Zebrafish husbandry
Stable transgenic zebrafish lines were kept at 28.5 °C in a dedicated aquatics facility with a 14 h on/10 h off light cycle. Casper fish with the following genotype were used for all experiments: mitfa-BRAFV600E;p53−/−;mitfa−/−. Fish were anaesthetized using Tricaine (MS-222; stock concentration of 4 g l−1), diluted until the fish were immobilized. All animal procedures were approved by the Memorial Sloan Kettering Cancer Center Institutional Animal Care and Use Committee (protocol no. 12-05-008).
Cloning of zebrafish CRISPR constructs
To generate hmgb2a and hmgb2b CRISPR guide RNA (gRNA) plasmids for use in vivo, three gRNAs for each gene were subcloned into Gateway entry vectors containing zebrafish-optimized U6 gRNA promoters. The resulting 3× gRNA plasmid was assembled through Gateway LR cloning. Validation of gRNA/Cas9 activity in vivo was performed using the Alt-R CRISPR-Cas9 system (Integrated DNA Technologies (IDT)) by injecting single-guide RNAs (sgRNAs) and purified Cas9 protein into one-cell-stage zebrafish embryos. Genomic DNA was isolated from five to ten embryos 24 h later, and mutation detection was performed using the Alt-R Genome Editing Detection Kit (IDT).
Zebrafish gRNA sequences:
hmgb2a sgRNA1: 5′-GAAAAGTTCACCGAGGTCCC-3′
hmgb2a sgRNA2: 5′-AAGGTGAAGGGCGACAACCC-3′
hmgb2a sgRNA3: 5′-GACAACCCGGGCATCTCTAT-3′
hmgb2b sgRNA1: 5′-CAAACCCAAGGGGAAGACGT-3′
hmgb2b sgRNA2: 5′-CTCAAACTTGACCTTGTCGG-3′
hmgb2b sgRNA3: 5′-AGAGAAGTTGACGGGCACGT-3′
NT sgRNA: 5′-AACCTACGGGCTACGATACG-3′
Zebrafish in vivo electroporation
Tumours were generated by means of TEAZ68. To generate hmgb2a/hmgb2b knockout melanomas, adult 3-month-old to 6-month-old fish were randomly assigned to groups and injected with the following plasmids: miniCoopR–GFP, mitfa:Cas9, Tol2, U6–sgptena, U6–sgptenb and either 394-zU6–3XsgRNA[hmgb2a] and 394-zU6–3XsgRNA[hmgb2b] or 394-zU6–3XsgRNA[NT]. Adult fish were anaesthetized using tricaine and injected with 1 µl of plasmid mixture below the dorsal fin, immediately electroporated and moved to fresh water to recover. Tumour growth was imaged every 1–2 weeks using a ZEISS Axio Zoom V16 fluorescence microscope. Male and female animals were used in equal proportions. No sample size calculation or blinding was performed.
Cell culture
The following cell lines were obtained from the American Type Culture Collection: A375 (CRL-1619), SKMEL5 (HTB-70), MIA-PaCa-2 (CRM-CRL-1420), Panc-1 (CRL-1469), HTB-4 (T24), HTB-9 (5637) and HEK293T. The cells were maintained in a 37 °C and 5% CO2 humidified incubator. The cell lines were authenticated by the American Type Culture Collection and routinely checked to be free from Mycoplasma. The cells were cultured in DMEM (Gibco; 11965) supplemented with 10% fetal bovine serum (GeminiBio; 100-500).
Transfection of siRNAs
SiRNAs targeting the following genes were obtained from Horizon Discovery: HMGB2 (L-011689-00-0005), SYNE2 (L-019259-01-0005) and non-targeting control (D-001810-10-05). DharmaFECT 1 Transfection Reagent (Horizon Discovery; T-2001) was used to transfect 250,000 A375 cells per condition. The medium was changed after 24 h, and experiments were performed 72 h after changing the medium. HMGB2 knockdown was validated by western blot with an antibody targeting HMGB2 (MilliporeSigma; HPA053314). For gel source data, see Supplementary Fig. 1. Downregulation of SYNE2 was validated by means of quantitative polymerase chain reaction (qPCR) with the following primers:
SYNE2 F: 5′-CAAAGCACAGGAAACTGAGGCAG-3′
SYNE2 R: 5′-AGACAGTGGCAACGAGGACATG-3′
β-Actin F: 5′-CACCAACTGGGACGACAT-3′
β-Actin R: 5′-ACAGCCTGGATAGCAACG-3′
Cloning of human CRISPR constructs
The lentiCRISPRv2 system69 was used to generate stable human knockout cell lines, and gRNAs targeting HMGB2 or ATAT1 were selected from the GeCKO2 library. Oligonucleotides containing each gRNA were obtained from IDT and cloned into the lentiCRISPRv2 backbone through restriction digest with BsmB1 and ligation with Quick Ligase (New England Biolabs). Ligated plasmids were transformed into Stbl3 bacteria (New England Biolabs) and sequenced to verify gRNA insertion. The final plasmids were used to create stable A375 and SK-MEL-5 lines using lentiviral transduction, as described below.
