Exploration of histone protein γ‐H2AX as a prognostic factor in soft tissue sarcomas and its association with biological behavior, immune cell environment and survival in leiomyosarcoma

Abstract

This study evaluates the H2AX/γ‐H2AX expression in soft tissue sarcomas (STS), its implications for biological behavior and immune environment, and its potential as a prognostic biomarker. RNA‐Seq data from 237 STS were obtained from The Cancer Genome Atlas project. Patients were stratified by H2AX mRNA expression using a survival‐associated cutoff. Differentially expressed genes and pathways as well as immune signatures between H2AX ^high‐ and H2AX ^low tumors were identified with DESeq2 analysis, gene set enrichment analyses (GSEA), Enrichr pathway analysis and CIBERSORTx. Tissue microarrays of a different cohort of 291 STS were generated for immunohistochemical staining to assess γ‐H2AX protein expression, followed by statistical evaluation. High H2AX mRNA expression was associated with shorter overall survival (OS) in STS (p = 0.02), particularly in leiomyosarcomas (LMS) (p < 0.001), and was a negative prognostic factor in LMS (HR 11.15, p < 0.001). H2AX ^high LMS tumors showed upregulation of cell cycle‐related pathways, while H2AX ^low LMS exhibited increased inflammatory activity, including elevated M1 macrophage signatures and resting mast cell signatures (both p < 0.001). High γ‐H2AX protein levels were an independent negative prognostic factor in the total LMS cohort (HR 12.12, p = 0.025) and in the subgroup of non‐uterine LMS (HR 153.80, p = 0.013). Consistent with CIBERSORTx analysis, γ‐H2AX^low LMS showed higher mast cell infiltration than γ‐H2AX^high LMS (p = 0.038). In conclusion, H2AX mRNA and γ‐H2AX protein expression are associated with distinct biological behavior, differences in the immune cell environment, and might serve as useful prognostic biomarkers in LMS.

What's new?

The remarkably poor prognosis of metastasized soft tissue sarcoma (STS) has motivated a search for novel biomarkers and therapeutic approaches. A promising candidate biomarker is H2A histone family member X (H2AX), a key player in DNA damage detection and repair pathways. Here, the authors investigated the expression of H2AX and its phosphorylated protein, γ‐H2AX, in STS, as well as possible associations between H2AX/γ‐H2AX and STS prognosis. For leiomyosarcoma, a common STS subtype, elevated H2AX mRNA levels were associated with worse overall survival. In particular, high H2AX/γ‐H2AX expression in leiomyosarcoma was linked to changes in immune pathways and the immune microenvironment.

1. INTRODUCTION

The group of adult soft tissue sarcomas (STS) currently includes around 80 different entities classified by the WHO, with undifferentiated pleomorphic sarcomas (UPS), leiomyosarcomas (LMS) and dedifferentiated liposarcomas (ddLS) being the most common subtypes among these rare cancers. ¹ Therapeutic options for high‐grade STS are limited, with surgical resection, often combined with perioperative radiation and/or chemotherapy, being the standard treatment modality. ² Despite these therapeutic efforts, about 40% of all patients with high‐grade STS develop distant metastases. ³ The prognosis of metastasized STS is dismal, with only around 25% surviving 5 years despite maximal treatment. ⁴ , ⁵ , ⁶ This indicates an urgent need for biomarkers that enable therapeutic monitoring and risk‐stratification, as well as the development of additional therapeutic approaches.

H2A histone family member X (gene: H2AX, phosphorylated protein: γ‐H2AX) is a protein within the H2A histone family. Together with other histone proteins including H2B, H3 and H4 it forms histone octamers, around which the double‐stranded DNA is wrapped. ⁷ H2AX plays a central role in the complex DNA damage detection‐ and DNA repair pathways: Following the occurrence of double strand breaks (DSB), several kinases of the PI3 kinase family are recruited to the DSB site, including the ATM serine/threonine kinase, which then phosphorylates H2AX to γ‐H2AX. ⁸ γ‐H2AX acts as a crucial binding hub for various mediator and repair proteins and the formation of so‐called DNA damage‐/DNA repair foci, leading to cell cycle checkpoint activation, DNA repair and, in the event of repair failure, to apoptosis (Figure S1). ⁷ , ⁹ Because of its rapid phosphorylation at the DSB site, γ‐H2AX has been established as a reliable marker protein for DNA damage and chromosomal instability within the last two decades. ¹⁰ In this study, we aim to elucidate the role of γ‐H2AX expression in a large cohort of soft tissue sarcomas, examining its association with tumor biology and aggressiveness, and evaluating its potential as a biomarker.

2. MATERIALS AND METHODS

2.1. Generation of the TCGA cohorts of soft tissue sarcomas

To analyze H2AX expression in STS, we obtained bulk‐mRNA‐sequencing data (raw counts, TPM counts with RSEM‐normalization) and clinicopathological data from The Broad Institute Firehose GDC portal (https://gdac.broadinstitute.org/). This patient cohort included 50 undifferentiated pleomorphic sarcomas (UPS), 104 cases of leiomyosarcomas (LMS; non‐uterine: n = 75, uterine: n = 29), 58 cases of dedifferentiated liposarcoma (ddLS) and 25 cases of myxofibrosarcoma (MFS), a sarcoma entity formerly assigned to the UPS spectrum ¹ (Table S1).

2.2. Cohort dichotomization by H2AX expression and cBioPortal data acquisition

The STS cohorts were dichotomized using the Cutoff Finder tool ¹¹ : This tool generated the optimal cutoff value of H2AX mRNA expression (RSEM‐normalized TPM count) for most significant stratification by overall survival (OS) (log rank test, p value ≤.05). The largest difference in OS was observed in leiomyosarcomas (see Results). Thus, we obtained genomic alteration‐ and mutation data from the cBioPortal repository (https://www.cbioportal.org/) to compare H2AX ^high ‐ and H2AX ^low LMS. Commonly dysregulated and mutated genes in LMS, according to the WHO classification of tumors, and other alternated genes were compared (TP53, RB1, ATRX), with an adjusted p value ≤.05 considered significant (Students' t‐test with Benjamini Hochberg correction for multiple comparisons).

