Identifies microtubule-binding protein CSPP1 as a novel cancer biomarker associated with ferroptosis and tumor microenvironment.

Centrosome and spindle pole-associated protein (CSPP1) is a centrosome and microtubule-binding protein that plays a role in cell cycle-dependent cytoskeleton organization and cilia formation. Previous studies have suggested that CSPP1 plays a role in tumorigenesis; however, no pan-cancer analysis has been performed. This study systematically investigates the expression of CSPP1 and its potential clinical outcomes associated with diagnosis, prognosis, and therapy. CSPP1 is widely present in tissues and cells and its aberrant expression serves as a diagnostic biomarker for cancer. CSPP1 dysregulation is driven by multi-dimensional mechanisms involving genetic alterations, DNA methylation, and miRNAs. Phosphorylation of CSPP1 at specific sites may play a role in tumorigenesis. In addition, CSPP1 correlates with clinical features and outcomes in multiple cancers. Take brain low-grade gliomas (LGG) with a poor prognosis as an example, functional enrichment analysis implies that CSPP1 may play a role in ferroptosis and tumor microenvironment (TME), including regulating epithelial-mesenchymal transition, stromal response, and immune response. Further analysis confirms that CSPP1 dysregulates ferroptosis in LGG and other cancers, making it possible for ferroptosis-based drugs to be used in the treatment of these cancers. Importantly, CSPP1-associated tumors are infiltrated in different TMEs, rendering immune checkpoint blockade therapy beneficial for these cancer patients. Our study is the first to demonstrate that CSPP1 is a potential diagnostic and prognostic biomarker associated with ferroptosis and TME, providing a new target for drug therapy and immunotherapy in specific cancers.

Introduction

Cancer is a leading cause of global health problems [1]. Although cancer treatment methods have recently improved, clinical outcomes remain unsatisfactory due to side effects and drug resistance issues. Therefore, it is urgent to identify new sensitive biomarkers for the diagnosis and treatment of these cancer patients.

Ferroptosis is a novel iron-dependent programmed cell death, that differs from typical cell death processes, mediated by lethal accumulation of lipid peroxides [2], [3]. It involves a series of metabolic pathways and lipid peroxidation signaling pathways and is characterized by increased lipid peroxidation and reactive oxygen species, smaller mitochondria, and higher mitochondrial membrane density, but the change in nuclear morphology is not obvious [4]. Ferroptosis is essentially an antitumor mechanism that suppresses tumor growth and kills these cells. Cancer cell ferroptosis not only promotes the antitumor response of immune cells but also affects the ferroptosis of immune cells themselves. Therefore, ferroptosis plays an important role in tumor occurrence, progression, and prognosis [5], [6].

Tumor microenvironment (TME) is the surrounding microenvironment for tumor cells, mainly including peripheral blood vessels, stromal cells (cancer-associated fibroblasts (CAFs), endothelial cells, etc.), immune cells, and non-cellular components (cytokines, growth factors, hormones, and the extracellular matrix) [7], [8], [9], [10]. Stromal components typically form a microenvironment conducive to tumor cell growth, including influencing metabolic pathways, inhibiting ferroptosis, inducing epithelial-mesenchymal transition (EMT), and regulating immune cell infiltration. Meanwhile, in the early stage, immune cells are recruited and activated by tumor cells to form an antitumor immune microenvironment and delay tumor development. With the continuous activation by tumor antigens, the relevant effector cells enter the depletion or remodeling stage, resulting in an immunosuppressive microenvironment. Different microenvironmental components interact and regulate each other, and are closely related to tumor progression and prognosis. Therefore, novel targets and biomarkers can be identified by identifying genes that influence ferroptosis and TME, leading to the selection of effective drugs and immunotherapy strategies to improve the prognosis of cancer patients.

Centrosome and spindle pole-associated protein (CSPP1), encoded by chromosome 8q13.2, is initially identified as a highly expressed proto-oncogene in diffuse large B-cell lymphoma (DLBC) [11]. It localizes to the interphase centrosome and mitotic spindle, migrates to the central spindle at the end of mitosis, and concentrates at the midbody during telophase and cytokinesis, thus functioning throughout cell cycle progression. Overexpression or suppression of CSPP1 causes cell-cycle defects [11], [12], [13], [14]. Interestingly, CSPP1 is not only localized to the centrosome and spindle in cycling cells but also interacts with Nephrocystin 8 to extend to the cilia axoneme in postmitotic resting cells, thus playing an important role in ciliogenesis. E3 ubiquitin-protein ligase UBR5-mediated ubiquitylation of CSPP1 is an underlying requirement for cilia localization. Meanwhile, interacting with the centrosomal protein of 104 kDa (CEP104), CSPP1 regulates axoneme length and cilia formation in the Hedgehog signaling pathway [15], [16], [17]. Mutation or loss of function in CSPP1 causes primary cilia abnormalities and ciliopathy, including Joubert syndrome and Meckel-Gruber syndrome [18], [19], [20], [21], [22]. Beyond cell cycle control and ciliogenesis, CSPP1 displays microtubule-independent but desmoplakin-dependent desmosome localization in apical-basal polarized epithelial cells and it is necessary for normal spheroid formation [23]. More importantly, CSPP1 has also been identified as a candidate oncogene in luminal breast cancer; meanwhile, nuclear CSPP1 expression can define subtypes and clinical subgroups of basal-like breast carcinoma [24]. In addition to DLBC, CSPP1 is also identified as a putative hallmark gene associated with the malignancy of oral squamous cell carcinoma [25]. However, no comprehensive analyses of the expression, function, and clinical significance of CSPP1 as well as its correlation with ferroptosis and TME components have been performed.