Human gRNA sequences:
HMGB2 sgRNA1: 5′-CTGCACGAAGAAGGCGTACG-3′
HMGB2 sgRNA2: 5′-AAGATCAAAAGTGAACACCC-3′
ATAT1 sgRNA1: 5′-CCAGAAGAACATCTACAGTG-3′
ATAT1 sgRNA2: 5′-CCTCACTGTAGATGTTCTTC-3′
NT sgRNA: 5′-AACCTACGGGCTACGATACG-3′
Cloning of HMGB2 overexpression and deletion constructs
To generate the HMGB2–GFP plasmid, the human HMGB2 coding sequence in a pENTR backbone (Horizon Discovery; OHS5898-202621565) was combined with a C terminus EGFP tag using In-Fusion Cloning. The HMGB2–GFP insert was then transferred into a lentiviral expression vector containing the cytomegalovirus promoter (pLX304; Addgene 25890) by means of Gateway cloning using LR Clonase II Plus (Thermo Fisher Scientific). Deletion constructs were generated through In-Fusion Cloning (Takeda Bioscience) using the HMGB2 open reading frame in the pENTR backbone as a template and were subsequently cloned into pLX304 by means of Gateway cloning, as described above. The primers were:
HMGB2-ΔA-box F: 5′-CCAACAAGCCTCCCAAAGGTGATAAGAAGGG-3′
HMGB2-ΔA-box R: 5′-TGGGAGGCTTGTTGGGGTCTCCTTTACC-3′
HMGB2-ΔB-box F: 5′-CCAATGCTGCCAAGGGCAAAAGTGAAGC-3′
HMGB2-ΔB-box R: 5′-CCTTGGCAGCATTGGGGTCCTTTTTCTTCCC-3′
HMGB2-Δacidic tail F: 5′-CATATCGTGACCCAGCTTTCTTGTACAAAG-3′
HMGB2-Δacidic tail R: 5′-CTGGGTCACGATATGCAGCAATATCCTTTTC-3′
Generation of stable cell lines
HMGB2OE, HMGB2–GFP, HMGB2KO, HMGB2del, ATAT1KO and FastFUCCI stable cell lines were generated by means of lentiviral transduction. The FastFUCCI reporter plasmid was obtained from Addgene (86849). The HMGB2–GFP reporter plasmid, HMGB2 gRNA+Cas9 plasmids, ATAT1 gRNA+Cas9 plasmids and non-targeting gRNA+Cas9 plasmids were assembled, as described above. The HMGB2OE plasmid was obtained from Horizon Discovery (OHS5897-202616132). Eight million HEK293T cells per condition were transfected with 1,200-ng lentiviral vector, 600-ng PAX2 plasmid and 300-ng MD2 plasmid using Effectene Transfection Reagent (QIAGEN). Virus was collected starting 24 h after transfection. Viral supernatant was filtered (0.45-µm filter) before adding to A375 cells at a 1:1 ratio with medium and 10 µg ml−1 of polybrene. Cells were infected for 72 h, allowed to recover for 24 h and then selected using blasticidin (5 µg ml−1; 7–10 days) or puromycin (1 µg ml−1; 3 days). For cell lines expressing a fluorescent reporter, cells were sorted using FACSAria III or FACSymphony S6 cell sorters (BD Biosciences). For HMGB2 overexpression and CRISPR lines, successful transduction was validated through western blot with an antibody targeting HMGB2 (MilliporeSigma; HPA053314). For ATAT1KO lines, knockdown was validated through qPCR. For gel source data, see Supplementary Fig. 1.
