Immunohistochemistry

The immunohistochemical staining was conducted as previously described [24]. The scoring of IHC was performed following a previously described method [24]. Briefly, the extent was scored by the percentage of nuclear (for p-p65, p-STAT3 and Ki67) or cytoplasmic (for MLKL, cGAS and ISG15) or membrane (for RAGE, E-cad and N-cad) staining: 0 = less than 1%, 1 = 1–10%, 2 = 11–50%, 3 = 51–80%, and 4 = more than 80%. The intensity was scored as 1 = weak, 2 = moderate, and 3 = strong. Scoring was performed by two pathologists without prior knowledge of the sample information. The overall score is then calculated as (1 + intensity/3) * extent. For each Sect. 5 HPF (200x) was analyzed and the mean score was calculated as the final score.

The scoring of p-MLKL were conducted as we previously described [24]. In brief, the extent of necroptosis was scored as the percentage of p-MLKL positive cells: 0, no positive staining; 1, < 10% positive staining; 2, 10–30% positive staining; and 3, > 30% positive staining. The scoring was than binarized as Absent/Focal (score = 0–2) and Extensive (score = 3) for correlation and survival analysis.

Cell culture and reagents

HNSCC cell lines SCC25, FaDu, colorectal adenocarcinoma cell line HT-29, pancreatic adenocarcinoma cell line BxPC-3 and AsPC-1, histiocytic lymphoma cell line U937, breast adenocarcinoma cell line MDA-MB-415 were purchased from American Type Culture Collection (ATCC, USA). SCC25 were cultured in DMEM/F12 1:1 (Gibco, USA) supplemented with 10% fetal bovine serum (FBS, Gibco, USA) and 400 ng/ml hydrocortisone (Solaria, Beijing, China). FaDu was cultured in MEM (Gibco, USA) supplemented with 10% FBS. BxPC-3, AsPC-1 and U937 were cultured in RPMI 1640 (Gibco, USA) supplemented with 10% FBS. HT-29 was cultured in McCoy’s 5a medium supplemented with 10% FBS. MDA-MB-415 was cultured in L-15 medium supplemented with 15% FBS, 10mcg/ml insulin and 10 mcg/ml glutathione.

Human recombinant TNF-α was purchased from PeproTech (USA). Smac mimetic AT-406 (SM-406), pan-caspase inhibitor zVAD-fmk, RIP1 inhibitor Necrostain-1 (Nec-1), RIP3 inhibitor GSK’872, and MLKL inhibitor Necrosulfonamide (NSA), cGAS inhibitor G150 and STING inhibitor C-170 were purchased from Selleck (Selleckchem, Houston, TX, USA). Human recombinant His6-ISG15 (Cat# UL-510), human recombinant GST-ISG15 (Cat# UL-600) and human recombinant ISG15 (Cat# UL-601) were purchased from R&D system (USA). Recombinant human HMGB1 (His Tag were purchased from Sino Biological (China). Small molecular inhibitor BAY 11-7082 (Cat# HY-13453), Stattic (Cat# HY-13818), FPS-ZM1 (Cat# HY-19370) were purchased from MCE (USA).

Antibodies against p65 (Cat# 8242), p-p65 (Cat# 3033), STAT3 (Cat# 9139), p-STAT3 (Cat# 9145), p-STAT5 (Cat# 931), MMP-9 ((Cat# 13667), E-cadherin (Cat# 3195), N-cadherin (Cat# 13116), Vimentin (Cat# 5741), ISG15 (Cat# 2758), IFIT1 (Cat# 1469), cGAS (Cat# 83623, #79978), STING (Cat# 13647), IRF3 (Cat# 11904), p-STING (Cat# 50,907), p-TBK1 (Cat# 5483), p-IRF3 (Cat# 37829), p-H2AX (Cat# 9718), Toll-like receptor 2 (Cat# 12276) were purchased from Cell Signaling Technology (USA). Antibodies against MLKL (Cat# ab184718), p-MLKL (Cat# ab187091), IL-6 (Cat# ab233706), ISG15 (Cat# ab285367), CD11a (Cat# ab52895), HMGB1 (Cat# ab79823) were purchased from Abcam (UK). Antibodies against IFIT2 (Cat# sc-390724), IFIT3 (Cat# sc-393512), RAGE (Cat# sc-365154) were purchased from Santa Cruz (USA). Antibodies against GST (Cat# 66001-2-Ig), His (Cat# 66005-1-Ig) were purchased from Proteintech (USA). Anti-GAPDH (CWBio Cat# CW0100M) antibody was purchased from CWBIO (China).

