Osteosarcoma (OS) is the eighth most common childhood cancer, representing approximately 20% of all primary bone cancers and 2.4% of all pediatric malignancies. Despite multi-modal treatment with surgery, chemotherapy, and neoadjuvant approaches, outcomes remain unsatisfactory: the 3-year survival rate ranges from 50 to 75%, the 5-year rate from 60 to 85%, and 25 to 50% of patients subsequently develop metastatic disease, which is the principal cause of death.
Hyperthermia therapy, in which tissue is heated above normal temperatures to therapeutic effect, has been used in cancer treatment since 1989. It can activate systemic anti-tumor immune responses and induces apoptosis in human OS cells by altering membrane characteristics, intracellular ion concentrations, and membrane potential. However, the molecular gene regulatory response of osteosarcoma cells to hyperthermia and the mechanisms driving its anti-cancer effects are not well characterized.
When cells are heat-stressed, heat shock transcription factor 1 (HSF1) is released from its inactive cytoplasmic monomer form, translocates to the nucleus, and drives transcription of heat shock proteins (HSPs). HSF1 binding to gene promoters therefore serves as a direct readout of which genes are being regulated during hyperthermia. An increase in HSPs can paradoxically render hyperthermia less effective, making the identification of HSF1 targets a priority for optimizing thermal therapy.
The study used the publicly available ChIP-seq dataset GSE60984 from the Gene Expression Omnibus (GEO). This dataset includes three osteosarcoma cell samples generated on the Illumina Genome Analyzer IIx platform: one treated with anti-HSF1 antibody after 10 minutes of heat shock (Peek-10), one treated with anti-HSF1 antibody after 20 minutes of heat shock (Peek-20), and one untreated control sample.
ChIP-seq (Chromatin Immunoprecipitation with next-generation sequencing) captures the genome-wide binding locations of a specific protein (here HSF1) by immunoprecipitating DNA bound to that protein and then sequencing the fragments. The GSE60984 dataset was originally generated by Janus et al. to study NF-kB signaling pathways during hyperthermia, but this study re-analyzed it using a different computational pipeline focused on promoter-region binding and downstream miRNA targeting.
For miRNA-gene network construction, the study used TargetScan, a widely used database that predicts miRNA target sites by searching for conserved 8mer, 7mer, and 6mer complementary sequences in the 3' untranslated regions (UTRs) of mRNA targets. The network was visualized and modularized using Cytoscape software. Functional enrichment was performed with DAVID using GO and KEGG pathway terms at a significance threshold of p-value less than 0.05.
Raw FASTQ sequencing reads were adapter-trimmed with Cutadapt and aligned to the hg19 reference genome using Bowtie (maximum 2 mismatches; only uniquely mapping reads retained). MACS2 (Model-based Analysis of ChIP-Seq) was used to identify genome-wide HSF1 binding peaks in the 10-minute and 20-minute heat-shock samples versus the control, at a significance threshold of p less than 10 to the power of -5 and a default fragment length of 300 bp.
The resulting peaks were annotated using the ChIPseeker R/Bioconductor package, which assigns each peak to genomic features (promoter, exon, intron, 3' UTR, 5' UTR, or intergenic) and identifies the nearest transcription start site. Only peaks located within promoter regions were retained for downstream analysis, since promoter binding is most directly linked to transcriptional regulation of the downstream gene.
Genes with HSF1 binding in promoter regions in both the 10-minute (Peek-10) and 20-minute (Peek-20) datasets were selected as the overlapping set, representing genes consistently regulated by HSF1 during hyperthermia regardless of exposure duration. TargetScan was queried for all miRNAs predicted to target these 889 overlapping genes, and all miRNA-gene pairs were assembled into a regulatory network. Clique and module analysis within Cytoscape was used to identify network communities.
MACS2 identified a total of 19,626 HSF1 binding sites (Peek-10) and 18,015 binding sites (Peek-20) genome-wide in osteosarcoma cells after 10 and 20 minutes of hyperthermia respectively. Of these, 10.017% of Peek-10 peaks and 7.921% of Peek-20 peaks were located in promoter regions, corresponding to 1,880 genes (Peek-10) and 1,283 genes (Peek-20) with HSF1 binding near their transcription start sites.
The overlap between Peek-10 and Peek-20 promoter-region targets yielded 889 genes consistently bound by HSF1 in their promoters at both time points. This set represents genes whose transcription is regulated by HSF1 regardless of whether hyperthermia exposure was brief (10 minutes) or more sustained (20 minutes). Among the 40 highlighted overlap genes are SGMS2, CGGBP1, SOD1, DNAJC7, MLH1, SMAD1, CDKN1A (p21 overlapping with KSHV-relevant targets from related studies), and JAK3.
GO enrichment analysis of the 889 overlapping genes identified 122 significant GO terms (p less than 0.05). The top 10 most significant GO terms include protein folding, response to oxidative stress, release of cytochrome c from mitochondria, response to peptide hormone stimulus, rhythmic process, cranial nerve development, regulation of programmed cell death, and regulation of cell death, with the latter two terms being the most directly relevant to hyperthermia's anti-cancer mechanism. Three KEGG pathways were enriched: p53 signaling pathway, methane metabolism, and viral myocarditis.