QPCR primer sequences:
ATAT1 F: 5′-CACAGTCCCACAGGTGAACA-3′
ATAT1 R: 5′-CTCCCTGCTTGGAGTCTTGG-3′
β-Actin F: 5′-CACCAACTGGGACGACAT-3′
β-Actin R: 5′-ACAGCCTGGATAGCAACG-3′
In vitro confinement and imaging
A375, HTB-4 and HTB-9 cells were subjected to overnight (approximately 16 h) confinement at a height of 3 µm using a static cell confiner (4Dcell). Pancreatic ductal adenocarcinoma cell lines (MIA-PaCa-2 and Panc-1) were confined at a 5-µm height owing to their larger size. The cells were plated 6 h before imaging in fibronectin-coated glass-bottom 35-mm dishes (FluoroDish) or glass-bottom six-well plates (MatTek). The cells were allowed to attach before confinement was applied. Confined cells were incubated at 37 °C and 5% CO2 overnight. For live imaging, dyes plus 10 µM verapamil were added to the plated cells 2–3 h before imaging. The dyes used for live imaging were SiR-tubulin (Spirochrome; 100 nM) and SiR-DNA (Spirochrome; 250 nM). Pharmacological inhibitors were added immediately before applying confinement. The inhibitors used were Taxol (Tocris; 1097), tubacin (Selleck Chemicals; S2239), nocodazole (Tocris; 1228) and trichostatin A (MilliporeSigma; T8552). Live imaging was performed on an LSM 880 (ZEISS) confocal microscope at 37 °C and 5% CO2, at ×63 magnification and 5–10 min of temporal resolution, using ZEN Black v.2.3 SP1 software (ZEISS). For immunofluorescence, cells were fixed with 4% paraformaldehyde for 15 min at room temperature before proceeding with staining and imaging, as described below.
In vitro proliferation and invasion assays
The CyQUANT Direct Red Cell Proliferation Assay (Thermo Fisher Scientific; C35013) was used to assay cell proliferation. Cells were plated at a density of 500 cells per well in 96-well plates and allowed to grow for 72 h. The cell number was quantified using the CyQUANT Direct Red Nuclei Acid Stain and Background Suppressor added at a 1:1 ratio to the cell culture medium, and the intensity was read out at 622 nm on a plate reader. For invasion assays, VitroGel Cell Invasion Assay Kit (TheWell Bioscience; IA-VHM01-1P) and Cultrex Collagen I Cell Invasion Assay kit (Bio-Techne; 3457-096-K) were used. The cells were serum-starved overnight in DMEM, plated in the upper chamber of the invasion assay insert and allowed to migrate for 18 h. The cell number was quantified using crystal violet or calcein staining.
Immunofluorescence staining and imaging
Cells were plated on glass CC2-coated chamber slides (Thermo Fisher Scientific) or fibronectin-coated glass-bottom dishes (FluoroDish) and allowed to attach for approximately 24 h. The cells were fixed with 4% paraformaldehyde for 15 min, permeabilized with 0.1% Triton in PBS and blocked in 10% goat serum (Thermo Fisher Scientific) for 1 h at room temperature. The primary antibodies used were rabbit anti-HMGB2 (Abcam; ab124670), rabbit anti-HMGB1 (Abcam; ab18256), rabbit anti-HMGA1 (Abcam; ab129153), mouse anti-α-tubulin (MilliporeSigma; CP06), chick anti-β-tubulin (Novus Biologicals; NB100-1612), mouse anti-acetylated tubulin (MilliporeSigma; 6793), rabbit anti-acetylated tubulin (Cell Signaling Technology (CST); 5335), rat anti-tyrosinated tubulin (MilliporeSigma; MAB1864-I), mouse anti-polyglutamylated tubulin (MilliporeSigma; T9822), mouse anti-GFP (Abcam; ab1218), rabbit anti-H3Ac (MilliporeSigma; 06-599), mouse anti-Annexin V (Santa Cruz Biotechnology; sc-74438), rabbit anti-cleaved caspase-3 (CST; 9661), rabbit anti-cleaved PARP (CST; 5625), rabbit anti-YAP (CST; 14074), mouse anti-Twist (Abcam; ab50887), rabbit anti-Snail (CST; 3879), rabbit anti-SMAD3 (Abcam; ab40854) and rabbit anti-SYNE2 (Abcam; ab204308). All primary antibodies were used at 1:200. The cells were incubated with primary antibodies overnight at 4 °C, washed in PBS and incubated with the appropriate fluorescently labelled secondary antibody (1:250). Alexa Fluor 488 conjugated phalloidin (CST; 8878S), when used, was added at 1:50, and Hoechst was added at 1:1,000. The cells were mounted in VECTASHIELD (Vector Laboratories) and allowed to cure overnight. Stained cells were imaged on a ZEISS LSM 880 confocal at ×40 or ×63 resolution using ZEN Black v.2.3 SP1 software (ZEISS). A Gaussian blur with a radius of 0.5–0.75 pixels was occasionally applied to images to reduce noise for visualization purposes only.
