Screening and validation of plasma cell-derived, purinergic, and calcium signalling-related genetic signature to predict prognosis and PD-L1/PD-1 blockade responses in lung adenocarcinoma

Abstract

Background

This study aims at  screening and validation of prospective genetic signature for lung adenocarcinoma (LUAD) prognosis and treatment.

Methods

The immune-related genes (IRGs) were obtained from The Cancer Genome Atlas (TCGA) dataset where a total of 535 LUAD and 59 control samples were included. A risk model was then developed for the risk stratification of LUAD patients.  The immune cell infiltration, clinical outcomes, and the therapeutic efficacy of programmed cell death protein 1 (PD-1) and its ligand (PD-L1) blockade were compared between high and low-risk groups. Gene set enrichment analysis (GSEA) and gene set variation analysis (GSVA) were used to explore the biological processes and signalling pathways associated with the IRGs. Finally, IRGs mRNA levels were assayed by reverse transcription quantitative real-time PCR (RT-qPCR) in LUAD and relevant cell lines.

Results

Two IRGs, P2RX1 (purinergic receptor P2X 1) and PCP4 (Purkinje cell protein 4), were screened from a module that possesses the highest correlation with plasma cells. RT-qPCR verified the expression of the two IRGs in plasmacytoma cell RPMI 8226 but not in LUAD cells. A higher risk score is associated with a lower infiltration of immune cells. Kaplan–Meier and nomogram analysis showed that the high-risk group has a lower survival rate than the low-risk cohort. Furthermore, the high-risk group had a worse response rate to PD-L1/PD-1 blockade. GSVA and GSEA-GO results indicated that a lower risk score is linked to signalling pathways and biological functions promoting immune response and inflammation. In contrast, a higher risk score is associated with signalling cascades promoting tumour growth.

Conclusion

The immune-related prognostic model based on P2RX1 and PCP4 is conducive to predicting the therapeutic response of PD-L1/PD-1 blockade and clinical outcomes of LUAD.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00432-023-05153-8.

Introduction

Globally, the estimated total new cases and deaths of lung cancer were 2.2 million and 1.8 million in 2020, accounting for 11.4% and 18%, respectively, of all cancer cases and cancer-related deaths (Thai et al. 2021). Pathology-defined subtypes of lung cancer include non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC), of which NSCLC has an overwhelming percentage of approximately 85% (Siegel et al. 2021). Lung adenocarcinoma (LUAD) develops from lung glandular cells and is the main subtype of NSCLC (~ 60%). LUAD poses a challenge to global health despite the considerable therapeutic achievements have been made in the past decades (Falzone et al. 2018). Thus, the characterization of molecular signatures predicting therapeutic responses will be conducive to optimize anti-tumour treatment and prognosis of LUAD (Calvayrac et al. 2017).

The tumour microenvironment (TME) is a complex system comprising multiple tumour-infiltrating immune/inflammatory cells that have been widely implicated in either tumorigenesis or tumoricidal effects (Pavlova et al. 2022). The well-orchestrated interaction between tumour cells and their surrounding TME determines disease progression. Thus, TME-consisting immune/inflammatory cells contain massive genetic and molecular information that is intimately associated with therapeutic efficacy and cancer prognosis (Sanegre et al. 2020). One of the tumour-infiltrating cells, plasma cells (PCs), were first reported in the 1980s in the stroma of medullary breast carcinoma (Ito et al. 1986), but have not been well-studied in lung cancer. Increased plasma cell infiltration has  only recently been connected to improved immunotherapy outcomes in NSCLC patients, according to a number of studies (Leader et al. 2021; Patil et al. 2022).

The genomic sequencing data deposited in public archives contribute to the identification of cancer-related hallmarks by licensing the screening of genetic alterations in the reprogrammed TME (Chen et al. 2021). In the present study, we screened and verified 66 hub genes and their associated immune cell infiltration in LUAD through a combination of CiberSort and weighted correlation network analysis (WGCNA) by using TCGA and GSE30219 databases, respectively. Based on a purple module that possesses the highest correlation with PCs, we further identified two key genes, P2RX1 and PCP4, and constructed a risk model to explore its predictive role in the therapeutic efficacy and prognosis for LUAD.

