Pan-cancer integrative profiling of immune-related lncRNAs unveils tumor microenvironment subtypes and prognostic biomarkers

Abstract

Background

Long non-coding RNAs (lncRNAs) have been widely implicated in modulating immune pathway activity and immune cell infiltration to influence cancer outcomes. However, only a few immune-related lncRNAs have been shown to play significant roles in cancer progression.

Methods

Therefore, we propose an algorithm, LncImmune, to identify immune cell related-lncRNAs, and integrate existing algorithms to comprehensively reveal candidate lncRNA regulators involved in immune pathways in specific immune cells across 32 cancer types.

Results

We demonstrated that immune cell-related lncRNAs exhibit expression perturbations in cancer and respond actively to immune pathway activity. The integrated algorithm further delineated the immune regulatory features of lncRNAs, revealing their tumor-associated roles. Utilizing core lncRNA regulators, we identified two molecular subtypes of lung adenocarcinoma, (C1 and C2 termed the “cold” and “hot” immune phenotypes, respectively), which were distinguished by divergent genetic mutation landscapes, immune cell infiltration levels, immune checkpoint expression profiles, inferred drug sensitivities, and prognostic outcomes.

Conclusions

In conclusion, the LncImmune serves as an important method in exploring the functional role of lncRNA in tumor immunity, which would shed light on the development of potential carcinogenic biomarkers and immunotherapeutic targets.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s12672-026-05470-7.

Introduction

It has been well demonstrated that the changes in the gene expression program are closely associated with the occurrence and development of cancer [1]. In fact, gene regulatory factors shape the tumor immune microenvironment by controlling the expression program of immune-related genes, thereby promoting or inhibiting tumor growth and dissemination [2]. As one of the important gene regulatory factors, long non-coding RNAs (lncRNAs) play a crucial role in the regulation of gene expression, including epigenetic modifications, transcriptional regulation, and post-transcriptional processes [3]. In the tumor immune microenvironment, lncRNAs have been proven to be widely involved in activating immune pathways and immune cell infiltration to influence cancer outcomes [4–6]. NEAT1, for instance, has been demonstrated to be involved in the activation of T cells, thereby affecting their immune function and inflammatory response [7, 8]. Similarly, lncRNA MEG3 has been shown to modulate the activation of macrophages by interacting with the STAT3 protein in cytoplasm [9].

However, so far, only a few lncRNAs have been revealed to play a role in tumor immunity. For most lncRNAs, it is currently unclear whether and how they play an important role in the tumor immune microenvironment, and which lncRNAs are involved in regulating specific immune-infiltrating cell types [10, 11]. Therefore, systematically deciphering the potential immune characteristics of lncRNAs is crucial for a deeper understanding of immune regulatory processes in tumors and provides important clues for identifying novel targets in cancer immunotherapy.

The ImmLnc algorithm has been proposed to identify immune pathway-related lncRNAs by using a large number of cancer samples from a pan-cancer perspective [12]. However, it remains unclear in which immune-infiltrating cells these lncRNAs exert their immune functions, which hinders further understanding of their immune characteristics. To identify lncRNAs related to immune cells, we propose the LncImmune pipeline based on multi-omics data from 32 cancer types. Additionally, we integrated the ImmLnc algorithm into the LncImmune pipeline to comprehensively reveal candidate lncRNA regulators involved in immune pathways in specific immune cells. We found that immune cell-related lncRNAs exhibit expression perturbations in cancer and actively respond to immune pathway activity, and this approach has also contributed to revealing tumor-related lncRNAs and classifying cancer subtypes based on immunological characteristics. All these findings substantiate that LncImmune can serve as an important method for exploring the functional role of lncRNAs in tumor immunity.

Materials and methods

Publicly available data collection and processing

We had compiled gene expression data for 32 types of cancer based on The Cancer Genome Atlas (TCGA) project (http://cancergenome.nih.gov/). Specifically, the raw RNA-sequencing read count from the R package “TCGAbiolinks“ [13] was converted to transcripts per million (TPM) and further log2-transformed. Based on the hg38 reference genome information from GENCODE, we obtained the gene annotations of lncRNA and mRNA and further excluded the genes with null reads in all samples. The clinical information of patients was also downloaded from the TCGA project.

LncImmune: reconstruct the immune cell related lncRNAs in cancer

To reconstruct the lncRNA expression in immune infiltrating cells across cancer types, based on the ideas of Joseph et al. [14], we proposed a two-step pipeline called LncImmune. This pipeline utilizes a linear regression framework for tumor purity adjustment, followed by correlation analysis to identify robust associations. First, we reasoned that each cell type present in a sample contributes varying levels of lncRNA expression to the observed bulk value. To conceptually model this, we constructed a linear regression framework where the contribution of each immune cell type is weighted by its proportion in the tumor sample, while explicitly adjusting for tumor purity [15]. The mathematical representation of this model is defined as follows:

Here, represents the observed bulk expression of gene g in sample s, and denotes the tumor purity. The model explicitly incorporates an intercept term, tumor specific expression, and purity-adjusted immune cell proportions to resolve cell-type-specific contributions. A comprehensive and detailed definition of each statistical parameter and term in this equation is provided in Supplementary Methods.