Staining of human tumour samples
Human melanoma tissue microarrays were obtained from TissueArray.Com (Me481f). Slides were baked at 60 °C for 20 min and deparaffinized in consecutive xylene and ethanol washes. Antigen retrieval was performed using 1X IHC Antigen Retrieval Solution (Thermo Fisher Scientific; 00-4955-58) heated at 95 °C for 20 min in a pressure cooker. After washing in PBS, the samples were blocked in 10% goat serum (Thermo Fisher Scientific) for 1 h at room temperature before incubation overnight at 4 °C in the following primary antibodies, all diluted in blocking buffer at 1:200: rabbit anti-HMGB2 (Abcam; ab124670) and mouse anti-acetylated tubulin (MilliporeSigma; 6793). After washing in PBS, the slides were incubated with the appropriate fluorescently labelled secondary antibody (1:250) and Hoechst (1:1,000). After washing in PBS, a final incubation was performed with a fluorescently conjugated rabbit anti-S100a6 antibody (Abcam; ab204028; 1:250) to label tumour cells before mounting the slides in VECTASHIELD. The slides were imaged on a Pannoramic slide scanner (3DHISTECH) using a ×20/0.8 numerical aperture objective, with higher-resolution images acquired on an LSM 880 confocal (ZEISS), as described above.
Image analysis
Images were analysed using CellProfiler70, TrackMate71 and MATLAB v.R2021b and R2023b (MathWorks). For images of fixed cells, the cells were segmented in CellProfiler using Hoechst staining to generate a nuclei mask and phalloidin or other cytoskeletal staining to generate a whole-cell mask. The mean intensity per cell/nucleus was quantified, and expression of nuclear-localized proteins was normalized to Hoechst intensity per nucleus. For quantification of live imaging data, HMGB2–GFP intensity per cell over time was quantified using TrackMate. The resulting intensity data were analysed in MATLAB by fitting a line to each curve and automatically removing curves in which more than four data points differed from the line of best fit by more than 0.2 a.u. In all cases, plotting and statistics were done in MATLAB. The images were assembled for figure preparation using Fiji (v.2.14).
Fluorescence recovery after photobleaching
A375 cells expressing HMGB2–GFP were confined for approximately 18 h before FRAP measurements. FRAP was done on an LSM 880 confocal at 37 °C with 5% CO2 using a ×63 oil immersion lens and ZEN Black v.2.3 SP1 software (ZEISS). A 5-µm circular diameter region of interest was defined within the nucleus of each cell before photobleaching at 405 nm and 488 nm wavelengths for ten pulses. One time point was acquired before photobleaching. Fluorescence recovery was imaged at 0.2-s intervals for a total of 20 s. All analyses were performed in MATLAB. For analysis, fluorescence within the region of interest was normalized to the fluorescence at the initial time point (before photobleaching). Samples in which the fluorescence within the region of interest was not bleached to at least 25% of the pre-bleaching value were automatically removed from the analysis. Each recovery curve was fitted with a two-component exponential using the function ‘fit’ with the ‘exp2’ parameter: F(t) = y0 + A1(1 − e−t/τ1) + A2(1 − e−t/τ2), where y0 represents the fluorescence immediately after photobleaching, A1 represents the amplitude of the fast-diffusing population, A2 represents the amplitude of the slow-diffusing population, t is time and τ1 and τ2 correspond to the time constants for the fast-diffusing and slow-diffusing populations, respectively.
Atomic force microscopy
Cells were plated on glass-bottom Petri dishes (FluoroDish FD35) and confined for 18 h, as described above. Immediately after removing the dish from the confiner, cell stiffness was measured using a NanoWizard V microscope (JPK Bruker) in QI Advanced mode. The samples were maintained at 37 °C during imaging using the PetriDishHeater (Bruker). For cell stiffness mapping, 1-μm-diameter spherical AFM probe (silicon nitride cantilever; nominal spring constant k = 0.2 N m−1; SAA-SPH-1UM; Bruker) was used. Each spring constant of the AFM probe was measured using the thermal noise method in liquid at 37 °C. For the stiffness mapping, a 2 nN set point was used (60 μm × 60 μm image size with 32 × 32 pixels of resolution) to ensure up to 10–20% sample indentation to avoid glass surface influence. The data were processed with JPK Data Processing software using the Hertz model with 0.5 Poisson ratio as a fit parameter. To calculate nuclear stiffness, force maps were segmented on the basis of the corresponding cell height measurements to extract the nuclear region.