Induction of cell death

The induction of apoptosis, necroptosis and passive necrosis were performed as previously described [24]. Briefly, combined treatment of TNF-α + Smac mimetic + zVAD-fmk (TSZ) was used to induce necroptosis, while TNF-α + Smac mimetic (TS) was used to induce apoptosis. Cells were pretreated with zVAD-fmk and Smac mimetic for 1 h. Then TNF-αwas added and treated cells for indicated time. DMSO was used as vehicle control. Classic freeze–thaw cycle (FT) was used to induce passive necrosis [7, 9, 64]. For inhibition of necroptosis, Necrostatin-1(Nec-1) were added together with S and Z before stimulation with TNF-α.

Transwell migration and invasion assay

Transwell migration and invasion assay are performed as previously described [24, 83, 87]. Briefly, 6.5-mm-diameter polycarbonate filters (8-μm pore size) and 24-well plates were used. For invasion assay, the filter was pre-coated with 1:20 diluted Matrigel (BD biosciences) for 2 h. Cells received indicated treatments were resuspended in 200ul serum-free medium and seeded in the top chamber and the bottom chamber was filled with medium containing 10% FBS. Cells were allowed to migrate for 24–48 h. The cells were fixed with 4% paraformaldehyde and stained by crystal violet. Five images per chamber were taken using inverted microscope under 100 × magnification (Axio, ZEISS, Germany) and the migrated cells were counted using ImageJ software.

PI staining

Propidium iodide (PI, BD Biosciences, USA) staining used for analysis of cell death rate as previously described [24]. Briefly, cells were harvested, washed twice with pre-cooled PBS and stained with PI solution for 5 min. The percentage of PI-positive cells was analyzed by CytoFlex (Beckman Coulter, USA).

Western blotting

Western blotting was performed as we previously described [24]. Briefly, cells were lysed in RIPA buffer (Beyotime, China). For released proteins, cultural supernatants were concentrated using AmiconUltra (Millipore, USA) centrifugal filters. Protein samples were quantified by BCA protein assay kit (CWBIO, China). 20ug proteins (10ug for proteins from the supernatants) were loaded for SDS-PAGE gel electrophoresis and then transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, USA). 2% skim milk or bovine serum albumin (BSA) were used for blocking unspecific bindings. The membranes were then incubated with primary antibodies at 4℃ overnight and then with secondary antibodies (EMAR, China) for 1 h at room temperature. Then the membranes were incubated in Immobilon ECL Ultra Western HRP Substrate (Millipore, USA) and the bands were detected using GeneGnome XRQ system (Syngene, USA).

Dox-inducible necroptosis model

Tet-on lentivirus carrying full length human MLKL with phospho-mimetic mutation at T357E/S358D was generated by Hanbio Technology (Shanghai, China) using pHBLV-TetOn-SV40-Puro-TRE3GS-MCS lentiviral vector. SCC25 and FaDu cells were infected and screened for stably transfected cells (MLKL-25/MLKL-FD) as previously described [24]. Dox-induced MLKL expression was validated by western blotting and qRT-PCR. For knockdown of ISG15 in this dox-inducible necroptosis model, MLKL-FD and MLKL-25 cells were further infected with shISG15 lentivirus and screened for stably transfected cells (shISG15-MLKL-FD/25).

Quantitative real-time PCR (qRT-PCR)

qRT-PCR was performed as previously described [24]. Briefly, total RNA was extracted using RNAzol (Molecular Research Center, Inc, USA) agent and quantified using NanoDrop2000. RNA reverse transcription process was performed using HiScript III RT SuperMix for qPCR kit (Vazyme, China) and qRT-PCR was performed using ChamQ SYBR qPCR Master Mix kit (Vazyme, China). Primer sequences are shown in Tables 2, 3.

