Systematic evaluation of the prognostic and immunological role of PDLIM2 across 33 cancer types

The protein PDLIM2 regulates the stability of various transcription factors and is required for polarized cell migration. However, the clinical relevance and immune infiltration of PDLIM2 in cancer are not well-understood. We utilized The Cancer Genome Atlas and Genotype-Tissue Expression database to characterize alterations in PDLIM2 in pan-cancer. TIMER was used to explore PDLIM2 expression and immune infiltration levels. We assessed the correlation between PDLIM2 expression and immune-associated gene expression, immune score, tumor mutation burden, and DNA microsatellite instability. PDLIM2 significantly affected the prognosis of various cancers. Increased expression of PDLIM2 was significantly correlated with the tumor grade in seven types of tumors. The expression level of PDLIM2 was positively correlated with immune infiltrates, including B cells, CD8+ T cells, CD4+ T cells, neutrophils, macrophages, and dendritic cells in bladder urothelial, kidney renal papillary cell, and colon adenocarcinoma. High expression levels of PDLIM2 tended to be associated with higher immune and stromal scores. PDLIM2 expression was associated with the tumor mutation burden in 12 cancer types and microsatellite instability in 5 cancer types. PDLIM2 levels were strongly correlated with diverse immune-related genes. PDLIM2 can act as a prognostic-related therapeutic target and is correlated with immune infiltrates in pan-cancer.

Prognostic potential of PDLIM2 in cancers. We investigated whether PDLIM2 expression was correlated with prognosis in patients with cancer. First, the impact of PDLIM2 expression on overall survival was analyzed using univariate survival analysis. As shown in Fig. 2a, multiple cancer types exhibited a significant association between patient prognosis and PDLIM2 expression, including breast, kidney, blood, brain, and esophageal cancer. We additionally employed the Kaplan-Meier method to assess how PDLIM2 expression is related to prognosis in various cancer types, revealing its elevation to be significantly linked with adenoid cystic carcinoma (ACC; P = 0.008), BLCA (P = 0.038), acute myeloid leukemia (P < 0.001), and LUSC (P = 0.037) (Fig. 2b-e). In contrast, reduced PDLIM2 expression was correlated with poor prognosis in thymoma (P = 0.04) (Fig. 2f).
To eliminate the influence of non-tumor-related death factors, we analyzed the relationship between gene expression and disease-specific survival. Notably, PDLIM2 expression significantly affected the prognosis of four types of cancers ( Fig. 3a-d), including ACC (P = 0.004), BLCA (P = 0.01), COAD (P = 0.008), and KIRP (P = 0.006). Therefore, high PDLIM2 expression may be an independent risk factor for poor prognosis in these four types of cancers. Correlation between PDLIM2 expression and immune cell infiltration. Immune cell infiltration affects survival and tumor metastasis in patients. In the previous analysis, PDLIM2 expression levels were found to be associated with prognosis and metastasis in BLCA, KIRP, and COAD. The correlation between the expression level of PDLIM2 and the six types of infiltrating immune cells in these three tumor types is shown in Fig. 5. PDLIM2 expression was negatively correlated with tumor purity. Tumor purity refers to the proportion of cancer cells in tumor samples. Tumors with low tumor purity tend to have a higher mutational load and stronger immune phenotypes 21,22 . The expression level of PDLIM2 was positively correlated with macrophages (R = 0.415, P = 1.17e−16) and dendritic cells (R = 0.301, P = 4.56e−09) in BLCA; CD4+ T cells (R = 0.338, P = 3.49e−12), macrophages (R = 0.329,P = 1.16e−11), neutrophils (R = 0.321, P = 4.62e−11), and dendritic cells (R = 0.348, P = 7.42e−13) in COAD; and B cells (R = 0.406, P = 1.45e−11), CD8+ T cells (R = 0.389, P = 9.44e−11), and dendritic cells (R = 0.355, P = 5.29e−09) in KIRP.
Correlation between PDLIM2 expression level and immune cell markers. High PDLIM2 expression was positively correlated with immune cell infiltration. We further investigated the correlation between PDLIM2 expression levels and immunological marker sets in BLCA, COAD, and KIRP ( Table 1). The results revealed that the PDLIM2 expression level was positively correlated with the expression of most immune cell  Correlation between PDLIM2 expression level and tumor microenvironment. The expression of PDLIM2 may be closely related to immune escape in BLCA and KIRP. The immune score predicts the response to tumor immunotherapy. Analysis of the relationship between PDLIM2 expression and immune scores revealed that higher PDLIM2 expression levels were associated with higher immune scores and stromal scores (Fig. 6). www.nature.com/scientificreports/ We next examined the relationship between the expression of immune checkpoint genes (including 11 immunostimulators and 7 immunoinhibitors) and PDLIM2. The expression of PDLIM2 was significantly associated with that of immune checkpoint genes ( Table 2).

