The oncogenic role of tubulin alpha-1c chain in human tumours

Thirty-three cancer-related RNA sequencing datasets and the clinicopathological and survival data for the corresponding patients were downloaded from the UCSC Xena website (https://​xena.​ucsc.​edu/​, derived from TCGA). Next, we analysed the mRNA expression of TUBA1C in 33 human malignancies using the online cancer microarray database Oncomine (https://​www.​oncomine.​org/​). We set the filter as we reported before [19]. Next, we used Perl software to extract and integrate the expression information of TUBA1C in 33 cancers in TCGA (https://​tcga.​xenahubs.​net) using the “wilcox.test” function. The R package “ggpubr” was applied to draw a box plot. In addition, the Tumor Immune Estimation Resource (TIMER) database (https://​cistrome.​shinyapps.​io/​timer/​) [20] and Gene Expression Profiling Interactive Analysis (GEPIA) (http://​gepia.​cancer-pku.​cn/​) database [21] were used for further profiling of TUBA1C expression in cancers. Moreover, we collected immunohistochemical data for cancer tissues using the HPA database (https://​www.​proteinatlas.​org/​) [22] and compared them with those for normal tissues to confirm the differential expression of TUBA1C in cancers at the protein level. In addition, the CGGA database (http://​www.​cgga.​org.​cn/​) [23] was used to further investigate the expression levels and prognostic value of TUBA1C in gliomas.

The survival information obtained from the TCGA database corresponding to each sample was used to further analyse the relationships between TUBA1C expression and clinical outcomes, including overall survival (OS), disease-specific survival (DSS), disease-free interval (DFI), and progression-free interval (PFI). The results of survival analysis were visualized using forest plots and Kaplan–Meier (KM) curves.

To analyse the association between TUBA1C and TMB and MSI, we first collated the information from the TCGA database using Perl software. Next, the command “cor.test”, based on Spearman’s method, was used, and the R package “fmsb” was utilized to plot radar plots. We then calculated the immune and stromal fractions using the ESTIMATE algorithm. Cell-type identification by estimating relative subsets of RNA transcripts (CIBERSORT) algorithm was applied to assess tumour purity and stromal/immune cell infiltration in tumour tissues (n = 33) according to expression file [24]. Subsequently, we evaluated the relationship between TUBA1C and TME or immune cell infiltration by applying the R packages “ggplot2”, “ggExtra” and ggpubr” (cut-off value p < 0.001). We further explored the associations between TUBA1C expression and immune cell infiltration with CIBERSORT, EPIC, TIMER, quanTIseq, xCELL and MCPcounter.

Next, CD274, CTLA4, HAVCR2, PDCD1, LAG3, SIGLEC15, PDCD1LG2, and TIGIT were chosen as immune checkpoint-related genes, and the expression data of these 8 transcripts were extracted. We used the R packages “ggplot2”, “immuneeconv” and “pheatmap” to assess immune checkpoint expression and the coexpression of TUBA1C with these immune checkpoints. The response to ICI therapy was predicted using the TIDE algorithm [25].

We downloaded the gene lists for immune activation, immunosuppression, chemokine receptor proteins, chemokines and Major histocompatibility complex (MHC) molecules from GSEA and processed the data in R. The results are shown in heatmaps.

We obtained the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) gene sets [26] from GSEA (http://​www.​gsea-msigdb.​org/​gsea/​downloads.​jsp) to investigate the functional pathways of TUBA1C in tumours and then used the R packages “limma,” “org.​Hs.​eg.db,” “enrichplot,” and “clusterProfiler” to perform of GO functional and KEGG pathway annotation as well as pathway enrichment analysis of TUBA1C.

We collected specimens for immunohistochemistry from low-grade glioma (LGG), glioblastoma (GBM) and normal tissues from Renmin Hospital of Wuhan University. After deparaffinization, hydration and antigen repair, endogenous peroxidase activity in the sections was inhibited with 3% hydrogen peroxide for 15 min and the sections were then blocked with BSA, incubated with an anti-TUBA1C primary antibody (ab222849) overnight at 4 °C, and incubated with a secondary antibody for 50 min. Colour was developed with DAB chromogen, and the sections were restained with haematoxylin and finally subjected to dehydration and sealing. Finally, images were acquired by microscopy. We used ImageJ to quantify the immunohistochemical results.