To identify the most statistically robust lncRNA-immune cell associations and avoid potential biases from multicollinearity in the regression coefficients (), we employed Spearman correlation analysis. We calculated the Spearman correlation coefficient (R) between the bulk lncRNA expressions () and the purity-adjusted immune cell infiltrations . The estimated relative proportions of immune cells were obtained from five published deconvolution algorithms, specifically CIBERSORT, CIBERSORT absolute mode, EPIC, quanTIseq, and TIMER [16–19]. To address potential inconsistencies in cell type nomenclature across different algorithms, the outputs were carefully harmonized into consensus immune cell lineages based on standard immunological classifications. To identify the significantly correlated core pairs of lncRNA-immune cell with the highest statistical robustness and to minimize algorithm-specific biases, the following selection criteria were applied: (i) based on the thresholds of P < 0.05 and |R| > 0.3, the lncRNA needs to be recognized by all five algorithms simultaneously; (ii) each lncRNA regulator is associated with more than 10 immune cell types in each cancer type; (iii) each immune cell type corresponds to more than 500 lncRNA across cancer types. Finally, we identified 26 immune cell types and obtained the lncRNA associated with each immune cell type in each type of cancer as described in the Results section.

To assess the effectiveness of the LncImmune, we collected a lung adenocarcinoma (LUAD) single-cell RNA sequencing (scRNA-seq) transcript dataset to validate the expression of lncRNAs in specific immune cell populations. Of note, the single-cell transcriptomic data is based on single-read sequencing using the 10X Genomics Chromium platform, resulting in the detection of expression for only approximately 10% of lncRNAs in these datasets. However, bulk RNA-Seq contains almost all lncRNA information. Therefore, by demonstrating LncImmune’s accuracy in identifying immune cell-specific lncRNAs among those detected by single-cell transcriptomics, we sought to infer the immune-related functions of the remaining lncRNAs missed by the single-cell approach. Fisher’s exact test and ROC analysis were utilized to assess whether the immune cell-related lncRNAs identified by the LncImmune were specifically expressed in the corresponding immune cell. The differentially expressed genes for each immune cell type were calculated based on the single-cell RNA-seq transcript data.

The ImmLnc algorithm was further integrated into the LncImmune pipeline to comprehensively reveal the candidate lncRNA regulators involved in immune pathways in specific immune cells. Figure 1B shows the workflow of ImmLnc algorithm, with specific parameter settings discussed in detail in the research by Li et al. [12].

Fig. 1.

Identification of immune cell-related lncRNAs across cancer types. A Schematic diagram of the process for the LncImmune to identify potential lncRNA regulatory factors related to immune cells. B Integrating the ImmLnc algorithm into the LncImmune pipeline. C The number of patients with each type of cancer and the average number of immune cells associated with lncRNAs, cancers with similar tissue origins are grouped together. D The number of lncRNAs associated with 26 immune cells. E The number of immune cell-related lncRNAs identified in each cancer type. The top y-axis shows the number of lncRNAs and the bottom y-axis shows the proportion of lncRNAs

Ranking analysis of tumor-related lncRNAs

Tumor-related lncRNAs have been demonstrated to exhibit expression perturbations in tumors and play an important role in tumor immunity [20, 21]. We considered that if lncRNA is associated with multiple immune cell infiltration across cancer types, and its expression perturbations can be observed in more cancer types, it’s more likely to play a crucial role in the occurrence and development of cancer. Therefore, we initially ranked each lncRNA based on the number of tumor types in which it regulates immune cells. Simultaneously, we also ranked each lncRNA within each immune cell based on the number of cancer types exhibiting expression disturbances. All rankings have been normalized using ‘0–1’ normalization. Consequently, for each lncRNA, we obtained ranking levels for both immune cells and expression perturbations. Finally, within each immune cell, the average ranking of each lncRNA is defined as:

The tumor-related lncRNAs, known to us, were obtained from Lnc2Cancer 3.0 [22] and LncRNADisease 2.0 [23]. We compared the normalized relative ranks of known tumor-related lncRNAs with those of other candidate lncRNAs by Wilcoxon rank-sum test.

Subtype classification of patients based on lncRNA regulators

To explore the potential of the identified lncRNAs for molecular cancer subtyping, we classified LUAD samples based on the integrated core lncRNA regulators. The candidate lncRNA regulators were filtered with the following selection criteria: (i) significant association with immune cell infiltration, (ii)involvement in immune pathway activity, and (iii) manifestation of expression perturbations in LUAD. Subsequently, we employed the ConsensusClusterPlus R package [24] to classify samples using the parameters ‘maxK = 9, reps = 1000, clusterAlg = km, distance = euclidean.’ The optimal number of clusters was determined synthetically using the consensus score matrix, cumulative distribution function (CDF) curve, and proportion of ambiguous clustering (PAC) score. To characterize subtypes, we systematically elucidated the differences between subtypes, considering clinical features, immune features, epigenetic regulatory features, mutation features, replication activation, and inferred drug sensitivity.

scRNA-seq data analysis

To validate our findings at the single-cell level, primary LUAD scRNA-seq data were obtained from the GSE131907 dataset. Following routine data preprocessing, quality control, and normalization, cell populations were annotated utilizing standard lineage markers. Furthermore, immune cell-specific lncRNAs were identified through differential expression analysis between specific immune clusters and other cell types.

