Factor cooperation for chromosome discrimination in Drosophila

Cell culture and RNAi

S2-DRSC (DGRC stock # 181), Kc167 (DGRC stock # 1) and L2-4 (S2 subclone, provided by P. Heun) cells were cultured in Schneider's Drosophila Medium (Thermo Fisher), supplemented with 10% heat-inactivated fetal bovine serum (Sigma-Aldrich), 100 units/ml penicillin and 0.1 mg/ml streptomycin (Sigma-Aldrich) at 26°C. RNAi against target genes in S2, L2-4 and Kc cells was performed as previously described (22). In brief, double-stranded RNA fragments (dsRNA) were generated with MEGAscript kit (Thermo Fisher) from polymerase chain reaction (PCR) products obtained using the following forward and reverse primers (separated by comma):

• clamp RNAi #1: TAATACGACTCACTATAGGGACGTCCAAACCCTTCAGTTGT, TAATACGACTCACTATAGGGATTGAGTGCAAAACGATCAGC;

• clamp RNAi #2 (DRSC29935): TTAATACGACTCACTATAGGGAGAGAAGACCTTACCAAAAACAT, TTAATACGACTCACTATAGGGAGAGCTTATGTTGGATATGGTGT;

• trl RNAi: TTAATACGACTCACTATAGGGAGAATGTCGCTGCCA, TTAATACGACTCACTATAGGGAGATTGCCTGGA;

• gst RNAi: TTAATACGACTCACTATAGGGAGAATGTCCCCTATACTAGGTTA, TTAATACGACTCACTATAGGGAGAACGCATCCAGGCACATTG; The sequence of Schistosoma japonicum GST was amplified from pGEX-6P-1 (GE Healthcare).

• gfp RNAi: TTAATACGACTCACTATAGGGTGCTCAGGTAGTGGTTGTCG, TTAATACGACTCACTATAGGGCCTGAAGTTCATCTGCACCA;

• msl2 RNAi: TTAATACGACTCACTATAGGGAGAATGGCCCAGACGGCATAC, TTAATACGACTCACTATAGGGAGACAGCGATGTGGGCATGTC.

Cells were washed with serum-free medium and 10 μg dsRNA per 10⁶ cells at a concentration of 10 μg/ml in serum-free medium (10⁶ cells in 6-well plate for ATAC-seq and immunofluorescence microscopy) or 4.2 μg dsRNA per 10⁶ cells at 6.3 μg/ml serum free medium (12 * 10⁶ cells in 75 cm² flask for ChIP-seq) was added, incubated for 10 min at room temperature (RT) with slight agitation and further 50 min at 26°C. Two volumes of complete growth medium were added and cells were incubated for 5 days at 26°C. Then, 1× of initial volume (6-well plate) or 0.75× of initial volume (75 cm² flask) of growth medium was added and cells were incubated for further 2 days at 26°C.

Sf21 cells (Thermo Fischer) were cultured in SF900 II SFM (Thermo Fisher), supplemented with 10% heat-inactivated fetal bovine serum (Sigma-Aldrich), 0.1 mg/ml gentamicin (Sigma-Aldrich) at 26°C.

Immunofluorescence microscopy

For immunofluorescence microscopy (IFM), 0.5 * 10⁶ L2-4 cells after RNAi treatment in 200–500 μl growth medium were seeded into each well of a 3-well (14 mm) object slide (Thermo Fisher) and incubated for 2 h at 26°C. Immunofluorescence staining was performed as described in (40) with slight modifications. Briefly, cells were washed in phosphate-buffered saline (PBS), fixed for 7.5 min with 2% (v/v) formaldehyde (FA) in PBS on ice and permeabilized for 7.5 min with 1% (v/v) FA in PBS with 0.25% (v/v) Triton-X 100 on ice. Cells were washed twice in PBS and blocked for 1 h with 200 μl Blocking Buffer (3% (w/v) bovine serum albumin (BSA), 5% (v/v) heat-inactivated fetal bovine serum (Sigma-Aldrich) in PBS) at RT in a humid chamber. Cells were incubated for 1 h with primary antibody diluted in 30 μl Blocking Buffer with 0.1% (v/v) Triton-X 100 at RT in a humid chamber. Cells were washed twice for 5 min in 200 μl PBS and incubated for 1 h with suitable secondary antibody diluted in 30 μl Blocking Buffer with 0.1% (v/v) Triton-X 100 at RT in a humid chamber. Cells were washed twice for 5 min in 200 μl PBS, stained for 2 min with 0.125 μg/ml DAPI in 200 μl PBS at RT and washed twice for 5 min in 200 μl PBS. Cells were mounted with 8 μl Vectashield (Vector Laboratories) and a coverslip was sealed to object slide with nail polish. Images were acquired with Axiovert 200 epifluorescence microscope (Zeiss) equipped with AxioCamMR CCD Camera (Zeiss).

Protein purification

Recombinant MSL2-FLAG was expressed in Sf21 cells and purified by FLAG-tag affinity chromatography, as previously described in (23) with minor modification. In brief, Sf21 cells at 10⁶ cells/ml (250 * 10⁶ cells) were infected 1:1000 (v/v) with baculovirus, expressing MSL2-FLAG. After 72 h, cells were harvested and washed once in PBS, frozen in liquid nitrogen and stored at −80°C. For lysis, cells were rapidly thawed, resuspended in 1 ml Lysis Buffer per 10 ml of culture (50 mM HEPES pH 7.6, 300 mM KCl, 1 mM MgCl₂, 5% (v/v) glycerol, 0.05% (v/v) IGEPAL CA-360, 50 μM ZnCl₂) supplemented with 0.5 mM TCEP and cOmplete ethylenediaminetetraacetic acid (EDTA)-free Protease Inhibitor Cocktail (Sigma-Aldrich) (PI). The suspension was passed thrice through Microfluidizer LM10 (Microfluidics). Cell extract was adjusted with Lysis Buffer containing 0.5 mM TCEP and PI to 2 ml per 10 ml of culture and incubated with end-over-end rotation for 1 h at 4°C. Cell debris were spun down at 4°C for 30 min at 50 000 g. The resulting supernatant was used for FLAG-tag affinity purification with 4 μl of a 50% slurry of FLAG-M2 beads (Sigma-Aldrich) per 1 ml of culture. Beads were first washed thrice in 20 bed volumes of Lysis Buffer, subsequently supernatant was added and incubated with end-over-end rotation for 3 h at 4°C. Beads were pelleted (4°C, 5 min, 500 g) and supernatant was removed. Beads were washed once with 20 bed volumes each of Lysis Buffer, Wash Buffer (Lysis Buffer with 1000 mM KCl), Lysis Buffer again, and finally twice with 20 bed volumes Elution Buffer (Lysis Buffer with 100 mM KCl). For protein elution, beads were incubated with 0.2 bed volumes of Elution Buffer containing 5 mg/ml FLAG peptide for 10 min at 4°C and subsequently 1.8 bed volumes of Elution Buffer with 0.5 mM TCEP and PI were added and incubated with end-over-end rotation for 30 min at 4°C. The elution step was repeated once. Elution fractions were combined, remaining beads were removed by passing through Corning Costar Spin-X centrifuge tube filters (Sigma-Aldrich) and the eluate was concentrated with 10 MWCO Amicon Ultra 0.5 ml (Merck). Protein concentration was determined using BSA standards on sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) with Coomassie brilliant blue G250 staining. To store at −80°C until use, concentrated protein was aliquoted and flash-frozen in liquid nitrogen.