The purinergic receptor P2RX₁ belongs to P2 receptors that are divided into ionotropic P2RXs (P2RX_1–7) and metabotropic P2RYs (P2RY_{1/2/4/6/11/12/13/14}). P2 and P1 (A₁/A_2A/A_2B/A₃R) receptors are expressed on most cell types and activated by extracellular adenosine 5′-triphosphate (ATP) and adenosine, respectively. However, triggering of P2 and P1 receptors produces opposite pro-inflammatory and anti-inflammatory effects, respectively (Eltzschig et al. 2012; Kaczmarek-Hájek et al. 2012). Ionotropic P2RXs are ATP-gated ion channels, and P2RX₁ can induce fast calcium influx which subsequently strengthens the activation of target cells (Kaczmarek-Hájek et al. 2012; Stojilkovic et al. 2005). TME is rich in ATP and adenosine; the former has shown anti-neoplastic activity whereas  adenosine-induced purinergic signalling may promote tumour growth (Lelièvre et al. 1998; Burnstock and Virgilio 2013; Gilbert et al. 2019). Purkinje cell protein (PCP) 4/peptide (PEP) 19 was  originally identified as a brain-specific polypeptide displays homology to calcium-binding proteins such as the β chain of S100 and intestinal calcium-binding protein (Ziai et al. 1986). The following studies demonstrated that PCP4 is expressed in cells with highly active Ca²⁺ dynamics, binding with Ca²⁺ and triggering calmodulin (CaM)-dependent kinase signalling pathway (Kleerekoper and Putkey 2009; Wang and Putkey 2016). PCP4 increases the response sensitivity of target cells to sub-saturating doses of ATP by increasing ATP-induced Ca²⁺ release (Wang et al. 2013).

Materials and methods

Data acquisition

The stepwise workflow of the study is presented in Fig. 1. All RNAseq and clinical data were obtained from TCGA database (The Cancer Genome Atlas Program—National Cancer Institute), and a total of 535 LUAD and 59 control samples were included. A transcripts per million (TPM) format was created from the FPKM (Fragments Per Kilobase of transcript per Million mapped reads) format for further analysis. The GSE30219 dataset for lung cancer, which had 307 samples, was retrieved from the Gene Expression Omnibus (GEO) database (Edgar et al. 2002) (GEO—NCBI (nih.gov)). Any missing data were removed from samples before analysis. In addition, four other individual GEO datasets including GSE68465, GSE101929, GSE37745, and GSE83845 were merged into a combined dataset and used as another validation cohort. The integration aims to reduce the likelihood of batch effects and magnitude harmonization.

Fig. 1.

The workflow of this study

Immune cell analysis

In TCGA database, the fraction of immune cells in stratified LUAD samples was examined using the CiberSort algorithm (Newman et al. 2015). The connections between multiple immune cell groups were displayed using the R package corrplot. As a result, a heatmap was constructed to visualize the distribution characteristics of 22 different kinds of immune cells that were seen infiltrating both cancerous and healthy tissue samples. The estimate package was used to compare stromal, immune, and ESTIMATE scores between high- and low-risk groups. Correlations between mRNA levels of the two immune-related genes (IRGs) and the infiltrating immune cells were analysed using the TIMER database.

Screening of IRGs

The weighted gene correlation network analysis (WGCNA) (Zhang and Horvath 2005) was employed (‘WGCNA’ package in R) to identify the target gene sets included in TCGA LUAD database. A β (soft-thresholding power) parameter was determined by the network’s scale-free topology for reconstruction with strongly correlated genes and elimination of weakly correlated genes. A purple module possessing the highest correlation with PCs was selected, followed by merging and verification by the GSE30219 database. Investigations of univariate COX regression, Least Absolute Shrinkage and Selection Operator (Lasso) regression, and multivariate COX regression were performed on the shared genes in order to derive IRGs.

GO enrichment analysis

Gene ontology (GO) enrichment analyses were performed by Metascape (Zhou et al. 2019) (Metascape), focusing on the functions of the shared genes.

Construction of IRGs model and prognostic evaluation

By using the univariate and multivariate COX regression analyses, the risk model was determined based on P2RX1 and PCP4 genes and used as an independent prognostic factor for LUAD. Based on the expression levels of the two IRGs, a formula has been developed to calculate each patient’s risk score: P2RX1 × (− 0.387) + PCP4 × (− 0.053). On the basis of the patients’ median risk score, patients were categorized into either a low-risk or high-risk group. R survival packages were utilized in order to carry out a Kaplan–Meier study of survival.