Evaluation of drug sensitivity

We assessed LUAD therapeutic responses using oncoPredict. Baseline cell line transcriptomes sourced from the Cancer Dependency Map (DepMap) were integrated alongside half-maximal inhibitory concentration (IC₅₀) metrics from the Genomics of Drug Sensitivity in Cancer (GDSC) and Cancer Therapeutics Response Portal (CTRP). We established separate ridge regression algorithms per medication, integrating automated hyperparameter refinement. Ultimately, applying these calculated algorithms to our LUAD expression profiles allowed us to determine individualized patient susceptibilities.

Statistical analysis

All statistical analyses were performed using R software. Where appropriate, multiple-testing correction was applied using the False Discovery Rate (FDR) method to adjust P-values across pathway enrichment, immune feature profiling, methylation probe evaluation, and drug-sensitivity analyses. An adjusted P-value < 0.05 was considered statistically significant unless specified otherwise. Significance thresholds were defined as: ns, not significant (p ≥ 0.05); * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001.

Results

Reconstruction of lncRNA expression in immune cell across cancer types

To identify potential lncRNA regulators exhibiting correlations with immune cells, we proposed a pipeline called LncImmune (Fig. 1A). The LncImmune can infer the lncRNA expression levels of each immune cell from bulk RNA-seq transcriptional samples. The workflow of the LncImmune was described in detail in the “Method” section. Briefly, LncImmune identifies lncRNA-immune cell pairs in three steps. Initially, for each tumor sample, the adjusted proportions of immune cell infiltration were obtained based on developed algorithms, accounting for tumor purity. Subsequently, deconvolution was performed to obtain the Spearman correlation coefficients between the lncRNA expression levels and immune cell infiltration. Finally, core lncRNA-immune cell pairs were selected based on predetermined screening criteria.

Utilizing the TCGA multi-omics data, we employed the LncImmune to generate a genome-wide identification of candidate lncRNA regulators in over 10,000 samples spanning 32 different cancer types (Fig. 1C). This effort resulted in the identification of over 100,000 significantly correlated lncRNA-immune cell pairs across cancer types, involving 26 core immune cell types (Fig. 1D). On average, each lncRNA was associated with at least 10 immune cell types, and every cancer type identified approximately 364 lncRNAs that were correlated with immune cells (Fig. 1E). A higher number of immune cell-related lncRNAs were determined in cancer types where a larger number of lncRNAs were expressed (Fig. 1B and E). Additionally, we noted a higher number of candidate lncRNA regulators that were correlated with CD8⁺ T cells and Treg cells (Fig. 1D). CD8⁺ T cells and Treg cells, as integral components of the immune system, play crucial roles in immunotherapy. Analyzing the functions of these lncRNA modulators will offer valuable insights into cancer immunity. Therefore, we further dissected the immune pathways in which these lncRNA molecules are involved and integrated them with immune cell populations to systematically reconstruct the immune characteristics of candidate lncRNA regulators. Furthermore, the immune-related lncRNAs identified by LncImmune were further validated at the single-cell level, demonstrating robust classification performance for specific immune populations, notably B cells [area under the curve (AUC) = 0.78], macrophages (AUC = 0.75), and neutrophils (AUC = 0.72) (Supplementary Fig. 1D). In conclusion, these findings suggest that the LncImmune can identify crucial lncRNAs implicated in immunological processes, making it a valuable resource for advancing precision medicine initiatives.

LncRNA regulators exhibit expression perturbation characteristics

To corroborate the findings obtained through the LncImmune, we investigated the potential significance of these identified lncRNAs in cancer and cancer immunity. First, we conducted the identification of all lncRNAs exhibiting expression perturbations in 16 cancer types, which have more than five normal samples. Remarkably, we observed that immune cell-related lncRNAs displayed a higher propensity for expression perturbation compared to other lncRNAs across various cancer types, particularly in cancer types that are amenable to immunotherapy (such as lung and kidney cancer) (Fig. 2A). For example, in kidney renal clear cell carcinoma (KIRC), over 90% of immune-related lncRNAs displayed expression perturbation. In cancer immunity, similar patterns were observed (Fig. 2A). Simultaneously, we delved deeper into the extent of lncRNA expression perturbation in specific immune-infiltrating cells in cancer (Fig. 2D). There are differences in the patterns of lncRNA expression perturbation in immune-infiltrating cells among different cancers (Fig. 2D). KIRC may have a unique immune microenvironment, exhibiting higher expression perturbation across various immune-infiltrating cells (Fig. 2D). However, the immune patterns in esophageal cancer (ESCA) may be opposite to those observed in KIRC, potentially linked to its poor response to immune therapies (Fig. 2D). These results suggested that the lncRNA molecules identified based on the algorithm can provide insights into analyzing differences in tumor immune patterns.