For purification of recombinant CLAMP by FLAG-tag affinity chromatography, the coding sequence of CLAMP (CG1832-RA) was fused to a C-terminal coding sequence of FLAG affinity tag and cloned into pFastBac1, using the following forward and reverse primer: AAGGATCCATGGAAGACCTTACCAAAAAC, CCTTTCTCGAGTTACTTGTCATCGTCGTCCTTGTAGTCTTCCCCGTCTGTATGCATCCG.

CLAMP-FLAG was expressed in Sf21 cells and purified by FLAG-tag affinity chromatography, like MSL2-FLAG, but with the following modifications. Lysed cells were resuspended in 1 ml Buffer C per 10 ml of culture (50 mM HEPES pH 7.6, 1 M KCl, 1 mM MgCl₂, 5% (v/v) Glycerol, 0.05% (v/v) IGEPAL CA-360, 50 μM ZnCl₂, 375 mM L-Arginine (according to (41)) supplemented with 0.5 mM TCEP and cOmplete EDTA-free Protease Inhibitor Cocktail (Sigma-Aldrich) (PI) and passed thrice through Microfluidizer LM10 (Microfluidics). Subsequently, the extract was adjusted with Buffer C containing PI to 2 ml per 10 ml of culture and supplemented with 0.1% (v/v) polyethylenimine by adding 2% (v/v) polyethylenimine (neutralized with HCl to pH 7.0) drop-by-drop while string in an ice bath, according to (42). Cell debris were spun down at 4°C for 30 min at 50 000 g. To the resulting supernatant 20 U Benzonase was added per 10 ml of culture and incubated with end-over-end rotation for 1 h at 4°C. The resulting extract was used for FLAG-tag affinity purification with 4 μl of a 50% slurry of FLAG-M2 beads (Sigma-Aldrich) per 1 ml of culture. Beads were washed thrice in 20 bed volumes of Buffer C, subsequently supernatant was added and incubated with end-over-end rotation for 3 h at 4°C. Beads were pelleted at 4°C for 5 min at 500 g and supernatant was removed. Beads were washed five times with 20 bed volumes of Buffer C. For protein elution, beads were incubated with 0.2 bed volumes of Buffer C containing 5 mg/ml FLAG peptide for 10 min at 4°C and subsequently 1.8 bed volumes of Buffer C containing 0.5 mM TCEP and PI were added and incubated with end-over-end rotation for 30 min at 4°C. The elution step was repeated once. The combined elution fractions were processed on flash-frozen in liquid nitrogen as for MSL2-FLAG.

Genomic DNA preparation

For genomic DNA (gDNA) extraction, 5 * 10⁷ S2 cells were harvested, washed in PBS and gDNA was extracted with NucleoSpin Tissue kit (Macherey-Nagel). Remaining RNA contaminants in eluted gDNA were digested with 0.1 mg/ml RNaseA (Sigma-Aldrich) for 1 h at 37°C. Subsequently, gDNA was sonicated with Covaris AFA S220 using microTUBEs at 175 W peak incident power, 10% duty factor and 200 cycles per burst for 430 s at 5°C to generate ∼150–200 bp fragments, as described in (22). Sheared gDNA was purified with MinElute kit (QIAGEN), concentration was determined using Qubit (Thermo Fisher) and fragment size was verified using 2100 Bioanalyzer (Agilent).

Antibodies

For immunoblotting, affinity-purified polyclonal α-H3 C-term antibody (Abcam, ab1791), affinity-purified polyclonal α-CLAMP antibody (Novus, 49880002), polyclonal rabbit α-MSL2^Gilfillan serum [‘α-MSL2^Gilfillan‘ (43)], polyclonal rabbit α-MSL2 serum [generated by Pineda Antikörper-Service here termed ‘α-MSL2^Pineda’ against the MSL2-fragment (amino acids 296–608) similar to the one used for α-MSL2^Gilfillan, polyclonal rabbit α-GAF serum [‘α-GAF’ (31)] and affinity-purified monoclonal α-FLAG M2 antibody (Sigma-Aldrich, F3165) were used. For immunostaining, α-MSL2^Gilfillan and culture supernatant containing monoclonal α-MSL3 (21) were used. For ChIP, α-MSL2^Gilfillan, α-GAF and α-CLAMP antibodies were used. For DIP, supernatant containing monoclonal α-MSL2 [generated by E. Kremmer (Helmholtz Zentrum Munich, Germany) against the same MSL2-fragment as used for α-MSL2^Gilfillan] and supernatant containing monoclonal α-CLAMP [generated by E. Kremmer (Helmholtz Zentrum Munich, Germany) against the peptide LATTDDNKTCYI] were used. For IP, serum containing polyclonal α-MSL2^Pineda and affinity-purified polyclonal α-CLAMP* antibody described in (44) (kind gift of E. Larschan) were used.

