3.1 SLC25A1 expression in pan-cancer

We first checked the SLC25A1 expression in normal tissues using the HPA database. As shown in Fig. 1A, SLC25A1 mRNA expression was prominent in the liver, breast, small intestine, duodenum, and adipose tissue, all of which have a high content of lipids. We next used cBioPortal to explore genetic aberrations of SLC25A1 based on mutation, fusion, amplification, deep deletion, and multiple alternations across 32 types of cancer. The results revealed that gene amplification accounted for the most common alteration, while gene mutation rates were generally low (Fig. 1B). Uterine carcinosarcoma (UCS), sarcoma (SARC), lung squamous cell carcinoma (LUSC), bladder urothelial carcinoma (BLCA), and ovarian serous cystadenocarcinoma (OV) had the highest amplification frequencies, i.e., 7.02, 3.92, 3.29, 2.92 and 2.74%, respectively. Therefore, we mainly focused on exploring the expression level and significance of SLC25A1 across human pan-cancers in the following analysis.

Fig. 1

The expression level of SLC25A1 in normal and tumor tissues. A SLC25A1 mRNA expression in normal tissues based on human protein atlas (HPA) database. B Alteration frequency of SLC25A1 across different cancer types in cBioPortal

The expression of SLC25A1 mRNA between normal tissues and tumors in pan-cancer was investigated by combining GTEx and TCGA data (Fig. 2A). The results showed that compared with normal tissues, the SLC25A1 expression in 20 of the 33 cancer types (BLCA, COAD, DLBC, ESCA, GBM, HNSC, KIRC, LGG, LIHC, LUAD, LUSC, PAAD, PRAD, READ, SKCM, STAD, THYM, UCEC, USC, and OV) was significantly higher. In comparison, its expression in 3 types (BRCA, LAML, and THCA) was lower. Meanwhile, the area under the curve (AUC) values for ROC analysis of SLC25A1 mRNA expression was performed in each cancer (Supplementary Fig. 1). The results indicated that SLC25A1 had certain accuracy (AUC > 0.7) in predicting 16 cancer types, including BLCA, COAD, DLBC, GBM, HNSC, LAML, LGG, LIHC, LUSC, PAAD, PRAD, READ, STAD, THYM, UCEC, and UCS. Among them, SLC25A1 had high accuracy (AUC > 0.9) in predicting GBM, LAML, LGG, and PAAD. We also explored the correlation between SLC25A1 expression and the pathological stages of cancers using TCGA data (Fig. 2B). Its expression was observed significantly increase in the advanced stage (stage III/ IV) relative to early-stage (stage I/ II) in ACC, BRCA, KIRC, and TGCT, suggesting a role for SLC25A1 in tumor suppression. Moreover, UALCAN bioinformatic analysis was used to analyze SLC25A1 protein expression comprehensively (Fig. 2C). There was elevated expression in the tumor tissues of COAD, LUAD, and UCEC (P < 0.001). Noteworthy, SLC25A1 protein expression in KIRC was lower than that in the normal tissues (P < 0.001), which was in contrast to the mRNA expression.

Fig. 2

SLC25A1 expression in pan-cancer and different pathological stages. A The mRNA expression levels of SLC25A1 in different types of cancers and normal tissues from the GTEx and TCGA databases. B The mRNA levels of SLC25A1 in different pathological stages (stage I/II and stage III/IV) of ACC, BRCA, KIRC, and TGCT from the TCGA. C The protein expression of SLC25A1 in four cancers (COAD, KIRC, LUAD, and UCEC) and normal tissues from the UALCAN database. *P < 0.05, **P < 0.01, ***P < 0.001

To further study the expression of SLC25A1, we performed immunohistochemical staining (IHC) in tissue microarrays containing several kinds of human normal tissues and cancers. High SLC25A1 protein expression was observed in normal liver tissue (Fig. 3A). This result was consistent with the mRNA expression data from the HPA database. Furthermore, the IHC showed that the expression levels of SLC25A1 in COAD and LUAD were significantly increased compared with adjacent normal tissues (Fig. 3B, C).

