Skip to main content

CSPG4 expression in soft tissue sarcomas is associated with poor prognosis and low cytotoxic immune response



Soft tissue sarcomas (STS) are heterogeneous and pro-metastatic tumors. Identification of accurate prognostic factors and novel therapeutic targets are crucial. CSPG4 is a cell surface proteoglycan with oncogenic functions. It recently emerged as a potential target for immunotherapy, including cell therapy based on CSPG4-specific chimeric antigen receptor (CAR)-redirected cytokine-induced killer lymphocytes (CSPG4-CAR.CIKs) in STS. However, expression of CSPG4 is poorly known in STS so far.


We analyzed CSPG4 gene expression in 1378 localized STS clinical samples, and searched for correlations with clinicopathological data, including disease-free survival (DFS), and with tumor immune features.


CSPG4 expression was heterogeneous across samples. High expression was associated with younger patients’ age, more frequent undifferentiated pleomorphic sarcoma and myxofibrosarcoma pathological subtypes, more frequent internal trunk tumor site, and more CINSARC high-risk samples. No correlation existed with pathological tumor size and grade, and tumor depth. Patients with high CSPG4 expression displayed 49% (95% CI 42–57) 5-year DFS versus 61% (95% CI 56–68) in patients with low expression (p = 3.17E−03), representing a 49% increased risk of event in the “CSPG4-high” group (HR = 1.49, 95% CI 1.14–1.94). This unfavorable prognostic value persisted in multivariate analysis, independently from other variables. There were significant differences in immune variables between “CSPG4-high” and “CSPG4-low” tumors. The "CSPG4-low" tumors displayed profiles suggesting higher anti-tumor cytotoxic immune response and higher potential vulnerability to immune checkpoint inhibitors (ICI). By contrast, the "CSPG4-high" tumors displayed profiles implying an immune-excluded tumor microenvironment, potentially induced by hypoxia, resulting from an immature chaotic microvasculature, and/or the presence of contractile myofibroblasts.


Patients with “CSPG4-high” STS, theoretically candidate for CAR.CIKs, display shorter DFS and an immune environment unfavorable to vulnerability to CAR.CIKs, which could be improved by combining anti-angiogenic drugs able to normalize the tumor vasculature. By contrast, “CSPG4-low” STS are better candidates for immune therapy involving ICI.


Soft-tissue sarcomas (STS) are rare and serious tumors of mesenchymal origin affecting children and adults [1]. These heterogeneous tumors encompass more than 100 distinct pathological subtypes associated with variable biological and clinical behaviors [2]. Despite the successful advances in surgical resection, radiotherapy, and chemotherapy, the outcome of patients with non-metastatic STS is still poor: the disease recurs in approximately 50% of patients, often with distant failure [3,4,5,6]. In the metastatic setting, the prognosis remains dismal with a 5-year survival rate inferior to 25% [7]. Currently, two issues represent major challenges in the management of STS. The first one is the improvement of prognostic factors that will help to better define the role, if any, of adjuvant chemotherapy. Historically associated with pathological grade [8] and size, and depth [9], the prognostic characterization is being refined, thanks to the contribution of genomic signatures such as CINSARC [10] or the immunologic constant of rejection (ICR) signature [11]. Clinical trials assessing prospectively the clinical utility of CINSARC (NCT03805022, NCT04307277) are ongoing. The second issue relies on the identification of novel therapeutic targets. The limited development of new chemotherapy drugs and targeted therapies during the last decade has not modified the overall prognosis, and doxorubicin remains the backbone of systemic treatment.

Chondroitin sulfate proteoglycan 4 (CSPG4), also called neural-glial 2 (NG2), is a cell surface proteoglycan, overexpressed in certain human cancers, with low expression in normal tissues and oncogenic roles in tumor growth and metastatic dissemination [12] via the promotion of cell proliferation, cell survival and drug resistance, angiogenesis, cell migration and invasion [13]. Inhibition of CSPG4 by gene deletion or treatment with anti-CSPG4 antibodies inhibits tumor growth in xenografts from some malignancies [14, 15]. Since CSPG4 is expressed in the mesenchymal progenitor cells [16, 17] and pericytes [18] from which STS are supposed to originate, its activation could play a role in sarcoma progression. Driving oncogenic mutations in Ng2/Cspg4-expressing cells leads to the formation of sarcomas [18]. During the last years, CSPG4 was described as a potential target of cellular immunotherapy in cancers [12]. CSPG4 is also known to influence activation, maturation, proliferation, and migration of different immune cell subsets suggesting likely interaction with immunotherapy efficiency [12].

Recent data suggested that the immune system might positively impact the outcome of patients with STS [11, 19, 20]. Several clinical trials testing immunotherapy based on immune checkpoint inhibitors (ICI) have been launched [21], but the results were relatively disappointing and remain controversial [22]. Identification of efficacy predictive markers, such as the presence of tertiary lymphoid structures [23, 24], is crucial in this so heterogeneous group of tumors. Another immunotherapy type is adoptive cellular therapy in which the T-cells are redirected by tumor antigen-specific chimeric antigen receptors (CAR-Ts). This approach, very effective in B-cell cancers [25,26,27], remains challenging in solid tumors [28]. One approach dedicated to improving the efficacy and safety of CAR-based therapies is the engineering of immune effectors different from αβT-lymphocytes, such as γδT-cells, natural killer (NK), NKT, or cytokine-induced killer (CIK) cells [29]. A recent study revealed the therapeutic potential of CSPG4-specific chimeric antigen receptor (CAR)-redirected cytokine-induced killer lymphocytes (CSPG4-CAR.CIKs) in STS [30]. In this study, the CSPG4-CAR.CIKs effectively targeted multiple STS pathological subtypes in vitro and in vivo. Antitumor activity against STS spheroids was associated with tumor recruitment, infiltration, and matrix penetration. In vivo, the CSPG4-CAR.CIKs delayed or reversed the tumor growth in three STS xenograft models (leiomyosarcoma, undifferentiated pleomorphic sarcoma, and fibrosarcoma).

Expression of CSPG4 is poorly known in STS. To our knowledge, only three studies [30,31,32] in the literature analyzed its expression in clinical samples, including respectively 251, 55, and 108 cases, but only the two smallest ones searched for correlations with tumor clinical features. To fill this gap and given the potential relevance of CSPG4 as a target for immunotherapy, we analyzed its expression in 1,378 localized STS clinical samples. We searched for correlations between expression and clinicopathological data, including disease-free survival (DFS), but also the components of the tumor immune landscape.


Patients and tumor samples

We collected clinicopathological and gene expression data of clinical STS samples from 15 public data sets through the National Center for Biotechnology Information (NCBI)/Genbank GEO, ArrayExpress databases, and authors' websites (Additional file 1: Table S1). The gene expression profiles had been generated using DNA microarrays or RNA-sequencing. The data sets were selected if the clinical and expression data, including CSPG4 expression level, were available. The final data set included 1,378 clinical samples. The study was approved by our institutional board. We also analyzed cancer cell line data from the Dependency Map (DepMap) portal (; accessed on 05 April 2022) to compare the RPPA-based protein expression levels versus the RNA-seq-based mRNA expression levels of CSPG4.