Table 2 Primers for qRT-PCRTable 3 Primers for genomic PCRImmunofluorescence

Treated cells were fixed with 4% paraformaldehyde for 30 min followed by membrane permeabilization by 0.1%Triton for 15 min. Goat serum was used to block nonspecific bindings for 30 min. Cells were incubated with primary antibody overnight at 4℃ and then with fluorescent secondary antibody for 1 h at room temperature. At last, nuclei were stained by DAPI and Antifade Mounting Medium (Beyotime, China) were added to prevent fluorescence quenching. Stained cells were visualized under the LSM980 laser scanning confocal microscope (Axio, ZEISS, Germany). Quantification was performed by calculating the percentage of positive staining cells in each field. For each group, at least 4 randomly selected fields are used for quantification. The quantification of colocalization was performed by analyzing Pearson’s correlation using Image-Pro Plus 6.0 software.

Collection of DAMPs

After inducing SCC25 and FaDu with different treatments for 12 and 24 h respectively, culture medium was renewed to eliminate the stimulation of drugs. Cells were cultured for additional 12 h and the culture medium were collected. The conditioned medium was centrifuged under 2000 × g for 10 min to eliminate cell debris before using for further experiments. For collection of the DAMPs from accidental necrosis, cells were cultured for the same time and went through three times frozen-thaw (F/T) cycles and centrifuge to eliminate cell debris. For DAMPs collection in Dox-inducible necroptosis model, MLKL-25 and MLKL-FaDu cells were induced with 1ug/ml Doxycycline for 12 h before supernatant were collected.

In vivo assay

For Dox-induce intratumoral necroptosis model, luciferase-containing FaDu cells (Luc-FD) and dox-inducible necroptosis cells (MLKL-FD/shISG15-MLKL-FD) cells were mixed at a ratio of 3:1 and reach a final concentration of 6 × 106 cells/ml. The mixing ratio of 3:1 was determined based on our pre-experiment to ensure an adequate amount of dox-induced necroptosis in the xenograft tumor which largely mimic the extensive necroptosis we previously observed in the clinical specimens [24]. Then 50ul of cells were orthotopically inoculated into the BALB/c nude mice. Five days after inoculation, 0.2 mg/ml Doxycycline was added to the drinking water to induce intratumoral necroptosis. For the inhibition of RAGE, 1 mg/kg FPS-ZM1 were intraperitoneally injected per day from Day 5. Mice in other groups were injected with same volume of saline. At day 14, in vivo imaging (IVIS system, PerkinElmer, USA) was used to observe cervical lymphatic metastasis before the mice were sacrificed, and then the primary tumors and cervical lymph nodes were collected for further analysis.

For intratumoral DAMPs injection model, Luc-FD cells were inoculated into the BALB/c nude mice. DAMPs were collected as above mentioned and concentrated by AmiconUltra (Millipore, USA) centrifugal filter. At day 1, 5 and 10, each mouse was intratumorally injected with 50ul of DAMPs. At day 15, cervical lymphatic metastasis was observed by live imaging (IVIS system, PerkinElmer, USA). Mice were then sacrificed, and the primary tumors and cervical lymph nodes were collected for further analysis.

The tumor volume was calculated as: volume = long diameter × short diameter^2 × 1/2.

The cervical lymphatic metastasis was analyzed by HE and pan-CK staining.

RNA-seq

SCC25 cells were treated by apoptotic (TS), necrotic (FT) and necroptotic (TSZ) DAMPs for 24 h, and total RNA were extracted using Trizol (Invitrogen, USA). The RNA-seq was performed by BGI (Shenzhen, China). Differential expression analysis was performed using the DESeq2(v1.4.5) with Q ≤ 0.05. GO (http://www.geneontology.org/) and KEGG (https://www.kegg.jp/) enrichment analysis of annotated different expression gene was performed by Phyper (https://en.wikipedia.org/wiki/Hypergeometric_distribution) based on Hypergeometric test. The significant levels of terms and pathways were corrected by Q value with a rigorous threshold (Q ≤ 0.05) by Bonferroni. Functional alterations were further evaluated by gene-set enrichment analysis (GSEA).

Label-free proteomic analysis

The apoptotic (TS), necrotic (FT) and necroptotic (TSZ) DAMPs from SCC25 cells were collected as aforementioned. The label-free proteomic analysis was conducted under the assistance of Applied Protein Technology (Shanghai, China). The experimental procedures were as follow:

Protein extraction and digestion

SDT (4%SDS, 100 mM Tris–HCl, 1 mM DTT, pH7.6) buffer was used for sample lysis and protein extraction. Proteins was quantified with the BCA Protein Assay Kit (Bio-Rad, USA) following by tryptic digestion according to the filter-aided sample Preparation (FASP) procedure.