In this study, we systematically analyzed CSPP1 expression and found that its aberrant expression is driven by genetic alterations, DNA methylation, and miRNAs. Phosphorylation of CSPP1 protein may regulate its activity, especially at Ser424. In addition, CSPP1 strongly correlated with ferroptosis and TME components, potentially serving as a diagnostic and prognostic biomarker. Patients with CSPP1-associated tumors may benefit from ferroptosis-based drug therapy and immunotherapy by modulating ferroptosis and TME in several types of cancer.

Material and methods

Data collection

Transcriptomic data containing mRNA, miRNA, and associated clinical information were retrieved from The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/repository) and UCSC XENA (https://xenabrowser.net) databases [26]. mRNA sequencing data in level 3 HTSeq-FPKM format was converted into TPM format, and miRNA sequencing data in level 3 BCGSC format was converted into RPM format; all data were downloaded from TCGA. UCSC XENA database contained both TCGA and Genotype-Tissue Expression (GTEx) data, which were processed by the Toil process into TPM format [27]. The Oncomine (https://www.oncomine.org/resource/login.html) image for CSPP1 expression was downloaded (threshold: p < 0.0001; fold change >2; gene rank: top 10%). R software (Version 3.6.3, https://cran.r-project.org/bin/windows/base/old/3.6.3/) and ggplot2 R package (Version 3.3.3, https://cran.r-project.org/web/packages/ggplot2/index.html) were used to statistical analyses and visualization, respectively.

Differential expression analysis

RNA sequencing data in TPM or FPKM format for CSPP1 expression in tissues and single cells were recorded from the Human Protein Atlas portal (HPA, https://www.proteinatlas.org/) and visualized by radar plots. To compare CSPP1 and miRNA expression in normal and tumor tissues, an unpaired Wilcoxon rank-sum test was performed using TCGA and GTEx datasets. Histograms of CSPP1 protein expression and site phosphorylation levels in normal and primary tumor tissues were downloaded from the UALCAN portal (https://ualcan.path.uab.edu/analysis-prot.html) using Clinical Proteomic Tumor Analysis Consortium (CPTAC) data [28].

Diagnostic analysis

The diagnostic value of CSPP1 was estimated using RNA-sequencing data from TCGA and visualized by a receiver operating characteristics (ROC) curve using the pROC R package (Version 1.17.0.1, https://cran.r-project.org/web/packages/pROC/index.html).

Genetic alterations and DNA methylation analysis

Histograms of CSPP1 mutation and copy number alteration (CNA) frequency in pan-cancer (TCGA, PanCancer Atlas) were downloaded from the cBioPortal (https://www.cbioportal.org/). Histograms of CSPP1 promoter methylation in normal and primary tumor tissues were downloaded from the UALCAN portal (https://ualcan.path.uab.edu/index.html) [29]. Correlations between CSPP1, CNAs, and DNA methylation (TCGA, Firehose) were recorded from cBioPortal and visualized by heatmaps. Dot size together with transition color represented the degree of correlation. The larger the dot, the stronger the correlation. Red and blue dots represented positive and negative correlations, respectively. Kaplan-Meier (KM) plots of these alterations on survival probability, including that of overall survival (OS), disease-specific survival (DSS), and progression-free survival (PFS), were downloaded from cBioPortal.

Correlations between CSPP1 and associated miRNAs

Spearman correlations between CSPP1 and associated miRNAs were recorded from the Encyclopedia of RNA Interactomes portal (ENCORI, https://rna.sysu.edu.cn/encori/index.php) (parameter setting: assembly, hg38; miRNA: all; CLIP-Data ≥ 3; pan-Cancer ≥ 1; programNum ≥ 2; target, CSPP1) and visualized by a heatmap [30].

Correlations between CSPP1 and clinical features

Correlations between CSPP1, pathologic stage, and histologic grade from TCGA database were analyzed using the Kruskal-Wallis test and visualized by violin plots. Correlations between CSPP1 and molecular subtype were also analyzed with the Kruskal-Wallis test and the violin plots were downloaded from Tumor-Immune System Interactions DataBase (TISIDB, https://cis.hku.hk/TISIDB/index.php) [31].

Correlations between CSPP1 and clinical features in brain low-grade gliomas (LGG) were analyzed using the Chi-squared test or Fisher’s exact test and visualized by a baseline datasheet.

Prognostic analysis of CSPP1 and associated miRNAs, model construction, and evaluation

Survival differences analyses of CSPP1 and associated miRNAs, including OS, DSS, and progression-free interval (PFI), were visualized by forest plots based on KM analyses. The Survivin R package (Version 3.2-10, https://cran.r-project.org/web/packages/survivalAnalysis/index.html) was used for statistical analysis, and the survminer R package (Version 0.4.9, https://cran.r-project.org/web/packages/survminer/index.html) was used for visualization.

Univariate and multivariate Cox regression analyses were visualized by forest plots. Based on multivariate Cox regression, risk score plots were constructed using the ggrisk R package (Version 1.3, https://cran.r-project.org/web/packages/ggrisk/index.html). Nomograms were also designed using the rms R package (Version 6.2-0, https://cran.r-project.org/web/packages/rms/index.html) and survival R package [26]. Calibration curves and the concordance index (C-index) were evaluated by comparing predicted probabilities with observed events.

Functional enrichment analysis in LGG

CSPP1-associated differentially expressed genes (DEGs) in LGG were identified using the limma R package (Version 3.40.2, https://bioconductor.org/packages/release/bioc/html/limma.html) and visualized by a volcano plot [32]. Spearman correlations between CSPP1 and the top 20 DEGs were assessed and visualized by a heatmap.