TargetScan analysis of the 889 overlapping genes generated 13,657 miRNA-gene regulatory pairs, which formed the complete miRNA-gene network. Because this network was too large and complex for direct biological interpretation, module analysis was conducted to identify densely connected subgraphs. The network was partitioned into four modules by Cytoscape community detection.
Module 1, the largest, had a cluster score of 8.59 with 62 nodes and 262 edges. Module 2 scored 6.625 with 33 nodes and 106 edges. Module 3 scored 4.556 with 37 nodes and 82 edges. Module 4 scored 4.0 with 29 nodes and 56 edges. SGMS2 (sphingomyelin synthase 2) and CGGBP1 (CGG triplet repeat-binding protein 1) emerged as the most highly connected nodes in Module 1, regulated by more miRNAs than any other gene in the network.
SGMS2 encodes sphingomyelin synthase 2, linked to liver steatosis, atherosclerosis, and NF-kB signaling. NF-kB pathway activation promotes proliferation and progression in many cancers. CGGBP1 is a nuclear and midbody protein that regulates cytokinetic abscission and whose expression is important for cell cycle progression in multiple cancer cell lines. Cell cycle control directly affects proliferation and apoptosis in OS progression.
The three KEGG pathways enriched in the overlap gene set illuminate distinct dimensions of hyperthermia's mechanism. The p53 signaling pathway has been reported to participate in OS cell proliferation, metastasis, and angiogenesis. Hyperthermia in mutant-p53 glioblastoma cells triggers nitric oxide as an initiator of intercellular signal transduction, and evidence suggests similar p53-pathway involvement in OS. The p53 pathway is also closely linked to the enriched GO terms for cell cycle and apoptosis regulation.
The enrichment of regulation of programmed cell death and regulation of cell death among GO terms aligns with published evidence that hyperthermia induces apoptosis in human OS cells through endoplasmic reticulum stress and reactive oxygen species pathways. These HSF1-regulated genes may represent the transcriptional response that either amplifies or limits this apoptotic response, which could explain both hyperthermia's therapeutic effect and the development of thermal resistance.
The enrichment of methane metabolism and viral myocarditis pathways was an unexpected finding. The authors note that few studies have explored these pathways in OS, making their biological significance under hyperthermia uncertain. These findings may represent off-target enrichment due to shared gene membership with better-characterized pathways, or could reflect previously unappreciated connections between OS biology and these metabolic or viral response programs that warrant further investigation.
This study is entirely computational: no in vitro or in vivo experiments were performed to validate that SGMS2 or CGGBP1 are functionally important for hyperthermia's effects on osteosarcoma. The miRNA-gene pairs from TargetScan are predicted interactions based on seed sequence conservation, not experimentally confirmed bindings. The initial 13,657 pairs include many that are likely indirect or non-functional, and even the module analysis cannot eliminate false-positive predictions.
The ChIP-seq dataset GSE60984 was generated from a single osteosarcoma cell type under two heat shock durations, limiting the generalizability of promoter-binding patterns across the heterogeneous landscape of osteosarcoma cell lines and primary tumors. The original purpose of GSE60984 was to study NF-kB and TNF-alpha signaling under hyperthermia, so the dataset may not be optimally designed for a general HSF1 binding survey.
The overlap strategy (selecting genes with HSF1 binding in promoters at both 10 and 20 minutes) is straightforward but does not account for dynamic changes in binding intensity over time, dose-response relationships at different temperatures, or HSF1 binding outside promoter regions (which represents 89 to 92% of all peaks) that could also influence gene expression through enhancers or other regulatory elements.
The most pressing next step is experimental validation of SGMS2 and CGGBP1 as functional mediators of hyperthermia's anti-osteosarcoma effect. siRNA or CRISPR knockdown of these genes in osteosarcoma cell lines subjected to hyperthermia would test whether reducing their expression alters apoptosis, cell death, or proliferation responses. Co-treatment with NF-kB pathway inhibitors could test whether SGMS2's NF-kB modulatory role is relevant in the thermal context.
The p53 signaling pathway's consistent appearance in both KEGG enrichment and in prior literature on osteosarcoma suggests that combining hyperthermia with p53-pathway-targeting agents (such as MDM2 inhibitors or p53-reactivating compounds) could produce synergistic anti-tumor effects. Trials combining hyperthermia with standard chemotherapy (doxorubicin, cisplatin, methotrexate) in osteosarcoma are ongoing; molecular biomarkers like SGMS2 and CGGBP1 expression could be evaluated as predictive markers in such settings.
Expanding this analysis to additional osteosarcoma ChIP-seq datasets capturing HSF1 binding at different temperatures and timepoints, and integrating RNA-seq data to identify which of the HSF1-bound promoter genes are actually transcriptionally activated during hyperthermia, would greatly sharpen the mechanistic insights. Integrating patient tumor methylation data could also reveal whether HSF1-target genes are epigenetically silenced in certain osteosarcoma subtypes, influencing sensitivity to hyperthermia.