Gene expression data analysis

The pre-analytic processing of data was done as previously described [11]. Briefly, a first step of normalization was applied to each data set separately, by using quantile normalization for the already processed non-Affymetrix data, and Robust Multichip Average (RMA) with non-parametric quantile algorithm for the raw Affymetrix data. Normalization was done in R using Bioconductor and associated packages. We then mapped hybridization probes and kept the most variant one in a given data set when multiple probes mapped to the same GeneID. The already normalized TCGA RNAseq data were log2-transformed. Finally, we corrected the batch effects across the 15 studies by using z-score normalization, in which mean and standard deviation were measured on leiomyosarcoma samples. CSPG4 expression was analyzed as discrete variable (high versus low) by using its median expression level of the whole series as cut-off.

Because CSPG4 is a potential target of immunotherapy, we searched for correlations between its tumor expression and different immune variables. We applied to each data set separately several immunity-, fibroblastic- or vasculature-related multigene classifiers/scores: the tumor-infiltrating lymphocyte (TIL) score [33], the 24 Bindea’s innate and adaptive immune cell subpopulations and two Bindea’s vessels signatures [34], two fibroblast subsets’ classifications [35, 36], hypoxia, acidosis and lactic acidosis Gatza’s scores [37], the Immunologic Constant of Rejection (ICR) classifier [38], and metagenes associated representative of T-cell-inflamed signature (TIS) [39], of tertiary lymphoid structures (TLS) signature [40], of cytolytic activity score [41], and the antigen processing machinery (APM) score [42]. We also applied the CINSARC signature, now recognized as the most relevant prognostic signature in STS [10].

To explore the biological pathways linked to CSPG4 expression in STS, we applied a supervised analysis to the largest transcriptomics data set [10] including 131 “CSPG4-high” and 179 “CSPG4-low” tumors (learning set), and used the remaining tumors as independent validation set, including 558 “CSPG4-high” and 510 “CSPG4-low” tumors. In the learning set (N=310), we compared the expression profiles of 18,606 genes between “CSPG4-high” and “CSPG4-low” tumors using a moderated t-test with empirical Bayes statistic (limma R packages) and false discovery rate (FDR) correction. Significant genes were defined by the following thresholds: p<5%, q<10% and fold change (FC) superior to |1.5x|. Ontology analysis of the resulting gene list was based on GO biological processes of the Database for Annotation, Visualization and Integrated Discovery (DAVID; We tested the robustness of the resulting gene list in the remaining data sets defined as validation set (N=1068) by computing the Pearson correlation distance between each sample and the CSPG4-high profile defined in the learning set as the average expression of each significant gene in the CSPG4-high class. Validation samples with negative correlation were classified as “CSPG4-low-like” and those with positive correlations were classified as “CSPG4-high-like”. Supervised analyses also searched for differentially altered genes between “CSPG4-high” and “CPG4-low” STS at the DNA level using the copy-number alteration (CNA), mutation and methylome public data available on TCGA portal ( We compared the CNA profiles using the available “GISTIC thresholded by genes” TCGA data between “CSPG4-high” (N=123) and “CSPG4-low” (N=132) tumors across 24,776 genes using a Fisher’s exact test and FDR correction. Significant altered genes at “one copy gain”, “amplification”, “one copy loss” and “homozygous deletion” were defined by the following thresholds: p<5% and q<10%. Next, we compared the mutation profiles between “CSPG4-high” (N=119) and “CSPG4-low” (N=126) tumors across 7492 filtered (genes mutated in at least one sample) using a Fisher’s exact test and false discovery rate (FDR) correction. Silent mutations were excluded from analysis. Significant mutated genes were defined by the following thresholds: p<5% and q<10%. Finally, we compared the methylation profiles between “CSPG4-high” (N=101) and “CSPG4-low” (N=118) tumors across 450,000 probes using a Student t-test and FDR correction. Significant methylation sites were defined by the following thresholds: p<5%, q<10% and fold change (FC) superior to |1.25x|.

Statistical analysis

The continuous variables were described using median and range, and the discrete values using number and percentage. The correlations between CSPG4 expression-based groups and clinicopathological variables and molecular signatures were measured using the Fisher’s exact test or Student’s t-test when appropriate. The endpoint of prognostic analysis was the disease-free survival (DFS), calculated from the date of diagnosis until the date of distant relapse or death from any cause, whichever occurred first. The follow-up was measured from the date of diagnosis to the date of last news for event-free patients. Survivals were estimated using the Kaplan-Meier method and curves compared with the log-rank test. Uni- and multivariate prognostic analyses were done using Cox regression analysis (Wald test). The variables tested in univariate analysis were patients’ age and gender, pathological tumor type (liposarcomas (LPS), leiomyosarcomas (LMS), undifferentiated pleomorphic sarcomas (UPS), myxofibrosarcomas (MFS), others), grade (3, 1-2), and size, tumor depth (superficial, deep) and site (extremities, head and neck, internal trunk, superficial trunk), CINSARC-based risk (high, low) and the CSPG4-based classification (high, low). Multivariate analysis incorporated all variables with a p-value inferior to 5% in univariate analysis. The correlations of molecular immune/stromal variables with “CSPG4-high” versus “CSPG4-low” status of samples were assessed by logistic regression analysis with the glm function (R statistical package; significance estimated by specifying a binomial family for models with a logit link). All statistical tests were two-sided, and the significance threshold was 5%. Analyses were done with the survival package (version 2.43) from R software (version 3.5.2).


Patients’ characteristics and CSPG4 expression

A total of 1378 localized STS samples were available for analysis. Their characteristics are summarized in Table 1. The median patients’ age was 61 years (range, 2–92), and 51% of cases were male. The pathological subtypes were represented by 476 (35%) liposarcomas (LPS), 311 (23%) leiomyosarcomas (LMS), 288 (21%) undifferentiated pleomorphic sarcomas (UPS), 100 (7%) myxofibrosarcomas (MFS), and 177 (13%) other subtypes. Fifty-one percent were grade 3. The median pathological tumor size was 10 cm (range 1.2-39.5), and the most frequent tumor sites were extremities in 42% of cases and internal trunk in 39% of cases. The tumor was deep-seated (under fascia) in 89% of cases. Sixty-two percent (62%) of samples showed a complex genetic profile and 53% a low-risk according to the CINSARC signature.

Table 1 Clinicopathological characteristics of patients and samples

CSPG4 mRNA expression varied among the 1,378 tumors with a range of intensities over four units in log2 scale (Additional file 2: Figure S1A), suggesting a heterogeneous expression across the samples. Using the available omics data of 343 cancer cell lines including four sarcoma cell lines, we showed an excellent correlation between the mRNA and protein expression levels of CSPG4, with a Spearman’s rank correlation coefficient (rho) mean equal to 0.85 (p=9E-102) (Additional file 2: Figure S1A).

CSPG4 expression correlates with clinicopathological features

We searched for correlations between CSPG4 expression, assessed as a discrete variable (high versus low) and clinicopathological features (Table 1). There was no significant correlation with patients’ gender, pathological tumor size, pathological grade, and tumor depth. A trend for correlation existed with the genetic profile (p=0.063), with more frequent complex profiles in “CSPG4-high” tumors (65% versus 60%, p=0.095). Significant correlations were found with patients’ age (p=3.21E−02), pathological subtype (p=7.52E−21), tumor site (p=7.11E−03), and the CINSARC class (p=4.59E−02). Briefly, we found younger median age in “CSPG4-high” tumors (57 versus 60 years), more UPS (24% versus 18%) and MFS (12% versus 3%) in “CSPG4-low” tumors and more LMS (33% versus 13%) in “CSPG4-high” tumors, more frequent extremity site for “CSPG4-low” tumors (47% versus 36%) and internal trunk site for “CSPG4-high” tumors (47% versus 33%), and more CINSARC high-risk samples in “CSPG4-high” tumors (50% versus 45%).