TurboID experiments
Generation of TurboID constructs
Cloning of TurboID constructs and validation was performed, as described in a previous study35. The cyto-TurboID plasmid was obtained from Addgene. For TurboID–HMGB2, the TurboID cassette was amplified by polymerase chain reaction and cloned into pENTR–HMGB2 at the N terminus of HMGB2 using In-Fusion Cloning. For nuclear localization signal–TurboID, an entry vector was assembled using In-Fusion Cloning, containing the TurboID cassette, followed by three consecutive nuclear localization signal sequences. The pENTR–TurboID–HMGB2 and pENTR–TurboID–3XNLS constructs were subcloned into the pLX304 backbone by Gateway cloning using LR Clonase II Plus (Thermo Fisher Scientific). Stable A375 cell lines were generated, as described above, and expression and localization of the TurboID fusion protein were confirmed by immunofluorescence targeting haemagglutinin (found at the N terminus of the TurboID cassette). The TurboID activity was validated by pulsing the cells with 10 mM biotin (MilliporeSigma; 1071508), followed by both western blotting and immunofluorescence using fluorescently labelled streptavidin. IRDye 800CW Streptavidin (LI-COR; 926-32230) was used for western blotting, and streptavidin conjugated to Alexa Fluor 488 (Thermo Fisher Scientific; S11223) or Alexa Fluor 555 (Thermo Fisher Scientific; S21381) was used for immunofluorescence.
Sample preparation
For mass spectrometry experiments, ten million cells per condition and replicate were plated in 15-cm dishes. The medium was removed from the dishes and replaced with 10 mM biotin for 1 h. The labelling reaction was stopped by placing the dishes on ice and washing the cells five times with ice-cold PBS. The cells were then detached by scraping in ice-cold PBS and then pelleted and resuspended in radio-immunoprecipitation assay buffer + protease inhibitors. The cells were lysed by means of sonication (10% amplitude; 2 s per cycle; six cycles), and Bradford assay was used to measure the protein concentration. For each sample, 1-mg protein was incubated with streptavidin magnetic beads (Thermo Fisher Scientific; 88817) in radio-immunoprecipitation assay buffer overnight with rotation at 4 °C. The next day, the beads were pelleted using a magnetic rack, the supernatant was removed and the beads were washed once in 50 mM Tris–HCl (pH 7.5) and twice in 2 M urea in 50 mM Tris–HCl (pH 7.5).
Protein digestion
The beads were resuspended in 80 ml of 2 M urea and 50 mM EPPS (pH 8.5) and treated with dl-dithiothreitol (1 mM final concentration) for 30 min at 37 °C with shaking (1,100 rpm) on a Thermomixer (Thermo Fisher Scientific). Free cysteine residues were alkylated with 2-iodoacetamide (3.67 mM final concentration) for 45 min at 25 °C at 1,100 rpm in the dark. The reaction was quenched using 3.67 mM dithiothreitol, and LysC (750 ng) was added, followed by incubation for 1 h at 37 °C at 1,150 rpm. Finally, trypsin (750 ng) was added, followed by incubation for 16 h at 37 °C at 1,150 rpm. After incubation, the digest was acidified to pH less than 3 with the addition of 50% of trifluoroacetic acid (TFA), and the peptides were desalted on Sep-Pak C18 cartridges (Waters). Briefly, the cartridges were conditioned by sequential addition of (1) 100% methanol; (2) 70% acetonitrile (ACN)/0.1% TFA; and (3) 5% ACN/0.1% TFA twice. After conditioning, the acidified peptide digest was loaded onto the cartridge. The stationary phase was washed with 5% ACN/0.1% formic acid twice. Finally, peptides were eluted using 70% ACN/0.1% formic acid twice. Eluted peptides were dried under vacuum in a SpeedVac centrifuge followed by reconstitution in 12 μl of 0.1% formic acid, sonication and transfer to an autosampler vial. Peptide yield was quantified using NanoDrop (Thermo Fisher Scientific).
