Deconvolution of Sarcoma Methylomes Reveals Varying Degrees of Immune Cell Infiltrates with Association to Genomic Aberrations
Total Page:16
File Type:pdf, Size:1020Kb
Simon et al. J Transl Med (2021) 19:204 https://doi.org/10.1186/s12967-021-02858-7 Journal of Translational Medicine RESEARCH Open Access Deconvolution of sarcoma methylomes reveals varying degrees of immune cell infltrates with association to genomic aberrations Malte Simon1,3†, Sadaf S. Mughal1†, Peter Horak2,5, Sebastian Uhrig1, Jonas Buchloh1, Bogac Aybey1, Albrecht Stenzinger4, Hanno Glimm6,7, Stefan Fröhling2,5, Benedikt Brors1,2,5 and Charles D. Imbusch1* Abstract Background: Soft-tissue sarcomas (STS) are a heterogeneous group of mesenchymal tumors for which response to immunotherapies is not well established. Therefore, it is important to risk-stratify and identify STS patients who will most likely beneft from these treatments. Results: To reveal shared and distinct methylation signatures present in STS, we performed unsupervised decon- volution of DNA methylation data from the TCGA sarcoma and an independent validation cohort. We showed that leiomyosarcoma can be subclassifed into three distinct methylation groups. More importantly, we identifed a com- ponent associated with tumor-infltrating leukocytes, which suggests varying degrees of immune cell infltration in STS subtypes and an association with prognosis. We further investigated the genomic alterations that may infuence tumor infltration by leukocytes including RB1 loss in undiferentiated pleomorphic sarcomas and ELK3 amplifcation in dediferentiated liposarcomas. Conclusions: In summary, we have leveraged unsupervised methylation-based deconvolution to characterize the immune compartment and molecularly stratify subtypes in STS, which may beneft precision medicine in the future. Keywords: Sarcoma, Deconvolution, Survival analysis, Tumor-infltrating leukocytes Background six subtypes including dediferentiated liposarcoma Soft-tissue sarcomas (STS) are rare cancers of mes- (DDLPS), leiomyosarcoma (LMS), undiferentiated enchymal origin that represent < 1% of adult solid pleomorphic sarcoma (UPS), myxofbrosarcoma (MFS), malignancies. Teir high diversity in terms of genetic malignant peripheral nerve sheath tumor (MPNST) aberrations and histological appearance results in and synovial sarcoma (SS). By analyzing genetic, epige- a subclassifcation into more than 70 subtypes [1]. netic, mRNA and protein expression data, the authors Recently, the TCGA consortium released a study com- stated that subtypes with complex karyotypes are prising the characterization of 206 sarcomas from mostly driven by copy number alterations instead of mutations, and that the presence of certain inferred immune cell types and methylation states associates *Correspondence: [email protected] †Malte Simon and Sadaf S Mughal contributed equally to this work with disease-specifc survival [2]. Several clinical tri- 1 Division of Applied Bioinformatics, German Cancer Research Center als on immunotherapies in STS have found low over- (DKFZ), Heidelberg, Germany all response rates, which highlights the importance of Full list of author information is available at the end of the article © The Author(s) 2021. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Simon et al. J Transl Med (2021) 19:204 Page 2 of 17 a more detailed characterization of immune infltrates We validated the immune-cell associated signature as and the development of robust predictors of clinical well as associated genomic aberrations in independent beneft in these tumors [3, 4]. While a global compari- cohorts. son of immunogenicity in diferent tumor types has been previously presented, to date, no study focused on Results sarcomas [5, 6]. Deconvolution of methylation data results In this study, we reanalyzed the TCGA-SARC dataset in subtype‑specifc patterns from an epigenetic perspective by employing unsuper- To identify shared methylation patterns (hereafter vised deconvolution of the methylation data to dis- referred to as latent methylation components, LMCs), cover shared as well as subtype-specifc methylation we analyzed the TCGA sarcoma methylation data aggre- profles. By correlating distinct methylation changes gated within equidistant and non-overlapping genomic with mRNA abundances, we derived gene signatures windows and