CSPG4 expression correlates with disease-free survival

The information on DFS and CSPG4 expression was available for 610 patients. In the whole population, the median follow-up was 28 months (range 1-222), 223 DFS events occurred, and the 5-year DFS was 56% (95% CI 52–61) (Figure 1A). The 5-year DFS was 61% (95% CI 56–68) in the “CSPG4-low” subgroup and 49% (95%CI 42-57) in the “CSPG4-high” group (p=3.17E−03; Figure 1B). In univariate analysis for DFS (Table 2), this DFS difference corresponded to a 49% increased risk of event in the “CSPG4-high” group (HR=1.49, 95% CI 1.14–1.94; p=3.36E−03, Wald test). The other variables associated with shorter DFS included the pathological subtype (p=2.06E−05) and tumor size (HR=1.05, p=1.39E−03), and the CINSARC risk (p=4.35E−09). A trend was observed for the pathological grade (p= 0.098) and the tumor depth (p=0.062). No correlation was found between DFS and patients’ age and gender, and tumor site. In multivariate analysis, the “CSPG4-high” group (HR=3.47, 95% CI 1.73–6.95, p=4.59E−04) and higher pathological tumor size (HR=1.05, 95% CI 1.01–1.09, p=1.98E−02) remained significant, suggesting independent poor-prognosis value. The same result was observed when the ICR signature was added in the multivariate analysis (Additional file 7: Table S2).

Fig. 1
figure 1

Disease-free survival in patients with localized STS after surgery. A Kaplan–Meier DFS curves in the 610 informative patients for DFS. B Similar to A, but according to CSPG4 expression (low and high). The p-value is for the log-rank test

Table 2 Uni- and multivariate prognostic analyses for DFS

CSPG4 expression correlates with immune features

Next, we investigated whether CSPG4 expression was associated with immunity-related parameters in STS samples (Figure 2). First, we compared the TIL scores in “CSPG4-high” and “CSPG4-low” tumors. These scores quantify the infiltration level of TIL, lymphoid and myeloid cells in tumor tissues, based on expression data. “CSPG4-low” STS displayed higher TILs scores (p=6.26E−05), notably with higher infiltration of lymphoid cells (p=5.11E−07) and myeloid cells (p=3.11E−04) than “CSPG4-high” STS, suggesting overall higher infiltration by immune cells in the former. Second, we looked at the composition and functional orientation of these tumor infiltrated cells, using immunity-related fibroblast-related and vasculature-related signatures/scores. Analysis of the 24 immune cell types defined as the immunome [27], fibroblast subsets and two vessels signatures showed significant differences between “CSPG4-high” and “CSPG4-low” tumors. “CSPG4-high” tumors displayed a higher infiltrate in NK cells (p=7.43E−05), notably the immunoregulatory ones (NK CD56bright cells: p=8.76E−03). This was associated with a highly suppressive microenvironment, as demonstrated by the enrichment in signatures for acidosis (p=1.82E−06), lactic acidosis (p=2.74E−10), and hypoxia (p=5.38E−12) despite a higher signature of blood vessels (p=3.13E−04). CSPG4-high” tumors were also enriched in myofibroblasts CAFs (Kinchen_Myofibroblast: p=2.88E−24) and myCAFs (Tuveson_myCAF: p=4.00E−04) subsets. By contrast, the “CSPG4-low” tumors showed enrichment in signatures of macrophages (p=2.64E−06), T cells (p=1.41E−05), aDC cells (p=1.70E−05), neutrophils (p=1.78E−04), γδ T cells (p=6.30E−04), B cells (p=1.45E-03), cytotoxic cells (p=1.97E−03) and CD8 T cells (p=6.97E−03), higher inflammatory CAFs (Tuveson_iCAFs: p=7.35E−05) and antigen-presenting CAFs (Tuveson_apCAFs: p=5.38E−04), and higher lymph vessels signature (p=3.15E−13). Thus, except for NK cells, most immune cell types were observed in higher proportions in “CSPG4-low” tumors, which also had a less immune-suppressive environment. In parallel, analysis of other immune functional signatures confirmed these results. “CSPG4-high” samples showed a lower immune cytolytic activity score (p=1.70E−06) than “CSPG4-low samples”, as well as lower antigen processing/presentation machinery (APM) score (p=5.81E−04). They also displayed lower scores for signatures associated with response to ICI: Immunologic Constant of Rejection (ICR) score (p=1.66E−04), reflect of an antitumor cytotoxic immune response, T cell-inflamed signature (TIS) (p=7.28E−06), and tertiary lymphoid structure (TLS) score (p=1.72E−05).

Fig. 2
figure 2

Correlations between CSPG4 expression and immune/stromal features. For each signature, the Z-score is displayed by color bars representing the 25 and 75 percentiles of “CSPG4-high” (red) and “CSPG4-low” (green). Average score of both CSPG4 classes are displayed by color coded boxes ranged from -3 (blue) to 3 (red) using white as mid color. Forest plots showing the Odds Ratios (log10) of immune and stromal signatures/scores in the “CSPG4-high” versus “CSPG4-low STS samples in univariate logistic regression analysis. The p-values are for the logit link tests: the black squares correspond to significant variables and the grey ones to non-significant variables

CSPG4 expression and correlations with biological variables

In order to further characterize the “CSPG4-high” and “CGSPG4-low” STS samples, we applied supervised analysis to transcriptomics, genomics and epigenomics data. Regarding transcriptomics data, the supervised analysis identified 379 genes differentially expressed between the “CSPG4-high” tumors and the “CSPG4-low” tumors of the learning set, including 229 genes upregulated and 150 genes downregulated in the “CSPG4-high” tumors (Additional file 8: Table S3, Additional file 3: Figure S2A). The robustness of this gene list was confirmed in the validation set including 1,068 tumors (Additional file 3: Figure S2B). Ontology analysis (Additional file 9: Table S4) revealed a net overrepresentation of immune ontologies among the genes overexpressed in the “GSPG4-low” tumors, whereas the genes overexpressed in the “GSPG4-high” tumors were associated with multiple ontologies such as those related to cell migration, cytoskeleton, collagen and extra-cellular matrix, response to stress, growth and development.

Using the TCGA data set [43], we then applied supervised analyses to identify the DNA copy number alterations (CNA), the gene mutations and the gene methylations differentially observed between the two sample groups. Regarding the comparison of CNA between the “CSPG4-high” and “CGSPG4-low” tumors, the profiles were very similar; no significant difference was observed regarding “one copy gains”, “amplifications”, and “homozygous deletions“, whereas 2255 genes were differentially altered regarding the frequency of “one copy loss”, including 255 genes more frequently lost in the ”CSPG4-high” samples and 2292 genes more frequently lost in the ”CSPG4-low” samples (Additional file 10: Table S5, Additional file 4: Figure S3). The comparison of mutational profiles showed no significant difference in term of mutation frequency between the “CSPG4-high” and “CGSPG4-low” tumors (Additional file 11: Table S6, Additional file 5: Figure S4). Three genes were more frequently mutated in the “CSPG4-high” samples (RB1, PTEN, PRKDC) and two were more frequently mutated in the “CSPG4-low” samples (PEG2, AHK2), but none of them remained significant after FDR correction. The comparison of methylation sites between the “CSPG4-high” and “CGSPG4-low” tumors identified 84 significant sites, including 35 more methylated in the “CSPG4-high” group and 49 in the “CGSPG4-low” group (Additional file 12: Table S7, Additional file 6: Figure S5).