Fig. 2.

The transcriptome perturbation of immune cell-related lncRNAs. A the left heat map shows the proportion of differentially expressed lncRNAs of all lncRNAs and of immune cell-related lncRNAs. The right panel shows the odds ratios in each cancer type. All p-value < 0.05, two-sided Fisher’s exact tests. The error bars were 95% confidence levels of odds ratios. B Box plots show the distribution of MIAT in normal and cancer samples. C Box plots show the distribution of HCP5 in normal and cancer samples. The Centre of the boxplots are median values, the bounds of the boxes are 25% and 75% quantiles. The minima are 25% quantile-1.5*interquartile range (IQR) and the maxima are 75% quantile + 1.5*IQR. D The proportion of immune related lncRNAs with expression disturbances in immune infiltrating cells. E The immune cell characteristics of MIAT across cancer types. (F) The immune cell characteristics of HCP5 across cancer types

Next, we further validated the analysis of individual lncRNA immune characteristics by applying the LncImmune. For instance, MIAT, which has been confirmed to exert a crucial influence in cancer initiation and progression [25], exhibits more than two-fold expression perturbation in 10 cancer types (Fig. 2B). HCP5, implicated in the regulation of immune responses and tumor immune surveillance [26], presents upregulation in 14 types of cancer (Fig. 2C). The LncImmune analysis revealed that both MIAT and HCP5 were associated with infiltration of multiple immune cells across different cancer types (Fig. 2E and F). Moreover, MIAT and HCP5 play roles in regulating distinct cancer immune patterns. For example, MIAT is linked with immune cell infiltration in cholangiocarcinoma (CHOL), whereas HCP5 is not associated with such infiltration (Fig. 2F). Concurrently, previous research has validated MIAT’s involvement in the proliferation and metastasis of tumor cells in CHOL [27], and the findings from the LncImmune offer new insights into the interplay between MIAT and CHOL in terms of immune regulation (Fig. 2E). Specifically, a significant correlation between HCP5 and CD4+/CD8 + T cell infiltration in various cancers was identified (Fig. 2F). Previous studies have demonstrated that HCP5 is predominantly expressed in immune cells and acts as a competitive endogenous RNA to regulate microRNAs, thereby participating in tumor immune regulation. The results obtained from the LncImmune underscore that potential lncRNA regulators are associated with immune cell infiltration in cancer, further influencing the involvement of different immune cells in constructing corresponding cancer immune patterns.

LncRNA modulators actively respond to immune pathway

Immune pathways have been established as pivotal players in the activation of immune cells and the regulation of immune responses. Hence, we hypothesize that if the lncRNAs identified by the LncImmune were genuinely significant in immune regulation, they would be more likely to be involved in specific immune pathways. The ImmLnc pipeline enables the inference of characteristics of candidate lncRNA regulators related to 17 core immune pathway activities, based on the gene expression pedigree of tumor samples (Fig. 1B). Therefore, we innovatively integrated the ImmLnc pipeline with the LncImmune algorithm to systematically screen the candidate lncRNA regulators associated to both immune cell infiltration and immune pathway activity from a vast number of matched specific tumor samples. Firstly, it was observed that a higher proportion of lncRNAs associated with immune cells exhibit immune pathway activity (Fig. 3A). Moreover, we further delved into the proportion of immune cell-related lncRNAs in each immune pathway (Fig. 3C and G). Notably, antigen processing and presentation pathways are significantly associated with immune cells across all cancers (Fig. 3C). As a bridge between innate immunity and adaptive immunity, the antigen processing and presentation pathway plays a crucial role in activating T cells. The second significantly enriched pathway is the interleukin receptor (IL) pathway, which plays a crucial role in cell communication and differentiation among immune cells (Fig. 3D). Therefore, these findings suggest that immune cell-related lncRNAs exhibit higher association in the immune pathway, thereby further validating the results of the LncImmune pipeline.

Fig. 3.

Immune cell-related lncRNA actively participates in immune pathways. A The proportion of immune cell-related lncRNAs that were correlated with immune pathway across cancer types. B Inferring the interaction between immune cells and immune pathways based on immune lncRNA. The immune cell-related lncRNAs were likely to be enriched in lncRNAs that were correlated with immune pathways across cancer types. The dots represent the odds ratio (OR) of Fisher’s exact test and the error bars show the 95% confidence intervals of the OR. The numbers inside bar plots show the proportion of cancers, where orange indicates cancers with p-values < 0.05, in two-sided Fisher’s exact test. C Antigen Processing and Presentation; D Interleukins Receptor; E Natural Killer Cell Cytotoxicity; F Antimicrobials; and G Chemokine Receptors. H The immune pathway characteristics of MIAT across cancer types. I The immune pathway characteristics of HCP5 across cancer types

To comprehensively identify the immune characteristics of candidate lncRNA regulators, we further utilized lncRNA regulators as bridges to connect immune infiltrating cells with immune pathways, revealing the core lncRNA regulators involved in the immune process across cancer types (Fig. 3B). Given that CD8 + T cells are important participants in immunotherapy [28, 29], lncRNAs associated with CD8 + T cells have a positive response to immune pathway activity, indicating their potential as immunotherapy targets (Fig. 3B). Simultaneously, most of the lncRNAs associated with CD8 + T cells exhibit T cell receptor (TCR) signaling pathway activity, further indicating that these lncRNAs play important roles in immune cell activation (Fig. 3B). In terms of immune pathways, the immune cell-related lncRNAs showed more cytokine receptors activity, suggesting their close involvement in regulation processes in the tumor immune microenvironment (Fig. 3B).