Fig. 3

Immunohistochemical analysis of the expression of SLC25A1 in normal and tumor tissues. A The protein expression of SLC25A1 in seven normal tissues (thyroid gland, esophagus, stomach, colon, liver, pancreas, and lung). B Typical results of one pair of samples in COAD and LUAD. C Statistical analysis of the staining score. Bars, means ± SD. *P < 0.05, ** P < 0.01

3.2 SLC25A1 prognostic value in pan-cancer

We examined the prognostic values of SLC25A1 across 33 cancer types in the TCGA database (Fig. 4A). Our study revealed that increased expression of SLC25A1 was significantly related with worse OS in LAML (HR 2.21 [95% CI 1.44–3.40], P < 0.001), ACC (HR 4.87 [95% CI 1.84–12.89], P = 0.001), LUAD (HR 1.68 [95% CI 1.23–2.28], p = 0.001), HNSC (HR 1.42 [95% CI 1.05–1.92], P = 0.023), LIHC (HR 1.57 [95% CI 1.05–2.34], P = 0.027), MESO (HR 1.81 [95% CI 1.03–3.18], P = 0.039) and SKCM (HR 1.32 [95% CI 1.01–1.73], P = 0.046) (Fig. 4B–H). Conversely, high SLC25A1 expression was correlated to better OS in LGG (HR 0.46 [95% CI 0.32–0.65], P < 0.001) and PCPG (HR 0.09 [95% CI 0.02–0.47], P = 0.004) (Fig. 4I-J). Additionally, PFI analyses indicated that high SLC25A1 expression predicted unfavorable outcomes in individuals with ACC, ESCA, HNSC, KIRC, PRAD, and TGCT, while the opposite trend was observed in individuals with LGG (Supplementary Fig. 2).

Fig. 4

Correlation between SLC25A1 expression and OS from TCGA database. A Analysis of the relationship between SLC25A1 expression and OS in 33 cancer types using univariate Cox regression and forest plot. The expression of SLC25A1 was related to the survival rate of LAML (B), ACC (C), LUAD (D), HNSC (E), LIHC (F), MESO (G), SKCM (H), LGG (I) and PCPG (J). OS, overall survival

Then, the relationship between SLC25A1 expression and tumor prognosis was checked based on GEO datasets. Using the gene chip data in Kaplan–Meier Plotter, we found that SLC25A1 was a high-risk gene in lung cancer (OS: HR = 1.6, log-rank P = 1.2e-08; FP: HR = 1.9, log-rank P = 2.3e-07), gastric cancer (OS: HR = 1.81, log-rank P = 7.2e-11; FP: HR = 1.75, log-rank P = 6.1e-08; PPS: HR = 2.42, log-rank P = 1.31e-15) and breast cancer (OS: HR = 1.4, log-rank P = 0.00043; PPS: HR = 1.34, log-rank P = 0.013) (Fig. 5A–G), but a gene of low risk in ovarian cancer (OS: HR = 0.8, log-rank P = 0.00083; PPS: HR = 0.76, log-rank P = 0.0016) (Fig. 5H, I). Intriguingly, further analysis showed that SLC25A1 was associated with a poor prognosis in lung adenocarcinoma, but was not correlated to OS or FP in lung squamous cell carcinoma (Supplementary Fig. 3). The impact of SLC25A1 on survival was also evaluated through PrognoScan. The results are summarized in Supplementary Table 1. Similar to our findings in Kaplan–Meier Plotter and TCGA database, SLC25A1 played a detrimental role in breast cancer, lung cancer (adenocarcinoma), and skin cancer (melanoma). Meanwhile, SLC25A1 had a protective role in brain cancer. Notably, patients with ovarian cancer displayed the opposite trend in DUKE-OC and GSE8841. These findings suggested that SLC25A1 expression is differentially related to the prognosis in pan-cancer and may be a potential prognostic marker in certain types of tumors.