Mass spectrometry
Peptides were separated on a 25-cm column with a 75-mm diameter and 1.7-mm particle size composed of C18 stationary phase (IonOpticks; Aurora 3 1801220) using a gradient from 2% to 35% B over 90 min and then to 95% B for 7 min (buffer A, 0.1% formic acid in high-performance liquid chromatography-grade water; buffer B, 99.9% ACN and 0.1% formic acid) with a flow rate of 300 nl min−1 using a nanoElute 2 system (Bruker). Mass spectrometry data were acquired on a timsTOF HT (Bruker) with a CaptiveSpray source (Bruker) using a data-independent acquisition parallel accumulation–serial fragmentation (PASEF) method (dia-PASEF). The mass range was set from 100 to 1700 m/z, and the ion mobility range was set from 0.60 V s cm−2 (collision energy of 20 eV) to 1.6 V s cm−2 (collision energy of 59 eV) with a ramp time of 100 ms and an accumulation time of 100 ms. The dia-PASEF settings included a mass range of 400.0–1,201.0 Da, mobility range of 0.60–1.60 and a cycle time estimate of 1.80 s. The dia-PASEF windows were set with a mass width of 26.00 Da, mass overlap of 1.00 Da and 32 mass steps per cycle.
Data analysis
Raw data files were processed using Spectronaut v.18.5 (Biognosys) and searched with the Pulsar search engine with a human UniProt protein database downloaded on 15 August 2023 (226,261 entries). Cysteine carbamidomethylation was specified as a fixed modification, whereas methionine oxidation, acetylation of the protein N terminus and deamidation (NQ) were set as variable modifications. A maximum of two trypsin missed cleavages were permitted. Searches used a reversed sequence decoy strategy to control peptide FDR, and 1% FDR was set as the threshold for identification. Unpaired t-test was used to calculate P value in differential analysis, and volcano plot was generated on the basis of log2FC and q value (multiple testing corrected P value using Benjamini–Hochberg method). A q value ≤ 0.05 was considered the statistically significant cut-off.
Mouse experiments
Mouse in vivo studies were performed in accordance with the guidelines approved by the Memorial Sloan Kettering Cancer Center Institutional Animal Care and Use Committee and Research Animal Resource Center. The mice were housed under pathogen-free conditions, in an environment with controlled temperature (21.5 °C ± 1.5 °C) and humidity (55% ± 10%) and under 12 h light/dark cycles. For the drug efficacy studies, 6-week-old to 8-week-old athymic female mice (The Jackson Laboratory) were injected subcutaneously with five million A375 cells in a 50:50 mix with Matrigel (Corning). Once tumours reached an average volume of 100 mm3, the mice were randomized into two treatment groups (n = 4–6 mice per group) to receive either a vehicle control or trametinib (1 mg kg−1) in combination with dabrafenib (30 mg kg−1). Both drugs were delivered through oral gavage daily five times for 3 weeks. The mice were observed daily throughout the treatment period for signs of morbidity/mortality. Tumours were measured twice weekly using calipers, and volume was calculated using the following formula: length × width2 × 0.52. Body weight was also assessed twice weekly. For the HMGB2 deletion construct growth curve studies, tumour cells were implanted, as described above, and tumour volume was measured twice weekly. The animals were monitored until their tumour size reached 1,500 mm3, at which point tumours were collected, fixed in 10% formalin for 24 h, transferred to 70% ethanol and processed for histology. In accordance with limits established by the Memorial Sloan Kettering Cancer Center (MSKCC) Institutional Animal Care and Use Committee, the mice were euthanized when tumour burden exceeded 1,500 mm3. These limits were not exceeded in any of the experiments. Histology was performed by HistoWiz using the following antibodies for immunohistochemistry: mouse anti-BRAFV600E (Abcam; ab228461) and rabbit anti-acetylated tubulin (Abcam; ab179484). Two biological replicates consisting of four to six mice per condition (for a total of 10–11 mice per group) were performed for both experiments. No sample size calculation or blinding was performed.
Bulk RNA-seq and analysis
For bulk RNA-seq of A375 and SKMEL5 cells overexpressing HMGB2, three replicates of approximately one million cells each were pelleted and resuspended in TRIzol before snap freezing. For bulk RNA-seq of confined A375 cells, 200,000 cells were plated in each well of a six-well plate. Three wells were confined for approximately 18 h using a six-well static confiner (4Dcell) at 3-µm height, whereas the remaining three wells were left unconfined. The cells were then collected in TRIzol, pooling the three wells for each condition to generate samples of approximately 600,000 cells each. This process was repeated for a total of three independent biological replicates per condition. Library preparation and sequencing were done by Azenta Life Sciences. Raw sequencing reads were processed using FastQC (Babraham Bioinformatics) and Trimmomatic72 before alignment to the human genome hg38. All downstream analyses were performed in R (v.4.3.1). Differential gene expression was analysed using DESeq2 (ref. 73) with the default parameters. GSEA was performed using the fgsea74 R package (v.1.26) with Gene Ontology biological process pathway sets from MSigDB75.