Yeast two-hybrid assay

Yeast two-hybrid assay was carried out using yeast strain pJ69-4A (MATa trp1-901 leu2-3,112 ura3-52 his3-200 gal4Δ gal80Δ GAL2-ADE2 LYS2::GAL1-HIS3 met2::GAL7-lacZ), with plasmids and protocols from Clontech. For growth assays, plasmids were transformed into yeast strain pJ69-4A by the lithium acetate method, as described by the manufacturer with some modifications. All cells were grown at 30°C in an orbital shaker at 250–300 rpm. Yeast colonies were transferred into a 15 ml culture tube with 6–7 ml of YPDA medium and grown for 1 day. The culture was 10-fold diluted in a 0.5 l culture flask with YPDA medium and cultivated for 3 h. Aliquots of 1.5 ml cell suspension were pelleted by centrifugation at 4000 g for10 s, and the supernatant was removed. Pelleted cells were resuspended in 1 ml 0.1 M LiAcO, incubated for 30 min and pelleted as above. To the pellet, 240 μl 50% (v/v) PEG 3380, 36 μl 1 M LiAcO and 50 μl of a mixture of two plasmids (the amount of each plasmid in a mixture of 400–800 ng) was added sequentially and the cells suspended to homogeneity. The tube was incubated at 30°C for 30 min, then at 42°C for 5 min and then placed on ice for 1–2 min. Cells were pelleted at 4000 g for 15–20 s and resuspended in 100 μl ddH₂O. The resuspended cells were plated on selective medium lacking Leu and Trp (‘medium-2’). The plates were incubated at 30°C for 2–3 days. Afterward, the colonies were streaked out on plates on selective medium lacking either Leu, Trp and His ‘medium-3’), or lacking adenine in addition to the three amino acids (‘medium-4’), or lacking the three amino acids but containing 5 mM 3-amino-1,2,4-triazole (‘medium-3 + 5mM 3AT’). The plates were incubated at 30°C for 3–4 days and growth assessed. Each assay was prepared as three independent biological replicates with three technical repeats. Fusion proteins were cloned into pGBT9 and pGAD424 vectors from Clontech (Supplementary Table S1) and verified by sequencing.

CLAMP-MSL2 co-immunoprecipitation

Per reaction, 100 nM recombinant MSL2-FLAG and CLAMP-FLAG in 50 μl Binding Buffer (2 mM Tris/HCl pH 7.5, 100 mM KCl, 2 mM MgCl₂, 10% (v/v) Glycerol, 10 μM ZnCl₂) were incubated with end-over-end rotation for 30 min at 26°C. For α-MSL2 immunoprecipitation, the reaction was added to 25 μl magnetic Dynabeads Protein G (Thermo Fisher) pre-coupled with antibodies. For α-CLAMP immunoprecipitation, the reaction was added to 5 μl beads pre-coupled with antibodies. For pre-coupling, beads were washed thrice with 1 ml Binding Buffer and incubated with end-over-end rotation for over-night at 4°C with 1 μl α-MSL2 antibody per 10 μl beads or with the corresponding pre-immune serum as control and with 20 μl α-CLAMP* antibody complemented with 0.75 μl of an irrelevant pre-immune serum per 10 μl beads or with 1μl irrelevant pre-immune serum as control. Beads were washed thrice with 1 ml Binding Buffer, the co-IP reaction was added and incubated with end-over-end rotation for 15 min at RT. Beads were washed thrice in 100 μl Binding Buffer, resuspended in SDS-Sample Buffer and analyzed together with the corresponding input and unbound sample by SDS-PAGE with Coomassie brilliant blue G250 staining, quantified using a ChemiDoc Touch Imaging System (Bio-Rad). Co-IP reactions were performed with three independent CLAMP and MSL2 preparations.

Mapping of interaction domain

Sf21 cells at 10⁶ cells/ml (20 * 10⁶ cells) were infected 1:1000 (v/v) with each baculovirus stock separately. While MSL2^ΔCXC-FLAG and MSL2-FLAG were described previously in (23), new constructs have been generated for CLAMP-FLAG (see above) and deletion mutants of MSL2 and CLAMP, cloned into pFastBac1 using the corresponding primer (Supplementary Table S2). After 72 h, cells were harvested and lysed in 1 ml Lysis Buffer (50 mM HEPES pH 7.6, 300 mM KCl, 1 mM MgCl₂, 5% (v/v) glycerol, 0.05% (v/v) IGEPAL CA-360, 50 μM ZnCl₂) with cOmplete EDTA-free Protease Inhibitor Cocktail (Sigma-Aldrich) (PI) per 20 ml of culture. The relative amounts of expressed protein were quantified by α-FLAG western blot. Per co-IP reaction, MSL2-FLAG and CLAMP-FLAG (or their mutants) containing extracts were mixed in 1:1 to 1:3 ratio and if required adjusted with cell extract from uninfected cells. Cell extracts were supplemented with 12.5 U Benzonase, 1 mM DTT, 0.2 mM MgCl2 and PI, incubated with end-over-end rotation for 10 min at RT and cell debris were spun down at 4°C for 10 min at 20 000 g. For α-MSL2 immunoprecipitation, the reaction was added to 15 μl magnetic Dynabeads Protein G (Thermo Fisher) pre-coupled with antibodies. For pre-coupling, beads were washed thrice with 1 ml Lysis Buffer and incubated with end-over-end rotation for over-night at 4°C with 1 μl α-MSL2 antibody per 10 μl beads or with the corresponding pre-immune serum as control. Beads were washed thrice with 1 ml Lysis Buffer, the extract mixture was added and incubated with end-over-end rotation for 45 min at 4°C. Beads were washed thrice in 500 μl Binding Buffer, resuspended in SDS-Sample Buffer and analyzed together with the corresponding input sample by SDS-PAGE and α-FLAG western blot.