Finally, we searched for correlation between the CSPG4 expression status and the five multi-omics subtypes defined in the TCGA set by iClustering integrating DNA copy number, DNA methylation, and expression of mRNA and miRNA [43]. As shown in Table 1, the correlation was highly significant : the “CSPG4-high” tumors were more frequently classified as iClusters 1 and 2, whereas the “CSPG4-low” tumors were more frequently classified as iCluster 3, 4 and 5 (p=6.51E−14).


In this series of 1378 STS clinical samples, high CSPG4 expression was an independent unfavorable prognostic factor for DFS and was associated with low cytotoxic immune response. To our knowledge, this is the largest study analyzing the expression of this new potential target for immune therapy in STS.

Our analysis was based on gene expression of CSPG4 in a very large series of clinical samples. The strong correlation between mRNA and protein expression levels of CSPG4 that we evidenced in 343 cancer cell lines suggests that CSPG4 protein expression parallels observations made at the transcriptomic level. Such mRNA level analysis allowed not only to avoid the classical limitations of immunohistochemistry (availability of antibodies, standardization, positivity cut-off, interpretation subjectivity…), but also to work on a large series of clinical samples and to search for correlations with expression of biologically and clinically relevant immune signatures. We found heterogeneous expression of CSPG4 in clinical STS samples, as reported in the three previous studies on STS, all also performed at the transcriptional level [30,31,32]. Benassi et al. profiled 55 samples [31], Cattaruzza et al. 108 samples including the 55 previous ones [32], and Leuci et al. analyzed 251 TCGA samples [30]. Benassi et al. focused their analysis on 55 deeply localized, >5-cm diameter and high-grade lesions [31] and did not find any significant correlation between the CSPG4 expression and clinicopathological variables. The same team [32] extended this series to a total of 108 cases and found higher expression in synovial sarcoma. Leuci et al did not search for eventual correlations [30]. In our present study, we found that expression was mainly associated with the pathological type, with higher expression in LMS and MFS and lower expression in UPS and LPS. Other significant correlations existed with age, tumor site, and CINSARC risk, the “CSPG4-high" samples being more frequently high-risk according to CINSARC than the “CSPG4-low” samples.

Our prognostic analysis included a large series of 610 patients informative for DFS, the largest prognostic study reported so far. In uni- and multivariate analyses, high CSGP4 expression was associated with a higher risk of DFS event, independently from other prognostic variables including pathological tumor size and grade and CINSARC. The 5-year DFS was 61% in the “CSPG4-low” subgroup versus 49% in the “CSPG4-high” group, representing a 49% increased risk of event in the “CSPG4-high” group. These results are consistent with the sole other prognostic study published in the literature [31]. In a series of 108 patients with STS deeply localized, >5-cm diameter and grade 2–3 [32], the authors found higher CSPG4 expression in the metastases when compared with paired primary lesions and when compared with normal lung and other tissues. They confirmed this result at the protein level using IHC based on a monoclonal antibody generated by their own team, with high expression on the surface of neoplastic cells and neovascular structures of primary and secondary tumor masses. Finally, they demonstrated in multivariate analysis the independent unfavorable prognostic value of high CSPG4 expression for MFS [32]. This prognostic impact in STS was investigated at the functional level. Using in vitro and in vivo models, the same team [32] showed that CSPG4 controlled the tumor progression (local growth, cell adhesion, and motility, and cell survival) by mediating the interaction of sarcoma cells with the host extracellular matrix, in particular with collagen 6 (Col VI) that accumulates in the peri- and intra-lesional stroma. In another study [44], Hsu et al. showed that the effects of CSPG4 on STS growth depended of the tumor developmental stage: in established murine and human STS, inhibition of CSPG4, using anti-CSPG4 antibody immunotherapy or gene deletion, decreased the cell proliferation and tumor size and increased apoptosis, whereas Ng2/Cspg4 deletion at the time of tumor initiation resulted in the opposite effect on tumor growth. The prognostic value of CSPG4 expression is likely tumor type-dependent. We recently showed the good-prognosis value of high CSPG4 expression in a series of 309 GIST [45], whereas higher expression was associated with poorer prognosis in melanoma [46], glioblastoma [47], breast cancer [48], head and neck squamous cell carcinomas [49], and hepatocellular carcinoma [50].

The transmembrane proteoglycan CSPG4 had been originally identified by Dr Ferrone‘s team as a highly immunogenic tumor antigen on the surface of melanoma cells and was named High Molecular Weight Melanoma-Associated Antigen [51]. This antigen was then characterized by the same team in several cancers: melanoma [46], TNBC [14], malignant mesothelioma [15], acute myeloid leukemia [52], chordoma [53], glioblastoma [54], and osteosarcoma [55]. It is now identified as a potential therapeutic target for immune therapy in different cancers, including anti-idiotypic antibodies in melanoma [56,57,58], monoclonal antibodies in triple-negative breast cancer [14] and melanoma [59], antibody-drug conjugate in melanoma [60], and CAR-T cells in many cancers [61]. In sarcomas, Leuci et al. recently demonstrated in vitro and in vivo the anti-tumor activity of CSPG4-CAR.CIKs in STS pre-clinical samples. We thus searched for eventual correlations between CSPG4 expression and immune/stromal features. Several of them were differentially enriched between "CSPG4-high" and "CSPG4-low" STS. "CSPG4-low" tumors had higher scores for immune signatures suggesting higher infiltration by immune cells, higher lymphatic vasculature and higher anti-tumor immune response. They were also associated with the presence of TLS and other signatures indicative of better response to ICI treatment [23], despite signs of exhaustion. In line with this, "CSPG4-low" tumors showed higher iCAFs and apCAFs, which respectively secrete cytokines to attract and entrap lymphocytes to turn them into harmless cells. Altogether, these data suggested that "CSPG4-low” STS could be better candidates for immune therapy involving ICI, which will fully unleash pre-infiltrated CD8 T cells cytotoxic potential [45]. By contrast, the "CSPG4-high" tumors displayed immune profiles suggesting an immune desert or immune-excluded tumor microenvironment, lacking all kinds of immune cells required to mount an effective anti-tumor response, except for the NKbright cells. This lack of immune infiltrated cells can be related to the presence of myofibroblasts and myCAFs, which are fibroblastic subsets endowed with contractile features, which affect the distribution of blood vasculature [62]. They also secrete a massive amount of matrix and prevent lymphocyte accessibility to tumor cells. For all these reasons, immune desert tumor microenvironments have been reported to have limited sensitivity to ICI. In the study reported by Leuci et al., the tumor elimination in vitro after treatment with CSPG4-CAR.CIKs was dependent on the expression level of tumor cells, suggesting that "CSPG4-high" STS should represent the most candidate population for such treatment [30]. But our present data suggest that a treatment based on CSPG4-CAR.CIKs infiltrating cells that will target "CSPG4-high" tumor cells could be interesting if some aspects responsible for the immune desert can be overcome first. Notably, immune desert can be induced by hypoxia. Hypoxia is a key determinant of tumor aggressiveness, therapy resistance and has a dampening effect on antitumor immune responses and immune cells recruitment. Hypoxia and lactic acidosis induce the functional suppression of NK cells. We found that both hypoxia and lactic acidosis were enhanced in "CSPG4-high" tumors. This observation is consistent with a certain overexpression of CSPG4 induced by the chronic hypoxia in vitro [63] and likely explains why only immunoregulatory NKbright cells are present in "CSPG4-high" tumors. A strategy based on CSPG4-CAR.CIKs would thus require to overcome the functional suppression. Hypoxia results from an immature chaotic microvasculature within the tumor. Strategies that seek to normalize the tumor vasculature, such as tyrosine kinase inhibitors (TKIs) that target pro-angiogenic receptors, should help reduce hypoxia, enhance tumor’s perfusion and optimize therapy uptake. This would be a key point to improve the efficacy of CSPG4-CAR.CIKs in STS.