Correlation between PDLIM2 expression level and tumor mutational burden and microsatellite instability.
Previous studies focused on the prognostic role of tumor mutational burden (TMB) in immunotherapy in many cancer types; here, we counted the TMB of each tumor sample and analyzed the relationship between PDLIM2 expression and TMB in 12 different cancer types. PDLIM2 was negatively correlated with the TMB in BLCA, cholangiocarcinoma, COAD, LIHC, LUAD, LUSC, paraganglioma, prostate adenocarcinoma, pulmonary enteric adenocarcinoma, stomach adenocarcinoma, and thymoma and positively correlated with ACC (Fig. 7a).
The microsatellite instability (MSI) status of different solid tumors was significantly different from that of checkpoint inhibitor drugs for the immune response rates. Several studies have shown that MSI is closely related to tumor occurrence. Analysis of the correlation between PDLIM2 expression and MSI revealed a correlation mainly in adenocarcinoma (Fig. 7b). PDLIM2 expression was positively correlated with MSI in breast invasive  Correlation between PDLIM2 expression level and immune checkpoint genes. Immune checkpoints are signaling pathways responsible for downregulating the immune response to avoid destruction of endogenous targets and temper the peripheral immune response. We analyzed the relationship between the expression of common immune checkpoint genes and PDLIM2 in 33 cancer types. The results showed that PDLIM2 expression was significantly associated with immune checkpoint expression in most tumor types, although there was no significant relationship in ACC, esophageal carcinoma, LIHC, uterine carcinosarcoma, and uveal melanoma (Fig. 8). Notably, PDLIM2 was related not only to the common targets of checkpoint inhibitors (PDCD1, CTLA4, etc.), but also to targets of stimulatory checkpoint molecules (CD27, ICOS, etc.).