DEGs were used for Gene Ontology (GO) enrichment analyses, including cellular components (CCs), molecular functions (MFs), and biological pathways (BPs). Gene Set Enrichment Analysis (GSEA) was also conducted to detect phenotypes and signaling pathways. Hallmark v7.2, GO c5 v7.2 (BPs, CCs, MFs), and Kyoto Encyclopedia of Genes and Genomes (KEGG) c2 v7.2 gene sets were used. Statistical analysis and graphical charting were performed using the clusterProfiler R package (Version 3.14.3, https://bioconductor.org/packages/release/bioc/html/clusterProfiler.html) [32], [33].

Gene mutation and ferroptosis correlation analysis

Somatic mutations in LGG from TCGA database were analyzed using the maftools R package (Version 3.14, https://bioconductor.org/packages/release/bioc/html/maftools.html) and visualized by an oncoplot [34], [35]. Ferroptosis-associated score was calculated with the gene set extracted from KEGG with the ssGSEA algorithm in the gene set variation analysis (GSVA) package (Version 1.34.0, https://bioconductor.riken.jp/packages/3.0/bioc/html/GSVA.html), and the difference between the driver score minus suppressor score was defined as the ferroptosis score to represent the ferroptosis status of samples [36]. Spearman correlations between CSPP1 and ferroptosis-related genes and ferroptosis scores were analyzed and visualized by heatmaps.

TME analysis and immune checkpoint blockade therapy prediction

TCGA datasets were used to analyze the Spearman correlations between CSPP1 and immune cells using the ssGSEA algorithm in the GSVA package. They were also used to estimate the stromal score, immune score, and ESTIMATE score using the ESTIMATE package (Version 1.0.13, https://bioinformatics.mdanderson.org/estimate/index.html) [37], [38]. Purity-adjusted Spearman correlations between CSPP1 and CAFs, endothelial cells, and immune checkpoints were recorded from the Tumor Immune Estimation Resource 2 portal (TIMER2, https://timer.cistrome.org) with XCELL or TIMER algorithm [39], [40]. The Spearman correlations between CSPP1 and major histocompatibility complex (MHC) molecules, immune stimulator genes, immune inhibitor genes, tumor mutation burden (TMB) score, and microsatellite instability (MSI) score from the TCGA database were analyzed [41], [42]. All corresponding correlations were visualized by heatmaps.

Tumor Immune Dysfunction and Exclusion (TIDE, https://tide.dfci.harvard.edu/) is a comprehensive score for tumor immune dysfunction and immune escape, including tumor-infiltrating cytotoxic T lymphocyte (CTL) dysfunction and rejection by immune checkpoints. RNA-sequencing raw count data and corresponding clinical information from TCGA database were estimated using the TIDE algorithm to predict the potential immune checkpoint blockade (ICB) response. A low score indicated good efficacy [43], [44].

Results

Aberrant expression of CSPP1 serves as a diagnostic biomarker among cancers

CSPP1 was widely present in all the tested tissues and cells. It was highly expressed in skeletal muscle, testis, and fallopian tube, as well as in respiratory ciliated cells, endometrial ciliated cells, and early spermatids; meanwhile, high expression was observed in testicular germ cell tumors (TGCT), stomach adenocarcinoma (STAD), and breast invasive carcinoma (BRCA) (Fig. 1A–C).

Fig. 1

Aberrant expression of CSPP1 serves as a diagnostic biomarker among cancers. (A) Radar Plot of CSPP1 expression in normal tissues basedon GTEx datasets from HPA portal. (B) Radar Plot of CSPP1 expression in single cells based on single-cell types dataset from HPA portal. (C) Radar Plot of CSPP1 expression in tumor tissues based on TCGA dataset from HPA portal. (D, E) Histogram of CSPP1 expression in 33 types of unpaired normal and tumor tissues from TCGA and TCGA plus GTEx database using Wilcoxon rank-sum test. ns: p ≥ 0.05, * p < 0.05, ** p < 0.01, *** p < 0.001. (F) Heatmap of CSPP1 expression from Oncomine portal. (G, H) ROC analyses of differential CSPP1 expression in 27 types of upregulated (G) and downregulated (H) cancer from TCGA and GTEx databases. AUC > 0.9 was considered a high diagnostic value, 0.9 ≥ AUC > 0.7 was median, and 0.7 ≥ AUC > 0.5 was low.

To compare CSPP1 expression in human adjacent normal versus 33 types of tumor tissues, TCGA datasets were used. CSPP1 was significantly upregulated in ten cancer types and downregulated in five from TCGA (Fig. 1D). In order to expand the sample size, we also introduced normal samples from the GTEx database. CSPP1 expression was increased in BRCA, cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), DLBC, esophageal carcinoma (ESCA), glioblastoma multiforme (GBM), head and neck squamous cell carcinoma (HNSC), acute myeloid leukemia (LAML), LGG, liver hepatocellular carcinoma (LIHC), pancreatic adenocarcinoma (PAAD), rectum adenocarcinoma (READ), STAD, and thymoma (THYM). In contrast, its expression was decreased in adrenocortical carcinoma (ACC), cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), kidney chromophobe (KICH), renal clear cell carcinoma (KIRC), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), ovarian serous cystadenocarcinoma (OV), prostate cancer (PRAD), skin cutaneous melanoma (SKCM), TGCT, thyroid cancer (THCA), endometrial cancer uterine corpus endometrial carcinoma (UCEC), and uterine carcinosarcoma (UCS) (Fig. 1E). Thereafter, we used the Oncomine datasets to validate CSPP1 expression patterns. Significantly elevated CSPP1 expression was observed in most cancer types; however, its expression was reduced in bladder cancer and kidney cancer (Fig. 1F). Combined with these databases, 27 cancers with differential CSPP1 expression from GTEx plus TCGA database were used in subsequent analyses.