200 μg of proteins were incorporated into 30 μl SDT buffer (4% SDS, 100 mM DTT, 150 mM Tris–HCl pH 8.0). The detergent, DTT and other low-molecular-weight components were removed using UA buffer (8 M Urea, 150 mM Tris–HCl pH 8.0) by repeated ultrafiltration (Microcon units, 10 kD). Then 100 μl iodoacetamide (100 mM IAA in UA buffer) was added to block reduced cysteine residues and the samples were incubated for 30 min in darkness. The filters were washed with 100 μl UA buffer three times and then 100 μl 25 mM NH4HCO3 buffer twice. Finally, the protein suspensions were digested with 4 μg trypsin (Promega) in 40 μl 25 mM NH4HCO3 buffer overnight at 37 °C, and the resulting peptides were collected as a filtrate. The peptides of each sample were desalted on C18 Cartridges (Empore™ SPE Cartridges C18 (standard density), bed I.D. 7 mm, volume 3 ml, Sigma), concentrated by vacuum centrifugation and reconstituted in 40 µl of 0.1% (v/v) formic acid. The peptide content was estimated by UV light spectral density at 280 nm using an extinctions coefficient of 1.1 of 0.1% (g/l) solution that was calculated on the basis of the frequency of tryptophan and tyrosine in vertebrate proteins.

SDS-PAGE

20 µg of protein for each sample were mixed with 5X loading buffer respectively and boiled for 5 min. The proteins were separated on 12.5% SDS-PAGE gel. Protein bands were visualized by Coomassie Blue R-250 staining.

LC–MS/MS analysis

LC–MS/MS analysis was performed on a Q Exactive mass spectrometer (Thermo Scientific) that was coupled to Easy nLC (Proxeon Biosystems, now Thermo Fisher Scientific) for 60/120/240 min. The peptides were loaded onto a reverse phase trap column (Thermo Scientific Acclaim PepMap100, 100 μm*2 cm, nanoViper C18) connected to the C18-reversed phase analytical column (Thermo Scientific Easy Column, 10 cm long, 75 μm inner diameter, 3 μm resin) in buffer A (0.1% Formic acid) and separated with a linear gradient of buffer B (84% acetonitrile and 0.1% Formic acid) at a flow rate of 300 nl/min controlled by IntelliFlow technology. The mass spectrometer was operated in positive ion mode. MS data was acquired using a data-dependent top10 method dynamically choosing the most abundant precursor ions from the survey scan (300–1800 m/z) for HCD fragmentation. Automatic gain control (AGC) target was set to 3e6, and maximum inject time to 10 ms. Dynamic exclusion duration was 40.0 s. Survey scans were acquired at a resolution of 70,000 at m/z 200 and resolution for HCD spectra was set to 17,500 at m/z 200, and isolation width was 2 m/z. Normalized collision energy was 30 eV and the underfill ratio, which specifies the minimum percentage of the target value likely to be reached at maximum fill time, was defined as 0.1%. The instrument was run with peptide recognition mode enabled.

Identification and quantitation of proteins

The MS raw data for each sample were combined and searched using the MaxQuant 1.5.3.17 software for identification and quantitation analysis. Related parameters and instructions are as follows:

Item

Value

Enzyme

Trypsin

Max missed cleavages

2

Fixed modifications

Carbamidomethyl (C),

Variable modifications

Oxidation (M),

Main search

6 ppm

First search

20 ppm

MS/MS Tolerance

20 ppm

Database

uniprot_Homo_sapiens_188433_20200217.fasta

Database pattern

Reverse

Include contaminants

True

protein FDR

 ≤ 0.01

Peptide FDR

 ≤ 0.01

Peptides used for protein quantification

Use razor and unique peptides

Time window (match between runs)

