Prostate cancer associated 3 (PCA3) is a potential indicator of the pathological grades of prostate cancer

Abstract

Prostate cancer (PCa) is a prevalent malignancy of the male reproductive system and ranks as the second most common cancer in men worldwide. The prognosis and clinical management of PCa are closely linked to its pathological grade; however, the high heterogeneity of the disease poses significant challenges in these assessments. Prostate cancer associated 3 (PCA3), an upregulated long non-coding RNA (lncRNA) in PCa, has been established as a urinary biomarker for PCa detection. Nevertheless, its potential role in distinguishing pathological grades remains unclear. In this study, we investigated the association between PCA3 expression and PCa pathological grades. First, we conducted a pan-cancer bioinformatics analysis of PCA3 expression across 32 cancer types, confirming its high specificity for PCa. Next, we performed least absolute shrinkage and selection operator (LASSO) regression, identifying 18 GS-associated lncRNAs. A Gaussian distribution analysis revealed that PCA3 exhibited the strongest PCa-specific expression among these lncRNAs. Furthermore, logistic regression analysis demonstrated that PCA3 expression was significantly higher in poorly differentiated PCa than in well-differentiated PCa. To validate these findings, we constructed a PCA3 promoter-driven transcriptional reporter system, which exhibited elevated activity in lower-grade PCa, reinforcing the differential expression pattern of PCA3 across pathological grades. Our findings suggest that PCA3 could serve as a robust and clinically valuable biomarker for predicting PCa pathological grade.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-04571-7.

Introduction

Prostate cancer ranks as the second most prevalent cancer in men globally, with approximately 1.4 million new cases and 375,000 fatalities annually [1]. It is a complex and heterogeneous disease that can be classified into aggressive and nonaggressive PCa, high-grade and low-grade PCa, or early-onset PCa and indolent PCa [2]. Patients with low-grade PCa have been shown in large trials to be at low risk of cancer-specific death and, thus, can be monitored with active surveillance and regular screening. In contrast, high-grade PCa is associated with a high risk of metastases and recurrence, which contribute to the majority of deaths from PCa. For these patients, aggressive treatment options are recommended to improve the prognosis, although these come at the cost of impaired erectile function and urinary continence. Therefore, considering the important role of risk assessment in guiding treatment decisions and predicting clinical outcomes, it is critical to accurately determine the grading of PCa in vivo.

Currently, the Gleason scoring system, which is mainly based on morphological/histological/pathological changes found in biopsies, is the most frequently used tool for PCa grading in the clinic. A Gleason score of 5 or less refers to a low-grade cancer. When the Gleason score is equal to 6 or 7, it is considered a medium-grade cancer. A Gleason score of 8 or above indicates advanced prostate cancer with a high risk of aggressive spreading and a poor prognosis. However, it should be noted that the accuracy of prostate biopsy analysis is influenced by multiple factors, such as sampling range and operative technique, which makes it difficult to reflect the pathological characteristics of all the cancerous tissues [3]. Consequently, false-negative cancer diagnoses and substantial grade misclassification may occur at the time of biopsy [4, 5], leading to the overtreatment of patients with low-grade disease because of concern that a high-grade cancer may have been missed [6], and undertreatment of patients with aggressive disease that is missed on biopsy [7, 8]. Furthermore, over 90% of prostate cancers present with multifocal lesions which could evolve independently [9, 10]. The high heterogeneity and multifocality of PCa, combined with the relatively small dimensions of the prostate, further compounds the challenges of obtaining enough homogenous material for accurate pathological grading. Therefore, it is necessary to find novel precise biomarkers which can complement the golden test for the accurate pathological grading of PCa.

