DOK1/3/6 Serve as Potential Prognostic Biomarkers and Immunotherapy Targets for Pancreatic Cancer


 Background: Downstream of kinase (DOKs), a family of adapter proteins, are frequently depicted as pivotal components of immune regulation complexes involved in the tumor progression in a wide range of cancers. Regrettably, little is known about the expression patterns and exact roles of 7 identified DOKs in pancreatic cancer (PC). Methods: In this study, we investigated the distinct expression and biological function of DOKs in PC using multiple public databases, including ONCOMINE, GEPIA, cBioPortal, and Kaplan-Meier plotter. The correlations between DOKs and cancer immune infiltrates was investigated via TIMER. In addition, we subsequently verified those in an independent cohort. Results: The expression levels of DOKs were found to be significantly upregulated in PC, interestingly higher DOK1/3/6 expressions were correlated with shorter overall survival (OS). Moreover, DOK1/2/3/5/6 had a dramatical positive correlation with the immune infiltration of PC and programmed cell death-ligand 1 (PD-L1). Conclusion: DOK1/3/6 may function as potential prognostic biomarkers and even promising immune checkpoints for PC immunotherapy.

complexes of protein tyrosine kinase (PTK) pathway, and mediate cytokine and immunoreceptor signaling in innate and adaptive immune systems [8][9][10] . Recent studies have noticed that abnormal expression of DOKs family members is associated with tumorigenesis. The elevated DOK1 in gliomas biopsies and cell lines positively facilitates platelet-derived growth factor (PDGF)-BB-mediated glioma cell motility via BCAR1-Rap1 axis [11] . DOK4/5/7 are over-expressed in acute myelocytic leukemia (AML), which may contribute to molecular classi cation of AML for precise treatment and prognostic assessment [12] . Nevertheless, the functions of distinct DOKs family members in PC still remains confused.
To evaluate the comprehensive value of DOKs in PC, multiple bioinformatic tools were used to anatomize the correlation between gene expression pattern and PC clinical features, and subsequently the data from bioinformatics analysis were well veri ed in our own cohort. Furthermore, the regulatory mechanisms of abnormal DOKs and the correlation between DOKs and immunologic microenvironment were also investigated. These results will provide a perspective to the unearth of prognostic biomarkers and novel molecular-targeted candidates for PC therapy.

Aberrant Expression of DOKs in PC
To investigate expression pro le of DOKs family in patients with PC, the mRNA expression of DOKs in pan-cancers were initially measured. From ONCOMINE analysis, DOK1/4/5 expression were dramatically uplifted in PC compared to normal tissues (Fig. 1). In Buchholz Pancreas dataset, 1.878-fold upregulation in DOK4 mRNA expression was found in PC samples (P = 2.63E-6) and Pei Pancreas showed that DOK4 was also elevated in PC tissues with a fold change of 2.444. Signi cant up-regulation of DOK5 transcription levels, with a fold change of 3.042, was similarly unveiled in Badea Pancreas dataset. Next, the mRNA expression patterns of 7 DOKs were further validated by GEPIA databases. The transcription levels of DOK1/2/3/4/5 were signi cantly up-regulated in PC tissues compared to normal samples, but no statistical difference in DOK6/7 expression ( Fig. 2A). Moreover, DOKs mRNA levels were higher in a panel containing 6 PC cell lines (MiaPaCa2, BxPC3, AsPC1, CFPAC1, HPAF-II, SW1990) than that in a HPDE cell line (Fig. 2B). ROC analysis was conducted to assess the power of DOKs in discrimination, and results demonstrated that DOKs have the promising ability to differentiate PC and non-cancerous tissues ( Figure S1).
IHC was also carried out in tumor tissues and their counterparts to assess DOKs protein expression. As shown in Fig. 3, DOKs protein primarily stained in the cell membrane and cytoplasm. Consistent with increase of transcription level, the protein level of DOKs was also elevated in PC, suggesting that DOKs protein was overexpressed in tumor compared to pericarcinomatous tissue.