2 min

protein quantification

LFQ

min. ratio count

1

Bioinformatic analysis

For Bioinformatic analyses, Cluster 3.0 (http://bonsai.hgc.jp/~mdehoon/software/cluster/soft-ware.htm) and Java Treeview software (http://jtreeview.sourceforge.net) were used to performing hierarchical clustering analysis. CELLO (http://cello.life.nctu.edu.tw/) which is a multi-class SVM classification system, was used to predict protein subcellular localization. Protein sequences are searched using the InterProScan software to identify protein domain signatures from the InterPro member database Pfam. The protein sequences of the selected differentially expressed proteins were locally searched using the NCBI BLAST + client software (ncbi-blast-2.2.28 + -win32.exe) and InterProScan to find homologue sequences, then gene ontology (GO) terms were mapped, and sequences were annotated using the software program Blast2GO. The studied proteins were blasted against the online Kyoto Encyclopedia of Genes and Genomes (KEGG) database (http://geneontology.org/) to retrieve their KEGG orthology identifications and were subsequently mapped to pathways in KEGG. Enrichment analyses were applied based on the Fisher’ exact test, considering the whole quantified proteins as background dataset. Benjamin-Hochberg correction for multiple testing was applied to adjust derived p-values, only functional categories and pathways with p-values under a threshold of 0.05 were considered as significant.

Gene knockdown by siRNA/shRNA

Small interfering RNAs were designed and synthesized by RiboBio (Guangzhou, China). The siRNA sequences are shown in Table 4. Additionally, siRNA targeting RAGE (Cat# sc-36374) and corresponding negative control (Cat# sc-37007) were purchased from Santa Cruz (USA). For siRNA transfection, Lipofectamine®3000 was used as transfection agent according to the user protocol.

Table 4 Sequences for siRNAs used in this study

shISG15 lentivirus were generated by GeneCopoeia (USA) using psi-LVRU6H lentiviral vector based on the si-ISG15 sequence. Lentiviral transfection was conducted as previously described [24].

Co-immunoprecipitation (Co-IP)

Protein A/G Magnetic Beads (MCE, USA) were pre-washed twice with 0.5% PBST and then incubated with primary antibodies on a rotator for 2 h at 4 ℃. The beads were then washed four times with 0.5% PBST and incubated with protein samples at 4 ℃ overnight. After washed four times with 0.5% PBST, the beads were resuspended in 1 × SDS-PAGE loading buffer (Cwbio, China) and incubated at 99 ℃ for 10 min to elute the immunoprecipitates for western blotting analysis.

For sequential IP of His-ISG15 and biotinylated surface protein, FaDu cells were treated with recombinant His-ISG15 for 2 h, then the surface protein were biotinylated using EZ-Link Sulfo-NHS-SS-Biotin (Thermo Fisher Scientific, USA) according to the manufacturer’s protocol, un-biotinylated FaDu cells were used as control. Cells were then lysed and centrifuged to remove the debris. The cell lysates were then incubated with anti-His Magnetic Beads (Cat#HY-K0209, MCE, USA) at 4℃ overnight. Captured proteins were eluted by Glycine (0,15M, pH = 2.5) following by incubating with Streptavidin Magnetic Beads (Cat#HY-K0208, MCE, USA) at 4 ℃ for 4 h. The beads were resuspended in 1 × SDS-PAGE loading buffer (Cwbio, China) and incubated at 99 ℃ for 10 min to elute the immunoprecipitates for western blotting analysis.

Molecular docking

The ligand-receptor docking between ISG15 and TLR1/2/RAGE/CD11a were conducted using Rosetta2020 (Cambridge, USA) software under the assistance of Shenzhen Shuli Tech Co., Ltd. (Shenzhen, China). Briefly, the protein conformations were extracted from Protein Data Bank (PDB), the PDB codes are as follow: 1lfa (CD11a), 1z2m (ISG15), 2z7x (TLR1/2), 3o3u (RAGE). Pretreatment of protein crystal structures included removing water molecules, original ligands, and other irrelevant protein conformations. Then global docking was performed by Rosetta 2020 and the snapshots were rendered by Pymol. For each pair of proteins, 10,000 conformations were generated. All conformations were then assessed by the InterfaceAnalyzer module of Rosetta 2020 and the conformations with packstate ≥ 0.65 and dG_separated/dSASAx100 ≤ −1.5 were selected. Finally, the conformation with the highest total score was selected as the most possible binding mode and was further visualized by Pymol and Ligplot software.