ROC analyses with TCGA or TCGA plus GTEx datasets indicated that the diagnostic values of CSPP1 were median or high in CHOL, DLBC, ESCA, HNSC, LAML, PAAD, STAD, THYM, ACC, CESC, KICH, LUAD, LUSC, OV, TGCT, THCA, and UCS (Fig. 1G, H). Our results indicated that aberrant CSPP1 expression serves as a diagnostic biomarker among cancers.

Multi-dimensional mechanisms involving genetic alterations, DNA methylation, and miRNAs underly CSPP1 dysregulation

To determine the cause underlying CSPP1 dysregulation, we comprehensively analyzed the factors related to the expression of CSPP1, including genetic variation, DNA methylation, and associated miRNAs. We first used cBioPortal to study the genetic variation and found that among these 26 cancer types (COAD and READ were combined into COADREAD in the Portal), 21 contained mutations and 21 had CNAs (Fig. 2A). Except for UCEC (>5%), CSPP1 mutation frequencies were relatively low in most cancers (<5%). These mutations caused poor PFS in SKCM and good PFS in UCEC (Fig. S1A). CNA is the genetic variation most closely associated with CSPP1 expression [45]. It occurred more frequently in UCS, PRAD, LIHC, BRCA, and OV (>5%). CSPP1 positively correlated with CNAs (Fig. 2B; Table S1) and had worse DSS and PFS in COADREAD, as well as worse OS, DSS, and PFS in PAAD, PRAD, STAD, and UCEC (Fig. S1B).

Fig. 2

Multi-dimensional mechanisms involving genetic alterations, DNA methylation, and miRNAs underly CSPP1 dysregulation. (A) Histogram of genetic alteration frequency of CSPP1 from cBioPortal portal. (B) A heatmap of correlations between CSPP1 and CNAs. Dot size together with transition color represented the degree of correlation. The larger the dot, the stronger the correlation. Red and blue represented positive and negative correlations, respectively. * p < 0.05, ** p < 0.01. (C) Histograms of CSPP1 promoter methylation in normal and primary tumors with significant differences from UALCAN portal. 0.7 ≥ Beta value > 0.5 was considered hyper-methylation, 0.3 ≥ Beta value > 0.25 was hypo-methylation. (D) A heatmap of correlations between CSPP1 and DNA methylation from cBioPortal portal. (E) A heatmap of correlation between CSPP1 and predicted miRNAs from ENCORI portal. Red and blue words indicated upregulated and downregulated cancers, respectively. (F, G) Differential expression of negatively associated miRNAs from TCGA database. Red stars represent negatively correlated miRNAs of CSPP1. (H) A forest plot of the correlations between CSPP1-negatively associated miRNAs expression and survival probability, including OS, DSS, and PFI. * p < 0.05, ** p < 0.01, *** p < 0.001. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Besides CNAs, DNA methylation also affects gene expression [46]. Promoter methylation was reduced with upregulated CSPP1 expression in BRCA, HNSC, and READ, while it was increased with downregulated CSPP1 expression in KIRC and LUSC from UALCAN portal (Fig. 2C; Fig. S2). CSPP1 expression negatively correlated with DNA methylation from cBioPortal (Fig. 2D and Table S2).

In addition to CNAs and DNA methylation, miRNAs also play important roles in regulating mRNA expression [47]. The ENCORI portal was used to search for miRNAs negatively associated with CSPP1. These miRNAs were found in most cancers except for ACC (Fig. 2E; Table S3). We further conducted differential expression analyses of these miRNAs across 14 cancers with data using TCGA datasets. CSPP1 upregulation was associated with the downregulation of miR-221-3p and miR-377-3p in BRCA; miR-145-5p and miR-125b-5p in STAD (Fig. 2F). Meanwhile, CSPP1 downregulation may be affected by upregulation of miR-222-3p in KICH; miR-425-5p, miR-221-3p, miR-222-3p, miR-340-5p, miR-150-5p, miR-708-5p in KIRC; miR-135b-5p, miR-222-3p, miR-27a-3p, miR-708-5p in LUAD; miR-135b-5p in LUSC; miR-135a-5p in PRAD; miR-105-5p, miR-221-3p and miR-222-3p in THCA; miR-135b-5p and miR-27a-3p in UCEC (Fig. 2G). A prognosis-related forest plot indicated that in CSPP1-overexpressed cancers, downregulated miR-221-3p in BRCA caused favorable DSS and PFI; meanwhile, downregulated miR-145-5p and miR-125-5p in STAD had favorable OS, DSS, and PFI. Among cancers with low CSPP1 expression, upregulated miR-425-5p, miR-221-3p, miR-222-3p, and miR-708-5p in KIRC caused poor OS, DSS, and PFI; whereas, upregulated miR-27a-3p in LUAD had poor OS and DSS (Fig. 2H; Fig. S3). Together, CSPP1 dysregulation involves multi-dimensional mechanisms, including genetic alterations, DNA methylation, and miRNAs.

Phosphorylation of CSPP1 at specific sites may play a role in tumorigenesis, especially at Ser424

With datasets available in BRCA, GBM, HNSC, KIRC, LIHC, LUAD, OV, PAAD, and UCEC, CSPP1 protein expression was significantly increased in LIHC, PAAD, and UCEC compared with normal tissues; meanwhile, it was decreased in BRCA and HNSC (Fig. 3A; Fig. S4A).