2.3. DESeq2 analysis

We further analyzed H2AX ^high ‐ and H2AX ^low LMS, dichotomized by the determined cutoff value, using the DESeq2 Bioconductor package (v3.16) to identify differentially expressed genes. Bulk‐mRNA‐Seq data from n = 30 samples of benign smooth muscle tissue from the openly available GTEx repository were used as control tissue (https://gtexportal.org/home/). For noise reduction and further normalization, a log fold change shrinkage method (apeglm package v.3.16) was applied. ¹² An adjusted p‐value of ≤0.05 was considered significant (Wald test with Bonferroni correction).

2.4. Gene Set Enrichment Analysis (GSEA) in H2AX ^high and H2AX ^low leiomyosarcomas

Significantly up‐ and downregulated genes were analyzed using the Gene Set Enrichment Analysis Software (GSEA, v.4.3.3, Broad Institute, Cambridge, USA). Genes with a log2 fold change (log2FC) of ≥1.0 were considered upregulated, those with a log2FC ≤−1.0 downregulated. These genes were analyzed for their overrepresentation in Reactome gene sets, retrieved from the MSigDB repository. ¹³ Gene sets with a p value of ≤0.05 and an FDR score <0.25 were visualized in an Enrichment Map Network (overlap = 0.04) using Cytoscape (v. 3.10.2).

2.5. Hub gene identification and pathway enrichment analysis

Genes exclusively up‐ and downregulated in H2AX ^high ‐ and H2AX ^low LMS (log2FC ≥2.5; log2FC ≤ −2.5) were analyzed using Cytoscape and the GENEMANIA plugin to identify closely related neighboring genes. ¹⁴ In the next step, we used the MCODE app to identify clusters within this network of query genes and their related neighboring genes (Degree Cutoff = 2, Node Score Cutoff = 0.2, K‐score = 2.0, max. depth = 100). ¹⁵ Hub genes were identified (by degree) using the CytoHubba app. ¹⁶ The clusters with the highest cluster scores were then analyzed with the Enrichr database to identify significantly up‐ or downregulated pathways. ¹⁷

2.6. CIBERSORTx compartmentalization of immune cells in H2Ax ^high ‐ and H2AX ^low leiomyosarcomas

To identify differences in the immune cell signatures between H2AX ^high ‐ and H2AX ^low LMs, we conducted CIBERSORTx analyses on the openly available digital CIBERSORTx platform (https://cibersortx.stanford.edu/runcibersortx.php). This machine learning approach deconvolutes bulk‐RNA sequencing data to estimate immune cell proportions. ¹⁸ As a reference, we used the standard LM22 Signature Matrix. The parameters included Batch correction (B‐mode) and 100 permutations. Relative proportions of the immune cell composition in H2AX ^high ‐ and H2AX ^low LMS were compared using a students' t‐test with three levels of significance (*p adjusted ≤.05, **p adjusted <.01, ***p adjusted <.001; Bonferroni correction for multiple comparisons).

2.7. Cologne University Hospital Cohort for soft tissue sarcomas

To assess the protein expression of phosphorylated H2AX (γ‐H2AX) in STS, we generated a second, independent cohort of various STS: All patients were treated at the University Hospital of Cologne between 2013 and 2024. The cohort included a total of n = 291 STS, including undifferentiated pleomorphic sarcoma (UPS, n = 122), leiomyosarcoma (LMS, n = 79), dedifferentiated liposarcoma (ddLS, n = 60) and myxofibrosarcoma (MFS, n = 30) (Table S2). Neo‐adjuvant or adjuvant treatment regimes, if applied for high‐grade soft tissue sarcomas, usually included up to 6 cycles of chemotherapy with doxorubicin and ifosfamide and/or radiation therapy, according to German national treatment guidelines for soft tissue sarcomas.

2.8. Generation of tissue microarrays immunohistochemical staining

For immunohistochemical analyses of proteins (γ‐H2AX, PLK1, TOP2A, ATRX) and the immune cell markers (CD68, CD163, mast cell tryptase), tissue microarrays (TMAs) were generated: Four tissue cores (1.2 mm each) from representative tumor areas were transferred to a recipient paraffin block using a semi‐automated precision instrument. Benign tonsil tissue and appendix tissue were used as controls. Immunohistochemical staining for γ‐H2AX, PLK1 and ATRX was conducted on a Leica Bond Max platform (Leica, Wetzlar, Germany) according to the manufacturer's instructions (γ‐H2AX: Rabbit mAb ab81299, Abcam, Cambridge, UK, 1:200, EDTA buffer; PLK1: Rabbit mAb 208G4, Cell Signaling, Danvers, Danvers, USA, 1:50, EDTA buffer; ATRX: Rabbit mAb #10321, Cell Signaling, 1:200, EDTA buffer). For TOP2A staining, a Ventana Benchmark platform (Roche, Basel, Switzerland) was used (TOP2A mAb Rabbit, D10G9, Cell Signaling, 1:50, EDTA). Each protein staining was independently assessed by two trained pathologists (AGS, SIL) using the H‐Score method: Staining intensity (negative = 0, weak = 1, moderate = 2, strong = 3) was multiplied by the percentage of stained cells for each intensity fraction. These fractions were then added to an H‐Score (minimum: 0, maximum: 300). The mean H‐Score of all four cores was used for further statistical analysis. If a patient had received neoadjuvant treatment (radiation and/or chemotherapy), only biopsy material prior to neoadjuvant treatment was included to exclude therapeutically induced high expression of γ‐H2AX. For further analysis of γ‐H2AX, the Cutoff Finder tool was used to dichotomize all STS and histological subgroups into γ‐H2AX^high‐ and γ‐H2AX^low tumors according to a survival‐associated cutoff as described above. For ATRX survival analyses, the tumors were dichotomized into tumors with retained expression of ATRX (H‐Score ≥ 20) and tumors with loss of ATRX expression (H‐Score < 20).

2.9. Assessment of immune cell infiltrate in tissue microarrays

To evaluate immune cell infiltration and the expression of key hub genes, the TMAs were stained using the Leica Bond staining platform (Leica, Wetzlar, Germany) with antibodies targeting CD68, CD163, and mast cell tryptase, following standard routine protocols of the Institute of Pathology at Cologne University Hospital. The stained slides were then digitized using a Leica Aperio GT450 scanner (Leica, Wetzlar, Germany). Immune cell infiltration was quantified with QuPath software (v.0.5.0), an open‐source image analysis tool ¹⁹ : After manual annotation of the tumor area, immunohistochemically stained cells were quantified as positive cells per mm² (detection parameters: detection image based on optical density sum; pixel size: 0.5 μm; background radius: 8 μm; median filter radius: 0 μm; sigma: 1.5 μm; minimum area: 6 μm²; maximum area: 400 μm²; intensity threshold: 0.1; maximum background intensity: 2). For M2 macrophages, CD163‐positive cells were counted. The difference between CD68‐ and CD163‐positive cells was used to approximate for M1‐phenotype macrophages. The results for each core were manually checked for plausibility and cores with artificial over‐staining were excluded. The mean count/mm² of each immune cell type was then used for further statistical analysis.