Conversely, we used MIAT and HCP5 as examples to illustrate the immune pathways involved in individual lncRNAs (Fig. 3H and I). MIAT appears to play an important role in the TCR signaling pathway, while it is also associated with CD4+/CD8 + T cell infiltration, indicating its potential involvement in immune processes related to T cells across cancer types (Fig. 3H). On the other hand, HCP5 is associated with multiple immune pathways across cancer types, with a notable emphasis on the natural killer (NK) cell cytotoxicity pathway, antimicrobials pathway, and antigen processing and presentation pathway (Fig. 3I). Furthermore, HCP5 exhibits a significant correlation with the infiltration of activated NK cells. These results suggest that HCP5 may play an essential role in the activation of NK cell and cytotoxicity related immune processes. In conclusion, these findings suggest that integrating the ImmLnc algorithm into the LncImmune pipeline can provide a more comprehensive understanding of the role of lncRNAs in tumor immune processes.

Ranking analysis of LncImmune reveals tumor-related lncRNA

Having confirmed that the LncImmune can effectively identify the immune characteristics of lncRNA and infer the immune processes involved in candidate lncRNA regulators, our subsequent investigation focused on the application of the LncImmune for unveiling tumor-related lncRNAs. Despite advancements in computational biology and biological experimental technology leading to the identification of more lncRNA functions [30, 31], the identification of novel tumor-related lncRNAs through large-scale experimental approaches still faces challenges such as high costs and time consumption longer time consumption. Tumor-related lncRNAs have been confirmed to exhibit expression perturbations in tumors and play an important role in tumor immunity. Therefore, our rationale was that if lncRNA expression perturbations can be observed in more cancer types, and this lncRNA is associated with multiple immune cell infiltration across cancer types, it is more likely to play a crucial role in the occurrence and development of cancer. We ranked each lncRNA based on the number of cancer types in which it regulates immune cell infiltration. Subsequently, the ranking of the candidate lncRNA regulators was adjusted based on the number of cancers showing transcriptome perturbations. Ultimately, we obtained a comprehensive ranking of lncRNA regulators associated with 26 types of immune infiltrating cells (Fig. 4A and B). Notably, we found that the rankings of known lncRNAs associated with tumors/diseases are significantly higher than those of other lncRNAs (Fig. 4C). Similar results were observed for each immune cell-related lncRNA (Supplementary Fig. 2A-2B). Furthermore, compared to other individual ranking methods, the integration of immune regulation demonstrated improved predictive performance. Specifically, receiver operating characteristic curve (ROC) analyses based on the Lnc2Cancer and Lnc2Disease databases revealed higher average AUC values for the integrated approach (average AUC = 0.69 and 0.61, respectively) (Supplementary Fig. 2C-2D). These results implied that the LncImmune has the capability to unveil tumor-/disease-related lncRNAs.

Fig. 4.

Prioritization of tumor-related lncRNAs based on immune cell infiltration and tumor-specific expression. A The distribution of average rank for lncRNAs. For each type of immune cell, the first column is ranked based on the number of tumor types regulated by lncRNA, while the second column is ranked based on the number of expression disorders in lncRNA regulated immune cells. The third column is the average ranking of each lncRNA. B Heat map shows the normalized rank scores for the ten top-ranked lncRNAs. C The distribution of the top 50% relative rank fractions of tumor-related lncRNAs and other lncRNAs. P-value = 3.5e-06, one-sided Wilcoxon’s rank-sum test. D Box plots show the distribution of TRG-AS1 in normal and cancer samples. E Box plots show the distribution of PCED1B-AS1 in normal and cancer samples. The Centre of the boxplots are median values, the bounds of the boxes are 25% and 75% quantiles. The minima are 25% quantile-1.5*interquartile range (IQR) and the maxima are 75% quantile + 1.5*IQR. F The correlation coefficient of the TRG-AS1-CD8 + T cell and the enrichment score for the TRG-AS1-TCR signaling pathway. G The immune characteristics of TRG-AS1 across cancer types. H The correlation coefficient of the PCED1B-AS1-Macrophage and the enrichment score for the PCED1B-AS1-Antigen processing and presentation. I The immune characteristics of PCED1B-AS1 across cancer types