The mRNA expression data of all genes in cancers were normalized using log2 transformation. Differences in data for each group were obtained using t tests or ANOVA. Significance was analysed with 0.05 as the threshold. Survival analysis was performed using Cox regression models, the KM method and the log-rank test. Moreover, Spearman’s or Pearson’s tests were used to identify relationships between two genes. All statistical procedures were completed in R software.

We used Oncomine to assess TUBA1C gene expression in 33 cancers. It was shown that TUBA1C expression was elevated in most tumours, including bladder cancer (BLCA), breast cancer (BRCA), colorectal cancer (COAD), gastric cancer, head and neck cancer (HNSC), kidney cancer, liver cancer (LIHC), lymphoma, melanoma, myeloma, ovarian cancer (OV), pancreatic cancer (PAAD), and sarcoma (SARC). In contrast, lower TUBA1C expression was also found in BRCA and cervical cancer (CESC) (Fig. 1A). We further explored the differential expression of TUBA1C in cancers in GEPIA. Similar to the results in Oncomine, TUBA1C was upregulated in most cancers, including BLCA, BRCA, CESC, COAD, large B-cell lymphoma (DLBC), GBM, LGG, LIHC, LUAD, lung squamous cell carcinoma (LUSC), OV, PAAD, rectal cancer (READ), stomach cancer (STAD), testicular cancer (TGCT), thymoma (THYM), endometrioid cancer (UCEC) and uterine carcinosarcoma (UCS). Intriguingly, lower expression of TUBA1C was noted in acute myeloid leukaemia (LAML) (Fig. 1B).

We then analysed the data from TCGA with R software. A total of 11,057 mRNA expression profiles were obtained for 33 cancer types comprising 730 normal samples and 10,327 tumour samples. We found that TUBA1C showed higher expression in tumour samples in BLCA, BRCA, CESC, bile duct cancer (CHOL), COAD, oesophageal cancer (ESCA), GBM, HNSC, kidney chromophobe (KICH), kidney clear cell carcinoma (KIRC), kidney papillary cell carcinoma (KIRP), LIHC, LUAD, LUSC, prostate cancer (PRAD), READ, STAD and UCEC. No low expression of TUBA1C was detected in 33 cancers (Fig. 1C). Subsequently, TUBA1C was found to be upregulated in BLCA, BRCA, CESC, CHOL, COAD, ESCA, GBM, KIRC, KIRP, LIHC, LUAD, LUSC, PRAD, READ, STAD, THCA and UCEC by analysis with TIMER. In addition, the TUBA1C expression level was higher in HNSC-HPV+ tumours than in HNSC-HPV- tumours, and it was higher in SKCM tumours than in SKCM metastatic tumours (Fig. 1D). In conclusion, TUBA1C functions as a tumour promoter in most cancer types.

Immediately thereafter, we investigated the differential expression of TUBA1C at the protein level in cancerous and normal tissues in depth using Human Protein Atlas (HPA, https://​www.​proteinatlas.​org/​). The IHC results of BRCA, UCEC, renal cancer, LUAD, CESC, GBM, LIHC, PRAD and normal tissues were showed in Fig. 2A-H and these results were consisted with the expression data of TUBA1C mRNA in TCGA.

Subsequently, we explored the association between TUBA1C expression and the prognosis of patients across cancers. Cox regression analysis showed a significant relationship between TUBA1C expression and OS in BRAC (P = 0.031, HR = 1.288), GBM (P = 0.018, HR = 1.287), KICH (P = 0.002, HR = 7.366), KIRC (P = 0.004, HR = 1.389), KIRP (P = 0.01, HR = 2.012), LAML (P = 0.01, HR = 1.801), LGG (P < 0.001, HR = 2.373), LIHC (P < 0.001, HR = 1.675), LUAD (P = 0.002, HR = 1.373), MESO (P = 0.001, HR = 1.854), PAAD (P < 0.001, HR = 1.779), READ (P = 0.026, HR = 0.477) and SKCM (P < 0.001, HR = 1.418) (Fig. 3A). Moreover, our results suggested that high TUBA1C expression was a high-risk indicator for KIRC, KIRP, LAML, LGG, LIHC, LUAD, MESO, PAAD, SARC, and SKCM—especially KICH (HR = 7.366)—but was not a low-risk indicator for any cancer. In addition, a negative relationship between TUBA1C expression and OS was observed in 8 cancers (Fig. 3B-I): BRCA (p = 0.011), KIRC (p = 0.016), LAML (p = 0.037), LGG (p < 0.001), LIHC (p < 0.001), LUAD (p = 0.003) and SKCM (p = 0.002).