ATAC-seq

ATAC-seq was performed as described earlier (37,38) with some modification for Drosophila cells. In brief, 50 000 S2 or Kc cells, also with prior RNAi, were used per reaction. Cells were washed in 100 μl PBS, resuspended in 100 μl Lysis Buffer (10 mM Tris/HCl pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.1% (v/v) IGEPAL CA-630) with cOmplete EDTA-free Protease Inhibitor Cocktail (Sigma-Aldrich) and incubated for 3 min on ice. Nuclei were spun down at 4°C for 10 min at 600 g, resuspended in 50 μl 1× TD buffer with 2.5 μl TD Enzyme (Illumina) and incubated with slight agitation for 30 min at 37°C. Tagmented DNA was purified with MinElute kit (QIAGEN) and eluted in 15 μl H₂O. To determine the cycle number for library amplification, quantitative PCR was performed in triplicates with 0.5 μl sample in 10 μl reaction containing 1× NEBNext HiFi PCR mix (NEB), 1.25 μM Ad1 and Ad2 primer each and 0.5x SYBRGreen (Thermo Fisher). Cycle number at quarter maximal intensity was used for library amplification (typically 11–13 cycles). Libraries were amplified by using 12.5 μl sample in 50 μl PCR reaction containing 1× NEBNext HiFi PCR mix (NEB), 1.25 μM Ad1 and Ad2 primer each. ATAC libraries were purified with MinElute kit (QIAGEN) and concentration determined using 2100 Bioanalyzer with High Sensitivity DNA kit (Agilent). Libraries were sequenced on HiSeq 1500 (Illumina) instrument yielding typically 25–35 million 50 bp paired-end reads per sample.

ChIP-seq

ChIP-seq was performed as previously described (39) with slight modification. S2 cells (∼10⁸ cells) after RNAi were harvested and chilled on ice. Cells were cross-linked with 1% formaldehyde for 55 min on ice by adding 1 ml volume 10× fixing solution (50 mM HEPES pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA) with 10% formaldehyde per 10 ml culture and reaction was stopped by adding 125 mM glycine and incubating for 10 min on ice. For nuclei isolation, cells were washed in PBS and resuspended in Buffer A (10 mM Tris/HCl pH 8.0, 10 mM EDTA, 0.5 mM EGTA, 0.25% (v/v) Triton-X 100) with cOmplete EDTA-free Protease Inhibitor Cocktail (Sigma-Aldrich) (PI) at 10⁸ cells/ml and incubated with end-over-end rotation for 10 min at 4°C. The cells were pelleted at 4°C for 10 min at 2300 g and resuspended in Buffer B (10 mM Tris/HCl pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.01% (v/v) Triton-X 100) with PI at 10⁸ cells/ml and incubated with end-over-end rotation for 10 min at 4°C. For chromatin fragmentation, nuclei were spun down at 4°C for 10 min at 2300 g, resuspended in RIPA (10 mM Tris/HCl pH 8.0, 140 mM NaCl, 1 mM EDTA, 1% (v/v) Triton-X 100, 0.1%(v/v) SDS, 0.1% (v/v) DOC) with PI at 10⁸ cells/ml in 1 ml for shearing with Covaris AFA S220 using 12 × 12 tubes at 100 W peak incident power, 20% duty factor and 200 cycles per burst for 25 min at 5°C to generate 150–200 bp fragments. Protein A and Protein G (GE Healthcare) beads (mixed in a 1:1 ratio) were washed thrice with 30 bed volumes RIPA. To remove cell debris, sheared chromatin was centrifuged at 4°C for 15 min at 15 000 g and 100 μl soluble chromatin in the supernatant was pre-cleared with 3 μl (6 μl 50% slurry) Protein A and Protein G beads mix by incubating with end-over-end rotation for 1 h at 4°C. Beads were pelleted at 4°C for 5 min at 500 g and supernatant was directly used for immunoprecipitation by adding antibody to 200 μl chromatin adjusted to 500 μl with RIPA including PI and incubating with end-over-end rotation for 16 h at 4°C. To remove precipitates, chromatin was centrifuged at 4°C for 15 min at 15 000 g and 100 μl supernatant was added to 3 μl Protein A and Protein G beads mix (RIPA-equilibrated as above) by incubating with end-over-end rotation for 4 h at 4°C. Beads were spun down and washed five times with 60 bed volumes RIPA including PI by incubating with end-over-end rotation for 10 min at 4°C. For DNA recovery, beads were spun down, resuspended in 6.7 bed volumes TE Buffer (10 mM Tris/HCl pH 8.0, 1 mM EDTA), RNA was digested with 50 μg/ml RNaseA (Sigma-Aldrich) for 30 min at 37°C, and after addition of 0.5% (m/v) SDS, proteins were digested with 0.5 μg/ml Proteinase K (Sigma-Aldrich) for 16 h at 65°C with agitation. DNA was purified with 1.8× AMPure XP beads (Beckmann Coulter). Quantitative PCR was performed using SYBR Green PCR Master Mix (Thermo Fisher) and 0.5 nM forward and reverse primer each (Supplementary Table S3), analyzed on LightCycler 480 (Roche). Libraries were prepared with NEBNext ChIP-Seq Library Perp Kit for Illumina (NEB) and analyzed with 2100 Bioanalyzer with DNA 1000 kit (Agilent). Libraries were sequenced on HiSeq 1500 (Illumina) instrument yielding typically 25–30 million 50 bp single-end reads per sample.

DIP-seq

DIP-seq was performed accordingly to (22), with modifications for combined DIP of two proteins. Sheared gDNA was diluted to 4 mg/ml in Binding Buffer (2 mM Tris/HCl pH 7.5, 100 mM KCl, 2 mM MgCl₂, 10% (v/v) Glycerol, 10 μM ZnCl₂). Ten microliter (corresponding to 10%) was taken as input material and adjusted to 50 μl with Binding Buffer for DNA purification. Per reaction, 100 nM recombinant MSL2 and/or 50 nM recombinant CLAMP was added in 100 μl diluted gDNA and incubated with end-over-end rotation for 30 min at 26°C. For immunoprecipitation, the reaction was added to 7.5 μl (15 μl 50% slurry) Protein G beads (GE Healthcare) pre-coupled with antibodies. For pre-coupling, beads were washed thrice with 100 μl Binding Buffer and incubated with end-over-end rotation for 3–4 h at 4°C with 1 ml culture supernatant containing monoclonal antibody or culture medium as control. Beads were spun down for 1 min at 500 g, washed thrice with 100 μl Binding Buffer, the DIP reaction was added and incubated with end-over-end rotation for 15 min at RT. Beads were spun down for 1 min at 500 g, washed twice in 100 μl Binding Buffer and resuspended in 50 μl Binding Buffer. Five microliter were taken for Western blot analysis. Proteins were digested by adding Proteinase K at 0.5 mg/ml and incubated for 1 h at 56°C with agitation and DNA was purified with 1.8× AMPure XP beads (Beckmann Coulter). Libraries were prepared with MicroPlex Library Preparation Kit v2 (Diagenode) and analyzed with 2100 Bioanalyzer with DNA 1000 kit (Agilent). Libraries were sequenced on HiSeq 1500 (Illumina) instrument yielding typically 20–40 million 50 bp single-end reads per sample. DIP reactions were performed with three independent CLAMP and MSL2 preparations.