Based on the comprehensive ranking of lncRNA regulators, several tumor-related lncRNAs have been identified. For instance, TRG-AS1, which has been demonstrated to play a crucial role in the process of bone metastasis in breast cancer [32] and lung cancer [33], exhibits a significant expression perturbation across all cancer types (Fig. 4D). Additionally, through the ranking analysis of the LncImmune, we discovered that TRG-AS1 is associated with multiple immune-cell infiltration and participates in various immune-related pathways across cancer types, underscoring its importance in pan-cancer immunity (Fig. 4G). Specifically, TRG-AS1 is notably associated with CD8⁺ T cell infiltration and exhibits remarkable enrichment in the TCR signaling pathway (Fig. 4F). Hence, TRG-AS1 might be a potential target involved in activating CD8⁺ T cells and exerting immune efficacy through the TCR signaling pathway. Furthermore, by utilizing the LncImmune, we identified several novel candidates potentially crucial in regulating the immune process of cancer. For instance, PCED1B-AS1 is ranked sixth among the identified lncRNAs and shows expression perturbation across most cancer types (Fig. 4E). The LncImmune suggests that PCED1B-AS1 could also be a core lncRNA regulator in cancer immunity, particularly playing a crucial role in the immune process involving macrophages in the antigenic processing and presentation pathway (Fig. 4H and I). These results implied that PCED1B-AS1 could be a novel tumor-related lncRNA. Based on the current findings, we propose that integrating the results from the LncImmune can help prioritize the identification of tumor-associated lncRNAs and further elucidate their key roles in immune infiltrating cells’ immune efficacy. This integrated approach provides novel insights into their regulatory functions in tumor immunity.

Tumor subtyping based on lncRNA regulators

The burgeoning research on tumor immunity has underscored the heterogeneity of the tumor immune environment [34]. Molecular subtyping of cancer, based on distinct tumor immune processes, holds promise for identifying patient-specific immune profiles and tumor phenotypes, thus enabling tailored treatment strategies for each patient [35]. Consequently, we delved deeper into the potential of the identified lncRNA regulators for molecular cancer subtyping. As lung adenocarcinoma is the primary histological subtype of non-small cell lung cancer (NSCLC), analyzing its intricate tumor immune processes can provide new insights for cancer treatment. We pinpointed 14 differentially expressed lncRNAs significantly associated with immune-cell infiltration, which also exhibited associations to immune pathway activity (Supplementary Fig. 3). Subsequently, we classified LUAD patients into two subtypes based on the expression of 14 candidate lncRNA regulators (Fig. 5A, Supplementary Fig. 4A-4B).

Fig. 5.

The genomic alteration, oncogenic pathways, replication stress and targeted inhibitors in subtypes. A The classification of LUAD patients based on key lncRNA regulons. B The Kaplan-Meier plot of LUAD patients-based classifications generated from consensus clustering. C1 patients showing a poorer prognosis than C2 patients. C Genomic alteration landscape in subtypes. Samples are sorted in descending order according to the cumulative contribution of APOBEC-relevant mutational signatures (i.e., SBS2 + SBS13) within each subtype. TMB, relative contribution of four mutational signatures, and selected differentially mutated genes and their corresponding copy number alteration (CAN) levels. D Distribution of fraction genome altered (FGA) and fraction genome gain/loss (FGG/FGL). Bar charts are presented as the mean ± standard error of the mean. Statistical p values were calculated by Wilcoxon’s rank-sum test. E The result of oncogenic pathways enrichment analysis shows C1 (blue) and C2 (yellow) biological processes. All p-values < 0.01. F Enrichment heatmap showing the signaling pathways that are involved in DNA maintenance and activated cell cycle regulation in DNA damage response and replication stress. The C1 is blue and C2 is yellow. G C1 subtype was inferred to be much more sensitive to both cell proliferation inhibitor and EGFR inhibitors by applying a ridge regression model using 727 human cancer cell lines

Based on the core lncRNA regulators, patients were classified into two distinct molecular subtypes: C1 (the “cold” immune phenotype) and C2 (the “hot” immune phenotype). Survival analysis revealed that C1 patients exhibited a significantly poorer prognosis compared to C2 patients (Fig. 5B). To understand the underlying biological distinctions, we systematically evaluated their molecular characteristics. The C1 subtype was primarily characterized by pronounced genomic instability and active proliferation. It exhibited a higher mutation abundance and tumor mutation burden (TMB) compared to the C2 subtype, along with enriched APOBEC-relevant mutational signatures (Fig. 5C). Furthermore, the landscape of lung cancer target genes revealed more frequent mutations in TP53, KRAS, STK11, and EGFR in the C1 group (Fig. 5C). This subtype also displayed high chromosomal instability, as evidenced by elevated fraction of genome altered (FGA), fraction of genome gained (FGG), and fraction of genome lost (FGL) (Fig. 5D). Consistent with these genomic features, enrichment analysis indicated the significant activation of multiple carcinogenic pathways in the C1 subtype, including the E2F, MYC, and mTORC1 signaling pathways (Fig. 5E). Consequently, the C1 group demonstrated heightened DNA replication stress activity, reflecting enhanced cell proliferation and differentiation capabilities (Fig. 5F). Inferred drug sensitivity analysis suggested that drugs targeting these pathways may have higher sensitivity in the C1 subtype (Fig. 5G, Supplementary Fig. 6). Conversely, the C2 subtype represented an immune-inflamed microenvironment. This “hot” phenotype was distinguished by higher immune scores, elevated immune checkpoint expression, and increased infiltration of effector cells, including CD4 + T cells, CD8 + T cells, and NK cells. Specifically, the majority of the identified core lncRNA molecules were highly expressed in C2 patients (Supplementary Fig. 5), functioning as key regulators of this active immune landscape. For instance, the high expression of USP30-AS1 (ENSG00000256262) in C2 correlates positively with the infiltration levels of CD8 + T cells, Th1 cells, and antigen-presenting dendritic cells [36]. Similarly, other enriched regulators in C2, such as PCED1B-AS1 (ENSG00000247774) and BZRAP1-AS1 (ENSG00000265148), are strongly associated with robust immune infiltration and favorable clinical outcomes [37, 38].