2.10. Data preparation, visualization, and statistical analysis

All data processing, visualization, and statistical analyses, including survival analysis, were conducted with R (v.4.2.2) and RStudio (v. 2022.12.0 + 353) using openly available packages including survival (v.3.4–0), survminer (v.0.4.9), and ggplot2 (v.3.4.0). Interdependencies between clinical data (patients' age, sex, application of adjuvant‐ or neoadjuvant therapy), histopathological data (site, FNCLCC grade and resection margin status [R]) and H2AX−/γ‐H2AX‐ or ATRX expression were evaluated using Fisher's exact test, Chi‐square test, and Spearman's correlation. FNCLCC grade (Grading according to Fédération Nationale des Centres de Lutte Contre le Cancer) included tumor cell differentiation, presence of necrosis and mitosis counts. ²⁰ Overall survival (OS) was evaluated from the date of surgery until the date of death (of any cause). Patients with a survival <30 days were excluded to account for immediate post‐surgical complications. For survival analysis, Kaplan Meier curves and a log rank test were used. Multivariable analyses were conducted using a Cox proportional hazard model with the Enter method (all variables entered simultaneously) with a p value ≤.05 considered significant.

3. RESULTS

3.1. High expression of H2AX mRNA is associated with shorter survival in leiomyosarcomas

When stratifying the complete TCGA STS cohort using the Cutoff Finder Tool, a cutoff value of 724 counts (TPM, RSEM normalized) showed the most significant difference in OS (<cutoff: n = 66 [27.8%]; >cutoff: n = 171 [72.2%]): High levels of H2AX mRNA expression were associated with shorter OS (log‐rank p = .02, Figure 1A). In univariate analysis, high H2AX expression was significantly associated with shorter OS (HR 1.84, 95%CI 1.09–3.08; p = .02), although this was not observed in multivariable Cox regression analysis (Table 1).

FIGURE 1.

H2AX mRNA expression and Gene Set Enrichment Analysis in TCGA leiomyosarcomas. (A) The Cutoff Finder Tool was used to identify mRNA cutoff values (TPM, RSEM‐normalization), and to stratify the cohort by the most significant difference in OS. High levels of H2AX mRNA expression were associated with significantly OS in all STS of the TCGA cohort (n = 237), including UPS, ddLS, LMS, and MFS (log rank test, p = .02). (B) This difference in OS was particularly noted in the LMS cohort (n = 104), where the Cutoff Finder Tool identified a cutoff value of TPM = 1952 counts (RSEM‐normalized). Based on this cutoff, we then divided the cohort into H2AX ^high ‐LMS (>cutoff) and H2AX ^low LMS (<cutoff). (C) Uterine LMS had significantly higher levels of H2AX expression (TPM, RSEM‐normalized) (t‐test, **p = .002); (D) GSEA of upregulated genes revealed that both H2AX ^high ‐ and H2AX ^low LMS shared a gene set enrichment in RNA metabolism, M phase, and reshaping and degradation of extracellular matrix. Additionally, H2AX ^high LMS had enriched gene sets involved in mitosis and cell cycle regulation, while H2AX ^low LMS had additional upregulated gene sets associated with collagen formation and ‐degradation and several immune signaling pathways. (E) Both H2AX ^high LMS and H2AX ^low LMS exhibited downregulation of MHC I‐mediated antigen presentation, along with reduced expression of gene sets regulating cellular starvation response, ubiquitination and IL10 signaling. Furthermore, regulation of proliferation, angiogenesis, and adhesion via SLIT‐ and ROBO‐pathways was decreased. H2AX ^high LMS additionally showed a reduced expression of autophagy, adaptive inflammatory immune response as well as cellular senescence. ddLS, dedifferentiated liposarcoma; LMS, leiomyosarcoma; MFS, myxofibrosarcoma; OS, overall survival; STS, soft tissue sarcoma; UPS, undifferentiated pleomorphic sarcoma.

TABLE 1.

Multivariable cox regression analysis of leiomyosarcomas in the TCGA cohort.

Variable	HR	95%CI	p value
All soft tissue sarcomas
H2AX high	1.72	0.90–3.29	.104
Age	1.03	1.01–1.05	.027
Sex	0.77	0.45–1.30	.325
Localization	0.93	0.81–1.07	.306
R status	2.73	1.60–4.68	<.001
Presence of necrosis	1.72	0.99–2.98	.053
Adjuvant therapy a	1.47	0.75–2.86	.26
LMS
H2AX high	16.14	4.07–63.92	<.001
Age	1.02	0.98–1.06	.347
Sex	0.80	0.19–3.26	.751
Localization	1.08	0.85–1.39	.527
R status	3.99	1.32–12.05	.014
Presence of necrosis	0.72	0.23–2.24	.573
Uterine LMS	0.37	0.07–2.01	.249
Adjuvant therapy a	2.54	0.72–8.95	.148
TP53mut status	0.22	0.07–0.67	.008
ATRXmut status	1.80	0.44–7.11	.487
RB1mut status	1.77	0.34–9.48	.420

Abbreviation: LMS, leiomyosarcoma.

^a Adjuvant therapy included radiation and/or chemotherapy.