Since factors unrelated to the tumours may also contribute to death during follow-up, the relationship between TUBA1C expression and DSS in patients was subsequently analysed. It was revealed that TUBA1C expression impacted DSS in GBM (p = 0.032, HR = 1.279), KICH (p = 0.002, HR = 5.227), KIRC (p < 0.001, HR = 1.772), KIRP (p = 0.010, HR = 2.191), LGG (p < 0.001, HR = 2.498), LIHC (p < 0.001, HR = 1.538), LUAD (p = 0.034, HR =1.279), MESO (p = 0.007, HR = 1.899), PAAD (p < 0.001, HR =1.872), PRAD (p = 0.019, HR = 0.320) and SKCM (p = 0.009, HR = 1.296) (Fig. 4A). In these analyses, KICH displayed the highest HR (HR = 5.227). In addition, KM curves showed that high expression of TUBA1C was correlated with poor DSS in patients with BRCA (p = 0.044), KIRC (p < 0.001), LGG (p < 0.001), LIHC (p = 0.002), MESO (p < 0.001), PAAD (p = 0.026), SKCM (p = 0.029) and UCEC (p = 0.040) (Fig. 4B-I).

In addition, we explored the association between TUBA1C expression and the DFI in patients with different cancer types and found that increased TUBA1C expression was correlated with a decreased DFI in LGG (p < 0.001, HR = 3.229), LUAD (p = 0.007, HR = 1.445), PAAD (p = 0.005, HR = 2.281) and SARC (p = 0.026, HR = 1.314) (Fig. 5A). Moreover, in LUAD (p = 0.031), PAAD (p = 0.030) and SARC (p = 0.025), a high level of TUBA1C predicted a shorter DFI and poor prognosis in patients (Fig. 5B-D). Subsequently, we evaluated the association between TUBA1C expression and PFI. The forest plots showed that a high level of TUBA1C was correlated with a shorter PFI in COAD (p = 0.041, HR = 0.780), GBM (p = 0.007, HR = 1.893), KICH (p < 0.001, HR = 6.417), KIRC (p < 0.001, HR = 1.447), LGG (p < 0.001, HR = 1.893), LIHC (p = 0.004, HR = 1.266), LUAD (p = 0.024, HR = 1.215), MESO (p = 0.029, HR = 1.524), PAAD (p < 0.001, HR = 1.758), SARC (p = 0.004, HR = 1.288) and UCEC (p = 0.035, HR = 1.238). Among these cancers, KICH showed the highest HR (6.417, Fig. 6A). The results also indicated that high expression of TUBA1C affected the PFI unfavourably in ACC (p = 0.027) (Fig. 6B), GBM (p = 0.022) (Fig. 6C), KICH (p = 0.026) (Fig. 6D), KIRC (p = 0.005) (Fig. 6E), LGG (p < 0.001) (Fig. 6F), LUAD (p = 0.019) (Fig. 6G), MESO (p = 0.010) (Fig. 6H), PAAD (p = 0.006) (Fig. 6I) and SARC (p = 0.022) (Fig. 6K) but favourably in READ (p = 0.024) (Fig. 6J).

We next assessed the relationship between TUBA1C expression and tumour stage. We observed that in BRCA (Fig. 7A, p = 0.033), HNSC (Fig. 7C, p = 0.0038), KICH (Fig. 7D, p = 0.00022), KIRC (Fig. 7E, p = 0.00025), LUAD (Fig. 7G, p = 0.034), MESO (Fig. 7H, p = 0.02) and READ (Fig. 7J, p = 0.025), significant differences in TUBA1C expression existed between stage I and stage IV tumours. In LIHC (Fig. 7F, p = 0.0067), LUAD (Fig. 7G, p = 0.0027) and MESO (Fig. 7H, p = 0.022), the expression of TUBA1C was higher in stage III than in stage I tumours. In ESCA, the TUBA1C expression in stage IV was higher than in stage III (Fig. 7B, p = 0.029). Intriguingly, in PAAD (Fig. 7I, p = 0.022), TUBA1C expression was higher in stage II tumours than in stage I tumours, but no dramatic difference existed in stage III or IV tumours. In conclusion, TUBA1C expression was correlated with tumour stage.