Subsequently, we investigated the potential role of candidate lncRNA regulators in cancer molecular subtyping from immunological and epigenetic perspectives. To explore the immune status reflected by the candidate lncRNA regulators, we compared the immune landscapes of patients in the two subtypes. Both immune scores and immune checkpoints, along with immunocyte infiltration, were significantly higher in C2 than in C1 (Fig. 6A, Supplementary Fig. 7). This robust immune response in C2 is supported by the enriched activity of antigen-presenting and immune-stimulating regulons, including HLA-B, B2M, and CD70. Conversely, the immune-suppressed microenvironment in C1 is potentially driven by LUAD-specific regulons such as STK11 and KEAP1. Features of methylation tumor-infiltrating lymphocytes (MeTIL) further hinted at the involvement of candidate lncRNA regulators in shaping the tumor immune microenvironment (Fig. 6A), indicating that these core lncRNA regulators may be potential targets for immunotherapy. At the epigenetic level, we observed unique differential methylation patterns in the two subtypes, possibly linked to the involvement of candidate lncRNA regulators in gene expression regulation. Methylation levels of all differential methylation probes (DMPs) were significantly higher in the C1 subtype than in the C2 subtype (Fig. 6B). The high stem cell index of C1 also indicated higher tumor heterogeneity (Fig. 6C). These pronounced disparities in promoter methylation and tumor stemness are deeply intertwined with the distinct activities of cancerous chromatin remodeling regulons. Specifically, the C1 subtype exhibited unique regulatory patterns driven by an extensive network of histone deacetylases (e.g., HDAC1, HDAC7, HDAC8) and Sirtuins (e.g., SIRT2, SIRT5), alongside key histone methyltransferases and demethylases (e.g., KMT2A/C/E, KDM1A/4 C/5B) [39–41]. These findings suggested that such epigenetically driven transcriptional networks highlight the influence of lncRNAs and may be pivotal factors in distinguishing these subtypes (Fig. 6D). Finally, the striking differences in the activity patterns of oncogenes (such as KRAS, TP53, EGFR, and ERBB2) and tumor suppressor gene regulons (including SMARCA1) between the two subtypes suggested fundamentally different regulatory mechanisms in cancer development and progression (Fig. 6D). In summary, LncImmune recognizes different cancer subtypes with significant molecular and immunological diversity, thereby enhancing personalized cancer management strategies.

Fig. 6.

Differential immune profile across subtypes and its molecular landscape. A Heatmap showing the immune profile in subtypes, with the top panel showing the expression of genes involved in immune checkpoint targets and the bottom panel showing the enrichment level of 24 microenvironment cell types. The immune enrichment score, stromal enrichment score and DNA methylation of tumor-infiltrating lymphocytes (MeTILs) were annotated at the top of the heatmap. B The distribution of differential methylation probe in subtypes. C The distribution of stem cell index in subtypes. P-values for two-sided Wilcoxon’s rank-sum tests. The Centre of the boxplots are median values, the bounds of the boxes are 25% and 75% quantiles. The minima are 25% quantile-1.5*interquartile range (IQR) and the maxima are 75% quantile + 1.5*IQR. D (A) Heatmap showing profiles of regulon activity for 15 regulators (top panel), potential regulators that are relevant with chromatin remodeling (middle panel), and potential regulatory factors associated with oncogenes and tumor suppressor genes (bottom panel)

Discussion

A growing body of research highlights the complex and vital roles lncRNAs play in the regulation of tumor immunity [42–44]. These lncRNAs have been shown to shape the anti-tumor immune response by influencing the functionality, infiltration, and activation states of immune cells, either directly or indirectly [45, 46]. Despite this, the functions of only a small fraction of lncRNAs in tumor immunity have been identified for their functions. To address the gap, we introduced the LncImmune pipeline, a systematic approach to identify lncRNAs associated with specific immune cells. By integrating this with the previously developed the ImmLnc algorithm, we were able to pinpoint the candidate lncRNA regulators involved in immune pathways within immune cell types. Our findings suggest that lncRNAs related to immune cells tend to exhibit expression perturbations in cancer and actively respond to immune pathway activity. Moreover, this approach aids in identifying tumor-related lncRNAs and classifying cancer subtype based on immunological profiles. These insights offer a fresh perspective on the potential immunological roles of lncRNAs in cancer, underscoring their importance in tumor immunity research.