When subdivided by sarcoma subtypes, the difference in OS remained only significant in the LMS subgroup (n = 104): A significant survival‐associated cutoff of 1952 counts (TPM, RSEM‐normalization) was identified. Patients with H2AX ^high LMS (n = 13, 12.5%) had a significantly shorter OS than those with H2AX ^low LMS (n = 91, 87.5%) (log‐rank p < 0.001, Figure 1B). In univariate analysis, high expression of H2AX was associated with reduced OS (HR 4.03, 95%CI 1.89–6.01). For multivariable analysis, genetic mutation data from the TCGA leiomyosarcoma cohort were obtained from the cBioPortal repository, including TP53‐, RB1‐ and ATRX mutation status, since these genes are commonly mutated in leiomyosarcomas and have been recently described as feasible markers for risk stratification in LMS. ¹ , ²¹ 48 LMS showed TP53 mutations (46.2%), 15 RB1 mutations (14.2%) and 13 LMS ATRX mutations (12.5%). Multivariable Cox regression analysis confirmed high H2AX mRNA expression as an independent factor significantly associated with shorter survival (Table 1). When examining the clinicopathological parameters, patients with uterine LMS had significantly higher counts of H2AX mRNA expression (TPM, RSEM‐normalized; Figure 1C; t test, p = .002). In accordance with that observation, H2AX ^high LMS were more often localized in the uterus than H2AX ^low LMS (Chi‐squared test p = .006, Bonferroni correction). No correlation with patients' age, sex and histopathological parameters was observed (Table S5). For mutations in TP53, RB1, and ATRX, no significant difference was noted between H2AX ^high ‐ and H2AX ^low LMS (Table S3). There was no significant difference in tumor mutational burden (TMB) between H2AX ^high ‐ and H2AX ^low LMS (Wilcoxon Test, p adjusted = 0.520).

3.2. GSEA enrichment analysis reveals distinct enrichment profiles in H2AX ^high ‐ and H2AX ^low LMS

DESeq2 analysis was performed to identify significantly up‐ or downregulated genes in H2AX ^high ‐ and H2AX ^low LMS (Table S4). Gene Set Enrichment analyses for both H2AX ^high ‐ and H2AX ^low LMS were conducted, with a log2 Fold Change of ≥1.0 (upregulated genes) and ≤ −1.0 (downregulated genes). Enrichment analysis using the Reactome repository showed that H2AX ^high LMS exhibited an upregulation of genes involved in cell cycle control, kinetochore formation and various other steps of mitosis, including CDK1, CCNB2, ZWILCH and ZWINT (Figure 1D). In H2AX ^low LMS, we observed an exclusive enrichment of gene sets regulating collagen biosynthesis, including COL1A1, COL1A2, COL4A1, COL4A2. H2AX ^high LMS displayed a downregulation of Reactome gene sets mediating adaptive immune responses, TNF‐ and interferon signaling and autophagy (Figure 1E). Additionally, genes involved in cell senescence were downregulated.

3.3. H2AX ^high ‐ and H2AX ^low leiomyosarcomas display distinct Hub genes and pathway enrichments

After identification of genes exclusively upregulated (log2FC of ≥2.5) and downregulated (log2 FC ≤−2.5), closely related genes were identified using the GENEMANIA plugin in Cytoscape. Gene clusters were formed applying the MCODE tool and pathway analyses was performed with the Enrichr database. Hub genes were identified using the CytoHubba app. In both H2AX ^high ‐ and H2AX ^low LMS, upregulated genes were allocated to Gene Ontology (GO) pathways regulating mitotic spindle formation, microtubule assembly, and chromatid segregation. The most prominent cluster (cluster 1) in both H2AX ^high ‐ and H2AX ^low LMS displayed a significant overlap in Hub genes, with Polo‐like kinase 1 (PLK1) being the central Hub gene (Figure 2A,C). To verify these findings, we assessed the protein expression of PLK1 in the second, independent LMS cohort (n = 79) of the University Hospital Cologne (Figure 2I). For this purpose, the Cutoff finder tool was used to dichotomize the cohort by a survival‐associated cutoff of γ‐H2AX protein expression into γ‐H2AX^high LMS (H‐Score >240) and γ‐H2AX^low LMS (H‐Score <240). As for PLK1 mRNA expression, we did not observe a significant difference between tumors with high or low γ‐H2AX expression (Figure 2K).

FIGURE 2.

Up‐ and downregulated gene clusters, Hub genes and Enrichr pathway analysis in H2AX ^high ‐ and H2AX ^low leiomyosarcomas. (A) In H2AX ^high LMS, Cluster 1 (Score 20.200), including exclusively upregulated genes, was significantly associated with pathways involved in mitotic spindle assembly and checkpoint signaling, Cyclin‐dependent kinase activity and nuclear processes (all p adjusted <.001, Bonferroni correction). Key Hub genes (by degree) in this cluster included PLK1, CCNB1, TOP2A, CDK4 and NME1. (B) Cluster 2 (Score: 8.897) of exclusively upregulated genes in H2AX ^high LMS was primarily associated with DNA‐binding as well as regulatory activity of RNA Polymerase II in Enrichr pathway analysis. Prominent Hub genes in this cluster included Zinc finger protein transcription factors such as ZNF582 and ZNF665. (C) In H2AX ^low LMS, Cluster 1 (Score: 70.548) shared many characteristics with Cluster 1 in H2AX ^high LMS, predominantly consisting of genes involved in microtubule formation and chromatin segregation. An overlap of Hub genes of those from Cluster 1 in H2AX ^high LMS was noted, including PLK1, CCNB1, BUB1B, PBK and CCNA2. (D) Cluster 2 (Score: 13.286) of exclusively upregulated genes in H2AX ^low LMS was associated with inflammatory responses, chemokine activity and MHC II antigen presentation in Enrichr pathway allocation. Key Hub genes in this cluster included CCR5, HLA‐DRA and CXCL9. (E) Similar to H2AX ^high LMS, Cluster 3 (Score: 13.000) in H2AX ^low LMS showed an upregulation of genes involved in DNA‐binding and regulatory activity of RNA Polymerase II. Hub genes in this cluster included Zink finger transcription factor ZNF100, ZNF716 and ZNF681. (F) Cluster 1 (Score: 30.258) of exclusively downregulated genes in H2AX ^high LMS were involved in muscle contraction, cytoskeleton formation and actin binding. Central Hub genes in this cluster were LMOD1 and TPM2. (G) The cluster 2 (Score: 17.320) in H2AX ^high LMS was involved in positive regulation of cell development. Additionally, these genes were allocated to pathways mediating clathrin‐coated vesicle transport and vitamin D metabolism. Key Hub genes in this cluster included TOP1, TBX5 and ABL1. (H) In H2AX ^low LMS, only a small fraction of downregulated genes was exclusively downregulated compared to H2AX ^high LMS. These genes in cluster 3 (Score: 16.500) were associated with translation processes and RNA binding in Enrichr pathway analysis. Key Hub genes in this cluster were RPL17, RPL7A, RPL6, RPL10A and RPS3A. (I, J) Protein expression of PLK1 and TOP2A in the LMS cohort of the Cologne University Hospital. Magnification ×200. (K) Between γ‐H2AX^high LMS and γ‐H2AX^low LMS, no significant difference in PLK1 protein expression was observed (Mann–Whitney U test). (L) LMS with high γ‐H2AX expression showed significantly higher levels of TOP2A protein expression (p = .042, Mann–Whitney U test); GO, gene ontology database/gene set repository; LMS, leiomyosarcoma.

