3.1 NRP Expression Analysis in PAAD and Other Cancers
First, data from tumor samples (TCGA database) and normal samples (GTEx database) were integrated to analyze the differences in the expression of NRPs in 33 tumors. The results showed that both NRP1 and NRP2 were significantly more expressed in all 17 tumors (Fig. S1A), including PAAD (P < 0.001, Fig, 1A), than in normal tissues. Second, the differential expression of NRPs at the protein level was explored via western blotting (Fig. 1B) and immunohistochemistry (Fig. 1C), which suggested that the expression of NRP proteins in PAAD tumor tissues (T) was higher than that in the adjacent normal tissues (N). In the HPA database, the expression of NRP1 was characterized by weak to moderate cytoplasmic positivity often with a granular pattern in most cancer tissues, but this was negative in carcinoids and several cases of skin and cervical cancer (Fig. S1B). NRP2 was characterized by skin cancer, several urothelial cancers, and a few lung cancers showing moderate to strong cytoplasmic and/or membranous positivity, but cervical cancers along with several colorectal, gastric, pancreatic, and liver cancers showed moderate cytoplasmic and/or membranous immunoreactivity with additional nuclear membranous staining in several cases (Fig. S1B). In addition, we created a PPI network and identified the top 20 relevant interacting proteins (Fig. S2A, B). Finally, to evaluate the expression of NRPs in the tumor microenvironment (TME), we conducted an mIHC experiment on human PAAD tumor samples, which showed that NRP1 and NRP2 were widely expressed in numerous cells types, including tumor cell (panCK+), cancer-associated fibroblasts (CAF, α-SMA+), tumor-associated macrophages (TAM, CD68+), CD8+ T cell (CD8+), and CD4+ T cell (CD4+) (Fig. 2A-B), and the specific statistics are shown in Figure 2C-D. Moreover, the Tumor Immune Single-cell Hub (TISCH) database was used to evaluate the expression of NRPs in TME from the perspective of single-cell RNA sequencing, which also showed that NRPs were expressed to varying degrees in tumor cells, immune cells, and stromal cells (Fig. 3).
3.2 NRPs Are Valuable Diagnostic and Prognostic Biomarkers
We further investigated the prognostic significance of NRPs in patients with cancer. In the log-rank test, the overall survival (OS) results revealed that NRP1 acts as a risk factor for patients with CSEC, HNSC, LIHC, LUSC, PAAD, PCPG, SARC, STAD, THCA, and CHOL, and as a protective factor in patients with KIRC. NRP2 acts as a risk factor for patients with BLCA, BRCA, HNSC, KIRP, OV, PAAD, STAD, and UCEC, and as a protective factor for patients with THCA and CHOL. Moreover, the recurrence-free survival (RFS) results revealed that NRP1 acts as a risk factor for patients with BRCA, CESC, ESCA, HNSC, LIHC, OV, PAAD, READ, STAD, THYM, and PRAD, and as a protective factor for patients with LUSC and UCEC. NRP2 acts as a risk factor for patients with BRCA, CESC, HNSC, KIRP, LUAD, PAAD, READ, and STAD, and as a protective factor for patients with ESCA, LIHC, and UCEC (Fig. 4A). It is noteworthy that the high expression of NRP1 and NRP2 was associated with a poor prognosis in PAAD, in terms of OS or RFS, and this result was obtained using the Kaplan–Meier method (Fig. 4B). Next, we estimated the diagnostic performance of NRPs using receiver operating characteristic (ROC) curves. As expected, NRPs showed significantly high sensitivity and specificity for the diagnosis of various cancers, especially PAAD, and NRP1 also showed high a diagnostic value in CHOL; this result is consistent with our previous research results [34] (Fig. 4C, Fig. S2C).
3.3 Associations of the NRP Family and Immune Infiltration
Recent studies have shown that NRPs are expressed in various subsets of immune cells and are important for regulating immune responses [14; 16; 18; 23; 26]. To further explore the role of NRPs in tumor immunology, we determined the correlation between NRP expression and immune cell infiltration in PAAD and pan-cancer. In the ssGSEA algorithm (Fig. 5), the expression of NRPs was significantly correlated with a variety of immune cell infiltrations in pan-cancer. Notably, the expression of NRPs was positively correlated with the infiltration of myeloid immune cells (e.g., monocytes, macrophages, mast cells, and myeloid dendritic cells) in nearly all cancers, including PAAD, suggesting that NRPs play a vital role in the regulation of innate immunity. For lymphoid immune cells, the expression of NRP1 was negatively correlated with the infiltration of B, T helper (Th) 1 cells, CD4+T central memory (Tcm), and NKT cells in almost all cancers, and NRP2 presented similar results. These lymphocytes may also be the main mediators of NRP-regulated tumor immunity.