Prognostic Value of DOKs in PC
In order to explore the critical effects of DOKs in the survival outcomes of PC patients, Kaplan-Meier Plotter analysis, a web-based tool, was utilized. As depicted in Fig. 4A, patients with higher DOKs expression had woeful outcome. The increased mRNA of DOK1 (HR = 0.59, 95% CI: 0.39-0.9, and P = 0.014), DOK3 (HR = 0.59, 95% CI: 0.36-0.98, and P = 0.041), DOK4 (HR = 1.86, 95% CI: 1.08-3.19, and P = 0.023) and DOK6 (HR = 0.56, 95% CI: 0.36-0.88, and P = 0.01) were strongly associated with poor overall survival (OS). However, the transcription levels of DOK2/5/7 were not related with OS. These results indicated that higher expression of DOK1/3/4/6 predictes worse outcome for PC patients. To corroborate these ndings, a prognostic evaluation was carried out in our cohort, and results were consistent with the aforementioned conclusion (Fig. 4B). DOK1/3/6 may have the potential to be clinical prognostic survival indicators as well as targeted therapies, which require further developed.

Predicted Functions and Pathways of DOKs in PC
We evaluated the mutation frequency of DOKs in PC using cBioPortal. DOKs were altered in 47 (27%) of the 177 patients queried. The percentages of genetic variations in DOKs family ranged from 3-7% (Fig. 5A). Next, the correlation of DOKs with each other was evaluated by analyzing their transcriptional expression. The results displayed a signi cant positive correlation between DOK1 and DOK2, DOK3 and DOK7, DOK2 and DOK3, DOK5 and DOK2, respectively (Fig. 5B). Further, a network of DOKs and their 20 related genes was constructed by GEPIA 2.0 (Fig. 5C).
To further appreciate the functions and regulatory mechanism of those related genes, GO annotation and KEGG analyses was conducted within clusterPro ler. Enrichments results revealed that biological process of those genes was mainly enriched in GTPase activity, regulation of leukocyte mediated immunity and T cell activation (Fig. 5D). Cellular component was largely focused on phagocytic cup, extrinsic component of cytoplasmic side of plasma membrane and extrinsic component of plasma membrane (Fig. 5E). Molecular function mostly participated in small GTPase binding, Ras GTPase binding and GTPase regulator activity (Fig. 5F). In KEGG analysis, we found that chemokine signaling pathway, Rap1 signaling pathway and B cell receptor signaling pathway were signi cantly associated with PC ( Fig. 5G).

DOKs Correlated with the Immune Microenvironment in PC
The number and activation state of TIICs affect the prognosis of tumor and the response to immune checkpoint blockers. The relationship between DOKs and TIICs was further probed by TIMER analysis. As depicted in Fig. 6A, elevated expression of DOK1-3, DOK5, DOK6 not only have a signi cant negative correlation with tumor purity, but signi cantly associated with TIICs ( Figure S2). However, there was no close connection between DOK4/7 and in ltrating immune cells, except a weak link of DOK4 with B cells and a weak negative link of DOK7 with macrophages. Moreover, we explored the relationship between DOKs and in ltration of various immune subtypes. The results revealed that DOK1-3/DOK5/DOK6 were closely related to CD8 + T cells, M2 macrophages, DC cells, and T exhaustion (Fig. 6B, Figure S3).