To date, several biomarkers have been introduced into the pathological grading process for PCa, including alpha-methylacyl-CoA racemase (AMACR), fatty acid synthase, and PCA3. Overexpression of AMACR in prostate cells is associated with a high Gleason score but occurs in only about 80% of PCa cases [11, 12]. Fatty acid synthase overexpression is indicative of aggressive tumors [13, 14], but its positive expression rate in PCa is less than 50% [15]. PCA3 (also known as DD3) is a non-coding RNA gene that is exclusively expressed in human prostate tissue [16]. Due to its restricted expression profile, PCA3 is regarded as the most specific and most promising biomarker for PCa diagnosis. In the clinic, PCA3 RNA has been utilized as a FDA-approved urinary biomarker for PCa screening [17, 18]. However, debate persists regarding whether the expression of PCA3 in tumors is correlated with the Gleason score [19, 20]. Previous studies have shown some correlation between the expression level of PCA3 and Gleason scores, which can be used to reflect the pathological level of PCa [21]. Some studies have reported no association between PCA3 and Gleason scores [22, 23]. Others have demonstrated higher levels of urinary PCA3 associated with more aggressive PCa based on the Gleason score [24, 25]. Therefore, the correlation between the expression of PCA3 in tumors and the Gleason score requires further exploration.

In this study, we investigated the variation in PCA3 expression across different pathological grades of PCa. First, we confirmed the prostate cancer specificity of PCA3 by means of a pan-cancer analysis and determined its expression level in prostate cancer using a differential analysis. Then, we identified 18 lncRNAs (including PCA3) related to the GS by LASSO regression and analyzed their distribution in normal and tumor tissues in a Gaussian distribution. Next, we constructed a logistic regression model to evaluate the correlation between the expression level of PCA3 and the GS. Given that previous research has indicated that the prostate cancer-specific expression of PCA3 is predominantly controlled by its promoter [26], a PCA3 promoter-driven transcriptional system was further constructed to verify the differential expression profile of PCA3 in PCa with varying pathological grades. We hypothesized that PCA3 and its promoter might be a potential biomarker that can help predict the grading of PCa and subsequently assist in treatment decision-making. In this context, a transcriptional system based on the PCA3 promoter could provide a promising and non-invasive approach for evaluating the risk stratification of prostate cancer in vivo by determining the expression of target genes.

Results

Pan-cancer expression landscape of PCA3 and differentially expressed LncRNAs in normal and tumor cells in TCGA-PRAD screening

First, we analyzed the expression of PCA3 lncRNA in 32 types of common cancers from the TCGA database, including ACC, BLCA, BRCA, CESC, CHOL, COAD, DLBC, ESCA, GBM, HNSC, KIRP, LGG, LIHC, LUAD, LUSC, OV, PAAD, PRAD, READ, SKCM, STAD, TGCT, THCA, THYM, UCEC, and UCS. In PRAD, the expression of PCA3 was significantly higher in tumors than in adjacent normal tissues, while other tissues displayed little PCA3 expression (Fig. 1A). In TCGA-PRAD, PCA3 was overexpressed in prostate cancer tissues compared to normal tissues (Fig. 1B). In addition, differentially expressed lncRNAs were identified in normal and tumor tissues from TCGA-PRAD, with 226 upregulated (p.adj < 0.001,log2FoldChange > 1) and 2127 downregulated (p.adj < 0.001,log2FoldChange < 1). As shown in the volcano plot showing the differentially expressed lncRNAs (Fig. 1C), levels of PCA3 were found to be 1.33-fold (log2) higher compared with normal tissues. In particular, the baseMean of PCA3 was the highest among all differentially expressed lncRNAs, reaching 19246.53, which suggests that its transcriptional elements were plentiful (Fig. 1D).

Fig. 1.

The specificity and high expression of PCA3 in prostate cancer. A PCA3 expression in tumors and the adjacent normal tissues in various TCGA cancer types. B Boxplot showing the difference in expression between normal and tumor tissues of the prostate. C Volcano map showing the results of lncRNAs’ differential expression in prostate cancer and normal prostate tissue (p.adj < 0.001, |log2(fold change)| > 1); PCA3 is represented with a yellow dot. D M-versus-A plot showing the expression difference among all differential genes (padj < 0.001, |logFC| > 1); PCA3 is represented with a yellow dot. Data presented as mean + SD. (Wilcox test; ns, p ≥ 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.)