Fig. 3

Phosphorylation of CSPP1 at specific sites may play a role in tumorigenesis, especially at Ser424. (A) Histograms of CSPP1 expression in nine types of the normal and primary tumors with significant differences using CPTAC samples from UALCAN portal. (B) The schematic diagram of CSPP1 phosphorylation sites. Red and blue words indicated high and low protein expression, respectively. (C) Histograms of the phosphorylation site of CSPP1 in normal and primary tumors with significant differences. p < 0.05 was considered statistically significant. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Post-translational modification (PTM) is a key molecular mechanism associated with the activity of the protein [48]. A higher S31 phosphorylation level was observed in BRCA, GBM, LIHC, and LUAD; meanwhile, a lower level was observed in KIRC and PAAD. S424 phosphorylation was increased in BRCA, HNSC, KIRC, and LIHC. S847 phosphorylation was increased in HNSC but decreased in PAAD. Increased S866 phosphorylation was observed in HNSC and LIHC but decreased in KIRC, LUAD, and PAAD. S885 phosphorylation was increased in LIHC and decreased in KIRC. S931 phosphorylation was increased in LIHC and decreased in KIRC and LUAD (Fig. 3B, C; Fig. S4B). No threonine or tyrosine phosphorylation was identified in these cancer types with UALCAN database. Together, these findings suggested that phosphorylation of CSPP1 at specific sites may play a role in tumorigenesis, especially at Ser424.

CSPP1 correlates with clinical features and outcomes in multiple cancers

Thereafter, we investigated CSPP1 expression at different pathologic stages, histologic grades, and molecular subtypes. CSPP1 overexpression significantly correlated with advanced pathologic stage in READ, ACC, and KICH, and advanced histologic grade in HNSC and LIHC; however, high CSPP1 expression correlated with low histologic grade in KIRC (Fig. 4A-D). In addition, CSPP1 expression significantly differed with respect to molecular subtypes in BRCA, COAD, ESCA, LGG, LIHC, READ, LUSC, and UCEC (Fig. 4E, F). However, no association was observed in other cancers (Fig. S5).

Fig. 4

CSPP1 correlates with clinical features and outcomes in multiple cancers. (A-D) Violin plots of correlation between CSPP1 expression and pathologic stage (A, B) and histologic grade (C, D) from TCGA database with significant differences using Kruskal-Wallis test. (E, F) Violin plots of correlation between CSPP1 expression and molecular subtype with significant differences from TISIDB portal. p < 0.05 was considered as a statistical difference between the two groups. (G) A forest plot of the correlations between CSPP1 expression and survival probability, including OS, DSS, and PFI. (H) A forest plot of univariate and multivariate Cox regression analyses with OS in LGG from TCGA database. p represented the overall difference. * p < 0.05, ** p < 0.01, *** p < 0.001. (I) Identification of CSPP1 as an independent risk factor in LGG. The upper portion scatters plot was survival time and survival status according to CSPP1 expression, and the middle portion scatters plot was risk score. (J) Construction of a prognostic nomogram in LGG. (K) Nomogram calibration analysis with prognostic data in LGG. C-index > 0.9 indicated highly accuracy, 0.9 ≥ C-index > 0.7 was median, and 0.7 ≥ C-index > 0.5 was low.

To monitor the clinical outcomes of CSPP1 differential expression, a Cox regression analysis was performed with respect to patients’ prognoses. Results indicated that in CSPP1-upregulated cancers, CSPP1 overexpression was associated with poor OS, DSS, and PFI in LGG and LIHC. In CSPP1-downregulated cancers, decreased CSPP1 expression was associated with favorable OS, DSS, and PFI in ACC (Fig. 4G; Fig. S6). Overall, excessive CSPP1 expression is unfavorable in several cancer types, especially LGG, LIHC, and ACC.

To further evaluate whether CSPP1 was an independent risk factor for prognosis, we used LGG as an example (p < 0.001 for OS, DSS, and PFI). The baseline datasheet showed that CSPP1 was significantly correlated with the WHO grade, IDH status, 1p/19q co-deletion, and histological type (Table S4). Univariate Cox regression analyses further indicated that CSPP1 correlated with poor prognosis. Furthermore, multivariate Cox regression analyses confirmed that CSPP1 overexpressed as an independent factor associated with OS, DSS, and PFI (Fig. 4H, I; Fig. S7A, B, E, F). Based on the multivariate Cox regression analyses, nomogram prediction models were established (Fig. 4J; Fig. S7C, G). We performed calibration analysis on the nomograms to verify the validity of the predictive models. The C-indexes of OS, DSS, and PFI indicate median accuracy (Fig. 4K; Fig. S7D, H). These results confirmed CSPP1 as an independent risk factor for LGG survival.

Functional enrichment indicates CSPP1 is potentially associated with ferroptosis and TME in LGG

Based on its unfavorable prognosis, LGG patients were divided into high and low CSPP1 expression groups, and mRNA expression patterns were compared. A total of 14 upregulated and 67 downregulated genes were identified (Fig. 5A). Correlations between CSPP1 and the top 20 DEGs were visualized by a heatmap (Fig. 5B; Table S5).

Fig. 5

Functional enrichment indicates that CSPP1 is potentially associated with ferroptosis and TME in LGG. (A) A volcano plot of CSPP1-related DEGs in LGG. Red and blue points indicated upregulated and downregulated genes, respectively. (B) A heatmap of correlation between CSPP1 and the top 20 DEGs. *** p < 0.001. (C) Bubble plots of GO enrichment. The X-axis represents the ratio of these DEGs, and the Y-axis represents the categories of DEGs. (D) Ridge plots of GSEA enrichment. p < 0.05 was considered the meaningful pathway. Red and blue indicated immune-related pathways and ferroptosis-related metabolic pathways, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