Fig. 5

Correlation between SLC25A1 expression and prognosis of different types of cancer in Kaplan–Meier Plotter database. OS (A) and FP (B) of lung cancer; OS (C), FP (D) and PPS (E) of gastric cancer; OS (F) and PPS (G) of breast cancer; OS (H) and PPS (I) of ovarian cancer. OS, overall survival; FP, first progression; PPS, post-progression survival

3.3 Relationship between SLC25A1 expression and MSI, TMB, and immune checkpoint genes in pan-cancer

Microsatellite instability (MSI) and Tumor mutation burden (TMB) were considered to be important biomarkers in predicting the response to ICIs, which are represented by anti–programmed death-1/ligand-1 (PD-1/PD-L1) and anti–cytotoxic T lymphocyte antigen-4 (CTLA-4) inhibitors. Hence, we analyzed the correlation between SLC25A1 expression and MSI/TMB in diverse cancer types of TCGA. The results revealed that SLC25A1 was positively associated with the MSI in DLBC (P = 1e-06), LUAD (P = 0.0053), PRAD (P = 0.00032), UCEC (P = 6.3e-11), TGCT (P = 2e-06), ESCA (P = 0.0079), STAD (P = 4.2e-05), KIRC (P = 0.0036) and HNSC (P = 2.8e-05) (Fig. 6A). We further found that SLC25A1 expression was positively correlated with TMB in KICH (P = 2.8e-08), LUAD (P = 6.4e-08), PRAD (P = 0.013), UCEC (P = 0.0044), BLCA (P = 0.0045) and STAD (P = 2.9e-06) (Fig. 6B). Interestingly, no significant negative correlation between MSI/TMB and SLC25A1 expression was observed in all these types of cancer.

Fig. 6

Association between SLC25A1 expression and microsatellite instability (MSI), tumor mutation burden (TMB), and immune checkpoint genes from TCGA database. A Association of SLC25A1 expression with MSI in different cancers. B Association of SLC25A1 expression with TMB in different cancers. C Association of SLC25A1 expression with immune checkpoint genes in different cancers. * P < 0.05

Subsequently, we evaluated the associations of SLC25A1 expression and common immune checkpoint genes. The heat map showed that SLC25A1 exhibited a positive correlation in OV and a negative correlation in LUSC, PRAD and BLCA with more than 40 significant co-expressed immune genes (Fig. 6C). In addition, SLC25A1 expression was positively correlated with CD276 and VEGFB but negatively associated with many known immune checkpoints, including CD274, PDCD1, and CTLA-4 in multiple cancers. These results indicated that the relationship between SLC25A1 and immune checkpoints varies by both tumor type and checkpoint-gene specificity.

3.4 Correlation between SLC25A1 expression and immune infiltration across cancers

Tumor-infiltrating immune cells are an essential part of the tumor microenvironment (TME) and have been identified as potential biomarkers for predicting prognosis and response to treatment in cancer patients. Following that, we searched the relationship between SLC25A1 expression and infiltration level of 6 immune cell types (CD4 + T cells, CD8 + T cells, neutrophils, myeloid dendritic cells, macrophage, and B cells) in pan-cancer from the TIMER database. The top three tumors where SLC25A1 expression showed the most relevance to immune infiltration levels were PRAD, PAAD, and LUSC (Fig. 7A). Among them, SLC25A1 expression showed positive relevance with CD8 + T cells (R = 0.128, P = 8.89e-03), macrophage (R = 0.268, P = 2.69e-08)and B cells (R = 0.186, P = 1.34e-04) but negative correlation with CD4 + T cells (R = − 0.217, P = 7.83e-06), neutrophils (R = − 0.213, P = 1.23e-05) and myeloid dendritic cells (R = − 0.192, P = 8.20e-05) in PRAD. In PAAD, SLC25A1 expression has a significant positive correlation with the immune infiltrating levels of CD4 + T cells (R = 0.221, P = 3.75e-03) but negative correlation with the infiltrating levels of CD8 + T cells(R = − 0.25, P = 9.98e-04), neutrophils (R = − 0.242, P = 1.40e-03) and myeloid dendritic cells (R = − 0.228, P = 2.74e-03). In addition, SLC25A1 expression was significantly and negatively correlated with infiltrating levels of CD8 + T cells (R = − 0.186, P = 4.17e-05), neutrophils (R = − 0.315, P = 1.99e-12), myeloid dendritic cells (R = − 0.189, P = 3.26e-05) and macrophage (R = − 0.128, P = 5.06e-03) in LUSC.

