriate stress responses in adult flies, we propose that this function partly relies on modulating NSL activity. This hypothesis is supported by the fact that genes with elevated H4K16ac levels at their TSSs upon Hang knockout are associated with stimulus response and intracellular signal transduction. Thus, it is plausible that phenotypic alterations in Hang loss-of-function animals (i.e. impaired stress response, neuronal malformations) might in part be caused by altered H4K16ac levels.

Concluding remarks

Environmental stimuli and stresses demand a complex genomic response. Multi-zinc finger proteins seem particularly suited to coordinate this task, since they have the potential to interact with RNA, chromatin, and transcriptional regulators [45, 55]. The Drosophila transcription factor Hangover is a striking example of this: it directly interacts with RNA [9], binds to thousands of sites on chromatin, and contacts many cofactors with diverse chromatin-related activities. How these different binding activities are integrated in response to external stimuli is still an open question.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

J.L. and A.B. were funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—TRR81 A01 and BR2102/8. Funding to pay the Open Access publication charges for this article was provided by the Open Access Publishing Fund of Philipps-Universität Marburg with support of the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation).

Data availability

Data generated in this study are available from the following repositories: European Nucleotide Archive (ebi.ac.uk/ena/browser/home, accession number: PRJEB88170) and ProteomeXchange Consortium via the PRIDE partner repository ([56, 57]; https://www.ebi.ac.uk/pride/, accession number: PXD062946).