Data analysis

Sequencing data were processed using SAMtools (45) version 1.3.1, BEDtools (46) version 2.26.0, R version 3.4.2 (http://www.r-project.org) and Bioconductor (http://www.bioconductor.org) using default parameters for function calls, unless stated otherwise.

Read processing

Sequence reads were aligned to the D. melanogaster release 6 reference genome (BDGP6) using Bowtie (47) version 1.1.2 (parameter –m 1) for ChIP-seq and DIP-seq samples and Bowtie2 (48) version 2.2.9 (parameters –very-sensitive, –no-discordant, –no-mixed, -X 100) for ATAC-seq samples considering read-pairs as non-nucleosome if ≤100 bp according to (37).

Peak calling, robust peak sets and HAS definition

Peaks were called using Homer (49) version 4.9.1 calling function findPeaks for DIP- and ChIP-seq samples using the corresponding input sample as control (parameters –style factor, -size 200, -fragLength 200, -inputFragLength 200 and –C 0) and for ATAC-seq samples without control (parameters -style dnase, -C 0, -gsize 137e6, -minDist 50). Peaks were defined as robust if the region was called in at least two replicate samples. HAS regions were used as defined by (22) with 309 HAS in total, of which 304 are located on the X chromosome and 5 on autosomes.

X chromosomal enrichment

The X chromosomal enrichment was defined as the ratio of X chromosomal peak density over autosomal peak density and peak density was calculated as the number of peaks divided by chromosome length as defined by (22).

De novo motif discovery and MRE definition

Enriched motifs in peak region were discovered using MEME (50) version 4.11.4 (parameters -dna, -mod zoops and -revcomp). For discovering the PionX motif, the list of PionX sites in (22) was used. As described in reference (22), we refer to MRE as the motif discovered within HAS, which is highly similar to the originally defined motif (20,21).

Motif search

Motif search in peak region was performed with FIMO (51) version 4.11.4 (parameters –qv-thresh, –thresh 0.2, –max-stored-scores 1e6) applying a fifth-order background model.

Browser profiles

Browser profiles were generated using tsTools (R) (https://rdrr.io/github/musikutiv/tsTools/) using mean per base read count per million mapped reads (rpm) of biologicals replicates, after aligned reads were extended to 200 bp fragments.

Assignment of peak regions to chromatin states

Peak regions within one of the chromatin states were assigned to the corresponding chromatin state defined by (52) and peaks overlapping with multiple chromatin states were assigned as ‘none’.

ChIP-seq analysis

For calculating ChIP-seq signal enrichment as log₂ ratio of IP over input, aligned reads were extended to 200 bp fragments and reads overlapping with target regions requiring a minimal overlap of half read length were counted and normalized to million mapped reads (rpm). For generating heat maps by calling the function pheatmap (R) and average profiles in a 2 kb window around HAS, coverage vectors were calculated as mean per base read count of IP samples and normalized to rpm. To test for statistical significant differences between gst RNAi and trl RNAi condition in CLAMP ChIP-seq, signal enrichment for the merged set of robust peak sets of both samples was calculated. Testing for difference was performed using limma (53) (R) including batch variables as random effect and calling the functions lmFit, eBayes (parameters trend = T, robust = T) and topTable (parameter adjust.method = ‘fdr’). Regions were defined as statistical significant different between conditions with false discovery rate (FDR) < 0.05.

DIP-seq analysis

As proxy for MSL2-FLAG in vitro genomic binding we used the previously published DIP-seq experiment by (22) (GSE75033), where MSL2 was immunoprecipitated using α-FLAG M2 beads (Sigma-Aldrich) directed against the FLAG-tag, here referring to as ‘MSL2 (α-FLAG) – Villa’. For calculating DIP-seq signal enrichment as log₂ ratio of IP over input, aligned reads were extended to 200 bp fragments and reads overlapping with target regions requiring a minimal overlap of half read length were counted and normalized to rpm. For generating heat map by calling the function pheatmap (R), euclidean distance between the enrichment for all biological replicates at combined robust peak set was measured by calling the function dist (R) and samples were hierarchical clustered using the ‘complete’ method by calling the function hclust (R). The row dendrogram was divided into 18 clusters by calling the function cutree (R) (parameter k = 18). For predicting the DNA roll between the base pairs at position +1 and +2 of the PionX motif, the best matching motif found by FIMO within the peak region was extended by 2 bp at each end. The roll was predicted using DNAshapeR (54) (R) by calling the function getShape. The DNA roll between the first two base pairs was assigned to the first base pair for simplicity, therefore referring to as ‘roll at position +1′. The number of GA:TC dinucleotides in the peak region was counted using rDNAse (R) by calling the function kmer considering also the reverse complement. We calculated the GA:TC dinucleotide density as the number of GA:TC-dinucleotides divided by the peak length. The length of GA:TC repeats was counted using Biostrings (R) by calling the function vcountPattern considering also the reverse complement.

ATAC-seq analysis

For calculating ATAC-seq signal intensities, fragments on the + strand were moved by +4 bp and on the − strand by −5 bp according to (37). Signal intensities in target regions were calculated by calling summarizeOverlaps (parameter ignore.strand = TRUE) from GenomicAlignments (55) (R). Further analysis was performed using DESeq2 (56) (R), including batch variables as random effect. Sites were considered as statistical significant different between conditions with absolute log₂ fold-change > 0.5 and FDR < 0.1 by calling the function results (parameters lfcThreshold = 0.5, altHypothesis = ‘greaterAbs’). For analyzing ATAC-seq signal at HAS, the size factor obtained by using the combined robust peak sets of the corresponding samples was used. For comparing ATAC-seq signals in Kc cells with S2 cells, only X chromosomal peak regions were considered.

MSL2 requires CLAMP for efficient binding to HAS in vivo

Larschan and colleagues previously reported that binding of MSL3 to the polytenic X chromosome and the crosslinking of MSL2 to some HAS required the presence of the CLAMP protein, which recognizes GA-rich sequences matching the MRE motif (26,27). On the other hand, recombinant MSL2, the male-specific DNA-binding subunit of the DCC, can select MRE sequences and is able to enrich for X chromosomal sites, especially those with PionX signature, in genomic DNA in vitro (22).