Bulk ATAC-seq and analysis
Samples containing approximately 100,000 cells each were centrifuged at 700g for 5 min at 4 °C before being resuspended in 500-µl growth medium supplemented with 10% DMSO. The cells were frozen at −80 °C overnight before library preparation and sequencing were performed by Azenta Life Sciences. Sequencing reads were trimmed and filtered for quality control using TrimGalore (v.0.6.7) with a quality setting of 15, Cutadapt76 (v.4.0) and FastQC v.0.12.1. Reads were aligned to the human genome assembly hg38 using Bowtie 2 (ref. 77) (v.2.3.5.1) and were deduplicated using MarkDuplicates from Picard (Broad Institute; v.2.16). Peaks were identified using MACS2 (ref. 78) with a P-value setting of 0.001 using a publicly available melanocyte dataset (GSM3191792) as control. To generate a global peak atlas, blacklisted regions were removed before merging all peaks within a 500-bp region and quantifying reads using featureCounts. Differentially enriched peaks were identified using DESeq2 (ref. 73). Peak gene mapping was done by assigning all intergenic peaks to that gene and, in other cases, by genomic distance to the transcription start site. Pathway was analysed using clusterProfiler79. Tornado plots were generated with deepTools80 (v.3.5.1) functions (computeMatrix and plotHeatmap), with genes annotated from the indicated pathway sets. Motif enrichment was analysed using Homer81 (v.4.11.1) functions (findMotifsGenome and annotatePeaks).
ChIP sequencing
Sample preparation and sequencing
For profiling of HMGB2 binding in A375 and SKMEL5 cells, freshly collected cells (approximately 20 million cells/replicate/condition) were crosslinked first with 1.5 mM of EGS (Thermo Fisher Scientific; 21565) for 20 min at room temperature and subsequently with 1% formaldehyde (Thermo Fisher Scientific; 28906) for 40 min at 4 °C. The reaction was quenched by the addition of glycine to the final concentration of 0.125 M. Fixed cells were washed twice with PBS and resuspended in SDS buffer (100 mM NaCl, 50 mM Tris–HCl (pH 8.0), 5 mM EDTA, 0.5% SDS and 1× protease inhibitor cocktail; Roche). The resulting nuclei were spun down, resuspended in the immunoprecipitation buffer (100 mM NaCl, 100 mM Tris–HCl (pH 8.0), 5 mM EDTA and 5% Triton X-100) at 1 ml per 0.5 million cells mixed in 2:1 ratio, with the addition of 1× protease inhibitor cocktail (MilliporeSigma; 11836170001). The nuclei were processed on a Covaris E220 Focused-ultrasonicator to achieve an average fragment length of 200–300 bp with the following parameters: peak incident power = 140, duty factor = 5, cycles per burst/burst per second = 200 and time = 20 min (for A375 cells) or 45 min (for SKMEL5 cells). Chromatin concentrations were estimated using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific; 23227) according to the manufacturer’s instructions. The immunoprecipitation reactions were set up in 500 µl of the immunoprecipitation buffer in Protein LoBind Tubes (Eppendorf; 22431081) and pre-cleared with 50 µl of Dynabeads Protein G (Thermo Fisher Scientific; 10004D) for 2 h at 4 °C. After pre-clearing, the samples were transferred into new Protein LoBind Tubes and incubated overnight at 4 °C with 5 µg of HMGB2 (Abcam; ab67282), V5 (Abcam; ab9116; used for SKMEL5 ChIP only) and H3K4me3 (Epicypher; 13-0041) antibodies. For normalization purposes, 5 µl of Drosophila spike-in chromatin (Active Motif; 53083) and 2 µl of spike-in antibody (Active Motif; 61686) were added to each reaction. The next day, 50 µl of BSA-blocked Dynabeads Protein G was added to each reaction and incubated for 2 h at 4 °C. The beads were then washed twice with low-salt washing buffer (150 mM NaCl, 1% Triton X-100, 0.1% SDS, 2 mM EDTA and 20 mM Tris–HCl (pH 8.0)), twice with high-salt washing buffer (500 mM NaCl, 1% Triton X-100, 0.1% SDS, 2 mM EDTA and 20 mM Tris–HCl (pH 8.0)), twice with LiCL wash buffer (250 mM LiCl, 10 mM Tris–HCl (pH 8.0), 1 mM EDTA, 1% Na deoxycholate and 1% IGEPAL CA-630) and once with TE buffer (10 mM Tris–HCl (pH 8.0) and 1 mM EDTA). The samples were then reverse-crosslinked overnight in the elution buffer (1% SDS and 0.1 M NaHCO3) and purified using the ChIP DNA Clean & Concentrator kit (Zymo Research; D5205) following the manufacturer’s instructions. After quantification of the recovered DNA fragments, libraries were prepared using the ThruPLEX DNA-Seq Kit (Takara Bio; R400676) following the manufacturer’s instructions, purified with SPRIselect magnetic beads (Beckman Coulter; B23318) and quantified using a Qubit Flex fluorometer (Thermo Fisher Scientific) and profiled using TapeStation (Agilent). The libraries were sent to MSKCC Integrated Genomics Operation core facility for sequencing on an Illumina NovaSeq 6000 (approximately 30–40 million 100-bp paired-end reads per library).