Construction of nomogram

The rms package in R language was used to construct a nomogram model based on P2RX1 or PCP4 in combination with T-, N-, and M categories and the pathological stage of LUAD patients. The nomogram model was used to forecast the survival rates of patients with LUAD after one, three, and five years.

Risk score-based immunotherapy response prediction

An immunotherapy response prediction tool, Cancer Immunome Atlas (TCIA) (https://tcia.at/home) (Allen et al. 2015; Hugo et al. 2016), was used to provide comprehensive immunogenomic analyses. As a quantitative measure of immunogenicity, immunophenoscore (IPS) was calculated from 0 to 10 and was used to forecast how immune checkpoint inhibitors would react.

Functional investigation of the gene and risk models

To explore biological processes and molecular mechanisms associated with P2RX1 and PCP4, GSEA-based Kyoto Encyclopaedia of Genes and Genomes (KEGG) and GO analyses were conducted by the R package (clusterProfiler). A normalized enrichment score was obtained by permuting gene sets 1000 times and filtering significant enrichments based on a P value of 0.05. GSVA was further introduced to characterize the specific signalling pathways involved in the two risk genes. The enrichplot package and GSVA package were used to build GSEA and GSVA, respectively.

Cell lines and cell culture

The Cell Center of FuHeng Biology (Shanghai, China) was the source from which we received the LUAD cell lines A549, H3255, and HCC827, as well as the plasmacytoma cell RPMI-8226. All of the cells were cultured in RPMI 1640 (Thermo Fisher Scientific, Waltham, Massachusetts, USA), which was augmented with 10% fetal bovine serum (FBS) (Lonza, Basel, Switzerland) and antibiotics in the form of penicillin (100 international units per millilitre) and streptomycin (100 µg per millilitre) (MedChemExpress, NJ, USA), and culture at a temperature of 37 degrees Celsius in an air atmosphere with 5% carbon dioxide. All cell lines underwent short tandem repeat (STR) profiling, sex-typing gene amelogenin detection, and authenticity checks against DSMZ STR cell line profiles before use.

Reverse transcription-quantitative real-time PCR

A total of 5 × 10⁶ A549, HCC827, H3255, and RPMI-8226 cells had their total RNA extracted using TRIzol reagent in accordance with the instructions provided by the manufacturer (Gibco, Grand Island, NY, USA). The transcript levels of P2RX1 and PCP4 were assayed by reverse transcription quantitative real-time PCR (RT-qPCR) technique. The forward and reverse sequences for P2RX1, PCP4, and GAPDH, and the thermal cycling conditions are shown in Table 1. The RT-qPCR was performed using a QuantStudio™ 5 Real-Time PCR System (ThermoFisher Scientific). Three separate replicates of each experiment were carried out, and the relative expression levels of P2RX1 and PCP4 mRNA were normalized with GAPDH level and calculated using the 2^−ΔΔCt method (Livak and Schmittgen 2001).

Table 1.

Primers and thermal cycling conditions used in amplification of P2RX1, PCP4 and GAPDH genes

Primers	Sequences	Thermal cycling conditions
P2RX1 (ID5023)-F	TCT GGA ATT GGC ATC TTT GGG	Pre-denatured: 95 °C, 30 s; Denaturation: 95 °C, 10 s Annealing: 60 °C, 30 s Extension: 60 °C, 30 s 40 cycles
P2RX1 (ID5023)-R	CCC ATG TCC TCA GCG TAT TTG
PCP4 (ID5121)-F	TGA CAT GGA TGC ACC AGA GA CAG
PCP4 (ID5121)-R	AGG ACT GAG ACC CAG CCT TCT T
GAPDH(ID2597)-F	AAG GTG AAG GTC GGA GTC AAC
GAPDH(ID2597)-R	GGG GTC ATT GAT GGC AAC AAT A

Statistics analysis

R version 4.1.3 (www.r-project.org) was used throughout all of the statistical studies that were conducted. An analysis of the survival rate was conducted using a Kaplan–Meier survival plot, and P values were calculated using the log-rank test. The levels of P2RX1 and PCP4 mRNA were recorded as the means standard error of the mean ± (SEM) of at least three sample trials, and the significance of the results was determined using a two-tailed Student’s t-test. Statistics were deemed significant at P < 0.05.