We first examined the association between TUBA1C expression level and TMB and MSI, two indicators that predict the sensitivity of cancer patients to ICIs. A positive association between TUBA1C expression and TMB was observed in 15 cancers, namely, UCS, UCEC, STAD, SKCM, SARC, PRAD, PAAD, LUAD, LGG, KIRC, KICH, COAD, CESC, BRCA and BLCA. In contrast, in THYM, a negative correlation between TUBA1C expression and TMB was observed (Fig. 8A). In ACC, UCEC, SARC and COAD, the TUBA1C expression level was positively associated with MSI, while in OV, LUSC, LUAD and LGG, a negative association existed between these factors (Fig. 8B).

Tumour development is associated not only with abnormal mutations in tumour cells but also with the composition of their microenvironment and stromal cell proportions or activation states [27]. Therefore, the association between TUBA1C expression and the TME was investigated by assessing the relationships of stromal and immune scores to TUBA1C expression in 33 cancers using the ESTIMATE algorithm. It was shown that TUBA1C expression was positively correlated with stromal scores in GBM and LGG, while a negative correlation was observed in ESCA, STAD and TGCT (Fig. 9A). In addition, TUBA1C expression was positively related to immune scores in GBM, LGG, PCPG and THCA and negatively related to immune scores in ESCA (Fig. 9B).

TIICs, an important component of the TME, are closely associated with the development, progression and metastasis of cancer. Herein, we examined the underlying relationship between the level of infiltration of 22 immune cell types and TUBA1C expression in different cancer types. We selected the TIICs with the highest correlation with TUBA1C expression in each cancer for demonstration (Fig. 10). Our results showed that TUBA1C expression was negatively related to the infiltration of memory CD4+ T cells in COAD (Fig. 10D), STAD (Fig. 10P) and TGCT (Fig. 10Q) but positively associated in BRCA (Fig. 10B). In ESCA (Fig. 10E), TUBA1C expression was observed to be negatively correlated with infiltration of regulatory T cells. In addition, TUBA1C expression was positively related to infiltration of neutrophils in BLCA (Fig. 10A), HNSC (Fig. 10G), KIRC (Fig. 10H) and SKCM (Fig. 10O) and positively correlated with infiltration of dendritic cells in THCA (Fig. 10R). In addition, the TUBA1C expression level was correlated with infiltration of several diverse subsets of macrophages in the TMB. TUBA1C expression was positively associated with M1 macrophage infiltration in CESC (Fig. 10C), KIRP (Fig. 10I), LGG (Fig. 10J) and UCEC (Fig. 10T), with M2 macrophage infiltration in LUSC (Fig. 10M) and with M0 macrophage infiltration in SARC (Fig. 10N). In addition, TUBA1C expression was negatively related to infiltration of mast cells in LUAD (Fig. 10L) and THYM (Fig. 10S). Moreover, TUBA1C expression was positively related to infiltration of gamma delta T cells in GBM (Fig. 10F) and LIHC (Fig. 10K). We performed further analyses with other algorithms, including CIBERSORT (Fig. S1), EPIC (Fig. S2), MCPcounter (Fig. S3), quanTIseq (Fig. S4), TIMER (Fig. S5) and xCELL (Fig. S6), and also found correlations between TUBA1C and immune cell infiltration.

Next, we investigated the correlations between TUBA1C expression and the expression of immune checkpoints, including CD274, HAVCR2, CTLA4, LAG3, PDCD1LG2, PDCD1, SIGLEC15, and TIGIT, in cancers. We found that TUBA1C was correlated with immune checkpoints in all cancers except UVM, KIRC and ACC. In LUSC and ESCA, TUBA1C expression was negatively related to the expression of most immune checkpoints, while in OV, LGG, BRCA and BLCA, a positive association of TUBA1C expression with immune checkpoints was revealed (Fig. 11A), suggesting that TUBA1C has promise as a predictive factor for immunotherapeutic response in patients with these cancers. Additionally, the group with high TUBA1C responded better to ICI therapy than the group with low TUBA1C in LGG (Fig. 11B, p < 0.001), LIHC (Fig. 11C, p < 0.001), LUAD (Fig. 11D, p < 0.001) and LUSC (Fig. 11E, p = 0.011), suggesting that TUBA1C has the potential to be an immunotherapy target.