Discussion
PDLIM2 is a member of the PDZ-LIM family, whose PDZ structural domain is a functional module for protein-protein inter-recognition and the LIM structural domain is a structural domain for protein-protein interactions. These structures play important roles in cell differentiation and signal transduction 23 . PDLIM2 is crucial for tumorigenesis and progression and is closely related to the immune system. We comprehensively characterized PDLIM2 across 33 cancer types and highlight the potential clinical utility of immunity therapy for PDLIM2 expression.
The expression of PDLIM2 is related to both tumor inhibition and tumorigenesis 11 . Our results support the dual role of PDLIM2 in cancer. PDLIM2 is a putative tumor-suppressor protein that is inhibited by epigenetics in different cancers and is highly expressed in most adjacent noncancerous tissues. PDLIM2, as a cytoskeleton component, can reverse the growth of cancer cells by regulating promoter methylation. Repression of PDLIM2 has been shown to continuously activate nuclear factor-κB and STAT3, eventually leading to tumorigenesis and tumor maintenance 24 . Sun et al. found that PDLIM2 is inhibited in lung cancer, which is associated with a poor prognosis 14 . However, in our study, PDLIM2 led to different prognoses in different types of lung cancer. As seen by clinical correlation analysis, the reason for this discrepancy may be that high PDLIM2 expression contributes to the poor prognosis of LUSC patients by affecting lymph node metastasis. In addition, PDLIM2 was also highly expressed in invasive cancer cells. Zhao et al. reported that PDLIM2 promotes ovarian cancer growth in vivo and in vitro through NOS2-derived nitric oxide signaling 25 . Another study showed that PDLIM2 inhibition effectively reduced the tumor growth and invasiveness of human castration-resistant prostate cancer cells 26 . A recent study revealed that PDLIM2 was highly correlated with tumor growth and metastasis in renal cell carcinoma in a mouse knockout model 27 . We also observed significant correlations between PDLIM2 expression and patient survival in thymic carcinomas, which has not been widely reported in previous studies. Additionally, gene expression was related to clinical stage in seven tumors, suggesting that PDLIM2 is involved in promoting cancer progression or metastasis. 1,25 (OH) 2D3-induced adhesion of cancer cells to the extracellular matrix is mediated by PDLIM2 28 . PDLIM2 promotes tumor angiogenesis by activating the MAPK/ERK pathway 26 depending on the tumor microenvironment 29 . These results confirm the prognostic value of PDLIM2 in some specific types of cancers and that increased and decreased PDLIM2 expression have different prognostic values depending on the cancer type.    www.nature.com/scientificreports/ Importantly, we found that PDLIM2 functions in the recruitment and activation of immune infiltrating cells. PDLIM2 participates in the differentiation of TAMs into M2 macrophages; TIMER database results showed positively correlated with the infiltration of B cells and macrophages in BLCA and KIRP. We showed that PDLIM2 expression was not related to M1 macrophages but was positively correlated with most M2 macrophage markers. A previous study revealed that PDLIM2 leads to the recruitment of M2 macrophages in ovarian cancer 25 . TAMs lead to poor prognosis of tumors by inhibiting the immune and secreting various factors that promote tumor growth 30 . M2 macrophages can promote tumors growth and secrete IL-10, transforming growth factor β, and other mediators that contribute to establishing a tumor-tolerant microenvironment and angiogenic factors 31 . IL-10 (a TAM marker) is often associated with decreased T-cell activation 32 . In addition, PDLIM2 may inhibit T cell-mediated immunity and is involved in the immune escape of tumors. The correlation between PDLIM2 expression and immune cell marker genes suggests that PDLIM2 can control the infiltration and interaction of immune cells in the tumor microenvironment, and Treg-like immunosuppression can be induced by inducing Foxp3 expression in naïve T cells 33 . Dendritic cells can lead to tumor metastasis by enhancing the Treg response and suppressing the cytotoxicity of CD8+ T cells 34 . Notably, PDLIM2 expression was strongly correlated with TIM-3 in bladder cancer and with PDCD1, CTLA4, GZMB, and LAG3 in KIRP. These genes are common markers for T cell exhaustion, which is one of the main factors of immune dysfunction in patients with cancer. PD-1-or CTL-4-mediated pathways suppress T cell function 35 . LAG3 can inhibit T cells without relying on CD4 36 . T cell exhaustion can be partially reversed clinically using immune checkpoint inhibitors 37 . Patients with high PDLIM2 expression showed higher T cell infiltration; thus, these patients are more likely to benefit from immune checkpoint blockade against PD-1 and CTLA-4. Generally, PDLIM2 can mediate the differentiation of M2 macrophages and T cell exhaustion, thus avoiding immune detection.
Another key finding of this study is that PDLIM2 can be used as a therapeutic target for epigenetic drugs combined with immune checkpoint inhibitors. Patients with high expression benefit more from immunotherapy. PDLIM2 expression was associated with the TMB in 12 cancer types and MSI in five cancer types. TMB [38][39][40] and MSI are important in the immunotherapy response. The TMB is a pan-cancer genomic biomarker related to the efficacy of checkpoint inhibitors. A higher TMB is an easier target for tumor immunotherapy. MSI 41 is also an indicator of the efficacy of immunotherapy and was first used in colon cancer. Interestingly, although MSI indicators are more suitable for digestive system tumors, there is no correlation between PDLIM2 and MSI. PDLIM2 is a potential target for combination therapy for cancer. Notably, the expression of PDLIM2 was significantly associated with immune checkpoint expression in most tumor types. Guo and Qu suggested that PDLIM2 can www.nature.com/scientificreports/ be used as an epigenetic drug with immunomodulatory potential in combined immunotherapy for cancer 42 . Administration of epigenetic drugs can also enhance the efficacy of immunological checkpoint treatments 43 . There were some limitations to this study. First, all analyses were performed using public datasets. Our findings should be verified in animal experiments and hospitalized patients. Second, although we found that the expression of PDLIM2 is associated with tumor immune cell infiltration and survival, we did not confirm that PDLIM2 affects the survival of patients through immune infiltration, and the prognostic value of this protein in tumor immune mechanisms and immune signals should be further explored. Finally, we only included total PDLIM2 RNA levels in this study, without considering RNA variants and protein modifications.
In summary, PDLIM2 can affect pan-cancer prognosis and participate in immune regulation. Particularly, in BLCA and KIRP, PDLIM2 is mainly associated with immunosuppression in tumor tissues. Therefore, PDLIM2 is useful as a prognostic-related biomarker and is correlated with immune infiltrates in the BLCA and KIRP. Moreover, PDLIM2 may be a valuable therapeutic target for tumor immunotherapy in 33 cancer types.  www.nature.com/scientificreports/ information from TCGA Pan-CANCER (pan can) including 41 datasets and 12,591 samples to obtain data on disease-specific survival and DFI. After eliminating cases with insufficient or missing data for age and the overall survival time, 11,057 samples were included in the study (Supplementary Table S2).