In H2AX ^high LMS, however, TOP2A, FOXM1, CDC20 and NME1 were the most important exclusively upregulated Hub genes. In the independent TMA cohort, LMS with high γ‐H2AX expression had significantly higher levels of TOP2A protein expression (Figure 2J,L) (p = .042, Mann–Whitney U test). Other upregulated genes in H2AX ^high LMS included various Zinc finger protein transcription factors, with ZNF582 being the most important Hub gene in this cluster (Figure 2B). In H2AX ^low LMS, a similar Zinc finger cluster was identified, with ZNF100 being a central Hub gene (Figure 2E). Notably, H2AX ^low tumors showed an increased expression of genes involved in chemokine activity, MHC II‐complex pathways, and inflammatory response (Figure 2D). Several Hub genes were identified, including CCR5, HlA‐DRA and CXCL9. H2AX ^high LMS showed an exclusive downregulation of genes associated with muscle contraction, cytoskeleton formation and actin binding (Figure 2F). Among these genes, LMOD1 and TPM2 stood out as Hub genes. The second downregulated gene cluster was allocated to cell development, clathrin‐coated endocytic vesicle membrane formation and vitamin D metabolism (Figure 2G), with Topoisomerase I (TOP1) and TBX5 identified as central Hub genes. In contrast, only 76 genes were exclusively downregulated in H2AX ^low LMS: These genes were involved in translational processes and RNA binding (Figure 2H).

3.4. H2AX^high ‐ and H2AX^low leiomyosarcomas exhibit differences in their immune cell mRNA signatures

Since GSEA analysis indicated a downregulation of adaptive immune cell responses in H2AX ^high LMS, whereas H2AX ^low LMs showed an upregulation in immune cell signaling, we performed CIBERSORTx analyses to compare the immune cell composition. The immune cell environment both in H2AX ^high ‐ and H2AX ^low LMS was predominantly characterized by macrophage signatures associated with the pro‐tumorigenic M2 phenotype, with CD8 T cell signatures nearly absent (Figure 3A). Both LMS subgroups showed additionally notable signatures of naïve B cells.

FIGURE 3.

CIBERSORTx analysis of immune cell compartment in H2AX ^high ‐ and H2AX ^low leiomyosarcomas. (A) This heatmap of immune signatures in H2AX ^high ‐ and H2AX ^low LMS of the TCGA cohort shows that M2 macrophages constituted the largest cell population in LMS. Few LMS additionally showed higher proportions of plasma cells and naïve B cells. (B) When compared, H2AX ^high LMS had significantly lower immune cell signatures of M1 macrophages, monocytes, and resting mast cells (all p adjuvest <.001, Bonferroni correction), with monocytes only accounting for a miniscule subset of the immune cell signatures. (C–E) Immunohistochemical staining was performed on the University Hospital of Cologne cohort for CD68 (a more general macrophage marker), CD163 (specific for M2 macrophages) and mast cell tryptase (a general mast cell marker) to assess the immune cell infiltrate in LMS; To approximate the infiltrate of M1 macrophages, the difference between CD68‐positive (total macrophages) and CD163 (M2 macrophages) was used. Magnification ×200. (F–H) For M1‐ and M2 macrophage infiltration, there was no significant difference observable between γ‐H2AX^high—and γ‐H2AX^low LMS. We observed significantly higher levels for (total) mast cell infiltration in γ‐H2AX^low LMS (p = .038, Mann–Whitney U test). (Students' t test: *p adjusted ≤.05; **p adjusted <.01; ***p adjusted <.001; Bonferroni correction for multiple comparisons); LMS, leiomyosarcoma.

H2AX ^low LMS showed significantly higher proportions of pro‐inflammatory‐ and anti‐tumorigenic M1 macrophage signatures (Students' t‐test, p adjusted <.001; Figure 3B). Additionally, higher proportions of resting mast cells (p adjusted = .001) and monocytes (p adjusted <.001) were observed in H2AX ^low LMS. Immunohistochemical staining for CD68, CD163 and mast cell tryptase, followed by automated quantification, confirmed that the immune cell infiltrate in LMS mainly consisted of CD163‐positive M2 macrophages, with M1 macrophages (approximated by CD68‐positive cells minus CD163‐positive cells) being mostly absent (Figure 3C,D). No significant difference between γ‐H2AX^high‐ and γ‐H2AX^low LMs was observed for M1‐ or M2 macrophage infiltrates (Figure 3F,G). Consistent with the CIBERSORTx analyses, LMS displayed high levels of mast cells (Figure 3E), with significantly more mast cells/mm² in γ‐H2AX^low LMS (p = .038, Mann–Whitney U test) (Figure 3H).

3.5. High γ‐H2AX expression correlates with shorter survival in leiomyosarcomas

To assess the γ‐H2AX protein expression in STS, we generated TMAs of a separate, independent cohort of 291 primarily resected STS (Table S2) and conducted immunohistochemical staining (Figure 4A,B). Using the Cutoff Finder tool, we dichotomized all STS and each histological subtype into H2AX^high‐ and H2AX^low tumors, based on the most significant survival‐associated cutoff. In the entire STS cohort, high levels of γ‐H2AX (> H‐Score 195; n = 88 [30.2%]) were linked to shorter OS (Figure 4F) and were a significant prognostic factor in univariate analysis (HR 1.80, 95%CI 1.21–2.67, p = 0.003), though not in multivariable cox regression analysis (Table 2). For UPS, a γ‐H2AX expression > cutoff (>H‐Score 195; n = 45 [36.9%]) was correlated to shorter OS (Figure 4G). In univariate analysis, high γ‐H2AX expression correlated with shorter OS (HR 2.28, 95%CI 1.36–3.83; p = .002), but not in multivariable analyses (Table 2). For MFS and ddLS, no significant survival difference was observed (Figure 4H,I). Despite the association of high H2AX mRNA expression with the uterine LMS subgroup (see above), significantly higher γ‐H2AX protein expression was observed in non‐uterine LMS (t‐test, p < .001) (Figure 4C). In LMS overall, a statistical trend towards shorter OS was observed for γ‐H2AX^high (p = .069) (Figure 4J). Non‐uterine LMS with high γ‐H2AX expression (n = 22, 39.3%) showed a significantly shorter OS (p = .004) (Figure 4K), and high γ‐H2AX expression was a negative prognostic factor in univariate analysis (HR 7.17, 95%CI 1.51–33.96, p = .013). In uterine LMS, no difference in survival was observed.