Construction and validation of the LncRNA signature

Usually, a Gleason score ranging from 5 to 7 indicate a moderately differentiated adenocarcinoma, while scores greater than 7 indicate a poorly differentiated adenocarcinoma. In the present study, the samples were divided into two groups according to the Gleason score, namely a low group (Gleason score ≤ 7) and a high group (Gleason score > 7). Next, the prostate cancer samples in TCGA were randomized into two groups, with a total of 386 samples in the training set and 97 samples in the test set. To identify candidate lncRNAs that could be used to predict the Gleason group from 2353 differentially expressed lncRNAs, we performed a LASSO regression analysis and screened out a total of 18 lncRNAs, including PCA3. It is worth noting that the model was constructed using a 10-fold cross-validation and the minimum lambda value was 0.074 (Fig. 2A, B). Additionally, a receiver operating characteristic (ROC) curve was used to evaluate how accurate the expression level of those genes was in predicting Gleason grouping, with AUC values of 0.833 and 0.792 in the training set and the test set, respectively (Fig. 2C and E). Interestingly, only the expression of PCA3 and PCAT14 exhibited two normal distributions corresponding to tumor and normal tissues, respectively, indicating their ability to clearly distinguish tumor tissues from normal tissues. Compared with PCAT14, the expression of PCA3 in normal tissues was far lower than that in tumor tissues, making it more in line with our expectations (Fig. 2F).

Fig. 2.

Construction and validation of the lncRNA signature: (A) 10-fold cross-validation LASSO logistic regression and extraction of non-zero coefficients. B, C ROC curve of the logistic regression model verified in the training set (AUC = 0.850). D ROC curve of the logistic regression model verified in the test set (AUC = 0.803). E Comparison of the ROC curve of the logistic regression model in the training set verification and test set verification. F Ridge plot showing the expression distribution of 18 signatures

Association between PCA3 expression and Gleason group

To examine the association between PCA3 expression level and Gleason group, we conducted both univariate and multivariate logistic regression analyses. As before, the cancer samples in TCGA-PRAD were randomly divided into a training set (n = 386) and a test set (n = 97). In the logistic regression analysis of PCA3, the ROC curve showed that PCA3 performed well in independently predicting the Gleason group, of which the AUC value reached 0.716 (Fig. 3A). However, the AUC values of the ROC curves for serum PSA, age, and tumor size were only 0.592, 0.580, and 0.594, respectively (Supplementary Table S1). The calibration curve showed that in the univariate logistic regression model, the larger the PCA3 value, the more likely it is to predict a low Gleason group (Fig. 3B). The data shows that pathological grade of prostate cancer is related to expression levels of PCA3(OR = 0.79, 95%CI:0.72–0.86), Age(OR = 1.06, 95%CI:1.02–1.09), Serum PSA(OR = 1.11, 95%CI:1.02–1.2) and T classification (OR = 1.45, 95%CI:0.86–2.47 for T2 vs. T1; OR = 8.30, 95%CI:3.47–19.82 for T3 vs. T1) (all p < 0.05) and the Estimate value are − 0.24(PCA3),0.05(Age),0.10(Serum PSA),0.37(T2 vs. T1) and 2.12(T3 vs. T1) respectively in univariate logistic regression(Supplementary Table S1). Next, we verified the above model on GEO datasets (GSE70768 and GSE59745), with AUC values of 0.716 and 0.719, respectively, indicating that PCA3 is stable in predicting the Gleason group (Fig. 3C, D).

Fig. 3.

Association between PCA3 expression and Gleason group. A ROC curve showing the effectiveness of PCA3 in predicting Gleason group in univariate logistic regression (AUC = 0.716). B Calibration curve for univariate logistic regression model. C, D Validation of the accuracy of the univariate logistic regression model through external datasets GSE70768 (AUC = 0.716) and GSE59745 (AUC = 0.719). E PCA3 combined with tumor size, serum PSA, and age to construct the ROC curve of the multivariate logistic regression model (AUC = 0.842). F Calibration curve for multivariate logistic regression model