Results

Immune cell infiltration in LUAD

According to TCGA information, the CiberSort algorithm was used to extract the data of 22 types of tumor-infiltrating immune cells in the TME of LUAD. Pearson correlation analysis showed no relationship or a weak to moderate correlation between these immune cell subpopulations. The maximum positive and negative correlation were observed between CD8⁺ T cells and activated memory CD4⁺ T cells (r = 0.53, P < 0.05), and PCs and macrophage M2 cells (r = − 0.41, P < 0.05), respectively (Fig. 2A). The heatmap showed that M1 macrophages, CD8⁺ T cells, activated memory CD4⁺ T cells, naïve B cells, PCs, follicular helper T cells, regulatory T cells, and resting dendritic cells are highly infiltrated in TME of LUAD (Fig. 2B). Thus, LUAD is characterized by the infiltration of immune cells which may contain specific genetic and molecular signatures for predicting therapeutic responses and prognosis.

Fig. 2 .

Immune cell infiltration pattern in LUAD. (A) Correlation matrix between the 22 immune cell types. (B) Differences of the 22 infiltrating immune cell subtypes between normal and LUAD tissue

WGCNA screening of immune cell-related gene panels

Next, we performed WGCNA analysis to screen immune cell-related genes in LUAD. The A β (soft-thresholding power) in WGCNA was determined based on a scale-free R² = 0.9 (Fig. 3A, B). Twelve modules were clustered based on a hierarchical clustering algorithm with a dynamic tree cut (Fig. 3C). The module-trait heatmap displays the correlations of the module eigengenes with infiltrating immune cells, of which the purple module has the highest correlation with PCs (Fig. 3D; r = 0.58, P = 1 × 10^–48). Kaplan–Meier survival analysis showed that higher infiltration of PCs is correlated with prolonged survival of LUAD patients (P = 0.002) (Supplementary Fig. 1). Therefore, PCs are intrinsically involved in the disease progression of LUAD.

Fig. 3 .

Scale-free network construction and gene clustering by WGCNA. (A) The scale-free topological indices at different soft-thresholding powers and the soft-thresholding parameter were determined on a scale-free R² = 0.9. (B) The correlation between soft-thresholding powers and the network's average connectivity. (C) Diagram showing gene clustering based on hierarchical clustering under optimized soft thresholds. (D) Heat map showing the correlation between gene modules and immune cells

GO enrichment analysis of the screened IRGs

For further screening and validation of the plasma cell-related gene panel, 307 lung cancer samples were separately obtained from the GSE30219 dataset, and the acquired gene profiles were merged with the TCGA-LUAD dataset. The intersected genes were further merged with the purple panel to obtain 66 shared genes. The GO enrichment analysis showed that the top 5 of the 66 gene sets are associated with major biological processes (BP) and molecular functions (MF) including adaptive immune response, cysteine-type endopeptidase regulatory activity involved in apoptotic processes, sarcoplasmic reticulum membrane, lymphocyte activation, and immunoglobulin production (Supplementary Fig. 2A). We then translated the representative terms from the GO analysis into networks (Supplementary Fig. 2B). Overall, based on these findings, it was hypothesized that the genes associated with plasma cells might be potential variables implicated in the immunological control of LUAD.

Identification of P2RX1 and PCP4 and risk model construction

The COX regression model was used in conjunction with Lasso to reduce the duplication of high-dimensional characteristics and choose the most beneficial prognostic genes. Univariate COX regression analysis was performed on the 66 shared genes, and it was discovered that there were 26 potential genes that were related to the overall survival (OS) of LUAD. Only one gene, BAIAP2L1 (BAR/IMD Domain Containing Adaptor Protein 2 Like 1) is positively correlated with OS, while the remaining 25 genes are negatively correlated with OS (Fig. 4A). The following Lasso regression analysis reduced the 26 genes to 7 candidates: BAIAP2L1; P2RX1; AMPD1 (Adenosine Monophosphate Deaminase 1); PCP4; CD19; RAB39B; and FCRLA (Fc Receptor-Like A) (Fig. 4B, C). Finally, multivariate COX regression analysis yielded two genes, P2RX1 and PCP4, as IRGs for the construction of a novel model for LUAD. Each patient’s risk score was determined using a method based on the amount of expression of the two genes: P2RX1 ×  (− 0.387) + PCP4 ×  (− 0.053). The estimate package calculation further demonstrated that the stromal, immune, and ESTIMATE scores are significantly lower in the high-risk group compared to  that in the low-risk cohort (Fig. 5A, all P < 0.001). Thus, the dysregulation of the two genes, P2RX1 and PCP4 in tumour-infiltrating immune cells, especially PCs, constitutes the genetic signature for LUAD progression.

