The gene expression profiles of PAAD samples and normal tissues were obtained from The Cancer Genome Atlas (TCGA) database (TCGA-PAAD, https://portal.gdc.cancer.gov/), the UCSC Xena project (http://xena.ucsc.edu/), and Gene Expression Omnibus (GEO) database (GSE28735 (12), GSE62452 (13), https://www.ncbi.nlm.nih.gov/geo/). The Xena project has recomputed all expression raw data from TCGA and GTEx (the Genotype-Tissue Expression, https://www.gtexportal.org/home/) based on a standard pipeline to minimize differences from distinct sources. The corresponding clinical information of PAAD patients from TCGA and methylation data were downloaded from the cBioPortal website (14) (https://www.cbioportal.org/). A total of 292 PAAD samples (TCGA: 178; GSE28735: 45; GSE62452: 69) and 277 normal tissues (TCGA: 4; GTEx: 167; GSE28735: 45; GSE62452: 61) were enrolled in this study.

Tumor Immune Estimation Resource (TIMER, http://timer.cistrome.org/) is a comprehensive resource for systematical analysis of immune infiltrates across diverse cancer types (15). By taking advantage of TIMER, we assessed the relationship between ASXL2 and the abundance of six types of immune infiltrating cells (B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils, and dendritic cells) in PAAD.

In addition, CIBERSORT algorithm was utilized to evaluate the fractions of 22 types of tumor-infiltrating cells to further confirm the effect of ASXL2 on tumor immune microenvironment (TIME) in PAAD (16). Only samples with P < 0.05 were included for further study.

Differentially expressed genes (DEGs) between two groups were screened by using “DESeq2” package in R software according to the thresholds of |log2FoldChange| > 1 and adjusted p-value (padj) < 0.05 (17). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed to annotate the biological functions of DEGs and ASXL2-related genes through “clusterProfiler” package (18). With the annotated gene sets in “h.all.v7.4.symbols.gmt” chosen as the reference gene sets, gene set enrichment analysis (GSEA) was conducted to investigate the potential regulatory mechanisms of ASXL2 (18, 19).

We estimated the IC₅₀ of common chemotherapeutic drugs via “pRRophetic” package (20). The potential response to immunotherapy was predicted and qualified by Tumor Immune Dysfunction and Exclusion (TIDE) algorithm via python (version 3.8.6) language (21). The lower TIDE score, the more sensitive to immune checkpoint blockade (ICB).

Biospecimens histopathologically confirmed PAAD were obtained from Shanghai East Hospital Biobank. A total of 10 PAAD samples and 5 para-cancerous tissues were collected for immunohistochemistry. All patients had signed informed consent for donating their specimens to Shanghai East Hospital Biobank. Biopsy samples were fixed with formalin followed by paraffin embedding. Immunohistochemistry was conducted using anti-ASXL2 antibody (1:200, Abcam). Images of slides were taken by a Leica microscope equipped with a digital camera at 10× and 10× magnification. Staining was independently evaluated by two experienced pathologists blinded to the clinical information of patients. The score for ASXL2 staining was based on the staining intensity and the percentage of positive cells. Staining intensity was quantified as follows: 0: no color; 1: yellow; 2: light brown; and 3: dark brown. The percentage of positive cells were scored as followings: 0: positive cells < 5%; 1: 6%- 25%; 2: 26%- 50%; 3: 51%- 75%; and 4: > 75%. The staining score was calculated as the score of staining intensity × the score of percentage of positive cells.

All statistical tests were conducted via R (version 4.0.3) software. Comparisons between two groups were conducted with Wilcoxon rank-sum test. Correlation analysis was performed using Pearson correlation test. Categorical variables between groups were compared with chi-square test. Kaplan-Meier survival curves were plotted to exhibit the overall survival (OS) for PAAD patients. Univariate and multivariate Cox regression analyses were employed to determine the independent prognostic significance of each variable enrolled in this study. P < 0.05 was considered statistically significant.

The RNA-seq data of 178 PAAD samples along with 4 adjacent normal tissues from TCGA and 167 normal samples from the GTEx were analyzed, and we found that the mRNA expression level of ASXL2 was distinctly high in PAAD samples (P = 1.31e-23, Figure 1A ). To investigate the prognostic significance of ASXL2 in PAAD patients, we divided the patients into high and low ASXL2 expression groups based on the optimal cutoff calculated via “survival” and “survminer” packages ( Supplementary Figure 1A ). The distribution of ASXL2 expression and survival status of PAAD patients were exhibited in Figure 1B . Kaplan-Meier survival curve demonstrated that PAAD patients with low ASXL2 expression levels showed longer survival time (P = 0.011, Figure 1C ). The ROC curve showed that ASXL2 expression had a certain power for predicting the 5-year-OS and PAAD patients with high ASXL2 expression tended to exhibit worse clinical outcomes ( Figures 1D, E ). Then, we investigate the correlation between ASXL2 expression and clinicopathological factors using logistic regression analysis and found that elevated ASXL2 expression levels in PAAD were correlated with tumor grade (G4 vs G1, P = 0.002) and ASXL2 DNA methylation (P < 0.001, Table 1 ). In addition, patients with PAAD were divided into high or low ASXL2 expression subgroups according to the median value and the chi-square test was conducted to further study the relationships between ASXL2 expression and clinical features. The results indicated that ASXL2 expression was significantly correlated with survival status (P = 0.049) and ASXL2 DNA methylation (P < 0.001, Supplementary Table 1 ). As shown in Figure 1F , the expression level of ASXL2 was negatively correlated with the methylation level (R = -0.33, P = 7e-06, Figure 1F ). Taken together, ASXL2 may be a potential prognostic biomarker for PAAD patients.