GO and GSEA analyses were performed to investigate the functional mechanisms of CSPP1. CSPP1-related DEGs were enriched in (i) CCs: presynapse, transport vesicle, and synaptic membrane; (ii) MFs: passive transmembrane transporter activity, channel activity, and substrate-specific channel activity; and (iii) BPs: signal release, vesicle-mediated transport in synapse, and regulation of transporter activity (Fig. 5C; Table S6). GSEA was also used to identify CSPP1-associated pathways. The results suggested that, in hallmark gene sets, CSPP1-related DEGs were positively related to cell cycle-related pathways (E2F targets, G2/M checkpoint, and mitotic spindle), cancer-related pathways (Notch signaling and TGF-β signaling), EMT, and inflammatory response; meanwhile, they were negatively related to KRAS signaling DN, ferroptosis-related metabolic pathways (fatty acid metabolism, cholesterol homeostasis, and oxidative phosphorylation). For GO and KEGG gene sets, CSPP1-related DEGs were positively associated with cell cycle-related pathways (cell cycle checkpoint, chromosome segregation, and microtubule cytoskeleton organization involved in mitosis), cancer-related pathways (Notch signaling pathway, TGF-β signaling pathway, and pathways in cancer), stromal-related pathway (ECM structural constituent, extracellular structure organization, ECM receptor interaction, and focal adhesion), and immune-related pathways (B cell-mediated immunity, adaptive immune response, positive regulation of T cell proliferation, T cell activation involved in immune response, complement and coagulation cascades, intestinal immune network for IgA production, cytosolic DNA sensing pathway, and toll-like receptor signaling pathway); meanwhile, they were negatively associated with ferroptosis-related metabolic pathways (steroid metabolic process, steroid biosynthetic process, response to metal ion, terpenoid backbone biosynthesis, and oxidative phosphorylation) (Fig. 5D; Table S7). These findings implied that CSPP1 may be involved in ferroptosis and TME.

CSPP1 dysregulates ferroptosis in LGG and other cancer types

TP53 is the most commonly mutated gene associated with cancers and its mutations have been reported to be closely associated with ferroptosis [49], [50], [51], [52], [53], [54], [55]. To verify the correlation between TP53 mutation and CSPP1 expression, somatic mutation analysis was performed according to CSPP1 expression in LGG. From the oncoplot, higher frequencies of TP53 and ATRX mutations and lower frequencies of CIC, FUBP1, NOTCH1, IDH2, and ZBTB20 mutations were observed in the high CSPP1 expression group (Fig. 6A, B). However, no association between ATRX, CIC, FUBP1, NOTCH1, IDH2, and ZBTB20 mutations and ferroptosis has been reported.

Fig. 6

CSPP1 dysregulates ferroptosis in LGG and other cancer types. (A) Oncoplot of somatic mutant landscape in high and low CSPP1 expression groups in LGG. * p < 0.05, ** p < 0.01, *** p < 0.001. (B) Histograms of gene mutants comparison in high and low CSPP1 expression groups by chisq.test with significant differences. (C, D) Heatmaps of correlation between CSPP1 and FAGs and ferroptosis-associated scores. Dot size together with transition color represented the degree of correlation. The larger the dot, the stronger the correlation. Red and blue dots represented positive and negative correlations, respectively. * p < 0.05, ** p < 0.01. (E) Histograms of ferroptosis scores between high and low CSPP1 expression groups from TCGA database. p < 0.05 was considered statistically significant. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

To further confirm the correlation between CSPP1 and ferroptosis, 30 ferroptosis-associated genes (FAGs) were extracted from KEGG, including 18 ferroptosis-driver genes (FDGs) of ACSL1, ACSL4, ACSL6, ALOX15, ATG5, ATG7, FTL, LPCAT3, MAP1LC3A, MAP1LC3B, NCOA4, SAT1, SLC39A14, TF, TFRC, TP53, VDAC2, VDAC3 and 12 ferroptosis-suppressor genes (FSGs) of ACSL3, FTH1, FTMT, GCLC, GCLM, GPX4, GSS, HMOX1, PCBP1, SLC3A2, SLC7A11, SLC40A1 [56], [57], [58], [59]. We found that CSPP1 positively correlated with most FAGs, but negatively correlated with two FDGs of FTL, MAP1LC3A, and three FSGs of FTH1, GPX4, and HMOX1 (Fig. 6C; Table S8).

Next, the overall scores of driver genes and suppressor genes were calculated by the ssGSEA algorithm, and the ferroptosis score obtained from driver score minus suppressor score was used to evaluate whether the function of CSPP1 was to activated or inhibited ferroptosis in cancers. The gene set and its corresponding algorithm have been proved to be able to predict ferroptosis status [56], [57], [58], [59]. In CSPP1-upregulated cancers, CSPP1 positively correlated with ferroptosis score in LAML; meanwhile, negative correlations were observed in BRCA, GBM, LGG, LIHC, and THYM. In CSPP1-downregulated cancers, positive correlations were exhibited in ACC and LUAD, while negative correlations existed in KIRC, OV, and PRAD (Fig. 6D; Table S8). These tumor samples were further divided into high and low CSPP1 expression groups, and the ferroptosis score was further compared between the two groups. From the histogram, as an antitumor mechanism, ferroptosis was overall inhibited in pan-cancer (ferroptosis score < 1), except for LAML. Lower scores represented lower ferroptosis levels in BRCA, COAD, GBM, LGG, LIHC, THYM, OV, and PRAD and higher scores represented higher ferroptosis levels in LAML and ACC were observed in the high CSPP1 expression group (Fig. 6E; Table S8). Thus, our findings revealed that CSPP1 dysregulates ferroptosis in LGG and other cancer types.