Fig. 7

Top 3 cancers related to immune infiltration, Stromalscore, Immunescore, and ESTIMATE score. A Association between SLC25A1 expression and the level of immune infiltration in PRAD, PAAD, and LUSC. B Top 3 cancers related with Stromalscore, Immunescore, and ESTIMATE score derived by ESTIMATE algorithm

The Estimation of Stromal and Immune cells in Malignant Tumor tissues using the Expression data (ESTIMATE) algorithm could help to determine the fraction of stromal and immune cells in the tumor microenvironment. The present study found that SLC25A1 gene expression was negatively correlated with Stromalscore, Immunescore, and ESTIMATE score (Fig. 7B). The expression of SLC25A1 was most associated with StromalScore in LUSC (r = − 0.35), STAD (r = − 0.33), and BLCA (r = − 0.30) and significantly negatively correlated with ImmuneScore in LUSC (r = − 0.47), TGCT (r = − 0.43), and CESC (r = − 0.43). Similarly, There were negative correlations between SLC25A1 expression and ESTIMATE score in LUSC (r = − 0.44), CESC (r = − 0.39) and STAD (r = − 0.38). These results indicated that SLC25A1 expression was closely related to the tumor microenvironment in certain tumor types.

3.5 SLC25A1-related proteins and genes enrichment analysis

To further uncover the biological significance of SLC25A1 in carcinogenesis at the pan-cancer level, we screened SLC25A1-related proteins and genes and conducted subsequent pathway enrichment analysis. By using GeneMANIA and STRING, we generated the protein–protein interaction network of SLC25A1. Figure 8A from GeneMANIA showed 20 genes most related to SLC25A1 and a PPI network. Furthermore, the STRING database was applied to acquire an interaction network of the top 50 SLC25A1-binding proteins (Fig. 8B). Then, we studied the top 100 genes positively correlated with SLC25A1 expression in pan-cancer of TCGA with the GEPIA2 tool. The scatter plots displayed that SLC25A1 expression was statistically related to that of the top 5 genes, including DGCR6L (R = 0.45), DDT (R = 0.44), PCYT2 (R = 0.42), TXNRD2 (R = 0.4) and COMT (R = 0.4) (Fig. 8C). The heat map from the TIMER database further confirmed a positive relationship between SLC25A1 expression and the above five genes in most cancer types of TCGA (Fig. 8D).

Fig. 8

SLC25A1-related protein and genes analysis. A Protein–protein interaction (PPI) network of SLC25A1 using GeneMANIA. B Protein–protein interaction (PPI) network of SLC25A1 using STRING. C The expression correlation between SLC25A1 and DGCR6L, DDT, PCYT2, TXNRD2, COMT utilizing GEPIA2. D Heatmap of the expression correlation between SLC25A1 and DGCR6L, DDT, PCYT2, TXNRD2, COMT utilizing TIMER database

According to the Venn diagram, three common genes of the STRING group and GeneMANIA group were screened out as ACLY, ACO2, and G6PD; while six common members were located in both STRING and GEPIA2 list, including DGCR2, DGCR14, MRPL24, MRPL40, UFD1L and TXNRD2 (Fig. 9A). Combined with the above three databases, we carried out GO and KEGG enrichment analyses. In GO enrichment analysis, these SLC25A1 interacted and correlated genes mainly enriched in the small molecule catabolic process in the biological process (BP), in the mitochondrial matrix in terms of cell component (CC), and in the coenzyme binding in terms of molecular function (MF) (Fig. 9B–D). The KEGG pathway enrichment data further suggested that the selected genes might be involved in the Carbon metabolism, Citrate cycle (TCA cycle), Biosynthesis of amino acids, 2-Oxocarboxylic acid metabolism, Glycolysis/Gluconeogenesis, etc. (Fig. 9E).