Methods
Gene expression analysis. We used "wilcox.test" to analyze the differential expression of total PDLIM2 RNA in normal and tumor tissue samples from 33 cancer types in TCGA database and drew a box plot using "ggpubr. " To compensate for the lack of normal organization in TCGA database, we added the GTEx(Genotype-Tissue Expression) database 46 for analysis; this part of the analysis was mainly performed in the Gene Expression Profile of the GEPIA2 database 47 (Gene Expression Profiling Interactive Analysis, http:// gepia2. cancer-pku. cn/# analy sis). The main parameters were as follows: gene, PDLIM2; differential methods, analysis of variance; q-value cutoff, 0.05; matched normal data: match TCGA normal and GTEx data.
Survival clinical analysis. Univariate survival analysis was performed using the Kaplan-Meier survival" package 48 . According to the median PDLIM2 expression level, patients with cancer were divided into low-and high-expression groups. Cox regression analysis was performed only for overall survival (OS). Kaplan-Meier analysis was conducted to compare the survival (OS, DDS, and DFI) differences between the low-and highexpression groups. A P value less than 0.05 was considered as the threshold in the Kaplan-Meier analysis results. The "survival" package was used for survival analysis. We used the R package "forestplot" to draw the forest map in Cox survival analysis and R package "survivminer" to draw the Kaplan-Meier survival curve.
A stage plot using the tumor stage as a variable was plotted to analyze the relationship between the expression level of PDLIM2 and tumor metastasis in different cancers. R package "limma" 49 was used to analyze the differential expression of PDLIM2 in different clinical stages, and the box diagram was drawn using "ggpubr. " TIMER database analysis. TIMER 50 (Tumor Immune Estimation Resource, https:// cistr ome. shiny apps. io/ timer/) can use RNA-sequencing expression profile data to detect the infiltration of immune cells in tumor tissue. We applied the Gene of Immune Association module to explore the correlation between PDLIM2 expression and the abundance of immune infiltrates in BLCA, KIRP, and COAD. We imported "PDLIM2" in gene symbol; selected BLCA, KIRP, and COAD for cancer types; and chose B cells, CD4+ T cells, neutrophils, CD8+ T cells, macrophages, and dendritic cells for immune infiltrates.
Mutation analysis. TMB is a useful biomarker for predicting the prognosis and efficacy of immunotherapy.
We calculated the TMB of the 33 tumor types using mutation data. Perl scripts were written to extract genomic alterations, and calculate the TMB of the patients.TMB refers to the number of somatic mutations per million bases in tumor tissue 51 . MSI values were derived from TCGA database. We then analyzed the correlation between PDLIM2, TMB, and MSI and designed a radar map using the R-package "fmsb. " Correlation analysis was performed using Spearman's correlation.
Immunological correlation analysis. In the tumor microenvironment, immune cells and stromal cells are two main types of non-tumor components and were suggested to be valuable in the diagnosis and prognosis of tumors. We used R package "estimate" 52 to calculate the immune score and stromal score for BLCA and KIRP. Correlation analysis was performed using Spearman's correlation. The correlation between PDLIM2 and the immune score and stromal score was analyzed and plotted using "ggplot2, " "ggpubr, " and "ggextra. " Co-expression analysis. We evaluated common immune checkpoint and immune marker genes. The coexpression relationship between PDLIM2 and these genes was calculated using limma package. A heat map of PDLIM2 co-expression with immune checkpoint genes was drawn using the R packages "reshape2" and "RColorBrewer. " All graphics and data analyses were completed on the R platform (3.6.3 version; The R Project for Statistical Computing, Vienna, Austria).

Statement.
All methods were carried out in accordance with relevant guidelines and regulations.

Data availability
The datasets obtained from public database.