Fig. 4.

 The univariate COX and lasso analyses of immune genes. (A) 26 IRGs were screened by the univariate COX analysis from 66 shared genes. (B, C) The screening of 7 immune genes using minimal λ in Lasso regression

Fig. 5.

 The infiltrating characteristics associated with the risk model. The comparisons of stromal, immune, and ESTIMATE scores (A) and the proportions of infiltrating immune cells (B) between high- and low-risk groups are shown by violin plot and box plot, respectively. *P  < 0.05; **P < 0.01;  ***P  < 0.001

Infiltrating characteristics associated with the risk model

To verify the infiltrating feature associated with the plasma cell-related IRGs in LUAD, further research was conducted to compare the 22 types of tumor-infiltrating immune cells between high- and low-risk score groups. In the low-risk score group, tumor-enriched immune cells comprise memory B cells (P < 0.05), PCs (P < 0.001), CD8⁺ T cells (P < 0.05), and activated CD4⁺ memory T cells (P < 0.05). In contrast, two types of immune cells including M2 macrophages (P < 0.01) and activated dendritic cells (P < 0.001) were abundant in the high-risk cohort (Fig. 5B). Thus, PCs are intimately associated with the novel risk model for LUAD, and the significant differences in the infiltration of immune cells between high- and low-risk score groups mirror pro- and anti-tumour immune responses, respectively.

P2RX1 and PCP4 expression in LUAD and plasmacytoma cell lines

To further confirm the cellular source of the two IRGs, we performed mRNA level analyses of P2RX1 and PCP4 in the plasmacytoma cell line RPMI 8266 and LUAD cell lines including A549, H3255 and HCC827. The findings demonstrated that the relative mRNA level of P2RX1 and PCP4 in RPMI 8266 cells was 39.585 ± 5.87 and 301.006 ± 53.182, respectively. In contrast, low P2RX1 and PCP4 mRNA expression was observed in A549, H3255 and HCC827 cells (all P < 0.001), as shown in Fig. 6A–F. Thus, these experimental results verified our bioinformatical analysis by ruling out the LUAD cell source of the two IRGs while highlighting their origination from PCs.

Fig. 6 .

P2RX1 and PCP4 mRNA level assay in LUAD and plasmacytoma cell lines. (A–D) Representative amplification plot of P2RX1 and PCP4 in A549 (A), HCC827 (B), H3255 (C), and RPMI-8226 (D) cell lines. (E, F) Statistical comparison of relative P2RX1 and PCP4 mRNA levels between A549, H3255, HCC827 and RPMI-8226 cells. ***P < 0.001

Prognostic prediction of the risk model for LUAD

The median risk score was used to divide all of the samples into low-risk and high-risk groups. The samples in TCGA and GSE30219 datasets were set as training and validation cohorts, respectively, for the prognostic risk model construction. The Kaplan–Meier survival analysis showed that the high-risk group has a lower survival rate than that of the low-risk cohort both in training (P = 0.004) and validation (P = 0.023) cohorts (Fig. 7A, B). Accordingly, the distribution of survival status (dead or alive) and the survival time for each patient in the training (Fig. 7C, E) and validation cohorts (Fig. 7D, F) were plotted. The results showed higher death events in high-risk cohorts. Moreover, the down-regulated expression of P2RX1 and PCP4 in the high-risk group compared to that in the low-risk patients was shown both in training (Fig. 7G) and validation (Fig. 7H) cohorts. Kaplan–Meier survival analyses based on the independent P2RX1 and PCP4 genes demonstrated that lower expression of both genes were correlated with significantly shorter survival in the TCGA (Fig. 7I, J) cohort. Finally, the prognostic prediction of the risk model for LUAD was further confirmed by the merged GEO datasets (GSE68465; GSE101929; GSE37745; GSE83845) (Supplementary Fig. 3A–D). Collectively, these results provide evidence that the constructed risk model by P2RX1 and PCP4 may serve as a predictor for LUAD prognosis.

Fig. 7.