Analysis of the cytosolic release of mitochondrial DNA and its binding with cGAS

SCC25 cells were treated for indicated time, and the nuclear, mitochondrial, and cytosolic fractions were isolated using Mitochondria Isolation Kit for Cultured Cells (Thermo Fisher Scientific, USA). DNA were then extracted from mitochondrial, cytosolic, and nuclear fractions using Quick-gDNA MiniPrep Kit (Zymo Research, USA). Genomic PCR and DNA electrophoresis were performed to analyze the cytosolic release of mtDNA. The primers were listed in Table 2.

For the PicoGreen-MitoTracker double staining, the mitochondrial DNA were stained by PicoGreen (Lumiprobe, USA) for 1 h and the mitochondria were stained by Mitotracker red CMXRos (Thermo Fisher Scientific, USA) for 15 min. The nuclei were stained by Hoechst 33342 and then cells were treated by DMSO/TSZ/TSZ + Nec-1 for 8 h. Treated cells were visualized and captured using LSM980 laser scanning confocal microscope (Axio, ZEISS, Germany), and the colocalization was analyzed by Image Pro Plus 6.0 software. For continuous imaging, cells were visualized and captured under PE Operetta CLS (PerkinElmer, USA).

To detect the binding between cGAS and mtDNA, SCC25 cells were treated by DMSO/TSZ for 8 h. The cytosolic fractions were isolated as aforementioned. Cytosolic cGAS were then pulled down by immunoprecipitation using Protein A/G magnetic beads (MCE, USA) as aforementioned. DNA were then directly extracted from the beads using Quick-gDNA MiniPrep Kit (Zymo Research). Genomic PCR and DNA electrophoresis were performed to detect cGAS-bound mtDNA.

For ethidium bromide (EB)-mediated mtDNA depletion, SCC25 cells were cultured in the presence of 450 ng/ml or 600 ng/ml EB for 4 days. Cells were then seeded in 6 well plate and cultured without EB for 24 h to be ready for downstream experiments. The depletion of mtDNA was validated by genomic PCR.

Patients’ cohort and tissue samples

A total of 108 patients that were diagnosed as head and neck squamous cell carcinoma (HNSCC) and received radical surgeries between 2015 and 2020 at the Department of Oral and Maxillofacial Surgery, Hospital of Stomatology, Sun Yat-sen University, were included in this study (Table 1). All patients had no history of pre-operative chemotherapy or radiotherapy. Paraffin-embedded primary tumor tissues were used for immunohistochemical and immunofluorescent staining. Clinicopathological parameters and follow-up data for all the study participants were collected. The starting point for patients’ survival was the date of surgery. The endpoint for OS, RFS and PFS was the date of patients’ death, locoregional recurrence, and tumor progression (locoregional recurrence or lymphatic/distant metastasis), respectively. Clinical staging and histological grading were based on the 8th edition of UICC/AJCC TNM classification. This study is approved by the Ethical Committee of the Stomatological Hospital of Sun Yat-sen University (No. KQEC-2024-78-01). Informed consents were obtained from all the participants.

Statistics

SPSS 20.0 software (SPSS, USA) and GraphPad Prism 9 (GraphPad Software, USA) were used for statistical analysis. Kruskal–Wallis’s test and Dunn’s multiple comparison test were used for comparison of the IHC scores of p-p65 and p-STAT3. Fisher’s exact test was used for comparison of metastatic rates. Unpaired student’s t-test was used for the comparison of two groups of quantitative data (mRNA relative expression, cell counts, etc.). One-way ANOVA was used for comparing multiple groups of quantitative data, Turkey’s multiple comparisons test was used for pairwise comparison between each group. Simple linear regression model was employed to analyze the correlation between the expression of two proteins. p < 0.05 was considered statistically significant. Quantitative data are showed as mean ± SD unless stated otherwise. For each experiment, data are representative of at least two replications, with similar results obtained.

For bioinformatic analysis, aside from those mentioned above, STRING software (http://string-db.org/) was used for protein–protein interaction (PPI) analysis and MCL clustering. GEPIA (http://gepia.cancer-pku.cn/index.html) and cBioPortal (http://www.cbioportal.org/) were used for analyses of human HNSCC datasets derived from The Cancer Genome Atlas (TCGA: http://cancergenome.nih.gov/). The OmicShare online platform (https://www.omicshare.com) was used for hierarchical clustering, GSEA and visualization of results.