performed a deconvolution using MeDe- for each profle and showed their biological relevance Com [7]. We chose a factorization into nine LMCs based and usability for subclassifcation. Importantly, we on a low cross-validation error and high stability of the identifed an immune cell-associated component that resulting methylation patterns (Additional fle 1: Fig- implies varying degrees of immune cell infltration in ure S1). Hierarchical clustering on the proportions of STS with enrichment in UPS, DDLPS and MFS cases, the LMCs, which represent the relative occurrence of whereas it was substantially lower in LMS and SS. In the respective patterns in the tumor samples, showed addition genomic aberrations could be identifed that clear associations to histopathological subtypes for sev- harbour the potential to infuence tumor infltration. eral components (Fig. 1). Te strongest association was found for LMC9 and synovial sarcoma (point biserial Fig. 1 MeDeCom deconvolves subtype-specifc methylation patterns. Proportions for the deconvolution of TCGA-SARC methylation data using 9 LMCs and λ 0.01 are shown. Unsupervised deconvolution based on non-negative matrix factorization resulted in methylation components = associated with histological subtypes and tumor tissue sites. LMS, leiomyosarcoma; DDLPS, dediferentiated liposarcoma; MFS, myxofbrosarcoma; UPS, undiferentiated pleomorphic sarcoma; SS, synovial sarcoma; MPNST, malignant peripheral nerve sheath tumor Simon et al. J Transl Med (2021) 19:204 Page 3 of 17 correlation coefcient (rpb) = 0.97) refecting the dramatic Figure S6B). In particular, we were interested in meth- changes to their methylome, which occur in this subtype ylation changes resulting in upregulation of gene expres- as a consequence of an SS18-SSX gene fusion [8]. We fur- sion, since these provide insights on gene activity in the ther observed a global hypomethylation in SS compared LMC-associated tumors and might constitute potential to the other subtypes (Additional fle 1: Figure S2). LMC1 new biomarkers. We calculated the Pearson’s correlation was associated with uterine LMS (ULMS, rpb = 0.72), between methylation and gene expression to further flter whereas LMC7 represented a methylation pattern com- VMRs. Te cutof for a signifcant correlation was greater mon in most nongynecological LMS (STLMS, rpb = 0.81). than 0.3 for hypermethylation-upregulation or smaller Although having a weaker association, LMC2 was pre- than -0.3 for hypomethylation-upregulation. dominantly shared among UPS cases (rpb = 0.36), and Following this approach, we identifed 100 genes for LMC4 showed the strongest association with DDLPS STLMS-associated LMC7 (Fig. 2), which we termed as (rpb = 0.46). Whereas, LMCs 3, 5, 6, and 8 were shared ‘STLMS core signature’. 42 of these genes had hypometh- among the diferent sarcoma subtypes. ylated and 58 had hypermethylated VMRs. 55% of the We validated our fndings using an independent meth- LMS tumors had the highest proportion in LMC7, with ylation dataset (MASTER) from 56 sarcoma samples the majority of the samples belonging to retroperitoneal/ (Additional fle 1: Figure S3). In concordance with the upper abdominal region and only one uterine sample deconvolution results from TCGA-SARC, subtype-asso- (LMS group 1). ciated methylation patterns were consistently observed Te STLMS core signature also contained the previ- for LMS and SS. In addition, we also found unique meth- ously known immunohistochemical markers such as ylation patterns for gastrointestinal stromal tumors MYLK and CASQ2 [9, 10]. Next, we performed func- (GIST), solitary fbrous tumors (SFT) and myxoid lipo- tional annotation of these genes and found enrichments sarcoma (MLS). Hierarchical clustering of the LMCs in gene sets associated with muscle function (Additional from the independent deconvolution of TCGA-SARC fle 2: Table S1) such as focal adhesion (P < 0.004), actin and HIPO showed a high similarity of the methylation binding (P < 0.007) and muscle contraction (P = 0.131) patterns associated with the subtypes present in both (Fisher’s exact test and Benjamini - Hochberg correction) datasets (LMS and SS, Additional fle 1: Figure S4). Tis indicating that LMS group 1 is mostly associated with confrms that these subtypes have distinct methylation smooth muscle function. changes, which are consistently observed