 Assessment of the predictive power of the risk model. TCGA and GSE30219 datasets were used as training and validation sets, respectively. (A, B) Kaplan-Meier analyses of OS in the high- and low-risk groups of the training (A) and validation (B) sets. (C–H) The distribution of risk scores (C, D), survival status (E, F) and expression levels of P2RX1 and PCP4 (G, H) in the high- and low-risk groups of the training (C, E, G) and validation (D, F, H) sets. (I, J) Kaplan-Meier analyses of OS in LUAD patients with high and low expression of P2RX1 (I) and PCP4 (J). OS, overall survival

Hazard assessment of the two IRGs for LUAD

By using four covariates, both univariate and multivariate COX regression analyses indicated that tumour staging and risk score, but not age and sex are independent hazard factors for LUAD (Fig. 8A, B). The association analysis visualized by the Complex Heatmap package showed that a higher risk score is correlated with a more advanced T-category and pathological stage, whereas age, sex, N-, and M-categories have no significant correlation with risk score (Fig. 8C). Thus, these findings further suggested that the two plasma cell-derived IRGs, P2RX1 and PCP4, are independent hazard factors for LUAD.

Fig. 8.

 Nomogram prediction based on the two IRGs. (A, B) Forest plots of the univariate (A) and multivariate (B) COX regression proportional hazards regression analysis of OS in TCGA LUAD cohort. (C) Relationship between risk score and demographic and pathological characteristics of LUAD. (D, E) Nomogram prediction of P2RX1 (D) and PCP4 (E) for 1-, 3-, and 5-year survival of LUAD patients in combination with T-, N-, and M-categories and pathological staging. (F, G) Calibration plots of the nomogram model based on P2RX1 (F) and PCP4 (G). OS, overall survival; *P < 0.05

Nomogram verification of P2RX1 and PCP4 as prognostic factors for LAUD

The nomogram prognostic models based on the stratified expression of P2RX1 or PCP4, T-, N- and M-categories, and pathological staging were used to verify the predictive role of the two IRGs for LUAD (Fig. 8D, E). The C-index of P2RX1 and PCP4 for OS prediction were 0.644 (95% CI 0.617–0.671) and 0.663 (95% CI 0.639–0.687), respectively. The calibration curves demonstrate optimal predictive accuracy between nomogram prediction and actual survival at 1-, 3-, and 5 years (Fig. 8F, G). Therefore, the prognostic value of the risk model for LUAD was further verified by the ability for predictive analysis shown by nomogram diagram of P2RX1 and PCP4.

Prediction of PD-L1/PD-1 blockade response

The immunogenomic analyses revealed that a significantly higher IPS is observed in the low-risk group of LUAD patients treated with PD-L1/PD-1 inhibitors (Fig. 9A). Of note, in the TME of LUAD, P2RX1 expression is positively correlated with the mRNA level of PDCD1 (encodes PD1) and CD274 (encodes PD-L1) (both P < 0.001), and PCP4 is positively correlated with PDCD1 (P < 0.05). Accordingly, the risk scores based on the two IRGs have a significant negative association with PDCD1 and CD274 mRNA levels (both P < 0.001) (Fig. 9B). Thus, these results indicated that LUAD patients with lower risk are more inclined to respond to PD-L1/PD-1 inhibitors, highlighting the risk model as a potential predictor for PD-L1/PD-1-based immune checkpoint blockade.

Fig. 9 .

 The association of risk score with PD-L1/PD-1 blockade. (A) The relative probabilities of responding to anti-PD-L1/PD-1 antibody in the low- and high-risk groups. (B) Correlation between P2RX1, PCP4, or risk score and PD-1 (PDCD1) and PD-L1 (CD274) (all at mRNA level except risk score)

GSEA and GSVA analysis

GSVA enrichment analysis showed that the toll-like receptor, T cell receptor, Nucleotide-binding and oligomerization domain (NOD)-like receptor, Janus kinase (JAK)-signal transducer and activator of transcription protein (STAT), Hedgehog, GnRH (Gonadotropin-releasing hormone), FcεRI, chemokine, calcium, and B cell receptor signalling pathways are negatively associated with the risk score. In contrast, the Wnt, PPAR (peroxisome proliferator-activated receptor), P53, mTOR (mammalian target of rapamycin), Insulin, and ErbB (also known as epidermal growth factor [EGF] receptor) signalling pathways are enriched in high-risk population (Fig. 10A). GSEA-GO enrichment analysis found that both P2RX1 and PCP2 are involved in biological processes highlighting B cell-mediated humoral immune responses (Fig. 10B, C). Collectively, these findings suggested that a lower risk score, represented by higher P2RX1 and PCP4 expression, is linked to signalling pathways and biological functions promoting immune response and inflammation. In contrast, a higher risk score is associated with signalling cascades promoting tumour growth.