We further incorporated tumor size, age, and serum PSA variables into the analysis of the multivariate logistic regression model. After removing samples that did not meet the inclusion criteria, there were 284 samples in the training set and 70 samples in the test set. Encouragingly, the predictive ability of the model was significantly improved, and the ROC value reached 0.842 (Fig. 3E). In univariate logistic regression the pathological grade of prostate cancer is related to expression levels of PCA3(OR = 0.81, 95%CI:0.73–0.91), Age(OR = 1.06, 95%CI:1.02–1.1) and T classification (OR = 1.17, 95%CI:0.67–2.05 for T2 vs. T1; OR = 5.24, 95%CI:2.16–12.74 for T3 vs. T1) (all p < 0.05) and the Estimate value are − 0.21(PCA3),0.06(Age), 0.16(T2 vs. T1) and 1.66(T3 vs. T1) respectively(Supplementary Table S2). Furthermore, the odds ratio (OR) of age and tumor size in T3 were both greater than 1, indicating that they were promoting factors for the occurrence of positive events (corresponding to the high group), while PCA3 expression with an OR less than 1 was a hindering factor. The calibration curves showed the same trend in PCA3’s predictive power, as demonstrated by the univariate logistic regression model (Fig. 3F).

Construction and verification of adenovirus transcriptional system based on the PCA3 promoter

According to the analysis in the previous section, there is a negative correlation between the expression level of PCA3 and the PCa’s Gleason score, indicating significant variations in the expression levels of PCA3 and its transcriptional elements across different pathological grades. Based on these findings, we hypothesized that the tissue specificity of the PCA3 promoter transcriptional elements in PCa can be utilized to reflect the pathological grade of PCa. Subsequently, we developed a transcriptional system utilizing the PCA3 promoter and validated it in PCa cell lines representing varying pathological grades.

First, we constructed a PCA3 promoter-based gene probe via the adenovirus. In brief, the PCA3 promoter was successfully inserted into the upstream of the target gene and green fluorescent protein (GFP) reporter expression cassettes to achieve overexpression in PCa. For this study, we chose the transferrin receptor (TfR), a broad-spectrum target expressed on cancer cell membranes, as the target gene to test the overexpression effect of the adenovirus we constructed [27]. Furthermore, a woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) enhancer was added to the adenovirus to further enhance expression efficiency. The adenoviral vector Ad.PPCA3-TfR-WPRE-PCMV-GFP (Ad.PCA3 for short) pattern is shown in Fig. 4A. We next transfected two prostate cancer cell lines, PC3 and 22Rv1, with Ad.PCA3 at different multiplicities of infection (MOIs). As shown by the results, the expression of GFP began to stabilize when the MOI was equal to 50 (Fig. 4B).

Fig. 4.

Construction and verification of an adenovirus transcriptional system based on the PCA3 promoter. A Schematic diagram of adenovirus based on the PCA3 promoter; TfR was used as the target gene for overexpression, and GFP was used to evaluate the expression effect. B Green fluorescence in gradient transfection among two prostate cancer cell lines, PC3 and 22Rv1

Validation of overexpression of PCA3 Promoter-based adenovirus in different prostate cancer cell lines

To demonstrate the ability of PCAs to predict different types of PCa, we selected three prostate cancer cell lines, DU145, PC3, and LNCap, as the verification objects. DU145 and PC3 were isolated from brain metastases and bone metastases, respectively, with a lower degree of differentiation, while LNCap retained the differentiation characteristics of early PCa [28]. Notably, an adenovirus vector package (Ad.Null) lacking a recombination gene was considered as a negative control.

First, we performed preliminary validation by qRT-PCR. After adenoviral transfection of DU145, PC3, and LNCap, the alteration in the mRNA level of TfR was verified. In the DU145 and PC3 cell lines, the mRNA level of TfR increased only slightly, while LNCap increased the most, by nearly 6 times (Fig. 5A). The findings suggest that the impact of the PCA3 promoter-based recombinant adenovirus on overexpression varies across prostate cancer cell lines with varying levels of differentiation.

Fig. 5.