Discussion
Tumorigenesis is a result of dysregulation of various signal pathways. DOK family members exist in multiple tumor types via providing a docking platform for the assembly of multi-molecular signal transduction [16] . The roles of DOKs in tumorigenesis are still emerging and controversial. Some researches indicated that DOKs functionally act as anti-oncogenes that positively correlated with cancer patient survival [17,18] . Conversely, others reported that DOKs perform as oncogenes in tumor progression and motility. Tong et al. demonstrated that DOK1 was signi cantly higher in EBV + gastric cancer tissues compared with EBV-tissues [19] . And higher DOK4 mRNA was found in clear cell renal cell carcinoma biopsies compared with matched normal tissue [20] . Little work, as far as we know, has been done on the effects and mechanisms of DOKs in PC. The present research revealed, in both the transcription and protein levels, that all 7 DOKs family members were over-expressed in PC tissues. Furthermore, DOKs mRNA levels were higher in PC cell lines than that in normal cells. The protein expression was consistent with mRNA expression level, suggesting DOKs may serve as oncogenes promoting PC carcinogenesis and progression. High expression of DOK1/3/6 mRNA forecasted a poor prognosis in PC, but increased of DOK4 expression indicated good prognosis. Notably, DOK4 protein did not show visible relations with survival. Further investigation of DOK4 is needed for validation.
Mounting evidence suggests that DOKs negatively modulate the biological activity of immune cells in the context of physiology in human. The speci c subtypes of TIICs in TME may affect tumor cell survival, metastasis and resistance to treatments [21,22] . Nevertheless, little is known about the impact of these DOKs on tumor immune in ltration, which dictates the functional orientation of DOKs. To determine why DOKs predicted poor survival, we performed enrichment analysis of DOKs and relevant 20 genes in PC which were mainly involved in immunomodulatory. Next, the relationships between DOKs and TIICs were anatomized. Results implied that DOK1/2/3/5/6 were prominently associated with CD8 + T cell, Treg, T exhaustion, macrophages and dendritic cells (DCs) in PC. T cell immune response is the central event in antitumour immunity [23] . Studies have illustrated that TIICs (especially CD8 + T cells) are associated with better outcomes in multiple malignancies, such as breast, lung, melanoma, colorectal, and brain cancers [24,25] . Although DOK1/2/3/5/6 was strongly associated with CD8 + T cell, high expression of DOKs contributed to poor prognosis. It is feasible that DOKs were mainly related to T cell exhaustion which presented a state of cytotoxic CD8 + T dysfunction in the TME. Moreover, our results also con rmed this hypothesis. Abundant Treg cell in ltration into TME is associated with poor clinical outcomes in PC [26] . Treg cell depletion in combination with conventional chemotherapy has been demonstrated to enhance CD8 + T activation [27] . DCs are took into account the most e cient antigenpresenting cells (APCs) and the only type of cells at stimulating CD8 + T cells [28] . Despite the presence of DCs in TME and their potential to generate anti-tumor immunity, tumor-in ltrating DCs (TIDCs) often exhibit impaired or dysfunctional functions [29,30] . DOKs may also negatively regulate TIICs in PC, leading to immune evasion and tumor progression.
Epidemiological evidence suggested that tumor-associated macrophages (TAMs) were often associated with poor outcomes in different forms of cancer (e.g. pancreatic and breast cancer) [31,32] . TAMs were the most common TIICs in PC, which can be divided into M1 and M2 subtypes according to their state. M1 cells secrete pro-in ammatory factors to activate the in ammatory response and anti-tumor effects. TGFβ, IL-6 and other immunosuppressive factors secreted by M2 cells induce the dysregulation of T cell receptor signaling pathway, thereby exerting a tumor-promoting effect [33] . Interestingly, in animal models of PC, M2 cells were the major source of PD-L1 in TME and negatively regulated CD8 + T cells [34] . Our results con rmed that DOK1/2/3/5/6 emerged a signi cant correlation with M2, however, it was not associated with M1. DOKs may promote PC development and metastasis by promoting macrophages polarization from M1 to M2 phenotype. However, causality between DOKs and TAMs cannot be inferred in this study and the speci c mechanism needs to be further explored.