To explore the effect of ASXL2 on TIME, we took advantage of TIMER website. As shown in Figure 2A , ASXL2 expression exhibited a positive correlation with the fractions of B cells (P = 6.87e-09), CD8+ T cells (P = 6.34e-26), macrophages (P = 2.01e-16), neutrophils (P = 2.84e-08), and dendritic cells (P = 4.56e-14). What’s more, to further confirm the role of ASXL2, we employed CIBERSORT algorithm to estimate the levels of 22 types of immune cells. After removing PAAD samples with P ≥ 0.05, a total of 124 samples remained and were assigned into a high or low ASXL2 expression group based on the median value. When compared to samples with low ASXL2 expression, the proportions of T cells CD4 memory resting, mast cells resting, and neutrophils were distinctly increased, however, macrophages M0 in PAAD decreased ( Figure 2B and Supplementary Figure 2 ). Relationships between 22 kinds of immune cells indicated that T cells CD4 memory resting and mast cells resting were weakly positively correlated and both of them were negatively correlated with macrophages M0 ( Figure 2C ). These findings suggested that ASXL2 played a crucial role in immune infiltration in PAAD.

Since ASXL2 contributes to the TIME, we next got an insight into the potential mechanisms underlying ASXL2. A total of 387 differentially expressed genes (DEGs) between low and high ASXL2 expression groups were screened. Among these DEGs, 123 were upregulated, while 264 were downregulated in the high ASXL2 expression group ( Supplementary Figure 1B and Supplementary Table 2 ). Functional annotation including GO term and KEGG pathway analyses of these DEGs were conducted. GO annotation indicated that ASXL2-associated DEGs were mainly involved in processes like “negative regulation of peptidase activity” and “receptor ligand activity” ( Figure 3A ). Meanwhile, KEGG pathway analysis showed that pathways including “neuroactive ligand-receptor interaction”, “pancreatic secretion”, and “protein digestion and absorption” were significantly enriched ( Figure 3B ). As the expression level of ASXL2 was correlated with tumor grade and the prognosis of PAAD patients, we hypothesized that increased expression of ASXL2 promotes tumor progression. We employed GSEA and found that hallmarks of tumor such as “epithelial mesenchymal transition”, “inflammatory response” and “mitotic spindle” were dynamically correlated with the high ASXL2 expression, while “oxidative phosphorylation” was significantly enriched in the low ASXL2 expression group ( Figures 3C, D and Supplementary Table 2 ).

Adjuvant chemotherapy can distinctly improve the survival rate of patients with curative resection of pancreatic cancer (22). Drugs including paclitaxel, erlotinib, gemcitabine, mitomycin C, and sunitinib are demonstrated to be helpful in pancreatic cancer treatment (23–26). On the other hand, the efficacy of immunotherapeutic strategies is under assessment. To figure out the potential role of ASXL2 in clinical treatment decisions, we estimated the IC₅₀ of common drugs recommended for pancreatic cancer therapy by AJCC guidelines. Interestingly, patients in the low ASXL2 expression group tended to be more sensitive to chemotherapeutic drugs such as paclitaxel, erlotinib, and sunitinib, while patients with high ASXL2 expression were more likely to benefit from gemcitabine and mitomycin C ( Figure 4A ). Interactions between tumor cells and tumor-infiltrating immune cells affect the effect of ICB (21, 27). Based on the therapy, we calculated the TIDE score of each sample using gene expression profiles. The results showed that the level of TIDE score in the high ASXL2 expression group was lower than that in the low ASXL2 expression group, which indicated that PAAD patients with high ASXL2 expression levels were more likely to respond to ICB ( Figure 4B ). Our results suggested that ASXL2 may help therapeutic strategies making for PAAD patients.

To further validate the aberrant expression of ASXL2 in pancreatic cancer, we obtained additional gene expression data from GSE28735 and GSE62452. Compared with the adjacent normal tissues, ASXL2 was significantly upregulated in tumor samples ( Figures 5A, B ). In addition, Kaplan-Meier survival analyses showed that patients with low ASXL2 expression survival longer than those with high ASXL2 expression, which highlights the prognostic value of ASXL2 in PAAD ( Figures 5C, D ). We conducted immunohistochemistry staining of ASXL2 in PAAD tissues and adjacent normal tissues, and the results agreed with our findings ( Figures 5E, F ).