Supervised analysis of transcriptomics data between the “CSPG4-high” and “CSPG4-low” tumors confirmed the immune desert observed in the “CSPG4-high” tumors and their enrichment in genes related to cell migration, collagen and extra-cellular matrix, response to stress, growth and development. That might explain in part their poorer prognosis when compared with the “CSPG4-low” tumors. No significant difference was observed regarding the frequency of gene amplification or gene deletion between both tumor groups, notably for CSPG4, suggesting that the DNA copy number is not responsible for the CSPG4 differential expression. Similarly, no gene showed a significant difference in term of mutation frequency between the “CSPG4-high” and “CSPG4-low” tumors. By contrast, 84 sites were differentially methylated between both tumor groups, calling for further investigations in order to assess an eventual functional link with CSPG4.


To our knowledge, our study is the largest one describing CSPG4 expression in clinical cancer samples, here in STS. We show that expression of CSPG4 in STS samples is heterogeneous and associated independently with shorter DFS and with an immune landscape not favorable to anti-tumor cytotoxic response. The main strength of our study lies in the high number of rare tumor samples analysed. Others include its originality and analysis of correlations with tumor immune variables. Limitations include the retrospective nature and associated biases, analysis of bulk samples, and absence of analysis at the protein level. Obviously, analysis of larger clinical series and protein analysis are warranted. Yet, our results suggest the prognostic value of CSPG4 expression in STS and describe the immune microenvironment of tumors candidate to specific immune therapy such as CSPG4-CAR.CIKs. Whether patients with STS, and what type of patients, will benefit from such immunotherapy deserve rapid assessment in prospective clinical trials.

Availability of data and materials

All data generated or analysed during this study are included in this published article [and its supplementary information file: Additional file 13: Table S8.



Soft tissue sarcomas


Chondroitin sulfate proteoglycan 4




Immunologic constant of rejection


Immune checkpoint inhibitors


Chimeric antigen receptors


CSPG4-specific chimeric antigen receptor (CAR)-redirected cytokine-induced killer lymphocytes


National Center for Biotechnology Information


Disease-free survival


Tumor-infiltrating lymphocyte


T-cell-inflamed signature


Tertiary lymphoid structures


Antigen processing machinery




Leiomyosarcomas (LMS)


Undifferentiated pleomorphic sarcomas




  1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2019. CA Cancer J Clin. 2019;69:7–34.

    Article  PubMed  Google Scholar 

  2. Karavasilis V, Seddon BM, Ashley S, Al-Muderis O, Fisher C, Judson I. Significant clinical benefit of first-line palliative chemotherapy in advanced soft-tissue sarcoma. Cancer. 2008;112:1585–91.

    Article  PubMed  Google Scholar 

  3. Correa R, Gómez-Millán J, Lobato M, Fernández A, Ordoñez R, Castro C, Lupiañez Y, Medina JA. Radiotherapy in soft-tissue sarcoma of the extremities. Clin Transl Oncol. 2018;20:1127–35.

    Article  CAS  PubMed  Google Scholar 

  4. Ray-Coquard I, Serre D, Reichardt P, Martín-Broto J, Bauer S. Options for treating different soft tissue sarcoma subtypes. Future Oncol Lond Engl. 2018;14:25–49.

    Article  CAS  Google Scholar 

  5. Ratan R, Patel SR. Chemotherapy for soft tissue sarcoma. Cancer. 2016;122:2952–60.

    Article  PubMed  Google Scholar 

  6. Ayodele O, Razak ARA. Immunotherapy in soft-tissue sarcoma. Curr Oncol. 2020;27:17–23.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Pollack SM, Ingham M, Spraker MB, Schwartz GK. Emerging targeted and immune-based therapies in sarcoma. J Clin Oncol. 2018;36:125–35.

    Article  CAS  PubMed  Google Scholar 

  8. Guillou L, Coindre JM, Bonichon F, Nguyen BB, Terrier P, Collin F, Vilain MO, Mandard AM, Le Doussal V, Leroux A, et al. Comparative study of the National Cancer Institute and French Federation of Cancer Centers Sarcoma Group grading systems in a population of 410 adult patients with soft tissue sarcoma. J Clin Oncol. 1997;15:350–62.

    Article  CAS  PubMed  Google Scholar 

  9. Casali PG, Abecassis N, Aro HT, Bauer S, Biagini R, Bielack S, Bonvalot S, Boukovinas I, Bovee JVMG, Brodowicz T, et al. Soft tissue and visceral sarcomas: ESMO-EURACAN Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann Oncol. 2018;29:iv51-67.

    Article  CAS  PubMed  Google Scholar 

  10. Chibon F, Lagarde P, Salas S, Pérot G, Brouste V, Tirode F, Lucchesi C, de Reynies A, Kauffmann A, Bui B, et al. Validated prediction of clinical outcome in sarcomas and multiple types of cancer on the basis of a gene expression signature related to genome complexity. Nat Med. 2010;16:781–7.

    Article  CAS  PubMed  Google Scholar 

  11. Bertucci F, Niziers V, de Nonneville A, Finetti P, Mescam L, Mir O, Italiano A, Le Cesne A, Blay JY, Ceccarelli M, et al. Immunologic constant of rejection signature is prognostic in soft-tissue sarcoma and refines the CINSARC signature. J Immunother Cancer. 2022;10: e003687.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Ilieva KM, Cheung A, Mele S, Chiaruttini G, Crescioli S, Griffin M, Nakamura M, Spicer JF, Tsoka S, Lacy KE, et al. Chondroitin sulfate proteoglycan 4 and its potential as an antibody immunotherapy target across different tumor types. Front Immunol. 2017;8:1911.

    Article  PubMed  Google Scholar 

  13. Nicolosi PA, Dallatomasina A, Perris R. Theranostic impact of NG2/CSPG4 proteoglycan in cancer. Theranostics. 2015;5:530–44.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  14. Wang X, Osada T, Wang Y, Yu L, Sakakura K, Katayama A, McCarthy JB, Brufsky A, Chivukula M, Khoury T, et al. CSPG4 protein as a new target for the antibody-based immunotherapy of triple-negative breast cancer. J Natl Cancer Inst. 2010;102:1496–512.