To explore the relationship between CLAMP and MSL2 in our system, we depleted CLAMP by RNA interference (RNAi) in male S2 cells and monitored the X chromosome binding of MSL3 and MSL2 by immunostaining. RNAi against clamp or msl2 is efficient and does not affect each other's protein levels, excluding indirect effects (Figure 1A). An irrelevant RNAi against green fluorescent protein (gfp) or Schistosoma japonicum glutathione-S-transferase (gst) served as control (‘control’ in all figures). Upon clamp RNAi, MSL2 and MSL3 no longer localize at a coherent X chromosome territory, but redistribute in smaller speckles, indicating a targeting deficiency (Figure 1B). The fact that both subunits still co-localize suggests that the DCC is still intact and rather mis-targeted as a complex.

To further investigate MSL2 binding in vivo in absence of CLAMP, we performed ChIP-seq of MSL2 in male S2 cells after depletion of CLAMP by RNAi (Supplementary Figure S1A). Upon depletion, the CLAMP ChIP-seq signal was robustly reduced and only minor residual binding was observable (Supplementary Figure S1B). Under those conditions MSL2 binding to HAS was massively reduced (Figure 1C and D). HAS with a PionX signature are the primary contacts for MSL2 in vivo (22). MSL2 binding to the subset of HAS which overlap with PionX sites (HAS-PionX) seemed to be slightly less affected by clamp RNAi, however the small amounts of remaining CLAMP after RNAi also tended to be enriched at HAS-PionX sites (Supplementary Figure S1C and D). The dependence of MSL2 on CLAMP for binding to all HAS in vivo was somewhat unexpected as MSL2 can target to PionX sites without the help of any additional factor in vitro.

As a control for these experiments we depleted the GAF. GAF (encoded by the trithorax-like [trl] gene) also binds to GA-rich sites genome-wide and co-localizes with CLAMP at many sites (27). Contrasting CLAMP, GAF is absent from most HAS (21,57,58). As expected, the depletion of GAF did not affect MSL2 binding (Figure 1C and D).

Since Larschan and colleagues suggested some competition in DNA-binding between CLAMP and GAF (59), we explored whether the two proteins compete for binding to HAS in vivo. Upon RNAi depletion of GAF in male S2 cells the CLAMP ChIP-seq signal was unchanged at HAS (Supplementary Figure S1B and D). Genome-wide, at an FDR < 0.05, only 93 out of 5983 CLAMP binding sites (1.55%) showed a statistically significant difference in CLAMP binding (Supplementary Figure S1E). These sites were mostly located in chromatin with enhancer- or promoter-related histone marks and CLAMP binding was increased or decreased (Supplementary Figure S2).

To analyze potential competition between GAF and CLAMP, we also monitored GAF binding at CLAMP binding sites after depletion of CLAMP (Supplementary Figure S1F). Nine robust CLAMP binding sites were monitored by ChIP-qPCR including three HAS (Supplementary Figure S1G). At none of these sites did GAF binding change upon removal of CLAMP.

We conclude that MSL2/DCC binding to HAS depends on CLAMP but not on GAF and that CLAMP and GAF do not generally compete for binding to GA-rich sequences.

Genomic binding sites of CLAMP defined in vitro by DNA immunoprecipitation

We previously assayed the DNA binding sites of MSL2 in vitro through genome-wide DNA immunoprecipitation (DIP) (22,35,36). In such experiments, purified and fragmented genomic DNA is incubated with protein of interest under conditions where competitive DNA binding occurs. The protein is immunoprecipitated and the bound DNA sequenced (22,35,36). For CLAMP, such an analysis had not been done. We expressed and purified recombinant full-length CLAMP via a baculovirus expression system (Supplementary Figure S3) and assayed it's in vitro DNA-binding property by genome-wide DIP-seq.

In vitro, CLAMP bound to 4037 sites in the Drosophila genome under our DIP conditions (Figure 2A), comparable in number to 5214 sites determined by our ChIP-seq approach. About one third (32.3%, n = 1307) of the in vitro CLAMP binding sites overlapped with in vivo binding sites (Figure 2B). Sequence analyses within in vitro binding sites yielded the core of the GA-repeat consensus motif defined in vivo by ChIP-seq, replicating the previously identified CLAMP motif, which has high similarity to the MRE motif (Figure 2C) (27). Furthermore, the CLAMP in vitro binding motif is highly similar to the previously identified motif for the CLAMP DNA-binding domain in vitro by protein-binding microarray, confirming the DNA binding specificity for our full-length recombinant protein (27,29). As expected, CLAMP bound sites on all chromosomes in vitro with a 2.4-fold enrichment of X chromosomal sequences similarly to in vivo with 1.7-fold enrichment (Figure 2D), reflecting the ∼2-fold enrichment of the MRE motif on the X chromosome (20,21). However, the binding intensities in vitro and in vivo were uncorrelated (Figure 2E). Whereas the in vitro binding sites solely reflect the intrinsic binding affinity of CLAMP to DNA, the in vivo interactions are obviously modulated by other factors (such as MSL2, see below).

Cooperation between CLAMP and MSL2 increases the efficiency of HAS recognition in the context of the genome

We recently described the in vitro genomic binding sites of MSL2 by DIP-seq (22). This study revealed the capacity of MSL2 to identify PionX sites and to enrich X chromosomal sequences, but the protein missed many HAS and in addition pulled out autosomal GA-rich sites that do not correspond to physiological binding sites (22). We now explored whether cooperation with CLAMP may improve the binding specificity and/or capacity of MSL2. Therefore, we expressed and purified recombinant full-length CLAMP and MSL2 via a baculovirus expression system (Supplementary Figure S3) and assayed their DNA binding properties by genome-wide DIP-seq.

In our previous study, MSL2 protein was immunoprecipitated with an α-FLAG antibody recognizing the epitope of its FLAG-tag. Because CLAMP protein is also FLAG-tagged to facilitate its purification, we could not use the α-FLAG antibody, but performed the DIP experiments using α-CLAMP and α-MSL2 antibodies specific for the proteins. The analysis is complicated by the fact that the α-MSL2 antibody available to us yielded a DIP profile for MSL2 without significant enrichments of sites under our conditions (Figure 3A). Remarkably however, the α-MSL2 antibody retrieved a robust profile for MSL2 in the presence of CLAMP. For reference, we included our previously published MSL2 DIP-seq data using the α-FLAG antibody for immunoprecipitation as proxy for MSL2 binding in vitro (22).