FIGURE 4.

Expression of γ‐H2AX and ATRX in the Cologne University Hospital sarcoma cohort. (A) γ‐H2AX was expressed in various STS (n = 291), including UPS (n = 122), MFS (n = 30), ddLS (n = 60) and LMS (n = 79), with a large variety of expression, assessed by the H‐Score; Magnification ×200. (B) The highest expression of γ‐H2AX was observed in UPS and LMS, with lower expression in MFS and ddLS. (C) When analyzing the LMS cohort, non‐uterine LMS (n = 57) had significantly higher levels of γ‐H2AX than uterine LMS (n = 22) (p < .001, t test). (D) Leiomyosarcomas with a loss of ATRX protein expression and retained ATRX expression (×200 magnification). (E) Although LMS with high γ‐H2AX expression had lower levels of ATRX expression, a significant difference between γ‐H2AX^high LMS and γ‐H2AX^low LMS was not observed. (F–K) The Cutoff Finder tool was used to dichotomize all STS and each histological subtype into γ‐H2AX^high‐ and γ‐H2AX^low sarcomas. Survival analyses revealed significant differences in overall survival (OS) for all STS, UPS and non‐uterine LMS (log rank test, p ≤ .05) as well as a statistical trend for all LMS (p = .069). (L–N) LMS with a loss of ATRX protein expression showed a significantly shorter survival in the total LMS cohort, in uterine LMS and a statistical trend in the non‐uterine LMS cohort (log rank test, p ≤ .05). ddLS, dedifferentiated;liposarcoma; LMS, leiomyosarcoma; MFS, myxofibrosarcoma; STS, soft tissue sarcoma; UPS, undifferentiated pleomorphic sarcoma.

TABLE 2.

Multivariable cox regression analysis of the University Hospital Cologne cohort.

Variable	HR	95%CI	p value
All soft tissue sarcomas
γ‐H2AXhigh	1.67	0.98–2.85	.058
Age	1.00	0.98–1.02	.740
Sex	0.56	0.32–0.97	.040
R	2.74	1.54–4.88	<.001
FNCLCC a	2.40	1.48‐4.00	<.001
Localization	1.06	0.91–1.23	.432
Neoadjuvant therapy b	0.29	0.09–0.97	.045
Adjuvant therapy b	0.54	0.31–0.95	.031
UPS
γ‐H2AXhigh	1.45	0.69–3.03	.324
Age	0.98	0.96–1.01	.232
Sex	0.54	0.20–1.42	.211
R	2.18	0.86–5.54	.101
FNCLCC a	2.97	0.68‐12.91	.147
Localization	1.00	0.75–1.32	.978
Neoadjuvant therapy b	0.78	0.06–1.28	.099
Adjuvant therapy b	0.48	0.20–1.33	.173
LMS
γ‐H2AXhigh	12.12	1.38–106.86	.025
Age	1.06	0.98–1.14	.154
Sex	0.57	0.11–3.09	.516
R	9.47	1.25–71.70	.030
FNCLCC a	12.75	1.82‐89.44	.010
Localization	1.21	0.82–1.77	.332
Adjuvant therapy b	0.25	0.02–2.64	.248
ATRX status	0.08	0.01–0.59	.013
Non‐uterine LMS
γ‐H2AXhigh	153.80	2.90–8.16e3	.013
Age	1.03	0.95–1.12	.424
Sex	0.20	0.02–2.74	.230
R	128.08	1.98–8.28e3	.033
FNCLCC	17.33	0.90–3.35e2	.059
Localization	2.08	0.99–4.39	.055
ATRX status	0.02	0.001–0.66	.028

Abbreviations: LMS, leiomyosarcoma; UPS, undifferentiated pleomorphic sarcoma.

^a FNCLCC grade = Grading according to Fédération Nationale des Centres de Lutte Contre le Cancer, including differentiation of tumor cells, presence of necrosis and mitosis count.

^b (neo)adjuvant treatment consisted of chemotherapy and/or radiation.

3.6. High expression of γ‐H2AX and loss of ATRX expression are prognostic factors in leiomyosarcoma

Since a recent study describes ATRX protein loss as a negative prognostic factor in LMS, ²² we conducted immunohistochemical staining for ATRX (Figure 4D,E). Among all LMS, 38 cases (48.1%) showed a loss of ATRX expression, 40 LMS retained expression (50.6%), one case was not assessable (no intact tissue in TMA cores, 1.3%). ATRX loss occurred in 11/22 uterine LMS (50%) and 27/56 non‐uterine LMS (47.4%). No correlation between ATRX loss and high γ‐H2AX expression was observed (Chi‐square test, p = .11). ATRX loss was significantly associated with shorter OS in all LMS (p = .019) (Figure 4L) and uterine LMS (p = .017) (Figure 4N) (log rank test), with a statistical trend observed for non‐uterine LMS (p = .09) (Figure 4M). In multivariable analysis, both high γ‐H2AX expression and ATRX loss were independent prognostic factors for shorter OS in all LMS and non‐uterine LMS (Table 2). No correlations were found between γ‐H2AX‐ or ATRX expression and patients' age, sex and histopathological parameters and treatment parameters (all p adjusted >.05) (Table S6).