    Article  PubMed  PubMed Central  Google Scholar 

  15. Rivera Z, Ferrone S, Wang X, Jube S, Yang H, Pass HI, Kanodia S, Gaudino G, Carbone M. CSPG4 as a target of antibody-based immunotherapy for malignant mesothelioma. Clin Cancer Res. 2012;18:5352–63.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  16. Kozanoglu I, Boga C, Ozdogu H, Sozer O, Maytalman E, Yazici AC, Sahin FI. Human bone marrow mesenchymal cells express NG2: possible increase in discriminative ability of flow cytometry during mesenchymal stromal cell identification. Cytotherapy. 2009;11:527–33.

    Article  CAS  PubMed  Google Scholar 

  17. Russell KC, Tucker HA, Bunnell BA, Andreeff M, Schober W, Gaynor AS, Strickler KL, Lin S, Lacey MR, O’Connor KC. Cell-surface expression of neuron-glial antigen 2 (NG2) and melanoma cell adhesion molecule (CD146) in heterogeneous cultures of marrow-derived mesenchymal stem cells. Tissue Eng Part A. 2013;19:2253–66.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Sato S, Tang YJ, Wei Q, Hirata M, Weng A, Han I, Okawa A, Takeda S, Whetstone H, Nadesan P, et al. Mesenchymal tumors can derive from Ng2/Cspg4-expressing pericytes with β-catenin modulating the neoplastic phenotype. Cell Rep. 2016;16:917–27.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  19. Lee C-H, Espinosa I, Vrijaldenhoven S, Subramanian S, Montgomery KD, Zhu S, Marinelli RJ, Peterse JL, Poulin N, Nielsen TO, et al. Prognostic significance of macrophage infiltration in leiomyosarcomas. Clin Cancer Res. 2008;14:1423–30.

    Article  CAS  PubMed  Google Scholar 

  20. Koumarianou A, Duran-Moreno J. The sarcoma immune landscape: emerging challenges, prognostic significance and prospective impact for immunotherapy approaches. Cancers. 2021;13:363.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Siozopoulou V, Domen A, Zwaenepoel K, Van Beeck A, Smits E, Pauwels P, Marcq E. Immune checkpoint inhibitory therapy in sarcomas: is there light at the end of the tunnel? Cancers. 2021;13:360.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  22. D’Angelo SP, Mahoney MR, Van Tine BA, Atkins J, Milhem MM, Jahagirdar BN, Antonescu CR, Horvath E, Tap WD, Schwartz GK, et al. Nivolumab with or without ipilimumab treatment for metastatic sarcoma (Alliance A091401): two open-label, non-comparative, randomised, phase 2 trials. Lancet Oncol. 2018;19:416–26.

    Article  PubMed  PubMed Central  Google Scholar 

  23. Italiano A, Bessede A, Pulido M, Bompas E, Piperno-Neumann S, Chevreau C, Penel N, Bertucci F, Toulmonde M, Bellera C, et al. Pembrolizumab in soft-tissue sarcomas with tertiary lymphoid structures: a phase 2 PEMBROSARC trial cohort. Nat Med. 2022.

    Article  PubMed  Google Scholar 

  24. Fridman WH, Meylan M, Petitprez F, Sun C-M, Italiano A, Sautès-Fridman C. B cells and tertiary lymphoid structures as determinants of tumour immune contexture and clinical outcome. Nat Rev Clin Oncol. 2022;19:441–57.

    Article  CAS  PubMed  Google Scholar 

  25. Jacoby E, Shahani SA, Shah NN. Updates on CAR T-cell therapy in B-cell malignancies. Immunol Rev. 2019;290:39–59.

    Article  CAS  PubMed  Google Scholar 

  26. Grupp SA, Kalos M, Barrett D, Aplenc R, Porter DL, Rheingold SR, Teachey DT, Chew A, Hauck B, Wright JF, et al. Chimeric antigen receptor-modified T cells for acute lymphoid leukemia. N Engl J Med. 2013;368:1509–18.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Vera J, Savoldo B, Vigouroux S, Biagi E, Pule M, Rossig C, Wu J, Heslop HE, Rooney CM, Brenner MK, et al. T lymphocytes redirected against the kappa light chain of human immunoglobulin efficiently kill mature B lymphocyte-derived malignant cells. Blood. 2006;108:3890–7.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Safarzadeh Kozani P, Safarzadeh Kozani P, Ahmadi Najafabadi M, Yousefi F, Mirarefin SMJ, Rahbarizadeh F. Recent advances in solid tumor CAR-T cell therapy: driving tumor cells from hero to zero? Front Immunol. 2022;13: 795164.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Rotolo R, Leuci V, Donini C, Cykowska A, Gammaitoni L, Medico G, Valabrega G, Aglietta M, Sangiolo D. CAR-based strategies beyond T lymphocytes: integrative opportunities for cancer adoptive immunotherapy. Int J Mol Sci. 2019;20:E2839.

    Article  PubMed  Google Scholar 

  30. Leuci V, Donini C, Grignani G, Rotolo R, Mesiano G, Fiorino E, Gammaitoni L, D’Ambrosio L, Merlini A, Landoni E, et al. CSPG4-specific CAR.CIK lymphocytes as a novel therapy for the treatment of multiple soft-tissue sarcoma histotypes. Clin Cancer Res. 2020;26:6321–34.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  31. Benassi MS, Pazzaglia L, Chiechi A, Alberghini M, Conti A, Cattaruzza S, Wassermann B, Picci P, Perris R. NG2 expression predicts the metastasis formation in soft-tissue sarcoma patients. J Orthop Res. 2009;27:135–40.

    Article  PubMed  Google Scholar 

  32. Cattaruzza S, Nicolosi PA, Braghetta P, Pazzaglia L, Benassi MS, Picci P, Lacrima K, Zanocco D, Rizzo E, Stallcup WB, et al. NG2/CSPG4-collagen type VI interplays putatively involved in the microenvironmental control of tumour engraftment and local expansion. J Mol Cell Biol. 2013;5:176–93.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Truntzer C, Isambert N, Arnould L, Ladoire S, Ghiringhelli F. Prognostic value of transcriptomic determination of tumour-infiltrating lymphocytes in localised breast cancer. Eur J Cancer. 2019;120:97–106.

    Article  CAS  Google Scholar 

  34. Bindea G, Mlecnik B, Tosolini M, Kirilovsky A, Waldner M, Obenauf AC, Angell H, Fredriksen T, Lafontaine L, Berger A, et al. Spatiotemporal dynamics of intratumoral immune cells reveal the immune landscape in human cancer. Immunity. 2013;39:782–95.

    Article  CAS  PubMed  Google Scholar 

  35. Kinchen J, Chen HH, Parikh K, Antanaviciute A, Jagielowicz M, Fawkner-Corbett D, Ashley N, Cubitt L, Mellado-Gomez E, Attar M, et al. Structural remodeling of the human colonic mesenchyme in inflammatory bowel disease. Cell. 2018;175:372-386.e17.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  36. Elyada E, Bolisetty M, Laise P, Flynn WF, Courtois ET, Burkhart RA, Teinor JA, Belleau P, Biffi G, Lucito MS, et al. Cross-species single-cell analysis of pancreatic ductal adenocarcinoma reveals antigen-presenting cancer-associated fibroblasts. Cancer Discov. 2019;9:1102–23.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  37. Gatza ML, Lucas JE, Barry WT, Kim JW, Wang Q, Crawford MD, Datto MB, Kelley M, Mathey-Prevot B, Potti A, et al. A pathway-based classification of human breast cancer. Proc Natl Acad Sci USA. 2010;107:6994–9.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  38. Hendrickx W, Simeone I, Anjum S, Mokrab Y, Bertucci F, Finetti P, Curigliano G, Seliger B, Cerulo L, Tomei S, et al. Identification of genetic determinants of breast cancer immune phenotypes by integrative genome-scale analysis. OncoImmunology. 2017;6: e1253654.