Data analysis
ChIP–seq reads were trimmed and filtered for quality and library adaptors using TrimGalore (v.0.4.5) with a quality setting of 15 and running cutadapt76 (v.1.15) and FastQC (v.0.11.5). Reads were aligned to human assembly hg38 using Bowtie 2 (v.2.3.4.1) (ref. 77) and were deduplicated using MarkDuplicates in Picard Tools (v.2.16.0). To ascertain enriched regions, MACS2 (ref. 78) was used with a P-value setting of 0.001 and run against a matched control for each condition. A peak atlas was created by combining the superset of all peaks using the ‘merge’ function in the BEDTools suite (v.2.29.2). Read density profiles were created using deepTools ‘bamCoverage’ (v.3.3.0), normalized to ten million uniquely mapped reads and with read pile-ups extended to 200 bp. The tool featureCounts (v.1.6.1) was used to build a raw count matrix, and DESeq2 was used to calculate the differential enrichment for all pairwise contrasts for experiments with replicates. For single-sample data, MACS2 was run by swapping bams of different conditions to find differential regions. Peak gene associations were created by assigning all intragenic peaks to that gene, whereas intergenic peaks were assigned using the linear genomic distance to the transcription start site. GSEA82 was performed with the pre-ranked option and default parameters, in which each gene was assigned the single peak with the largest (in magnitude) log2FC associated with it. Composite and tornado plots were created using deepTools80 (v.3.3.0) by running computeMatrix and plotHeatmap on normalized bigwigs with average signal sampled in 25-bp windows and flanking region defined by the surrounding 2 kb. Motif signatures were obtained using Homer81 (v.4.5).
Reanalysis of human melanoma scRNA-seq data
Human melanoma scRNA-seq data from a previous study7 were downloaded from Gene Expression Omnibus (GEO) (GSE115978). All analyses were performed in R using Seurat83 (v.4.4.0 and v.5.0.1). The count matrix was normalized using sctransform. Clustering was done using Seurat functions (FindNeighbors and FindClusters) with a resolution of 0.8. Cell types and treatment status were annotated using metadata from the original publication7. Cell types were classified using gene lists from a previous study2 and the Seurat function AddModuleScore with default parameters. Module scores were scaled between 0 and 1. Cells were classified by differentiation state on the basis of the highest expression score for the given gene modules. Differentially expressed genes were calculated using the Seurat function FindMarkers with default parameters. GSEA was performed using fgsea, as described above.
Statistics and reproducibility
All statistical analyses and plotting were performed in either R (for RNA-seq and ATAC-seq data; v.4.3.1) or MATLAB (for imaging data; v.R2021b). For scRNA-seq data, P values were calculated using the Wilcoxon rank-sum test with Bonferroni’s correction for multiple groups (R function pairwise.wilcox.test). Pearson correlation coefficients and corresponding P values were calculated using the R function cor.test. For differential expression analyses of bulk RNA-seq and bulk ATAC-seq data, P values were calculated in DESeq2 using the Wald test. For image analysis, P values were calculated using MATLAB functions (anova1 and multcompare) using the Tukey post hoc test. To calculate the cell migration velocity, the Euclidean distance travelled by individual cells across time points was measured using the MATLAB function pdist, and the velocity was calculated by dividing by the time step. For mouse experiments, we performed a series of likelihood ratio tests to investigate growth rate differences. A biexponential model was fit to the growth curve using maximum likelihood estimation to obtain estimates for the early-time and late-time growth rates, with a single exponential fit to vehicle data. For all representative images shown, the images represent at least three independent replicates.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Source link