4. DISCUSSION

This study evaluates the role of H2AX/γ‐H2AX expression in various soft tissue sarcomas. The most significant difference in overall survival between H2AX ^high ‐ and H2AX ^low tumors was observed in the TCGA leiomyosarcoma subgroup. High H2AX mRNA expression correlated with enrichment of gene sets regulating cell cycle checkpoints, spindle assembly and sister chromatid segregation. This was expected, as high levels of γ‐H2AX indicate extensive DNA damage, triggering cell cycle checkpoint activation, and, if DNA repair fails, apoptosis. ⁷ , ⁹

Gene set enrichment‐ and pathway analyses revealed considerable overlap between H2AX ^high ‐ and H2AX ^low leiomyosarcomas, including overexpression of genes regulating mitotic activity, spindle formation, and cell cycle. However, H2AX ^high tumors showed a unique upregulation of CDC20 and TOP2A. Consistent with this finding, we observed that leiomyosarcomas with high γ‐H2AX protein expression showed significantly higher levels of TOP2A protein expression than tumors with low γ‐H2AX expression. TOP2A (Topoisomerase 2A), essential for maintaining DNA integrity and chromosome dynamics, introduces transient double strand breaks to facilitate chromosomal plasticity during replication, transcription, chromosome segregation, and DNA repair. ²³ High expression of TOP2A has been correlated to poor prognosis in lung adenocarcinomas, hepatocellular carcinoma, and ovarian cancer. ²⁴ , ²⁵ , ²⁶ In sarcomas, the role of TOP2A and its prognostic and therapeutical potential are unknown. TOP2A has been primarily investigated in breast cancer: In breast cancer patients with Her2/neu amplification, high TOP2A protein expression was associated with a higher response on anthracycline chemotherapy. ²⁷ The subgroup of H2AX ^high /γ‐H2AX^high leiomyosarcoma patients might therefore show higher response rates to anthracycline‐based chemotherapy as well. CDC20 (protein: Cell division cycle 20, Cdc20) is an oncoprotein essential for cell cycle progression via activation of anaphase promoting complex. ²⁸ It is linked to aggressive behavior and poor prognosis in various cancers (breast cancer, hepatocellular carcinoma, gastric cancer) and has been investigated as a therapeutical target in vitro models of breast cancer and osteosarcoma. ²⁸ , ²⁹ , ³⁰

Another crucial therapeutic target extensively investigated in cancer is the tumor microenvironment, including the immune microenvironment. H2AX ^high leiomyosarcomas exhibited downregulation of TNF signaling, interferon signaling, adaptive immune cell response and autophagy, while H2AX ^low tumors showed enrichment of several pathways mediating inflammatory response in general, and chemokine signaling. CIBERSORTx analyses of immune cell signatures showed that H2AX ^low leiomyosarcomas had significantly higher signatures for M1 macrophages, which create a pro‐inflammatory, anti‐tumorigenic environment associated with better prognosis in various cancers, including sarcomas. ³¹ , ³² Both leiomyosarcoma subgroups exhibited high immune signatures of tumor‐ associated M2 macrophages, which are associated with a pro‐tumorigenic, immunosuppressive shaping of the tumor microenvironment and decreased the efficiency of immune based therapy, including immune checkpoint inhibition and CAR‐T cell therapies. ³³ , ³⁴ , ³⁵ , ³⁶ Thus, several ongoing studies investigate M2 macrophages as potential targets in patient‐individualized settings and evaluate the possibilities of re‐programming these M2 macrophages towards a M1 phenotype. ³⁶ While higher M1 macrophage signatures in H2AX ^low leiomyosarcomas might therefore be a partial explanation for the longer survival compared to patients with H2AX ^high leiomyosarcomas, we could not confirm significant differences in the macrophage infiltrate between leiomyosarcomas with high or low protein expression of γ‐H2AX. This can be explained by a relatively small sample size of 79 patients. Additionally, a recent study by Bill et al. demonstrated that the strict separation of tumor‐associated macrophages by M1‐ and M2‐immunohistochemical markers like CD68 and CD163 does not capture the complexity of these immune cells, with tumor‐associated macrophages having a much more complex biological plasticity and extensive interactions with other crucial immune cell populations, one of them being mast cells. ³⁷ Mast cells play a complex, multi‐faceted role within the tumor microenvironment, including pro‐tumorigenic functions such as support of angiogenesis and T cell suppression and anti‐tumorigenic factors like innate immune cell activation and direct inhibition of tumor growth. ³⁸ The role of mast cells in soft tissue sarcoma is mostly unknown. In this study, H2AX ^low leiomyosarcomas showed higher immune cell signatures of resting mast cells that H2AX ^high tumors. While there are, to our knowledge, no well‐established immunohistochemical markers for separation and between activated and resting mast cells in paraffin‐embedded tumors, we could demonstrate that leiomyosarcomas with high γ‐H2AX levels had significantly more extensive infiltration of total mast cells. In context with the CIBERSORTx analyses, in which activated mast cells only accounted for a very miniscule proportion of the immune cell environment, we thus might postulate that most of the assessed and counted mast cells are resting mast cells – however, more refined immunohistochemical markers must be established and validated in the future to differentiate between these two cell states. A recent study by Panagi et al. demonstrated that mast cell stabilization led to increased response to immunotherapy and doxorubicin in vitro. ³⁹ Further investigation of mast cells in the tumor microenvironment might therefore allow an assessment of their potential as a therapeutic target in sarcomas.