    Article  PubMed  PubMed Central  Google Scholar 

  39. Ayers M, Lunceford J, Nebozhyn M, Murphy E, Loboda A, Kaufman DR, Albright A, Cheng JD, Kang SP, Shankaran V, et al. IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Invest. 2017;127:2930–40.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Coppola D, Nebozhyn M, Khalil F, Dai H, Yeatman T, Loboda A, Mulé JJ. Unique ectopic lymph node-like structures present in human primary colorectal carcinoma are identified by immune gene array profiling. Am J Pathol. 2011;179:37–45.

    Article  PubMed  PubMed Central  Google Scholar 

  41. Rooney MS, Shukla SA, Wu CJ, Getz G, Hacohen N. Molecular and genetic properties of tumors associated with local immune cytolytic activity. Cell. 2015;160:48–61.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  42. Thompson JC, Davis C, Deshpande C, Hwang W-T, Jeffries S, Huang A, Mitchell TC, Langer CJ, Albelda SM. Gene signature of antigen processing and presentation machinery predicts response to checkpoint blockade in non-small cell lung cancer (NSCLC) and melanoma. J Immunother Cancer. 2020;8: e000974.

    Article  PubMed  PubMed Central  Google Scholar 

  43. Abeshouse A, Adebamowo C, Adebamowo SN, Akbani R, Akeredolu T, Ally A, Anderson ML, Anur P, Appelbaum EL, Armenia J, et al. Comprehensive and integrated genomic characterization of adult soft tissue sarcomas. Cell. 2017;171:950-965.e28.

    Article  PubMed Central  Google Scholar 

  44. Hsu S-HC, Nadesan P, Puviindran V, Stallcup WB, Kirsch DG, Alman BA. Effects of chondroitin sulfate proteoglycan 4 (NG2/CSPG4) on soft-tissue sarcoma growth depend on tumor developmental stage. J Biol Chem. 2018;293:2466–75.

    Article  CAS  PubMed  Google Scholar 

  45. de Nonneville A, Finetti P, Picard M, Monneur A, Pantaleo MA, Astolfi A, Ostrowski J, Birnbaum D, Mamessier E, Bertucci F. CSPG4 expression in GIST is associated with better prognosis and strong cytotoxic immune response. Cancers. 2022;14:1306.

    Article  PubMed  PubMed Central  Google Scholar 

  46. Price MA, Colvin Wanshura LE, Yang J, Carlson J, Xiang B, Li G, Ferrone S, Dudek AZ, Turley EA, McCarthy JB. CSPG4, a potential therapeutic target, facilitates malignant progression of melanoma. Pigment Cell Melanoma Res. 2011;24:1148–57.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  47. Svendsen A, Verhoeff JJC, Immervoll H, Brøgger JC, Kmiecik J, Poli A, Netland IA, Prestegarden L, Planagumà J, Torsvik A, et al. Expression of the progenitor marker NG2/CSPG4 predicts poor survival and resistance to ionising radiation in glioblastoma. Acta Neuropathol (Berl). 2011;122:495–510.

    Article  CAS  Google Scholar 

  48. Hsu NC, Nien P-Y, Yokoyama KK, Chu P-Y, Hou M-F. High chondroitin sulfate proteoglycan 4 expression correlates with poor outcome in patients with breast cancer. Biochem Biophys Res Commun. 2013;441:514–8.

    Article  CAS  PubMed  Google Scholar 

  49. Warta R, Herold-Mende C, Chaisaingmongkol J, Popanda O, Mock A, Mogler C, Osswald F, Herpel E, Küstner S, Eckstein V, et al. Reduced promoter methylation and increased expression of CSPG4 negatively influences survival of HNSCC patients. Int J Cancer. 2014;135:2727–34.

    Article  CAS  PubMed  Google Scholar 

  50. Lu L-L, Sun J, Lai J-J, Jiang Y, Bai L-H, Zhang L-D. Neuron-glial antigen 2 overexpression in hepatocellular carcinoma predicts poor prognosis. World J Gastroenterol. 2015;21:6649–59.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Wilson BS, Imai K, Natali PG, Ferrone S. Distribution and molecular characterization of a cell-surface and a cytoplasmic antigen detectable in human melanoma cells with monoclonal antibodies. Int J Cancer. 1981;28:293–300.

    Article  CAS  PubMed  Google Scholar 

  52. Fenton M, Whiteside TL, Ferrone S, Boyiadzis M. Chondroitin sulfate proteoglycan-4 (CSPG4)-specific monoclonal antibody 225.28 in detection of acute myeloid leukemia blasts. Oncol Res. 2015;22:117–21.

    Article  PubMed  PubMed Central  Google Scholar 

  53. Schoenfeld AJ, Wang X, Wang Y, Hornicek FJ, Nielsen GP, Duan Z, Ferrone S, Schwab JH. CSPG4 as a prognostic biomarker in chordoma. Spine J. 2016;16:722–7.

    Article  PubMed  Google Scholar 

  54. Pellegatta S, Savoldo B, Di Ianni N, Corbetta C, Chen Y, Patané M, Sun C, Pollo B, Ferrone S, DiMeco F, et al. Constitutive and TNFα-inducible expression of chondroitin sulfate proteoglycan 4 in glioblastoma and neurospheres: Implications for CAR-T cell therapy. Sci Transl Med. 2018;10:eaao2731.

    Article  PubMed  PubMed Central  Google Scholar 

  55. Riccardo F, Tarone L, Iussich S, Giacobino D, Arigoni M, Sammartano F, Morello E, Martano M, Gattino F, Maria RD, et al. Identification of CSPG4 as a promising target for translational combinatorial approaches in osteosarcoma. Ther Adv Med Oncol. 2019;11:1758835919855491.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  56. Mittelman A, Chen ZJ, Yang H, Wong GY, Ferrone S. Human high molecular weight melanoma-associated antigen (HMW-MAA) mimicry by mouse anti-idiotypic monoclonal antibody MK2-23: induction of humoral anti-HMW-MAA immunity and prolongation of survival in patients with stage IV melanoma. Proc Natl Acad Sci USA. 1992;89:466–70.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  57. Mittelman A, Chen ZJ, Liu CC, Hirai S, Ferrone S. Kinetics of the immune response and regression of metastatic lesions following development of humoral anti-high molecular weight-melanoma associated antigen immunity in three patients with advanced malignant melanoma immunized with mouse antiidiotypic monoclonal antibody MK2-23. Cancer Res. 1994;54:415–21.

    CAS  PubMed  Google Scholar 

  58. Mittelman A, Chen GZ, Wong GY, Liu C, Hirai S, Ferrone S. Human high molecular weight-melanoma associated antigen mimicry by mouse anti-idiotypic monoclonal antibody MK2-23: modulation of the immunogenicity in patients with malignant melanoma. Clin Cancer Res. 1995;1:705–13.