CSPP1-associated tumors are infiltrated in different TMEs, improving ICB therapeutic efficacy in specific cancers

Functional enrichment analysis also implied that CSPP1 may regulate the TME by influencing the immune response and stromal response. Therefore, we performed a pan-cancer analysis of the correlation between CSPP1 and these two components. We first focused on CSPP1 and 24 types of immune cells using the ssGSEA algorithm. The results showed that CSPP1 negatively correlated with most immune cells across cancers but positively correlated with T helper cells, central memory T (Tcm) cells, and T helper 2 (Th2) cells. Of note, Tcm cells are also immunosuppressive cells. Next, stromal cell infiltration was assessed using the XCELL algorithm from the TIMER2 portal, mainly including CAF cells and endothelial cells [60]. There was a positive correlation between CSPP1 and CAFs in HNSC, LGG, LIHC, THYM, KICH, SKCM, and THCA, whereas an inverse correlation was observed in STAD, KIRC, LUSC, and TGCT. Moreover, CSPP1 was negatively correlated with endothelial cells in BRCA, DLBC, LGG, LIHC, STAD, THYM, PRAD, TGCT, and UCEC, whereas a positive correlation was noted in OV. Thereafter, we comprehensively calculated the TME score using the ESTIMATE package. CSPP1 was negatively associated with the stromal score, immune score, and ESTIMATE score in most cancers, while positively associated with these scores in LGG (Fig. 7A; Table S9). To sum up, CSPP1 comprehensively regulates the TME from both immune cell infiltration and stromal cell infiltration.

Fig. 7

CSPP1-associated tumors are infiltrated in different TMEs, improving ICB therapeutic efficacy in specific cancers. (A) A heatmap of correlations between CSPP1 and 24 types of immune cells from TCGA database, CAFs and endothelial cells from TIMER2 portal using XCELL algorithm, and TME scores from TCGA database using the ssGSEA algorithm. (B) A heatmap of correlations between CSPP1 and immunomodulators, including MHC molecules, immune stimulator genes, and immune inhibitor genes from TCGA database. (C) A heatmap of correlations between CSPP1 and immune checkpoints from the TIMER2 portal. (D) A heatmap of correlations between CSPP1 and TMB score, MSI score from TCGA database. Dot size together with transition color represented the degree of correlation. The larger the dot, the stronger the correlation. Red and blue dots represented positive and negative correlation, respectively. * p < 0.05, ** p < 0.01. (E, F) Histograms of CSPP1-associated ICB therapeutic effect between high and low CSPP1 expression groups from TCGA database by TIDE algorithm with a significant difference. A low score indicated good efficacy. p < 0.05 was considered statistically significant. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

To further study the regulatory mechanism of CSPP1-related tumor infiltration, correlations between CSPP1 and three types of immunomodulators were investigated with TCGA datasets [61]. CSPP1 negatively correlated with MHCs and positively with immune stimulators and immune inhibitors in most cancers (Fig. 7B; Table S9). Among immune inhibitors, CD274 (PD-L1), CTLA4, HAVCR2, LAG3, PDCD1 (PD1), PDCD1LG2 (PD-L2), TIGIT, and SIGLEC15 are known immune checkpoints responsible for tumor immune escape. Combined with these immune checkpoint results and subsequent analysis using the TIMER2 portal, it was further confirmed that CSPP1 was positively correlated with immune checkpoints in BRCA, DLBC, ESCA, HNSC, LGG, LIHC, PAAD, READ, KICH, and KIRC, and negatively correlated with them in COAD, LAML, CESC, THCA, and UCEC (Fig. 7B, C; Table S9).

TMB and MSI are two emerging biomarkers associated with immunotherapy response. Tumor cells with high TMB or MSI scores have strong antigenicity and more neoantigens, thus promoting immune cell infiltration. Results showed that CSPP1 positively correlated with TMB in LGG, STAD, and PRAD, but inversely correlated with it in COAD, LIHC, THCA, and UCEC. The correlation between CSPP1 and MSI was then investigated. LGG, READ, STAD, LUAD, and LUSC showed positive correlations, whereas DLBC presented a negative correlation (Fig. 7D; Table S9).

Tumor immunotherapy is a treatment that controls and eliminates tumors by reactivating and maintaining the tumor-immune cycle and restoring the normal antitumor immune response, including ICB and cell therapy. The effectiveness of ICB therapy depends not only on immune cell infiltration but also on immune checkpoints, TMB, and MSI. The close correlations between CSPP1 and immune checkpoints, TMB, and MSI implied that these CSPP1-associated tumor patients may respond well to immunotherapy. Therefore, the TIDE algorithm was used to predict the therapeutic effect of ICB from TCGA database. Results revealed that in CSPP1-upregulated cancers, the high CSPP1 expression group exhibited a lower TIDE score, including BRCA, DLBC, LGG, and STAD; meanwhile, the low CSPP1 expression group exhibited a lower TIDE score in LIHC (Fig. 7E; Fig. S8A). In CSPP1-downregulated cancers, the high CSPP1 expression group exhibited a lower TIDE score, including CESC, KIRC, LUSC, PRAD, SKCM, TGCT, THCA, and UCES (Fig. 7F; Fig. S8B). It is suggested that these patients with low TIDE scores may benefit from ICB therapy.

Discussion

Cancer is the leading cause of morbidity and mortality worldwide. CSPP1 is a centrosome and microtubule-binding protein that plays a role in cell cycle-dependent cytoskeleton organization and cilia formation. Although there is increasing evidence that CSPP1 may play a role in tumorigenesis, its specific role across different cancers remains unclear. This study systematically analyzed CSPP1 expression and demonstrated that its aberrant expression in 27 cancer types is driven by multi-dimensional mechanisms. CSPP1 correlates with clinical features and serves as a potential diagnostic and prognostic biomarker as well as the target for ferroptosis-based drug therapy and immunotherapy.