Fig. 9

SLC25A1-related genes enrichment analysis. A Venn diagram of the SLC25A1- interacted and correlated genes of STRING, GeneMANIA and GEPIA2 database. B–D GO enrichment analysis of SLC25A1-interacted and correlated genes. E KEGG pathway analysis of SLC25A1-interacted and correlated genes

3.6 SLC25A1 gene dependency in pan-cancer

In order to comprehensively study SLC25A1 expression on cancer cell growth, we analyzed the vulnerability to CRISPR/Cas9-mediated perturbation of SLC25A1 in large panels of human cancer cell lines using the DepMap project data (DepMap 22Q2 Public + Score, Chronos). Negative scores of Gene Effect (dependency scores) reflect reduction in cell growth and survival after depletion of a particular gene and the score of -1 corresponds to the median of all common essential genes.

Notably, SLC25A1 knockout showed an inhibitory effect in most cancer cell lines of all cancer types with a median dependency score < 0 (Fig. 10A). The Top10 cell lines sensitive to SLC25A1 gene knockout were lymphoma cell lines (SUDHL10, SUDHL5, and WSUNHL), endometrial/uterine cancer cell line (JHUEM1), sarcoma cell lines (RH30 and RHJT), leukemia cell lines (ROS50 and HB1119) and breast cancer cell line (KPL1), all of which exhibited a dependency score less than − 0.85 (Fig. 10B). Among them, 2 DLBCL cell lines SUDHL10 and SUDHL5 were the most sensitive cell lines for SLC25A1-KO, and both of their dependency scores were lower than—1. Furthermore, we queried the Top Co-dependency Pearson correlations of SLC25A1, and PDHA1, MDH2, PDHB, SH3GL1 and DLAT were identified as the top five dependencies (Fig. 10C). Taken together, these findings suggest a potential role of SLC25A1 in clinical applications.

Fig. 10

Dependency of cancer cells on SLC25A1 gene in the DepMap project. A Gene effect of CRISPR/Cas9 knockout of SLC25A1 gene across various human cancer cell lines. More negative scores indicate greater sensitivity to knockout. Each cell line is represented by a circle symbol. B Top 10 cancer cell lines sensitive to SLC25A1 gene knockout. C Top 5 Co-Dependencies upon SLC25A1 deletion by CRISPR/Cas9 in tumor cell lines

3.7 Effect of SLC25A1 knockdown on cell proliferation

To confirm the function of SLC25A1 in tumor growth, we selected lung adenocarcinoma cell lines to examine the effect of SLC25A1 knockdown on cell proliferation in vitro. We used small interference RNA (siRNA) to knock down the expression levels of SLC25A1 in A549 and H1299 cell lines (Fig. 11A, B). The MTS assays showed a significant decrease in the viability of SLC25A1 knockdown cells, compared with the negative control cells (Fig. 11C). The colony formation assays demonstrated that SLC25A1 knockdown markedly suppressed the proliferation of lung adenocarcinoma cells (Fig. 11D). These observations suggest that SLC25A1 may play a role in LUAD progression.

Fig. 11

Knockdown of SLC25A1 attenuated tumor growth in LUAD cells. A, B Knockdown of SLC25A1 was confirmed by qRT-PCR and western blot assays in A549 and H1299 cells. C Cell viability was determined by MTS assay in A549 and H1299 cells with SLC25A1 knockdown. D Colony formation assay in A549 and H1299 cells with SLC25A1 knockdown. Bars, means ± SD. *P < 0.05, **P < 0.01, ***P < 0.001