Further gene coexpression analysis was carried out between TUBA1C and immune-associated genes, including those encoding MHC molecules, chemokines, chemokine receptors, and proteins related to immune activation and immunosuppression, in 33 cancers. The heatmaps revealed that most immune-associated genes except CCL27 were coexpressed with TUBA1C and that the main immune-associated genes were positively related to TUBA1C in LGG, LIHC, PCPG and THCA (Fig. 12). TUBA1C was coexpressed with MHC genes in several cancer types, particularly BLCA, ESCA, LGG, LICH and THCA (Fig. 12A). We also found that TUBA1C was coexpressed with immune activation genes in all cancer types and with immunosuppressive genes in most cancers except uveal melanoma (UVM) (Fig. 12B, E). TUBA1C was coexpressed with most chemokines except CCL27, and coexpression of TUBA1C with chemokines and chemokine receptors was observed in most cancers except UVM (C, D).

To explore the underlying mechanism of action of TUBA1C, we conducted GO term and KEGG pathway enrichment analyses of TUBA1C (Fig. 13). Our results suggested that TUBA1C participated in several immune-associated pathways in the GO database, such as regulation of lymphocyte activation, positive regulation of cytokine production, and T-cell activation, in PCPG. In addition, it influenced immunoglobulin in KICH and MESO (Fig. 13A). In ESCO, TUBA1C was associated with lymphocyte differentiation in GO analysis and with chemokines in KEGG analysis. In LGG, there was a tight link between TUBA1C and the GO term regulation of immune receptor processes and between TUBA1C and the KEGG pathways chemokine signaling pathway and T-cell receptor signaling pathway (Fig. 13A, B). KEGG pathway analysis also indicated that TUBA1C mediated RIG-I-like receptor signalling in OV and mediated the regulation of autophagy, antigen processing and presentation and RIG-I-like receptor signalling in LUSC (Fig. 13B). Furthermore, TUBA1C was implicated in many biological pathways, including miRNA binding and gene silencing by RNA, and in many other pathways, such as allograft rejection, drug metabolism cytochrome p40, retinol metabolism, and neuroactive ligand receptor interaction.

We further investigated the expression of TUBA1C in gliomas and its impact on the prognosis of patients with gliomas. Figure 14A shows TUBA1C expression in gliomas of different histologies in the CGGA dataset. TUBA1C expression was highest in primary glioblastoma (pGBM) and recurrent pGBM (rGBM) and lowest in astrocytoma (A). In anaplastic oligodendrocytoma (AO) and anaplastic astrocytoma (AA), TUBA1C was expressed at moderate levels. Overall, TUBA1C expression increased in gliomas as WHO classification increased. In addition, TUBA1C expression was elevated in relapsed A, O, AO, and AA, similar to the results shown in Fig. 14I and Fig. 14J. Figure 14B further confirmed this phenomenon. We further found that TUBA1C expression was lower in the IDH mutant state than in the wild type (Fig. 14C), and similar results were found in WHO II, III and IV (Fig. 14D). In addition, TUBA1C expression was lower in patients with IDH mutation than in those with wild-type IDH (Fig. 14E), and similar results were found for WHO grades III and IV (Fig. 14F). In IDH wildtype LGG, the TUBA1C expression was higher than in LGG with IDH mutation and 1p/19q codeletion (Fig. 14G). Moreover, TUBA1C expression was higher in patients older than 42 years than in those aged < 42 years (Fig. 14H). Subsequently, we collected normal, LGG and GBM tissues for immunohistochemical analysis. The results showed that the IHC staining intensity was strongest in GBM (Fig. 14K, right), moderate in LGG (Fig. 14K, middle) and weakest in normal tissues (Fig. 14K, left). In primary gliomas, the probability of survival was higher in the TUBA1C low expression group than in the TUBA1C high expression group in all WHO grades (Fig. 14L, p < 0.0001) and in the WHO grade II (Fig. 14N, p = 0.032), WHO grade III (Fig. 14P, p = 0.032) and WHO grade IV (Fig. 14R, p = 0.017) subclassifications. In recurrent glioma, the probability of survival was higher in the TUBA1C low expression group in all WHO grades (Fig. 14M, p < 0.0001) and the WHO grade III (Fig. 14Q, p = 0.012) subclassification. In WHO II and IV recurrent glioma, the expression of TUBA1C was not related to the probability of survival (Fig. 14O, p = 0.98; Fig. 14S, p = 0.1).