Fig. 10 .

Gene set variation analysis (GSVA) and gene set enrichment analysis (GSEA). (A) GSVA estimates the enrichment of signalling pathways associated with a risk score. (B, C) GSEA-GO for the biological processes related to P2RX1 (B) and PCP4 (C)

Discussion

In this study, two IRGs, P2RX1 and PCP4, were screened from the infiltrating PCs by comprehensive high-dimensional data exploration. The following experimental results verified bioinformatical analysis by ruling out the LUAD cell source of the two IRGs while highlighting their association with PCs. A novel risk model was then constructed based on the two IGRs which constitute a remarkable signature for PC-related immune cell infiltration in LUAD tissues. The abnormal distribution of the risk score was negatively correlated with memory B cells, CD8⁺ T cells, and activated CD4⁺ memory T cells while positively correlated with M2 macrophages and activated dendritic cells. It is now appreciated that PCs exert anti-tumour effects on LUAD progression. Lohr et al. showed that higher tissue CD138 levels are associated with longer survival of LUAD but not for NSCLC patients as a whole (Lohr et al. 2013). This was confirmed by Backman et al. who demonstrated that higher stroma and total (stroma and tumour area combined) tumour infiltration of PCs were associated with longer survival in LUAD but not in the lung squamous cell cancer subgroup (Backman et al. 2021). Our results are consistent with these observations by showing the increased infiltration of PCs and prolonged survival of the LUAD cohort with abundant PCs. Moreover, the intra-tumoral enrichment of PCs is associated with the presence of tertiary lymphoid structures (TLSs) and/or lymphoid aggregates, supporting an intimate interaction between PCs and other infiltrating immune cells in the specific organization within the tumours (Patil et al. 2022). The long-lived PCs and memory B cells are the two main types of memory B cells that have collaborative effects on humoral immune responses (Akkaya et al. 2020). Activation of memory CD4⁺ T cells has an important facilitator role by producing early effector cytokines that enhance the B cells/PCs and CD8⁺ T cells responses (Stockinger et al. 2006). It is, therefore, very likely that the predilection of PCs for these subpopulations contributes to fulfilling tumoricidal effects. Both macrophages and dendritic cells (DCs) are mononuclear phagocytes closely related to immune responses. M2 macrophages contribute to tumour growth and immunosuppressive function via the secretion of anti-inflammatory cytokines such as arginase-I, IL-10, and TGF-β (Yunna et al. 2020). With high potential in promoting antitumour responses, dendritic cells are indispensable for generating responses to anti-PD-1 inhibitors (Garris et al. 2018). Collectively, the risk model-related characteristic immune cell infiltration in LUAD is associated with its prognosis and immune checkpoint blockade.

Ionotropic P2RX1 is an ATP-gated Ca²⁺ channel capable of inducing a quick calcium influx (Kaczmarek-Hájek et al. 2012; Stojilkovic et al. 2005). PCP4 is a calmodulin regulator protein acts by increasing the response sensitivity of target cells to sub-saturating doses of ATP, consequently enhancing ATP-induced Ca²⁺ release (Wang et al. 2013). Therefore, the two IRGs act in a syngenetic way to increase cytoplasmic Ca²⁺ concentration. Being a crucial second messenger, intracellular Ca²⁺ homeostasis and the process of Ca²⁺ signalling is a universal signal transduction mechanism that triggers a series of molecular and biophysical events (Berridge et al. 2003). Our findings suggested that a lower risk score is linked to signalling pathways promoting an immune response and inflammation. In contrast, a higher risk score is associated with signalling cascades regulating cell proliferation and apoptosis. GSEA-GO results further indicated that both P2RX1 and PCP4 are involved in biological processes highlighting B cell-mediated humoral immune responses. Thus, the two IRGs are involved in the purinergic signalling which induces intracellular Ca²⁺ release and subsequent chain of events, making the risk model an eligible genetic signature for LUAD.