Validation of overexpression of PCA3 promoter-based adenovirus in different types of prostate cancer cell lines. A RT-qPCR analysis of TfR expression in DU145, PC3, and LNCap cells transfected with PCA3 promoter-based adenovirus. B Immunofluorescence of TfR expression in DU145, PC3, and LNCap cells transfected with PCA3 promoter-based adenovirus. C Quantification of immunofluorescence. Data presented as mean + SD. (T test; ns, p ≥ 0.05; *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001.)

In order to observe these changes more intuitively, we performed an immunofluorescence analysis to investigate the overexpression of target genes in prostate cancer cell lines with varying differentiation levels. Consistent with the results of the mRNA expression, the fluorescence overexpression of LNCap increased more significantly than that of DU145 and PC3, and expression was primarily localized to the cell membrane (Fig. 5B). We also quantified the mean fluorescence intensity of the immunofluorescence, the results are shown in Fig. 5C.

PCA3 epigenetic modification and transcription factor binding prediction

To further explore the mechanism of action of the PCA3 promoter, we obtained the PCA3 promoter sequence from the NCBI website. As shown in Supplementary Figure S1, a high expression peak of H3K27me3 was observed at the PCA3 promoter in PC3 cells, which is consistent with the lower expression level of PCA3 in PC3 cells compared to LNCaP cells (Supplementary Figure S2). In addition, analysis was performed using the PCATools platform [29]. Across independent experiments, we observed clear enrichment of transcription-associated marks, including H3K27ac and H3K4me1/2, in the regions upstream and downstream of the PCA3 locus, together with peaks indicative of open chromatin. In contrast, repressive histone marks such as H3K9me2/3 and H3K27me3 did not show strong signals at this locus. These chromatin features are consistent with a transcriptionally permissive state at the PCA3 locus and provide supportive, although indirect, evidence relevant to the question of PCA3 expression (Supplementary Figure S3). Next, we used the JASPAR database to predict the binding sites and binding sites of the PCA3 promoter and transcription factors. The PCA3 promoter can bind to multiple transcription factors, such as AR and TFAP4. The specific binding sites are shown in Figs. 6A, B. At the same time, we also analyzed the expression of AR and TFAP4 in different prostate cancer cell lines using the CCLE database. As shown in Figs. 6C, D, AR and TFAP4 expression in LNCaP cells was significantly higher than that in DU145 and PC3 cells, which may explain the highest efficiency of the PCA3 promoter-based overexpression adenovirus in LNCaP.

Fig. 6.

Transcription factor binding prediction of PCA3. A AR binding site prediction in the PCA3 promoter region using JASPAR. B TFAP4 binding site prediction in the PCA3 promoter region using JASPAR. C AR expression in various prostate cancer cell lines using CCLE database. D TFAP4 expression in various prostate cancer cell lines using CCLE database

Discussion

The pathological grade of PCa plays a pivotal role in determining the appropriate treatment and prognosis [3], as well as in the development of targeted imaging and precise gene therapy, which differ depending on the grade. Hence, conducting in-depth investigations into the differences in expression and specificity of PCA3 across various pathological grades of PCa has become an especially crucial area of research.

Through our pan-cancer analysis, we observed that PCA3 was highly expressed only in PCa tumors and not in other tissues, which is consistent with previous research conclusions [30]. This outcome implies that PCA3 holds significant potential as a biospecific marker. Moreover, our analysis revealed conspicuous upregulation of PCA3 expression in prostate cancer tissues, suggesting that its transcriptional elements are abundantly present in PCa. Taken together, these findings signify promising potential for the application of the PCA3 promoter.

Based on these results, we decided to further explore the role of PCA3 in PCa. We identified 18 effective lncRNAs, including PCA3, from a pool of DEGs in PCa that are capable of predicting the Gleason group of PCa. Among the 18 lncRNAs that contributed to the prediction of Gleason group, the expression distribution showed that, in addition to PCA3, other lncRNAs were also highly expressed in normal prostate tissue, or there was little difference between normal prostate tissue and PCa. Some studies have proposed that there is no association between PCA3 and Gleason score, and the main reason for this may be that the PCA3 was derived from non-tumor tissue [20].

Next, we used univariate and multivariate logistic regressions to construct models for the PCA3 prediction of Gleason group. In the univariate logistic regression model, the AUC value of the ROC curve reached 0.716, which was further validated using external datasets, confirming the robustness of the model. Upon adding tumor volume, age, and serum PSA to construct a multivariate logistic regression model, the predictive performance of the model was significantly enhanced, resulting in an AUC value of 0.842. Notably, both tumor size and age proved to be valuable factors in constructing the model. The above results demonstrated that the expression level of PCA3 in tumor tissue was negatively associated with the pathological grade of PCa.

The lncRNA PCA3, which is specifically expressed in PCa, exhibits excellent performance in predicting the pathological grade of PCa. Given this intriguing discovery, we contemplated whether it would be feasible to devise a transcriptional system for the assessment of pathological grades in PCa. Drawing upon these findings, we subsequently developed an adenovirus utilizing the PCA3 promoter to examine its overexpression efficacy across various types of PCa. To assess the overexpression effect of the adenovirus utilizing the PCA3 promoter, we selected three distinct prostate cancer cell lines, each representing different levels of differentiation: DU145, PC3, and LNCap. Validation of the overexpression effect was carried out through qRT-PCR, and immunofluorescence analyses. The results of these experiments were highly consistent, demonstrating that the overexpression effect was less pronounced in poorly differentiated prostate cancer cell lines as compared to highly differentiated prostate cancer cell lines, with a roughly linear relationship between differentiation level and overexpression efficacy. We demonstrated that the transcription system based on the PCA3 promoter has the ability to differentially express genes in PCa of different pathological grades and can effectively reflect the pathological grades of PCa. Furthermore, we found that the PCA3 promoter has multiple epigenetic modification sites and can bind to multiple transcription factors, which partly explains the different efficiency of the PCA3 promoter-based overexpression adenovirus in PCa cell lines. For example, in PC3 cells, there are high H3K27me3 peaks on the PCA3 promoter, a modification that inhibits gene expression; however, in LNCaP cells, the upstream and downstream regions of the PCA3 gene locus are enriched with significant transcription-related markers, including H3K27ac and H3K4me1/2, along with peaks indicating an open chromatin state.

By capitalizing on the high efficiency of the PCA3 promoter and the differential expression of target genes in different pathological grades of PCa, we may be able to achieve more precise gene therapy [31, 32] or specific targeted imaging in the future [33, 34]. In the field of targeted imaging, this approach has the potential to not only address the limitations of inadequate targets and poor specificity [35, 36] but also indirectly indicate the pathological grade of PCa via the expression of overexpressed targets.

Our study demonstrated that PCA3 promoter-driven gene expression differs in prostate cancers of different pathological grades. By combining the detection of target genes, such as protein quantification, fluorescence detection, and visualization of targeted imaging probes, it is possible to evaluate the classification of prostate cancer in vivo. A limitation of this study is the lack of exploration of the specific mechanism underlying PCA3 expression differences across the various pathological grades of PCa. Most of the data came from public databases. To avoid the analysis bias caused by the current retrospective study, we will continue to conduct prospective studies in the future. In addition, all PCA3 expression analysis was based on bulk-RNA seq, lacking the validation of PCA3 expression in living tumors. Finally, we have not verified the specificity and actual expression effect of the transcriptional system in vivo, and we need to continue to improve it in the future. Additionally, further investigation is needed to elucidate the specific biological functions of PCA3 in PCa.

Materials and methods

Data collection and processing

RNA-seq data were collected from The Cancer Genome Atlas (TCGA), including 32 cancer types (adrenocortical carcinoma (ACC), bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), cervical squamous cell carcinoma (CESC), cholangiocarcinoma (CHOL), colon adenocarcinoma (COAD), lymphoid neoplasm diffuse large B cell lymphoma (DLBC), esophageal carcinoma (ESCA), glioblastoma (GBM), brain lower grade glioma (LGG), head and neck squamous cell carcinoma (HNSC), kidney chromophobe (KICH), kidney renal clear cell carcinoma (KIRC), kidney renal papillary cell carcinoma (KIRP), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), mesothelioma (MESO), ovarian serous cystadenocarcinoma (OV), pancreatic adenocarcinoma (PAAD), pheochromocytoma and paraganglioma (PCPG), prostate adenocarcinoma (PRAD), rectum adenocarcinoma (READ), sarcoma (SARC), skin cutaneous melanoma (SKCM), stomach adenocarcinoma (STAD), testicular germ cell tumors (TGCT), thyroid carcinoma (THCA), thymoma (THYM), uterine corpus endometrial carcinoma (UCEC), uterine carcinosarcoma (UCS), and uveal melanoma (UVM)). Microarray human series were collected from the Gene Expression Omnibus (GEO) database using a keyword search.

We downloaded the RNA-seq data of 32 cancers from the official website of the TCGA database (https://portal.gdc.cancer.gov/). GSE59745 (Platform: GPL5175) and GSE70768 (Platform: GPL10558) were downloaded from GEO and used for validating the univariate logistic regression model. The programming language R (https://www.r-project.org/) was used to write scripts for data sorting and analysis. The data were standardized to a log scale. The function “normalizeBetweenArrays” in “Limma” R package was utilized to further normalize the log-scaled GEO data.

Identification of differentially expressed LncRNAs

The DEseq2 package was utilized to conduct a differential expression analysis in R. The lncRNAs expressed differentially in normal and tumor groups were identified as those having a threshold value of p.adj < 0.001 and |log fold change (FC)| > 1. A total of 2353 differentially expressed lncRNAs were screened out. The volcano plots and MA plots were created by R packages ggplot2 and ggrepel.

Construction of the LncRNA signature

We divided the samples into a low Gleason group (Gleason score ≤ 7) and a high Gleason group (Gleason score > 7) according to the clinical information. We used a 10-fold cross-validation LASSO logistic regression model to determine the minimum lambda value and screened 18 lncRNAs to determine which Gleason group they were associated with. Then, we randomly divided the samples into a training set and a test set, built a logistic regression model on the training set, and verified it on the test set. This was all carried out using the R package glmnet. A ridge plot was used to display the Gaussian distribution of normal and tumor tissues in these lncRNAs.

Logistic regression

We analyzed PCA3 alone and in combination with tumor size, age, and serum PSA to predict the Gleason group using the glm function in R and implemented a logistic regression analysis to obtain the true positive (TP), false positive (FP), true negative (TN), and false negative (FN) results. The training set and test set were randomly split according to K-fold (K = 5). The ROC curves were plotted based on the sensitivity and specificity. We used GEO datasets as external datasets to verify the stability of the model.

Construction of Ad.PPCA3-TfR-WPRE-PCMV-GFP

The plasmids carrying the minimal PCA3 promoter (AF279290 nt309-522) and TfR gene were gifts from Dr. Xinyuan Liu, Xinyuan Institute of Medicine and Biotechnology, and Dr. Hong Shan, Third Affiliated Hospital of Zhongshan University, respectively. The pLV-shRNA plasmid harboring the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) and the adenoviral shuttle vector pHBAdU6-GFP were supplied by Hanbio (Shanghai) Biotechnology Co., Ltd. First, the sequence of the PCA3 promoter, TfR fragment, and WPRE regulatory element were amplified via PCR reaction with forward and reverse primers that are listed in Supplementary Table S3. The reaction conditions were as follows: pre-denaturing at 95 °C for 2 min, followed by 35 cycles of amplifications by denaturing at 94 °C for 20 s, annealing at 55 °C for 30 s, and extension at 72 °C for 30 s. After a final extension at 72 °C for 2 min, DNA sequencing was utilized to analyze the amplified products. Following sequencing, the obtained fragments were sequentially inserted into the pHBAd-U6-GFP plasmid in order to generate the recombinant shuttle plasmid, namely pHBAd-PPCA3-TfR-WPRE-PCMV-GFP, and the new shuttle vector pAdeasy-PPCA3-TfR-WPRE-PCMV-GFP was generated via the homologous recombination of the adenoviral shuttle vector pHBAd-PPCA3-TfR-WPRE-PCMV-GFP and the adenovirus backbone plasmid pBHGIox(delta)E1, 3Cre. After being packaged in HEK-293 cells, the construction of the recombinant adenovirus vehicles at the DNA level was confirmed by PCR. After authentication procedures, the constructed adenovirus was amplified and purified via double cesium chloride density gradient ultracentrifugation and dialyzed in PBS with 10% glycerol solution. A plaque assay was used to measure the titers of the recombinant adenovirus. Viral aliquots were stored at − 80 °C for further use.