To explore how CSPP1 influences the progress and prognosis of cancer, its effects on ferroptosis and TME were studied. Function enrichment demonstrated that CSPP1 was involved in ferroptosis-related metabolic pathways. Mutation analyses further indicated that CSPP1 was closely associated with TP53 mutation, which has been reported to be associated with cancer and ferroptosis, thus speculating that CSPP1 might correlate with ferroptosis. At present, studies on ferroptosis-associated gene mutations are mainly limited to the oncogenes and tumor suppressor genes, including oncogenes of PIK3CA, KRAS, NEDD4, VDAC2/3, DJ1, PDK4, and tumor suppressor genes of TP53, BAP1, KEAP1, ARF. Activating mutations in oncogenes and inactivating mutations in tumor suppressor genes regulate the expression of FAGs, and generally tend to inhibit ferroptosis and promote tumor progression. However, mutations in tumor suppressors of the E-cadherin-NF2-Hippo axis, VHL, and oncogenes of EGFR and IDH1 render cancer cells vulnerable to ferroptosis in one or more cancer types [62], [63]. Our correlation analyses between CSPP1 and FAGs and ferroptosis-associated scores further confirmed that CSPP1 was indeed involved in the regulation of ferroptosis in pan-cancer. Combined with the prognostic data, we believed that in CSPP1-upregulated cancers, CSPP1 overexpression inhibited ferroptosis, including BRCA, GBM, LGG, LIHC, and THYM, thus promoting tumor cell growth and leading to poor prognoses in BRCA, LGG, and LIHC. Meanwhile, its overexpression promoted ferroptosis in LAML and led to cancer with a good prognosis. These results were consistent with the progression and prognosis results of CSPP1. However, there were some exceptions. For example, patients with lower expression of CSPP1 showed lower ferroptosis level but better prognosis in ACC, and higher ferroptosis levels but still led to cancer in OV and PRAD. Thus, there should be other mechanisms influencing the progression and prognosis of these cancers. For CSPP1-associated tumors with suppressed ferroptosis, drug-induced ferroptosis through the CSPP1-FDGs or CSPP1-FSGs axis may inhibit tumor progression and thus improve prognosis. Therefore, the possibility of drug therapy for CSPP1-associated tumor patients by regulating ferroptosis is proposed.

In addition to regulating ferroptosis, CSPP1 may have a regulatory role in the TME by affecting EMT, stromal-related pathways, and immune-related pathways. In recent years, an increasing number of studies have linked microtubule-associated genes to immune infiltration. High expression of Targeting Protein for Xenopus kinesin-like protein 2 (TPX2) and Tubulin alpha-1C chain (TUBA1C) increases immune cell infiltration in LIHC, LUAD, and LGG, respectively, and is associated with poor prognosis [64], [65], [66]. A high level of microtubule-associated protein Tau is inversely correlated with the vascular and immune contents, delaying tumor growth in gliomas [67]. Increased expression of microtubule interacting and trafficking domain containing 1 (MITD1) indicates a poor prognosis with decreased NK cell infiltration in LIHC and increased CD8+ T cells infiltration in KIRC [68], [69]. Spindle and kinetochore-associated protein (Ska) complex negatively and positively correlate with immune cell infiltration in BRCA and LIHC, respectively [70], [71]. In our study, CSPP1 comprehensively regulated the TME from both immune cell infiltration and stromal cell infiltration. CSPP1 was negatively correlated with immune scores, stromal scores, and TME scores for most cancers. Moreover, it was also negatively correlated with MHCs and positively associated with immune stimulators and immune inhibitors, including immune checkpoints. CSPP1 expression also significantly correlated with TMB and MSI in specific cancers. ICB therapy prediction confirmed that these cancer patients could benefit from ICB therapy, thus promoting a favorable prognosis. Specifically, in CSPP1-upregulated cancers, low TME scores and high levels of immune checkpoints expression indicated immune infiltration was greatly suppressed, leading to tumor growth and poor prognosis. Therefore, ICB therapy promoting immune infiltration is effective for patients in the high CSPP1 expression group of BRCA, DLBC, LGG, and STAD, but more effective for patients in the low CSPP1 expression group of LIHC. In CSPP1-downregulated cancer, low CSPP1 expression had higher levels of immune cell infiltration and lower levels of immune checkpoints expression, thus these cancers themselves were in a favorable prognostic immune microenvironment, and ICB therapy may be more effective for the early treatment of these patients, including CESC, KIRC, LUSC, PRAD, SKCM, TGCT, THCA, and UCEC.

However, several limitations still remain. At present, our study on the regulation of CSPP1 on ferroptosis and TME, as well as the subsequent potential drug treatment and ICB therapy are limited to bioinformatics, which provides a reference to basic experiments, but basic experiments are still necessary for follow-up research. In addition, the relatively small sample size is also one of the main reasons for data deviation. For example, as described above, the regulation of CSPP1 expression involves multiple factors, which leads to the inconsistency between any individual factor and CSPP1 expression, and the small sample size increases this inconsistency. In addition, the inconsistent expression between CSPP1 mRNA and protein is the same case. Therefore, further basic experiments and more clinical samples are required to explore the direct functional mechanism of CSPP1 affecting cancer progression and prognosis through ferroptosis and TME function.

Conclusion

In conclusion, our study is the first to demonstrate that CSPP1 is a potential diagnostic and prognostic biomarker associated with ferroptosis and TME, providing a new target for ferroptosis-based drug therapy and immunotherapy in specific cancer types.

Author contributions

SZ contributed to the conception of the study. WW designed the study and wrote the manuscript. WW, JZ, YW, and YX had full access to all of the data in the study and take responsibility for the integrity of the data. WW, JZ, YW, and YX performed the statistical analyses. SZ obtained funding. All authors contributed to the article and approved the submitted version.

Funding information

This work was supported by the (81773242), (LY19H160032), and Major Project of Hangzhou Science and Technology Bureau (20180417A01).

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.