    CAS  PubMed  Google Scholar 

  59. Uranowska K, Samadaei M, Kalic T, Pinter M, Breiteneder H, Hafner C. A chondroitin sulfate proteoglycan 4-specific monoclonal antibody inhibits melanoma cell invasion in a spheroid model. Int J Oncol. 2021;59:70.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  60. Hoffmann RM, Crescioli S, Mele S, Sachouli E, Cheung A, Chui CK, Andriollo P, Jackson PJM, Lacy KE, Spicer JF, et al. A Novel Antibody-Drug Conjugate (ADC) delivering a DNA mono-alkylating payload to Chondroitin Sulfate Proteoglycan (CSPG4)-expressing melanoma. Cancers. 2020;12:E1029.

    Article  PubMed  Google Scholar 

  61. Harrer DC, Dörrie J, Schaft N. CSPG4 as target for CAR-T-cell therapy of various tumor entities-merits and challenges. Int J Mol Sci. 2019;20:E5942.

    Article  PubMed  Google Scholar 

  62. Tomlinson J, Barsky SH, Nelson S, Singer S, Pezeshki B, Lee MC, Eilber F, Nguyen M. Different patterns of angiogenesis in sarcomas and carcinomas. Clin Cancer Res. 1999;5:3516–22.

    CAS  PubMed  Google Scholar 

  63. Keleg S, Titov A, Heller A, Giese T, Tjaden C, Ahmad SS, Gaida MM, Bauer AS, Werner J, Giese NA. Chondroitin sulfate proteoglycan CSPG4 as a novel hypoxia-sensitive marker in pancreatic tumors. PLoS ONE. 2014;9: e100178.

    Article  PubMed  PubMed Central  Google Scholar 

Download references


Our work was supported by Institut Paoli-Calmettes, la Ligue Nationale Contre le Cancer (Label Ligue EL2022: FB), and Le Prix Ruban Rose 2020 (FB). JYB is supported by NetSARC (INCA), LYRICAN (INCA-DGOS-INSERM 12563), InterSARC (INCA), LabEx DEvweCAN (ANR-10-LABX-0061), and EURACAN (EC 739521).

Author information

Authors and Affiliations



Conceptualization, LB, AN, PF, and FB; methodology, LB, AN, PF and FB; validation, FB; formal analysis, PF and FB; data curation, PF; writing—original draft preparation, LB, AN, PF, EM, and FB; writing—review and editing, all authors; supervision, FB. All authors read and approved the final manuscript.

Corresponding author

Correspondence to François Bertucci.

Ethics declarations

Ethics approval and consent to participate

Our in silico study is based upon public data from published studies in which the informed patient's consent to participate and the ethics and institutional review board were already obtained by authors. The study was approved by our institutional review board.

Consent for publication

Not applicable.

Competing interests

ALC received consulting fees from Pharmamar, Bayer, and Deciphera. The other authors have declared no conflict of interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1:

Table S1. (File format .xls). List of soft tissue sarcoma data sets included.

Additional file 2:

Figure S1. (File format .ppt). CSPG4 expression in clinical STS samples and cancer cell lines. A/ Box plot of mRNA expression levels in the 1,378 STS clinical samples. B/ Spearman’s rank correlation coefficient (rho) between mRNA (RNAseq data) and protein (RPPA) expression in 343 cancer cell lines (grey) including four sarcoma cell lines (orange).

Additional file 3: Figure S2.

(File format .ppt). Identification and validation of the CSPG4 gene expression signature in STS samples. A/ Identification of the signature in the Chibon’s learning set (N=310). Left, Volcano-plot showing the 379 genes differentially expressed between “CSPG4-highversusCSPG4-low” STS samples. Middle, box plot of classification score (Pearson correlation, r) between both CSPG4 classes (Student t-test) associated with cross-table between observed and predicted CSPG4 groups (Fisher’s exact test). B/ Validation in the remaining public data sets (N=1,068). Box plot of classification score (Pearson correlation, r) between both CSPG4 classes (Student t-test) associated with cross-table between observed and predicted CSPG4 groups (Fisher’s exact test).

Additional file 4: Figure S3.

(File format .ppt). Comparison of the CNA profiles between the “CSPG4-high” (N=123) and “CSPG4-low” (N=132) tumors. A/ Heatmap of CNA in which the 24,776 genes were sorted by chromosomal location and the 255 STS samples were sorted by their CSPG4 mRNA expression level. B/ Left, frequency plots of CNA in both CSPG4 classes following four CNA levels: one-copy-gain (red), amplification (dark red), one-copy loss (green) and homozygous deletion (dark green). Loss alteration frequencies were negatively weighted. Right, supervised analysis of CNA frequencies between “CSPG4-high” vs. “CSPG4-low” tumors. Plotted values represent the –log10 corrected p-values of the Fisher’s exact test weighted by the sign of the odds ratio for each CNA level. The vertical orange lines represents the significance thresholds (i.e. q<0.1).

Additional file 5:

Figure S4. (File format .ppt). Distribution of mutations of the top 45 mutated genes in STS samples and comparison between the “CSPG4-high” (N=119) and “CSPG4-low” (N=126) tumors. A/ Heatmap of mutations across 45 genes with mutation rate greater than 3% over all patients. Genes were sorted according their global mutation rate (blue bar plot to the right) and the 245 tumors were sorted by their CSPG4 mRNA expression level. B/ Bar plot of mutation frequency according to the “CSPG4-high” (red) or “CSPG4-low” (green) status. Orange plotted values represent the -log10 p-values of the Fisher’s exact test. The vertical dashed orange line represents the significance threshold (i.e. p<0.05), but no gene is significant after FDR correction (q-value).

Additional file 6:

Figure S5. (File format .ppt). Comparison of methylation site profiles between the “CSPG4-high” (N=101) and “CSPG4-low” (N=118) tumors. A/ Heatmap of methylation sites in which the 450,000 probes were sorted by chromosomal location and the 219 STS samples were sorted by their CSPG4 expression level. B/ Supervised analysis of methylation levels between the “CSPG4-high” vs. “CSPG4-low” tumors. The bar plotted values represent the –log10 corrected p-values of the Student t-test of the 84 significant probes with greater methylation in “CSPG4-high” (red) or in “CSPG4-low” (green).

Additional file 7:

Table S2. (File format .xls). Uni- and multivariate prognostic analyses for DFS including the ICR signature.

Additional file 8:

Table S3. (File format .xls). List of 379 genes differentially expressed between the “CSPG4-high” and “CSPG4-low” tumors.

Additional file 9:

Table S4. (File format .xls). Ontologies associated with the 379 genes differentially expressed between the “CSPG4-high” and “CSPG4-low” tumors.

Additional file 10:

Table S5. (File format .xls). Supervised analysis of CNA between the “CSPG4-high” (N=123) and “CSPG4-low” (N=132) tumors.

Additional file 11:

Table S6. (File format .xls). Supervised analysis of mutation rates between the “CSPG4-high” (N=119) and “CSPG4-low” (N=126) tumors.

Additional file 12:

Table S7. (File format .xls). List of the 84 sites differentially methylated between the “CSPG4-high” (N=101) and “CSPG4-low” (N=118) tumors.

Additional file 13:

Table S8. (File format .xls). Expression data of each tumor normalised and used for the paper.

Rights and permissions

Open Access 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 The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Boudin, L., de Nonneville, A., Finetti, P. et al. CSPG4 expression in soft tissue sarcomas is associated with poor prognosis and low cytotoxic immune response. J Transl Med 20, 464 (2022).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: