Overexpression of a subset of long intergenic noncoding RNAs in uterine serous carcinoma predicts poor prognosis

Abstract

The evaluation and prediction of uterine serous carcinoma (USC), a type of endometrial cancer that is more severe than endometrioid adenocarcinoma, remain challenging. Long noncoding RNAs (lncRNAs) are frequently dysregulated in human cancers. This study assessed the expression patterns and prognostic values of long intergenic noncoding RNAs (lincRNAs) in USC. RNA sequencing, copy number variation (CNV), and clinical data from The Cancer Genome Atlas were used to investigate various lncRNAs in endometrial cancer. LincRNAs, a major subclass of lncRNAs, exhibit specific expression patterns modulated by CNVs and act as predictors of poor prognosis, survival, and recurrence in USC. Functional analyses were conducted to investigate the roles of lncRNAs in USC. Finally, the expression of these lincRNAs was verified in 32 pairs of USCs collected from the hospital over 3 years. A series of lincRNAs were found to be specifically expressed in USC compared with other lncRNAs and regulated by CNV. Moreover, these specific upregulated lincRNAs, particularly ENSG00000281406, ENSG00000226791, ENSG00000269903, and ENSG00000204277, demonstrated poor prognoses for survival and recurrence in USC. Functionally, our analysis showed that ENSG00000281406 positively correlated with the Wnt signaling pathway, whereas ENSG00000226791, ENSG00000269903, and ENSG00000204277 negatively correlated with the T-cell receptor signaling pathway. Importantly, we confirmed that ENSG00000204277 negatively correlated with CD8⁺ T-cell immune infiltration in USC. Our results highlight that these lincRNAs can serve as new biomarkers for the prognostic prediction of USC. In particular, ENSG00000204277 may be used as a therapeutic target for USC.

1. Introduction

Endometrial cancer (EC) is the second most common cancer of the female genital system, and its incidence is increasing with the growing prevalence of obesity.^[1–3] EC can be classified as endometrioid, serous, or carcinosarcoma according to the histopathological subtype. Approximately ≥85% of newly diagnosed cases were classified as endometrioid EC, with only 3% to 10% classified as uterine serous carcinomas (USCs). Although USC accounts for a lower proportion of EC cases, it causes a disproportionate number of deaths owing to its aggressive behavior.^[4,5] Accordingly, it is necessary to identify the EC subtype and discover ways to improve the prognosis of USC.^[6] However, current prognostic markers and therapeutic targets remain limited during the diagnosis and treatment of USC; therefore, there is an urgent need to search for molecular-level indicators specifically targeting the aggressive subtype.

Human coding genes account for <2% of the total genome. However, 70% of the human genome is transcribed into RNA, yielding thousands of noncoding RNAs.^[7] Broad-term long noncoding RNAs (lncRNAs) are transcripts >200 nt in length that do not contain a protein-coding sequence.^[8] Based on genome positioning, lncRNA genes can be further grouped into subclasses, including long intergenic noncoding RNAs (lincRNAs), natural antisense transcripts, sense intronic RNAs, sense overlapping RNAs, and processed transcripts.^[9] Among these subclasses, lincRNAs regulate neighboring protein-coding genes and represent the largest and most extensively studied group. LincRNAs are transcribed from genomic regions independent of protein-coding genes and function in trans, ensuring that they regulate distant genes and broader cellular networks. Additionally, lincRNAs generally exhibit high sequence conservation and tissue-specific expression patterns, making them attractive candidates. LincRNAs retain exon–intron structure and are encoded in highly regulated regions of the genome.^[10] They can serve as transcriptional regulators, improve gene transcription, and act as decoys to bind proteins or miRNAs.^[7,11] In recent years, many studies have reported that lincRNAs play tumor-suppressive or tumor-promoting roles. For instance, lincRNAs correlated with EC include FER1L4, GAS5, and Linc-RoR.^[12] To the best of our knowledge, no study has specifically illustrated the relationship between lincRNAs and EC subtypes, particularly USC, nor has any lincRNA been regarded as a subtype-specific prognostic marker for the aggressive form of EC.

Therefore, to fill this gap, this study aimed to identify lincRNA-based prognostic markers for USC and uncover the mechanisms underlying the function of lincRNAs in USC. Clinically, the findings of this study address the urgent need for subtype-specific biomarkers to improve risk stratification, guide treatment, and potentially target the tumor immune microenvironment in this aggressive EC subtype.

2. Materials and methods

2.1. Raw data

RNA sequencing, copy number variation (CNV), and clinical data from The Cancer Genome Atlas (TCGA) database were downloaded from https://portal.gdc.cancer.gov/. RNA-Seq data were presented as fragments per kilobase million (FPKM). Genotype classification was based on the Ensembl dataset (http://asia.ensembl.org/index.html). RNA-Seq expression data were downloaded from the TCGA/GDC portal as FPKM values and log2-transformed as log2(FPKM + 1) before downstream analyses. As all expression profiles were generated using the same TCGA/GDC processing pipeline, no additional batch correction was performed. Count-based methods are generally preferred for formal differential expression analysis; therefore, the FPKM-based analyses in this study were mainly used for comparative expression profiling and prognostic modeling within a uniformly processed cohort.

2.2. Enrichment analysis

Gene set enrichment analysis (GSEA) software was used to identify the enrichment degree of lincRNAs in EC and USC based on their expression levels.^[13] For functional enrichment analysis, mRNAs were arranged from largest to smallest based on their correlation with target lincRNAs. The correlation level is presented as the Pearson correlation coefficient (R value). The listed mRNAs and corresponding R values were analyzed for gene ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Reactome enrichment using GSEA in ClusterProfiler.^[14,15] GSEA was performed using the ClusterProfiler package with 1000 permutations; gene sets with an adjusted P-value or false discovery rate (FDR) q-value < 0.05 were considered significantly enriched.

2.3. Expression level of lincRNAs and their relationship with CD8 expression

Over a 3-year period, 32 pairs of endometrial serous carcinoma and paracancerous tissues were collected from the Second Affiliated Hospital of South China University of Technology and the First Clinical Medical College (affiliated with the Medical College of Three Gorges University). All experimental procedures were approved by the Ethics Committees of both institutions, and written informed consent was obtained from all patients prior to the study. The patients’ clinical information is presented in Table 1. Total tissue mRNA was extracted using the TRIzol reagent. After testing the RNA concentration, purity, and integrity, a reverse transcriptase kit was used to convert the RNA to cDNA (Takara Bio; Kitayama, Higashiyama-ku, Kyoto, Japan). Prior to quantitative real-time polymerase chain reaction (qRT-PCR), GAPDH, β-actin, and 18S rRNA were identified as candidate housekeeping genes. GAPDH showed the most stable expression among all samples and was therefore selected as an endogenous control. All reactions were performed in triplicate, and the average cycle threshold (Ct) values were used for the analysis; ΔCt values were calculated by normalizing target gene Ct values to GAPDH, and relative expression levels were calculated using the 2^−ΔΔCt method. Target gene expression was detected using real-time fluorescence quantitative PCR amplification (Applied Biosystems 7300; Applied Biosystems, Waltham, MA). The analysis of TCGA data revealed a correlation between ENSG00000204277 and CD8 expression, which was verified in 32 cases of endometrial serous carcinoma. The lincRNA and CD8A mRNA primers used are listed in Table 2. Data were analyzed using Step One Software (v2.3; Applied Biosystems).

Table 1.

The patients’ clinical information (n = 32) in this study.

No.	Age	Menstrual history	Maternity history	CA125	CEA	AFP	TNM	Stage
1	63	4–6/28–32 d	G2P1	5.2	2.87	3.2	T1N0M0	ⅠC
2	71	4–7/26–28 d	G2P2	8.2	7.26	2.9	T1N0M0	ⅠC
3	68	3–4/30 d	G3P1	10.6	1.83	0.98	T1N1M0	ⅢC
4	83	6–7/28–32 d	G5P3	40.7	0.8	1.2	T2N1M0	ⅢC
5	52	3–5/28–30 d	G4P1	5.1	0.8	3.1	T1N0M0	ⅠC
6	73	3–5/30 d	G1P0	4.6	2.8	1.5	T2N1M0	ⅢC
7	64	2–3/30–60 d	G0P0	2.5	1.96	2.4	T1N0M0	ⅠC
8	69	5–6/25–28 d	G4P1	10.1	0.7	2.6	T2N0M0	ⅡC
9	72	6–7/28–32 d	G5P3	8.9	0.8	3.2	T3N0M0	ⅢA
10	66	3–4/35–45 d	G1P1	3.4	1.57	2.2	T1N0M0	ⅠC
11	74	4–6/28 d	G4P1	8.1	2.7	1.4	T2N0M0	ⅡC
12	75	4–6/32 d	G2P2	7.6	2.7	1.5	T2N1M0	ⅢC
13	60	4–6/28 d	G3P1	8.5	5.87	3.6	T1N0M0	ⅠC
14	72	5–6/24–30 d	G3P2	8.9	1.4	2.5	T1N0M0	ⅠC
15	63	3–4/30–37 d	G3P1	58.1	1.24	1.08	T3N0M0	ⅢB
16	54	7/30–32 d	G4P2	10.5	1.8	3.5	T2N1M0	ⅢC
17	42	4–5/25–35 d	G1P1	4.2	2.5	1.3	T1N0M0	ⅠC
18	52	4–6/32 d	G3P2	12.5	2.7	6.7	T2N1M0	ⅢC
19	57	3–5/30–60 d	G0P0	1.3	4.06	4.1	T1N0M0	ⅠC
20	69	5–6/25–28 d	G4P1	10.1	0.7	2.6	T2N0M0	ⅡC
21	54	3–6/28 d	G3P2	5.1	2.9	1.3	T2N0M0	ⅡC
22	72	4–6/30 d	G2P1	5.8	2.6	1.9	T2N1M0	ⅢC
23	60	3–4/30 d	G3P1	9.5	1.01	1.1	T1N1M0	ⅢC
24	43	5–7/32 d	G5P2	10.1	3.2	4.1	T2N1M0	ⅢC
25	82	4–5/28–35 d	G4P2	4.7	5.7	3.9	T1N0M0	ⅠC
26	69	3–5/32–36 d	G1P0	50.4	0.8	1.9	T2N1M0	ⅢC
27	68	3–5/30–60 d	G0P0	43.7	0.9	4.8	T1N0M0	ⅠC
28	57	5/26–28 d	G4P1	1.1	1.8	2.5	T2N0M0	ⅡC
29	76	5/32 d	G6P2	10.7	3.4	3.9	T3N0M0	ⅢB
30	69	3–4/35–45 d	G2P1	3.7	1.54	2.7	T1N0M0	ⅠC
31	70	4–6/28–30 d	G4P0	4.7	2.1	1.6	T2N1M0	ⅢC
32	65	4–6/32–34 d	G2P2	10.7	2.1	7.4	T2N1M0	ⅢC

AFP = alpha fetoprotein, CA125 = carbohydrate antigen 125, CEA = carcinoembryonic antigen, TNM = tumor node metastasis.

Table 2.

The information of lincRNA and CD8 mRNA primers in this study.

No.	Gene	Forward primer	Reverse primer
1	ENSG00000281406	GCTTAGCATCTAAACTATCTCTGGT	GTCTCCCGCTCTCTGTGAAC
2	ENSG00000226791	GGGCTGGCTGTTGGATGAAT	GACACCTGATGGGCCACTAC
3	ENSG00000269903	CCAAGGACGAAGAGCCTGAG	GCATTCGAATTCGGGTCACG
4	ENSG00000204277	GGACGGATTGAGGACCTTGG	CCCTGTGCCATAAGCCAGAT
5	ENSG00000228784	ATGTTCCAGCATCCCACAGG	ACCACGTGTGCTCCAAGAAT
6	ENSG00000224635	TCGTCCCAAGCGTCCTACTA	CTCAGGCTCTTCGTCCTTGG
7	ENSG00000271151	AATCGCCCTCCCCTTGTT	TTCAATTAATTCGTTTTGGCAGAAG
8	CD8	ACTTGTGGGGTCCTTCTCCTGT	TGTCTCCCGATTTGACCACAGG
9	GAPDH	GTCTCCTCTGACTTCAACAGCG	ACCACCCTGTTGCTGTAGCCAA

LincRNA = long intergenic noncoding RNA.

2.4. Inclusion and exclusion criteria

The inclusion criteria were histologically confirmed EC with a clear subtype classification; available high-quality RNA-seq expression data and complete clinical information, including age, stage, grade, survival status, and follow-up data; and primary tumor samples with sufficient RNA quality and sequencing depth for reliable lincRNA quantification. The exclusion criteria were missing survival data or essential clinical information, ambiguous or mixed histological diagnoses that could not be assigned to a single subtype, and low RNA quality or insufficient sequencing depths.

2.5. Statistical analysis

The Mann–Whitney U test was used to compare differences in expression between the 2 groups. For comparisons involving multiple lincRNAs or survival endpoints, P-values were adjusted using the Benjamini–Hochberg FDR, where appropriate. For GSEA and other enrichment analyses, FDR-adjusted q-values were used to assess statistical significance. Unless otherwise specified, an adjusted P-value or FDR q-value < 0.05 was considered statistically significant. Overlap analysis was performed using Venny 2.1 software (http://bioinfogp.cnb.csic.es/tools/venny/index.html). Bivariate and Cox proportional hazard regression models were used to evaluate the prognostic value of USC-specific lncRNAs. For survival analyses, the optimal cutoff value for each lincRNA expression level or lincRNA-based risk score was determined using the surv_cutpoint function in the R package survminer, which applies the maximally selected rank statistics. Patients with values above the optimal cutoff were assigned to the high-expression or high-risk groups, whereas those with values below the cutoff were assigned to the low-expression or low-risk groups. Kaplan–Meier survival curves were generated, and survival differences between the groups were evaluated using the log-rank test. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated using Cox proportional hazard regression models.

Receiver operating characteristic (ROC) curve analyses were performed to evaluate the predictive performance of lincRNA-based prognostic models. The area under the curve (AUC), P-values, and 95% CIs are reported. Statistical analyses were performed using SPSS (version 18.0; IBM, Armonk, NY) and R software (v3.5.0; R Core Team, Vienna, Austria), and plots were generated using GraphPad Prism 8.0 (GraphPad Software, San Diego, CA) and R. Statistical significance was defined as a 2-sided P-value < .05, wherein multiple-testing correction was not applicable.

For high-throughput enrichment analyses, FDR correction was applied, and FDR q-values < 0.05 were considered significant. For qRT-PCR validation and survival analyses, nominal P-values are reported because these analyses were hypothesis-driven and focused on selected candidate lncRNAs.

3. Results

3.1. LincRNAs specifically expressed in USC compared with other types of lncRNAs

To identify dysregulated lncRNAs in EC, their expression was profiled in a cohort of 521 EC tumor samples (including 114 serous and 407 endometrioid samples) and 35 normal samples from the TCGA dataset. Notably, lincRNAs constitute the main type of lncRNA and were overexpressed in EC tumor samples compared with normal samples (Fig. 1A and D). Furthermore, although lincRNAs are specifically expressed in USCs and not in endometrioid samples, antisense lincRNAs were overexpressed in normal samples compared with tumors (Fig. 1B). Only weakly significant differences were observed in the expression of other types of lncRNAs between EC tumors and normal samples (Fig. 1C).

Figure 1.

Landscape of lncRNA expression in EC. (A–C) Bar plot of lincRNA (A), antisense (B), and other lncRNA (C) expression levels in different types of EC samples. Statistics were obtained from the Mann–Whitney U test. *P < .05, **P < .01, ***P < .001, ***P < .0001. (D) Pie plot of the lncRNA types. (E) Venn plot of lincRNAs whose expression levels are not zero in endometrioid cancer, uterine serous carcinoma, and normal samples. EC = endometrial cancer, lincRNA = long intergenic noncoding RNA, lncRNA = long noncoding RNA.

Accordingly, lincRNAs are specifically expressed in USC compared with other types of lncRNAs. The expression levels of numerous lincRNAs were very low in the cancer samples, with some lincRNAs showing no expression. To reduce background noise and detect specific lincRNAs expressed in USC, the lincRNAs expressed in USC, EC, and normal samples were identified (Fig. 1E). Thirty-nine lincRNAs were expressed in USC samples and used for subsequent analyses.

3.2. LincRNAs specifically expressed in USC compared with normal samples

Using principal component analysis, the correlation between USC and normal samples was demonstrated based on the expression of 39 lincRNAs (Fig. 2A) that were shown to be differentially expressed between USC and normal samples. To understand the dysregulation of these lincRNAs in USC, a differential expression analysis was conducted. Most of these lincRNAs (87.18%) were upregulated in USCs (Fig. 2B–C). However, 12.82% of lincRNAs were not upregulated in USCs. To identify the degree of enrichment of the 39 lincRNAs in EC, GSEA was performed based on their expression levels. According to GSEA, these 39 lincRNAs were significantly enriched in EC (Fig. 2D) and USC (Fig. 2E) samples compared with normal samples. In general, these analyses suggest that these 39 lincRNAs were overexpressed in USC; therefore, they were defined as USC-specific lincRNAs.

Figure 2.

LincRNAs are specifically expressed in uterine serous carcinoma. (A) PCA plot showing the difference in expression of 39 lincRNAs between uterine serous carcinoma and normal samples. (B) Volcano plot showing significant dysregulation of 39 lincRNAs. (C) Pie plot of different degrees of dysregulation of 39 lincRNAs. (D–E) GSEA demonstrating the enrichment of 39 lincRNAs in the ranked gene list of EC (D) and uterine serous carcinoma (E) versus normal samples. EC = endometrial cancer, GSEA = gene set enrichment analysis, lincRNA = long intergenic noncoding RNA, PCA = principal component analysis.

3.3. CNV amplification promotes lincRNA expression in USC

To investigate the mechanisms underlying the upregulation of USC-specific lincRNAs in USC, CNV profiles of lincRNAs were analyzed using the TCGA dataset. The CNVs of these lincRNAs were primarily amplified in USCs (Fig. 3A). Moreover, correlation analysis suggested a positive correlation between CNV and the expression of most lincRNAs (74.35%; Fig. 3B). Collectively, these data suggest that CNV amplification mediates the overexpression of USC-specific lncRNAs in USC.

Figure 3.

CNV amplification promotes lincRNA expression in uterine serous carcinoma (USC). (A) Density plot showing the distributions of CNV amplification and deletion for each USC-specific lincRNA. (B) Dot plot (upper) and heatmap (bottom) showing the significant degree and correlative degree between CNV and expression level for each USC-specific lincRNA. CNV = copy number variation, lincRNA = long intergenic noncoding RNA.

3.4. USC-specific lincRNAs can serve as predictors of poor prognosis

Overall survival (OS) is considered a vital endpoint; however, the use of OS as the only endpoint may weaken clinical studies of death statistics. Determining the OS or disease-specific survival (DSS) requires a longer follow-up period; thus, disease-free interval (DFI) and progression-free interval (PFI) are used in many clinical trials.^[16] To test whether the expression of all USC-specific lincRNAs can predict USC survival as a group, bivariate regression and multivariate Cox models of USC samples in the TCGA dataset were developed. Models were built using 4 outcome measures (OS, DSS, DFI, and PFI), and ROC analyses were performed to predict the survival of these lincRNAs based on bivariate (Fig. 4A) and Cox (Fig. 4B) regression models. The results of both analyses verified the use of these lincRNAs as prognostic markers for significant survival prediction regarding OS, DSS, and PFI in USC. The Kaplan–Meier curves showed that the high-risk group in both models was associated with poor OS, DSS, DFI, and PFI in patients with USC (Figs. 4C, D). Hence, this 39 USC-specific lincRNA-based model is a significant prognostic indicator.

Figure 4.

LincRNAs specifically expressed in uterine serous carcinoma (USC) can serve as a poor prognosis. (A, B) Receiver operating characteristic analyses demonstrate survival prediction of USC-specific lincRNAs based on bivariate regression (A) and Cox (B) models. (C, D) Kaplan–Meier curves showing the OS, DSS, DFI, and PFI based on bivariate regression (C) and Cox (D) models. DFI = disease-free interval, DSS = disease-specific survival, lincRNA = long intergenic noncoding RNA, OS = overall survival, PFI = progression-free interval.

ROC analyses were conducted to evaluate the prognostic performance of the lincRNA-based risk scores. In the bivariate regression model (Fig. 4A), the AUC values for predicting OS, DSS, DFI, and PFI were 0.7863, 0.7888, 0.6642, and 0.7199, respectively (all P < .05; 95% CI provided in Fig. 4A). In the Cox regression model (Fig. 4B), the AUC values were 0.7044, 0.6977, 0.6006, and 0.6577, respectively (DFI, P > .05; OS/DSS/PFI, P < .05).

Kaplan–Meier survival analyses revealed that patients in the high-risk group had significantly shorter survival times than those in the low-risk group. In the bivariate regression model (Fig. 4C), the HRs were 3.556 (OS), 4.614 (DSS), and 3.489 (PFI; all P < .01). In the Cox model (Fig. 4D), the HRs were 4.639 (OS), 5.592 (DSS), 4.645 (DFI), and 4.004 (PFI; all P < .001). These results confirm that the USC-specific lincRNA signature is significantly associated with poor prognosis.

To verify whether the expression of these lincRNAs is related to prognosis in USC, the OS, DSS, DFI, and PFI of these lincRNAs were analyzed in USC samples from the TCGA database (Fig. 5). Notably, ENSG00000281406 served as a marker of poor prognosis for all 4 survival types. ENSG00000226791 and ENSG00000269903 can serve as markers of poor prognosis for 3 of the 4 survival types; ENSG00000204277 can serve as a marker of poor prognosis regarding DFI and PFI; and ENSG00000224635, ENSG00000228784, and ENSG00000271151 can serve as markers of poor prognosis regarding DFI. These 7 lncRNAs were significantly upregulated in USC, except for ENSG00000269903. ENSG00000225489 showed a favorable prognosis regarding DSS and DFI, but was upregulated in USC. ENSG00000214049 showed favorable prognoses for DSS and DFI, but poor prognoses for OS and PFI. ENSG00000226644 showed favorable prognoses for OS and DFI, but poor prognoses for DSS and PFI.

Figure 5.

FC, OS, DSS, DFI, and PFI for each uterine serous carcinoma (USC)-specific lincRNA. Heatmap showing the FC between USC and normal samples, OS, DSS, DFI, and PFI for each USC-specific lincRNA. Black box indicates P < .05. DFI = disease-free interval, DSS = disease-specific survival, FC = fold change, lincRNA = long intergenic noncoding RNA, OS = overall survival, PFI = progression-free interval.

3.5. Oncogenicity and function of USC-specific lincRNAs

To investigate the expression of the 7 USC-specific lincRNAs that were significantly correlated with poor prognosis, their expression was analyzed using the TCGA database (Fig. 6A). The expression level of USC was ranked among the top 3 of these 7 lincRNAs, except for ENSG00000269903. This result was consistent with the above finding that ENSG00000269903 was not significantly upregulated in USC.

Figure 6.

Oncogenicity and function of uterine serous carcinoma-specific lincRNAs. (A) Bar plot showing the expression levels of 7 lincRNAs in human cancers. (B) GSEA demonstrating the enrichment of KEGG (upper), GO term (middle), and Reactome (bottom) Wnt signaling pathways in the ranked gene list of ENSG00000281406 correlated mRNAs. (C) GSEA demonstrating the enrichment of the KEGG T-cell receptor signaling pathway in the ranked gene list of ENSG00000226791 (left), ENSG00000269903 (middle), and ENSG00000204277 (right) correlated mRNAs. GO = gene ontology, GSEA = gene set enrichment analysis, KEGG = Kyoto Encyclopedia of Genes and Genomes, lincRNA = long intergenic noncoding RNA.

ENSG00000281406, ENSG00000226791, ENSG00000269903, and ENSG00000204277 can serve as markers of poor prognosis for at least 2 of the 4 types of survival in USC. To identify the function of these lincRNAs in USC, GSEA was conducted using the KEGG, GO, and Reactome datasets based on the expression levels of these lincRNAs. According to GSEA, ENSG00000281406 was positively correlated with the Wnt signaling pathway in all 3 datasets, whereas ENSG00000226791, ENSG00000269903, and ENSG00000204277 were negatively correlated with the T-cell receptor signaling pathway (Fig. 6B, C).

3.6. Expression of lincRNA in cancerous and paracancerous tissues and its relationship with CD8 gene expression in cancer

The expression levels of ENSG00000281406, ENSG00000226791, ENSG00000204277, and ENSG00000228784 were higher in cancerous tissues than in the adjacent normal tissues (P < .05). No significant differences in the expression of ENSG00000269903, ENSG00000224635, or ENSG00000271151 between cancerous and adjacent normal tissues were observed (P > .05; Fig. 7A). According to the TCGA EC dataset, a negative correlation was observed between the expression of ENSG00000204277 and CD8 in carcinoma (P < .05; Fig. 7B). Furthermore, a negative correlation was found between the expression of ENSG00000204277 and CD8 in the 32 investigated cases of carcinoma (r = −0.626 and P = .005), thus further verifying the results of the TCGA data analysis (Fig. 7C).

Figure 7.

Expression of lincRNAs in cancer and paracancerous tissues and their relationship with CD8 expression in cancer. The expression levels of ENSG00000281406, ENSG00000226791, ENSG00000204277, and ENSG00000228784 in cancer tissues compared with those in adjacent normal tissues (P < .05) (A). Combined with TCGA data, a negative correlation was found between the expression of ENSG00000204277 and CD8 in carcinomas (P < .05) (B). A negative correlation was found between the expression of ENSG00000204277 and CD8 in 32 cases of carcinoma (r = −0.626 and P = .005), which further verified TCGA data (C). LincRNA = long intergenic noncoding RNA, TCGA = The Cancer Genome Atlas.

4. Discussion

USC is a rare type of EC with a high mortality rate owing to its aggressive behavior.^[17] Most genes encode noncoding RNAs, including many types of lncRNAs that comprise lincRNAs and antisense RNAs. Although several studies have reported that lincRNAs play tumor-suppressive or tumor-promoting roles in EC,^[18] no study has reported the relationshipbetween lincRNAs and USC. In this study, we first identified the expression levels of different lncRNAs in different EC subtypes. LncRNAs were found to be overexpressed in USC compared with other types of EC and represented the bulk of lncRNAs (50.49%) expressed in USC. These results suggest that lincRNAs are specific types of lncRNA associated with USC. Thirty-nine lincRNAs were found to be expressed only in USC samples, and CNV analysis was performed to identify the mechanisms by which these 39 were specifically overexpressed in USC. Furthermore, this group of lincRNAs was identified as predictors for OS, DSS, DFI, and PFI in USC. Hence, these 39 lincRNAs, defined as USC-specific lincRNAs, play vital roles in USC and can be used to predict patient survival.

Some of these lincRNAs were reported to be associated with cancer progression.^[19] For instance, ENSG00000214049 (urothelial cancer-associated 1, UCA1) promotes tumorigenesis in gastric, bladder, and breast cancers;^[20–22] ENSG00000228784 can diagnose melanoma and predict the prognosis of patients;^[23] and ENSG00000226791 promotes the upregulation of MS4A6A expression in glioblastoma, which is closely linked to shorter survival times through the miR-139-5p/MS4A6A regulatory axis.^[24] Subsequent studies on ENSG00000226791 revealed high expression in clear cell renal cell carcinoma and significant correlation with cancer cell migration and invasion.^[25]

Finally, the survival and function of each EC-specific lncRNA were investigated. ENSG00000281406 (bladder cancer-associated transcript 1) was associated with poor prognosis regarding OS, DSS, DFI, and PFI and was found to be overexpressed in USC compared with other cancers. Functional enrichment analysis of USC showed that ENSG00000281406 positively correlated with the Wnt signaling pathway in the KEGG, GO, and Reactome datasets. The Wnt signaling pathway plays a role in human cancers, including the regulation of cancer stem cells and metastasis.^[26–29] ENSG00000281406 has been reported to promote tumorigenesis and indicates poor prognosis in small cell lung cancer and colorectal cancer.^[30,31] Recently, ENSG00000281406 was identified as an oncogenic lncRNA that promotes proliferation, migration, and invasion by modulating the Wnt signaling pathway.^[32,33] These results are similar to those of the present study. Additionally, ENSG00000226791, ENSG00000269903, and ENSG00000204277 can result in a poor prognosis for at least 2 of the 4 types of survival in USC. According to KEGG functional enrichment analysis, these lincRNAs were negatively correlated with the T-cell receptor signaling pathway in USC. The upregulation of these 3 lincRNAs may affect T-cell receptor signaling to reduce the tumor immune response in the EC microenvironment and is associated with a poor prognosis.

Although the expression of ENSG00000269903 was higher in cancerous tissues than in the adjacent normal tissues in 32 cases, the difference was not statistically significant (P > .05). However, this study may lack statistical power owing to the small sample size.^[34] In this study, we verified for the first time a negative correlation between ENSG00000204277 and CD8⁺ expression in 32 cases of USC (P < .05). ENSG00000204277 is a functional partner associated with the carcinogenic effects of SOCS3 in tumors. The epigenetic alterations of SOCS3 render T cells dysfunctional, and this phenotype can be assessed using the Tumor Immune Dysfunction and Exclusion framework (including B cells, CD8⁺ T cells, CD4⁺ T cells, DC cells, neutrophils, and macrophages) in different types of tumors.^[35] Knockdown of these lincRNAs allows detection of downstream target expression, shedding light on the underlying mechanism. These findings may lead to the development of new potential EC immunotherapeutic approaches.

This study holds promise for advancing precision medicine, guiding targeted treatment strategies, and improving patient outcomes, similar to the previous findings on ovarian cancer research.^[36] Collectively, these insights provide actionable information to support clinicians in making personalized diagnostic and therapeutic decisions for USC.

However, this study has some key limitations. First, it largely relied on TCGA data from a retrospective cohort, which may limit the generalizability of the results. Second, the validation sample size was small (n = 32), which may have reduced the reliability of the experimental validation. All conclusions are therefore associative, rather than causal, and independent external validation and mechanistic studies are required to confirm their clinical and functional relevance.

Acknowledgments

We would like to thank the TCGA Project. We thank the developers of each method and package. We thank the First People’s Hospital of Guangzhou for providing us with the experimental platform. We would like to thank Editage (www.editage.com) for English language editing.

Author contributions

Conceptualization: Hong Ye.

Data curation: Yong Zhang, Qiu Yun Huang.

Methodology: Yong Zhang, Lan Lan Ma.

Resources: Qiu Yun Huang, XueYuan Qin.

Validation: Yong Zhang.

Writing – original draft: Yong Zhang.

Writing – review & editing: Yong Zhang, Hong Ye.