By using the samples in TCGA as training cohort, and GSE30219 and merged (GSE68465; GSE101929; GSE37745; GSE83845) GEO datasets as validation cohorts, the Kaplan–Meier survival analysis showed that the high-risk group has a lower survival rate than that of the low-risk cohort. This was further confirmed by the distribution plot of risk score and outcome, the survival analysis based on independent P2RX1/PCP4 gene and PCs, and the nomogram prognostic models. Till now, many prognostic models have been reported for LUAD (Luo et al. 2020; Zhao et al. 2020; Zuo et al. 2020; Yi et al. 2021). Our study differs from previous findings in the following aspects: (1) we focus on tumour-infiltrating PCs that have been postulated as an attractive target for anticancer therapies (Patil et al. 2022; Sakaguchi et al. 2021; Weiner et al. 2022). It is now appreciated that PCs have far-reaching effects on cancer progression independent of canonical antibody secretion (Pioli 2019); (2) For the first time, our findings suggest a possible collaboration between P2RX1 and PCP4 that engage an increased susceptibility and Ca²⁺ influx in PCs in response to ATP stimulation; (3) Besides the prognostic prediction, the results of this study highlighted the predictive role of the risk model for PD-L1/PD-1 immunotherapy that is emerging as a front-line treatment for LUAD. Taken together, the novel prognostic model developed based on P2RX1 and PCP4 has unique immune characteristics and functional potential. However, further studies and external validation are required to determine the clinical usefulness of this model.

Our study revealed that LUAD patients with lower risk are more inclined to respond to PD-L1/PD-1 inhibitors. During the last decade, anti-PD-L1/PD-1 therapies have proved to be of significant clinical responses in patients with different types of advanced cancer including NSCLC (Christofi et al. 2019). However, the adaptive immune resistance (AIR) at the tumour site may block the efficacy of immunotherapy. AIR is derived from either the lack of PD-L1 expression on tumour cells or the low frequency of tumour-infiltrating lymphocytes (TILs) or both (Kim et al. 2022). Thus, the qualification of the risk model as a potential indicative biomarker for anti-PD-L1/PD-1 treatment may result from the negative association of risk score with the following two factors: (1) TILs including memory B cells, CD8⁺ T cells, and activated CD4⁺ memory T cells; (2) mRNA levels of PDCD1 and CD274. Given that the two IRGs are screened from plasma cell-related gene modules, the effects of PCs should be noted as increased plasma cell per se could be predictive of the OS in NSCLC patients treated with a PD-L1 inhibitor, atezolizumab (Patil et al. 2022). Furthermore, the extended OS associated with PD-L1 blockade is independent of CD8⁺ T cell signals (Patil et al. 2022). Further characterization of the effects of the two IRGs on plasma cell activation in experimental design may elucidate the roles of plasma cells played in PD-L1/PD-1 immunotherapy for LUAD.

Conclusion

In conclusion, the novel prognostic model constructed based on the two IRGs, P2RX1 and PCP4, is a prospective predictive biomarker for LUAD. The two IRGs are screened from plasma cell-related gene modules, and their function is associated with purinergic and Ca²⁺ signalling. Furthermore, the IRGs-based risk model has the potential to predict PD-L1/PD-1 immunotherapy for LUAD. Taken together, this study identified a novel genetic signature for the prediction of prognosis and efficacy of PD-L1/PD-1 blockade in LUAD, and may enhance our understanding of LUAD progression.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Junfeng Huang was involved in the project design; Bingqi Hu analysed the data; Bingqi Hu and Xingyu Fan contributed to the data visualization; Junfeng Huang contributed to writing—original draft; Liwen Chen assisted in writing—review and editing. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by The Natural Science Foundation of Anhui Province (Grant No. 1808085MH229) and Key Research and Development Program of Anhui Province (Grant No. 202004j07020027).

Availability of data and materials

The gene expression profiles and the clinical data of the subjects were downloaded from TCGA (The Cancer Genome Atlas Program-National Cancer Institute). The GSE30219 dataset was downloaded from the Gene Expression Omnibus (GEO) database (GEO-NCBI (nih.gov)).

Declarations

Conflict of interest

The authors declare that they have no competing interests.

Ethics approval and consent to participate

All public datasets utilized were generated by others who obtained ethical approval; therefore, no ethical approval was required for this work.

Consent for publication

Not applicable.