Targeting lncRNAs that are integral to immune processes presents a formidable challenge due to their complexity [47, 48]. Through the application of the LncImmune, our study has been able to delineate lncRNAs associated with immune mechanisms across a wide pan-cancer spectrum by deconvoluting tumor immune microenvironment from bulk cancer samples. This approach has enabled the identification of specific lncRNA targets that play a crucial role in regulation of cancer immunity, enhancing our understanding of the tumor immune microenvironment significantly. Furthermore, the insights into lncRNA biology provided by the LncImmune served as a solid foundation, offering robust guidance for the subsequent experimental validation of lncRNA functions. These insights not only bolster our understanding but also aid in streamlining the research process. Moreover, the discoveries made through the LncImmune have the potential to expedite the creation of innovative therapeutic interventions. By applying these findings, researchers can formulate tailored approaches aimed at modulating immune responses. Such strategies have the promise to improve the effectiveness of immunotherapies, thereby pushing the boundaries of precision medicine within the field of oncology forward.

This study acknowledges several limitations that warrant consideration. Firstly, the identification of immune cell-related lncRNAs with confirmed functional roles in tumor immunity remains limited, which poses a challenge in directly assessing the accuracy of the LncImmune pipeline. To address this, we collected a single-cell RNA sequencing data to systematically validate the capability of the LncImmune to identify lncRNAs associated with specific immune cells in LUAD. However, the predominate use of single-end sequencing in most single-cell RNA sequencing datasets limited the complete capture of all lncRNAs, suggesting that future studies could benefit from more advanced sequencing techniques for a more accurate identification of immune cell-specific lncRNAs. Secondly, to investigate the immune pathways involved in lncRNA associated with specific immune cells, we integrated the ImmLnc algorithm into the LncImmune pipeline. Although both algorithms generate results from mutually matched bulk samples, the inherent variations at the bulk RNA-Seq level may not be entirely captured by these methods, necessitating cautious interpretation of the findings. Thirdly, the tumor immune microenvironment is an intricately complex system that encompasses a variety of pivotal regulatory factors [49], including transcription factors, RNA binding proteins and microRNAs, which interact to maintain a dynamic balance affecting the tumor evolution and immune response. Focusing solely on the immune-modulating properties of lncRNAs falls far short of the comprehensive perspective required. We posit that the utility of the ImmLnc pipeline extends beyond merely analyzing the regulation of lncRNAs; it can also be leveraged to uncover other crucial regulatory factors instrumental in tumor immunity. Identifying lncRNAs associated with immune processes marks a significant stride, yet the challenge remains in determining their molecular mechanisms for regulating gene expression persists. While we have gained insights into lncRNAs’ roles in gene regulation, including epigenetic modifications, transcriptional control, and post-transcriptional processes, unraveling their precise functions within complex gene regulatory networks still necessitates overcoming obstacles. It is worth noting that the inherent technical characteristics of current scRNA-seq platforms may influence the comprehensive detection of certain transcripts. Nevertheless, the stable expression patterns observed in our targeted clusters still provide supportive evidence for their roles within the tumor microenvironment. Additionally, the drug sensitivities reported in this study were computationally inferred from transcriptomic profiles and should not be interpreted as direct therapeutic evidence without further in vitro and in vivo validation. The present study is mainly computational and hypothesis-generating. While LncImmune offers valuable insights for target discovery, further experimental validations are generally warranted to fully translate these findings into clinical or therapeutic applications. With the continuous advancement of RNA-centric technologies, we are better positioned to explore lncRNA interactions more comprehensively [50, 51]. Identifying these interaction targets will enrich our comprehension of lncRNAs’ roles in modulating immune pathway, thus shedding light on their pivotal regulatory functions in the tumor immune landscape.

In conclusion, our study underscores the crucial regulatory capacity of lncRNAs in immune modulation. This contributes to an enhanced understanding of the tumor immune microenvironment and supports the pursuit of novel immune therapeutic strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1. ^{(17.6KB, docx)}

Supplementary Material 2. ^{(3.8MB, pdf)}

Author contributions

Guohua Zhu was responsible for the study design, revision or critical review of the article and gave final approval of the version to be published; Yun Wang took part in the drafting, data analysis and interpretation. All authors have agreed on the journal to which the article has been submitted and agree to be accountable for all aspects of the work.

Funding

This study recieved no funding support.

Data availability

We had compiled gene expression data for 32 types of cancer based on The Cancer Genome Atlas (TCGA) project (http://cancergenome.nih.gov/). To validate our findings at the single-cell level, primary LUAD scRNA-seq data were obtained from the GSE131907 dataset. The data underlying this article, including the input data sources and processed lncRNA-immune cell association tables, as well as the custom code and parameter settings supporting the computational pipeline, are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Consent for publication

Not applicable as no identifiable data or images are present within the manuscript.

Competing interests

The authors declare no competing interests.