MSL2 alone binds in vitro to 288 sites of which 54 are HAS (out of 309 HAS in total), as reported previously (22) (Figure 3B). In the presence of CLAMP, MSL2 bound now to 1927 sites including 234 HAS. A similar situation was observable for CLAMP binding in vitro. CLAMP alone bound 4037 sites including 160 HAS. In presence of MSL2, CLAMP bound 7032 sites including 278 HAS (Figure 3C). This reveals an intrinsic cooperativity between MSL2 and CLAMP through stabilizing each other's interactions at many common binding sites in the genome under these conditions. This cooperativity allows both factors together to select most HAS from genomic DNA in vitro, but at the cost of an increased number of non-physiological binding sites. However, those additional binding sites all bear sequence determinants similar to the physiological ones showing that the intrinsic binding properties of both proteins contribute to an increased affinity for GA-rich sequences in general (see below, Supplementary Figure S4). Expectedly, many of those sites are autosomal and thus lead to a lower X chromosomal enrichment of bound sequences (see ‘Discussion’ section).

To explore whether different modes of binding cooperativity exist we performed hierarchical clustering of the in vitro binding intensities at the combined robust peak set of all DIP-seq samples (Figure 4). We identified 12 major clusters defined by the different binding behaviors of both proteins. We categorized the clusters into four binding scenarios (Figure 5A and B; Supplementary Figures S5 and 6A): (i) independent: CLAMP and MSL2 bind alone and the presence of the other factor makes little difference [clusters 10 and 11]; (ii) CLAMP-dependent: CLAMP binds largely independently of MSL2 and recruits MSL2 [clusters 6–9 (clusters 2–4 show similar behavior with lower signal enrichment)]; (iii) MSL2-dependent: MSL2 binds largely independently of CLAMP and recruits CLAMP [cluster 12]; (iv) interdependent: CLAMP and MSL2 do not bind alone, but show cooperative binding [cluster 5 (cluster 1 shows similar behavior with lower signal enrichment)]. As expected, MSL2 binding to PionX sites did not depend on CLAMP (Figures 4 and 5A). CLAMP alone is absent from many PionX sites and is recruited by MSL2 to these sites.

De novo discovery of most-enriched sequence motifs within each cluster yielded variations of the GA-rich MRE motif of variable length and regularity (Supplementary Figure S4). We attempted to stratify the binding sites further by extracting additional sequence features within each binding site (Figure 5C; Supplementary Figures S4 and 6B): (i) the score of the best-matching MRE motif; (ii) the score of the best matching PionX motif; (iii) the DNA roll at position +1 of the best-matching PionX motif (high roll at position ‘+1’ is a defining feature of the PionX signature (22); for simplicity we assigned the inter-base structural feature roll to the lead base pair); (iv) density of GA:TC-dinucleotides; (v) length of the longest (GA:TC)_x-repeat (as x repeats).

Interestingly, the ‘independent’ sites have the highest MRE scores, the highest GA:TC-dinucleotide density and the longest GA:TC-repeats (Figure 5C and Supplementary Figure S6B). This is in good agreement with previous findings, as the first motif discovered in MSL2 in vitro binding sites is a long, low-complexity GA:TC-repeat, which was suggested to be bound through the MSL2 C-terminus (22), and CLAMP in vitro binding intensities correlate with GA:TC-repeat length (29). The ‘CLAMP-dependent’ sites tend to have higher MRE scores, higher GA:TC-dinucleotide density, longer GA:TC-repeats, but low PionX scores as well as low roll at position +1. By contrast, the ‘MSL2-dependent’ sites tend to have lower MRE scores, lower GA:TC-dinucleotide density and shorter GA:TC-repeats, but the highest PionX scores including the ‘high roll at position +1’ (22). Interestingly, ‘interdependent’ sites lack the features that characterize CLAMP or MSL2 binding sites in vitro (high PionX and MRE scores, high GA:TC-dinucleotide density and long GA:TC-repeats length), with the exception of ‘high roll at position +1’. Apparently, neither MSL2 nor CLAMP alone bind well to these sites, but together they mount an interaction surface able to recognizes a degenerate GA-rich MRE consensus motif (Supplementary Figure S4). We speculate that the CXC domain of MSL2 reads out the DNA shape at the 5′ end of these binding sites.

The cooperativity between MSL2 and CLAMP expands each other's binding repertoire in vitro. Together both proteins are capable to identify nearly all physiological functional MRE sequencing (HAS) in vitro. They also bind to many autosomal sites, which may be occluded by nucleosomes in vivo. Accordingly and in agreement with earlier findings (22), the enrichment of X chromosomal sequences in vitro is almost entirely due to the action of MSL2. MSL2-dependent sites which harbor the PionX signature are 10.4-fold enriched on the X chromosome, recapitulating the 9.8-fold enrichment of PionX sequences (Figure 5D and Supplementary Figure S6C) (22).

CLAMP and MSL2 form a stable complex

The results of the DIP experiments suggest cooperativity between CLAMP and MSL2. Such synergism may be due to physical interaction of the two proteins. Previously, CLAMP has been found associated with the DCC after crosslinking in vivo (60).

As a direct test for protein interactions we employed a Yeast Two-Hybrid assay (Y2H). MSL2 (and deletion derivatives) were expressed in fusion with a Gal4 DNA binding domain (DBD) along with CLAMP (and deletion derivatives) fused to the Gal4 activation domain (AD). If co-expressed in yeast, the association of the two proteins reconstitutes the function of the transcription factor GAL4 that activates the his3 gene in the yeast strain pJ69-4A that is auxotroph for histidine (61). Yeasts in which DBD and AD fusion proteins interact can grow on plates lacking histidine, leucine and tryptophan. The assay revealed a robust and reproducible interaction of MSL2-DBD and AD-CLAMP (Supplementary Figure S7A) even in stringent conditions posed by absence of adenine and presence of 3-amino-1,2,4-triazole, a competitive inhibitor of the enzyme encoded by his3. MSL2-DBD or AD-CLAMP alone did not support growth in the presence of the unfused AD or DBD, respectively (Supplementary Figure S7A). Using appropriate deletion constructs we found that the first 153 amino acids of CLAMP are sufficient to interact with DBD-MSL2 but further deletion to the first 123 amino acids abolished the interaction (Supplementary Figure S7A). This assay locates the MSL2 interaction site to an N-terminal fragment of CLAMP, which harbors the first ZnF domain. The remaining six C-terminal ZnF domains are involved in binding to GA-dinucleotide repeats (27,29,59). Using various C-terminal deletion constructs of MSL2 showed that the ∼200 C-terminal amino acids downstream of the CXC domain are required for interaction with CLAMP (Supplementary Figure S7A). As a control for proper MSL2 folding we confirmed that MSL2^1-573-DBD was still able to interact with MSL1 via the N-terminal RING domain in our assay (Supplementary Figure S8B) as reported previously (62).