In this study, high H2AX mRNA levels and high γ‐H2AX protein expression were an independent prognostic factor associated with shorter survival in leiomyosarcomas. This is in concurrence with two soft tissue sarcoma studies which previously investigated γ‐H2A: Perez et al. demonstrated that high expression of γ‐H2AX was an independent adverse prognostic factor for survival in a smaller cohort of 69 patients. However, this study only included around 40 undifferentiated pleomorphic sarcomas, liposarcomas and leiomyosarcomas, and did not assess these histological subtypes separately. ⁴⁰ Kim et al. also showed that γ‐H2AX expression negatively correlated with disease‐specific survival and event‐free survival in a smaller patient cohort of 112 soft tissue sarcomas, with no distinguished assessment of histological subtypes. ⁴¹ This study therefore is, to our knowledge, the largest study investigating the role of H2AX mRNA‐ and γ‐H2AX protein expression in two larger independent cohorts of sarcomas, not only confirming the findings of these previous studies, but also showing that γ‐H2AX is especially of prognostic significance in distinct soft tissue sarcoma entities like non‐uterine leiomyosarcomas. We additionally included commonly mutated genes like TP53, RB1 and ATRX in our multivariable model, which have been described to be associated with worse prognosis in particularly leiomyosarcomas before: Dermawan et al. recently developed a risk classifier for leiomyosarcomas, demonstrating that tumors with RB1 mutation and ATRX mutation or chromosome 12q deletion had a significantly shorter survival. ²¹ When inserted in multivariable analysis, H2AX mRNA expression remained a significant prognostic factor associated with shorter survival. Additionally, Denu et al. recently described that ATRX protein expression was associated with shorter survival in uterine leiomyosarcomas. ²² In another study, high expression of ATRX protein was correlated with longer metastasis‐free survival and overall survival in a mixed sarcoma cohort, including 17 leiomyosarcomas. ⁴² We therefore included ATRX protein expression in our model. We confirmed the findings of these previous studies, with loss of ATRX being correlated to significantly shorter survival in Kaplan Meier survival analysis in the total leiomyosarcoma cohort and uterine leiomyosarcomas, with a statistical trend towards shorter survival in non‐uterine leiomyosarcoma. We could additionally demonstrate that in multivariable analysis, both γ‐H2AX and ATRX loss were significant adverse independent factors towards shorter overall survival in all leiomyosarcomas and non‐uterine leiomyosarcomas as well. The ATRX gene encodes for a SWI/SNF‐like chromatin‐remodeling protein, which, in combination with its partner gene product DAXX, is involved in histone H3.3 deposition in DNA strands especially in telomere regions. ⁴³ ATRX mutations and loss of ATRX are associated with increased DNA‐ and telomere instability and the development of several cancers, including pediatric neuroblastoma, gliomas, and pancreatic neuroendocrine tumors. ⁴⁴ , ⁴⁵ While we observed that leiomyosarcomas with high γ‐H2AX expression had lower levels of ATRX expression, a significant difference was not observed, probably due to low sample size. However, a mechanistical overlap between loss of ATRX, consecutive DNA damage and upregulation of γ‐H2AX seems likely—the exact role of ATRX remains poorly understood, however, and its interaction with H2AX remains to be further elucidated.

This study is retrospective by design and has therefore several limitations: The size of the patient cohort and size of histological subgroups was limited due to the rarity of these cancers, especially in case of myxofibrosarcoma and uterine leiomyosarcomas. This led to partially inconclusive results between mRNA and protein expression: While uterine leiomyosarcomas expressed higher levels of H2AX mRNA in this study, we observed significantly lower levels of γ‐H2AX protein expression in uterine leiomyosarcomas than in non‐uterine tumors. A Cancer Genome Atlas Research network study recently investigated various soft tissue tumors and described higher mRNA levels of DNA damage response in uterine‐LMS, which supports or findings in the TCGA cohort. ⁴⁶ , ⁴⁷ Regarding γ‐H2AX protein expression specifically in uterine leiomyosarcomas, no data exist for comparison. Since the number of uterine leiomyosarcoma in our TMA cohort is limited to only 22 cases, further studies need to evaluate, if and how high H2AX mRNA levels correlate with γ‐H2AX protein expression in these cancers, and if our results can be verified in larger cohorts.

5. CONCLUSION

This study extensively analyzed the expression of H2AX mRNA and γ‐H2AX in various soft tissue sarcomas and its biological implications in cellular pathway regulation and immune microenvironment. We demonstrated that high levels of H2AX are associated with distinct biological pathways and Hub genes and worse overall survival in leiomyosarcomas. Protein expression of γ‐H2AX was associated with shorter survival in leiomyosarcomas, therefore it might serve as a valuable prognostic biomarker in these rare cancers. However, the exact function of γ‐H2AX within DNA maintenance and tumorigenesis as well as its potential as a therapeutical target remain to be elucidated in further studies.

AUTHOR CONTRIBUTIONS

Adrian Georg Simon: Conceptualization; writing – original draft; methodology; validation; visualization; writing – review and editing; formal analysis; data curation. Su Ir Lyu: Formal analysis; visualization; validation; methodology; writing – review and editing; conceptualization. Anne Maria Schultheis: Resources; supervision; validation; visualization; methodology; writing – review and editing; conceptualization. David Stahl: Data curation; writing – review and editing. Nora Wuerdemann: Data curation; writing – review and editing. Sebastian Walter: Data curation. Lena Hieggelke: Data curation. Reinhard Buettner: Supervision; resources. Christiane Josephine Bruns: Supervision. Peer Eysel: Supervision. Lars Mortimer Schiffmann: Data curation. Karl Knipper: Data curation. Peter Mallmann: Supervision. Alexander Quaas: Supervision; resources. Roland Ullrich: Supervision; resources.

FUNDING INFORMATION

David Stahl was supported by a MD Research Stipend of the Else Kröner Forschungskolleg Clonal Evolution in Cancer, University Hospital Cologne, Cologne, Germany.

CONFLICT OF INTEREST STATEMENT

Reinhard Buettner has received honoraries for in Lectures and Advisory Boards for AbbVie, Amgen, AstraZeneca, Bayer, BMS, Boehringer‐Ingelheim, Illumina, Janssen, Lilly, Merck‐Serono, MSD, Novartis, Qiagen, Pfizer, Roche, Sanofo, Targos MP Inc. He is a Co‐Owner and Scientific Board Member of Gnothis Inc. (SE), Timer Therapeutics (DE). Funding was provided by the German Cancer Aid (DKH), German Research Society (DFG), the German Ministry of Research and Sciences (BMBF) and the European Union (UN). All other authors declare no competing interests.

ETHICS STATEMENT

This study was approved by the Ethics committee of the Faculty of Medicine of the University of Cologne (23–1339). Consent was obtained from all patients included in this study.

Supporting information

DATA S1: Supporting Information.

Simon AG, Lyu SI, Schultheis AM, et al. Exploration of histone protein γ‐H2AX as a prognostic factor in soft tissue sarcomas and its association with biological behavior, immune cell environment and survival in leiomyosarcoma. Int J Cancer. 2025;156(11):2237‐2250. doi: 10.1002/ijc.35310

DATA AVAILABILITY STATEMENT

All data of the TCGA cohort are openly available in the TCGA Firehose Institute repository (https://gdac.broadinstitute.org/). The CIBRERSORTx analyses were conducted on the official, openly available digital platform (https://cibersortx.stanford.edu/runcibersortx.php). All data for the CIBERSORTx analyses, including the R code for further analyses, are openly available on GitHub (https://github.com/adrsimon-pathology/IJC-24-2092-Repositor_H2AX-in-soft-tissue-sacoma). Other data that support the findings in of this study are available from the corresponding author upon request.