PAAD has an intrinsically complex TME. To verify and supplement the above results, we further investigated the correlation between NRP expression and immune cell infiltration in PAAD using the other two algorithms, MCP-counter (Fig. S3A, B) and QuantTIseq (Fig. S3C, D) [30]. These chordal results showed a positive correlation between the expression of NRP and infiltration of macrophages, natural killer (NK) cells, dendritic cells (DCs), CAF, regulatory T (Treg) cells, and CD8+ T cells (R > 0.2, P < 0.05), but negatively correlated with the infiltration of CD4+ Tcm cells (R < -0.2, P < 0.05). Interestingly, the infiltration of type 1 macrophages (M1), CD8+ T central memory (Tcm) cells, and mast cells (R = -0.288, P = 0.031) was only significantly correlated with NRP1 expression (Fig. 5A, Fig. S3A, C). In comparison, the infiltration of Th2 cells was only significantly correlated with NRP2 expression (Fig. 5B, Fig. S3B, D). Given the important influence of non-tumor components on PAAD, we also investigated the total abundance of immune and stromal cells in individual PAAD samples using the ESTIMATE method, and found that the immune, stromal, and estimated scores were higher in the high NRP1 expression group (P < 0.001, Fig. S3E), but NRP2 expression was only significantly correlated with the stromal score (P < 0.05, Fig. S3F). These findings suggest that the NRP family has a complex effect on the immune microenvironment of PAAD from innate to adaptive immunity, shifting the balance between immunosuppression and activation.
3.4 Association between NRPs and Immune Checkpoints
Of further interest is whether NRP expression is associated with immune checkpoints. We collected 11 common immune checkpoint genes, the correlation of which with NRP expression was assessed using the TIMER2.0 database [15]. In nearly all cancers, NRP1 expression was positively correlated with the expression of most immune checkpoint genes. The results in the NRP2 group were similar to those in the NRP1 group, but in SARC and UVM, NRP2 expression was negatively correlated and non-statistically correlated, respectively, with the expression of most immune checkpoint genes (Fig. 6A). Figure S4 shows the details of the correlation between NRP expression and immune checkpoint gene expression in PAAD. These results suggest that NRP1 and NRP2 are involved in tumor immune evasion by interacting with immune checkpoints.
3.5 Association of NRPs and Tumor Immunotherapy
ICB, the most important tumor immunotherapy, has been shown to considerably improve antitumor efficacy, but the response of pancreatic cancer to ICB is not promising. Here, we predicted the response of different NRP expression levels to immune checkpoint inhibitors in PAAD using the TIDE algorithm, a tool to evaluate the dysfunction of tumor-infiltrating cytotoxic T lymphocytes and the rejection of it by immunosuppressive factors based on gene expression [11]. Interestingly, the high NRP1 and NRP2 expression groups had higher tide scores, which means that the curative effect of ICB was poor and the survival time after ICB treatment was short (Fig. 6B, C). These findings highlight the possibility that NRP1 and NRP2 can predict the efficacy of tumor immunotherapy.
3.6 In Vitro and In Vivo Experiments and Functional Enrichment of NRPs in PAAD
As described in the Methods section, we constructed the NRP1 knockdown PANC-1 cell line and NRP2 knockdown CFPAC-1 cell line in vitro (Fig. S5A-D). Compared with normal PAAD cells, the proliferative (Fig. 7A, B), invasive (Fig. 7C, D), and migratory (Fig. 7E, F) capacity of NRP-knockdown PAAD cells was impaired. Furthermore, the effects of the NRP1 inhibitor on the in vivo growth and progression of PAAD were investigated using the orthotopic pancreatic tumor-bearing nude (immunodeficient) and C57BL/6 (immunocompetent) mouse models (Fig. 8). The results revealed that the NRP1 inhibitor suppressed tumor growth (P < 0.001, Fig. 8B) and prolonged survival (P = 0.01, Fig. 8D) compared with those in the control group of immunocompetent C57BL/6 mice. In immunodeficient nude mice, although no significant difference was observed in the survival time between the NRP1 inhibitor group and control group (P = 0.059, Fig. 8C), some degree of improvement was observed, and the NRP1 inhibitor group showed some degree of reduction compared with the control group (P < 0.05, Fig. 8A). These findings of in vivo experiments suggest that NRP1 depletion exerts anti-tumor effects and improves survival mainly via immune-related pathways.
In addition, we performed DEG analysis in silico using the DESeq2 R package. There were 929 and 683 DEGs for NRP1 and NRP2 expression, respectively, and the upregulated and downregulated genes are shown in volcano plots (|log2(FC)| > 2, adjusted P < 0.05) (Fig. 9A, B). Given a large number of DEGs, only the top 20 upregulated and downregulated genes with the greatest differences have been presented in heat maps (Fig. 9C, D). To explore the potential functions of NRP1- and NRP2-interactive DEGs, we performed functional enrichment analyses using GO/KEGG and GSEA. In addition to several known functions, such as nervous system development, angiogenesis, and lymphangiogenesis [24; 25], some immune-related functions were also revealed. Specifically, NRP1 was significantly associated with primary immunodeficiency, myeloid leukocyte migration (Fig. 9E), interleukin interactions, and inflammatory response regulation (Fig. S6). NRP2 was found to be associated with the TGF-β signaling pathway (Fig. 9F). Interestingly, NRP1 was also involved in the PI3K-Akt signaling pathway and epithelial-mesenchymal transition (EMT) (Fig. S6). NRP2 was also involved in the Notch signaling pathway (Fig. S6) and transmembrane signal transduction (Fig. 9F). Further in vivo studies are needed to verify whether NRPs can affect PAAD progression via these pathways, including, but not limited to, proliferation, invasion, and migration.