To probe whether this interaction was direct we tested the recombinant purified proteins used in DIP in the absence of DNA. The two proteins were probed at equimolar concentration (100 nM) for interaction by co-immunoprecipitation (co-IP) with α-MSL2 and α-CLAMP antibodies. Both proteins were quantitatively (∼90%) immunoprecipitated with the α-MSL2 antibody (Figure 6A and B). While with the α-CLAMP antibody only ∼50% of each protein was immunoprecipitated, presumably because the antibody amount was limiting. Of note, the IP of CLAMP was more efficient in the presence of MSL2 (Supplementary Figure S9A and B), perhaps due to conformational stabilization.

To map the interaction surfaces within MSL2 and CLAMP, we co-expressed deletion mutants of both recombinant FLAG-tagged proteins in Sf21 cells and immunoprecipitated them from total cell extracts. To map the CLAMP-interaction site in MSL2, we used a series of C-terminal deletions systematically lacking conserved regions identified within MSL2. In short, alignment of MSL2 sequences from 12 Drosophila species revealed five conserved regions (CR) (Supplementary Figures S9B and 10): the highly conserved RING and CXC domains; CR2 consisting of 66 amino acids with 50%-90% conservation; a conserved PAKKFR motif, as part of a stretch of 25 residues enriched in basic amino acids; and CR3, which corresponds to the 28 C-terminal amino acids of the D. melanogaster MSL2 (Supplementary Figure S9B). The 20 amino acid proline-rich region within MSL2′s C-terminus separates the CR2 from the PAKKFR motif. Alignment of CLAMP sequences from 12 Drosophila species revealed that CLAMP is conserved throughout its entire length (Supplementary Figure S11).

MSL2 derivatives harboring CR2 (MSL2^1-726 and MSL2^1-688) showed quantitative binding of CLAMP, whereas fragments lacking CR2 (MSL2^1-619 and MSL2^1-567) did not bind CLAMP (Figure 6C and D). CLAMP binding was unaffected by deletion of the CXC domain, but internal deletion of CR2 abolished all CLAMP binding. We therefore refer to the conserved region between amino acids 620–685 as ‘CLAMP-binding domain’ (CBD) of MSL2.

To interrogate the MSL2 interaction surface in CLAMP we used conditions where CLAMP was in excess over MSL2 to favor the interaction with MSL2 in α-MSL2 co-IP experiments. While full-length CLAMP reproducibly bound to MSL2, none of eight CLAMP deletion mutants, including the N-terminal CLAMP^1-153 fragment, which interacted with MSL2 in the Y2H assay (Supplementary Figures S7A and 8A), bound to MSL2 above a background, even at protein concentrations in the extracts approaching 100 nM (Supplementary Figure S9C–F).

In summary, our data document a stable interaction between MSL2 and CLAMP and we map the interaction domains to the N-terminus of CLAMP including the first zinc finger domain and the CBD of MSL2 just downstream of the CXC domain.

CLAMP and MSL2 cooperate to keep HAS nucleosome-free

The genome-wide DIP revealed the potential for extensive cooperation between MSL2 and CLAMP to bind shared binding sites, but the physiological X/autosome discrimination of MSL2 was not improved in free DNA. Conceivably, exclusive X chromosome binding requires a chromatin environment.

To survey the contribution of either factor to HAS accessibility in chromatin, we performed ATAC-seq after RNAi against clamp or msl2 in male S2 and female Kc cells (Figure 7A and Supplementary Figure S12A). Depletion of CLAMP in S2 and Kc cells by RNAi caused only few significant changes in accessibility genome-wide (258 of 8913 sites and 102 of 9767 sites in S2 and Kc cells, respectively; Figure 7B and Supplementary Figure S12B). This indicates that most sites bound by CLAMP are kept accessible by other factors. Most sites affected by CLAMP depletion became less accessible, showing that CLAMP contributes to keeping particular loci open. These sites, that depend on CLAMP to be accessible, include especially many HAS (61 HAS and 6 HAS-PionX) in S2 cells. Focusing on the 309 HAS showed that HAS are commonly accessible in male S2, but not in female Kc cells (Figure 7C and Supplementary Figure S12B) suggesting that the DCC is required for HAS accessibility. Indeed, depletion of MSL2 in S2 cells caused only few significant changes in accessibility genome-wide (61 of 8913 sites), but these sites were nearly exclusively HAS [49 HAS and 7 HAS-PionX] (Figure 7B).

Remarkably, most HAS are inaccessible in the absence of either CLAMP or MSL2 in S2 cells (Figure 7D). Apparently, CLAMP and MSL2 contribute equally to HAS accessibility (Figure 7E). The minority of HAS that remained accessible in the absence of MSL2 or CLAMP are also accessible in Kc cells (Figure 7C and D; Supplementary Figure S12B), indicating that these HAS are kept open by other factors.

Broadening the view, we found that CLAMP promotes access to other loci in the genomes of both cell lines (Figure 7B and Supplementary Figure S12B). According to the 9-state chromatin model (52) (Supplementary Figure S12C) many of these sites bear promoter or enhancer signatures. Unexpectedly, we found that sites marked by H3K27me3, the polycomb signature, are most enriched within the sites affected by CLAMP depletion.

These experiments identify a second layer of DNA binding cooperativity between CLAMP and MSL2. Faced with purified genomic DNA both proteins cooperate to bind to essentially all sites with appropriate DNA sequence features, regardless of chromosomal origin. By contrast, in a physiological chromatin environment extensive cooperativity between both factors is restricted to X chromosomal HAS.