Cell culture

All cells were purchased from Procell Life Science & Technology Co., Ltd. Cell lines were cultured in a humidified incubator at 37 °C and 5% CO2 conditions. The medium was RPMI-1640 medium (Procell, PM150110) supplemented with 10% fetal bovine serum (Procell, 164210-50).

Quantitative RT-PCR analysis

The expression of mRNA was detected by means of an RT-PCR. RNA was isolated from cells using Trizol reagent (Invitrogen, China) and reverse transcribed using the Reverse Transcription System (Roche, China) according to the manufacturer’s instructions. After cDNA extraction, PCR was performed using the Applied Biosystems 7900 Real Time PCR System (Thermo Scientific, China) and SYBR Green PCR Master Mix (Roche) according to the manufacturer’s instructions. GAPDH was used as an internal control. All primers used are listed in Supplementary Table S3.

Immunofluorescence

To create a subculture, cells were inoculated into the culture dish with treated coverslips placed in advance; the coverslips were removed after the cells had nearly grown into a monolayer. The cells were then washed twice with PBS. Samples were fixed with 4% PFA overnight at 4 °C. The fixed specimens were washed three times with PBS, for 5 min each time, then the blocking solution was added dropwise to completely cover the tissue and allowed to react at room temperature for 30 min. The blocking solution was shaken off the slices and the diluted anti-TfR (1:200, Invitrogen, 13-6800) solution was dripped on directly to cover the tissue, which was then placed in a humid box overnight at 4 °C. The incubated specimen was dipped and washed four times in TBST, for 3 min each time. A sufficient quantity of secondary antibody solution was added to completely cover the tissue, which was then incubated at a constant temperature of 37 °C for 1 h (from this step onwards, it was protected from light). The specimen was dipped in TBST four times, for 3 min each time. DAPI staining solution was added dropwise and the specimen was incubated at room temperature for 2 min in the dark. The specimen was dipped in TBST four times, for 3 min each time. Immunofluorescence was quantified by means of ImageJ (https://onlinelibrary.wiley.com/doi/10.1002/ar.22641).

Statistical analysis

R software(x86_64, R version 4.2.1 ) was used for all statistical analyses. A probability value of p < 0.05 was taken to establish statistical significance. We used LASSO regression to screen out lncRNAs related to GS. Univariate and multivariate logistic regression were used to determine the association between PCA3 expression and Gleason Group. Student’s t-test was used to determine the significance of differences between two groups. A p < 0.05 was considered statistically significant.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization, Y.N and Y.Z.; methodology, Y.G. and W.C.; software, Y.G.; validation, W.C.; formal analysis, J.Z. and X.W.; investigation, Y.C and L.D.; resources, Y.M. and Y.B.; writing—original draft preparation, Y.G.; writing—review and editing, Y.Z. and Y.N.; funding acquisition, Y.Z. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by Scientific and Technological Research Program of Tianjin Municipal Education Commission (No. 2019ZD025).

Data availability

Publicly available datasets were analyzed in this study. The tumor cell line data were downloaded from the CCLE database (https://portals.broadinstitute.org/ccle). RNAseq data of 32 cancers were obtained from the data portal of TCGA database (https://gdc.cancer.gov/support/gdc-webinars/tcga-resources-available-gdc). The external validation cohort was downloaded from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE59745 and GSE70768. Other raw data generated in the study are available from the corresponding author on reasonable request.

Declarations

Consent for publication

Not application.

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

Not applicable.