Numerous studies have revealed that bright prospects of immune checkpoint blockers (ICBs), such as CTLA-4 and PDL-1 blockade, to promote T-cell responses by preventing T-cell exhaustion, have bright
prospects. However, only part of patients with PC can bene t from ICBs clinically [35] . ICBs exert a crucial part in the activation and in ltration of cytotoxic CD8 + T cells after tumor antigen recognition. The classical TLA-4 and PD-1/PDL-1 axis modulate different inhibitory pathways and have non-overlapping mechanisms. Clinical trials have shown that more than 80% of patients with advanced melanoma have a better outcome after ICBs combination therapy [36] . Therefore, it is worth studying the immunoserotyping and regulation mechanisms of PC. In our research, DOK1/2/3/5/6 were co-expressed with CD247 and PDCD1, and high DOK1/2/3/5/6 expression in tumor accompanied with a high positive rate of PDL-1 protein. DOK1/2/3/5/6 may have the potential to be a new immune checkpoint molecule similar to PDL-1, but its role requires further research in future.

Conclusions
To summarize brie y, ndings gained in the present study for the rst time corroborate that increased DOK1/3/6 in PC predicted poor prognosis, and may mechanically act as a negative regulator of TIICs.
The importance of DOK1/3/6 in PC prognosis, and as potential immune checkpoints for immunotherapy with PC has been highlighted in current work, which has improved our understanding of DOKs' functions in PC. However, as a retrospective analysis still has limitations. Given a lack of effective evaluation of DOK in PC prognosis and treatment, a multi-center study with a large sample size is required. Moreover, biases need to be further veri ed by prospective experiments.

Oncomine Analysis
Transcriptional levels of DOKs in distinct cancers were con rmed by excavating the ONCOMINE database (http://www.oncomine.org), the largest database of cancers integrated microarray data from published literatures. To compare the differential expression of DOK1-7 between tumors and normal tissues, Students t test was conducted. The cut-off criteria were restricted as follows: P < 0.05, a 2-fold change and top 10% gene rank.  Table S2.

The Kaplan-Meier Plotter
The prognosis value of DOKs in PC was appraised using the Kaplan-Meier plotter (www.kmplot.com), a user-friendly database contained gene expression pro les and follow-up data, which is widely applied in tumor prognosis analyses. On the basis of median values of speci c gene expression as the cut-off criteria in this tool, patients were divided into 2 groups: the high expression group and the low expression group. The results were presented in Kaplan-Meier survival plots with HRs, 95% CIs, P-values and numberat-risk cases [15] . cBioPortal and GEPIA 2.0 The online software cBioPortal (www.cbioportal.org) is based on other authoritative databases and subjected to scheduled update. In this study, Mutation maps polifts of DOKs were generated using cBioPortal. A network of DOKs and their 20 related genes was further constructed by GEPIA 2.0 (www.gepia2.cancer-pku.cn/).

Enrichment Analysis
In this study, Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were performed on DOKs and their 20 related genes using clusterPro ler in R (17). The assigned GO terms were mainly classi ed into 3 categories, including biological process (BP), cellular component (CC), and molecular function (MF), which predicted the function of DOKs and their 20 related genes. The P value < 0.05, adjusted by Benjamini-Hochberg method was used.

TIMER Analysis
Tumor in ltrating lymphocytes are associated with tumor genesis and outcome. TIMER, a reliable database for online analysis (https://cistrome.shinyapps.io/timer/), integrates high-throughput sequencing data of 32 diverse cancers in TCGA to calculate the abundance of TIICs in tumor tissues according to gene expression pro le. We used the TIMER to assess the correlation of DOKs with 6 different typical tumor-associated immune cells. Heatmap showed the purity-corrected partial Spearman's correlation and statistical signi cance. In view of the speci c genetic markers expressed by immune cells in each state, we then calculated the association between DOKs and each TIICs based on the genetic markers set, as described in a previous study (Assessment of the expression of the immune checkpoint molecules PD-1, CTLA4, TIM-3 and LAG-3 across different cancers in relation to treatment response, tumor-in ltrating immune cells and survival).

Statistical Analysis
Receiver operating characteristic (ROC) analysis using pROC package generated a score to differentiate PC and non-cancerous tissues. The prognosis value of DOKs protein was analyzed by Kaplan-Meier analysis using survminer R package (https://CRAN.R-project.org/package=survminer). All hypothetical tests were two-tailed and P values < 0.05 were considered statistically signi cant. University. This study was rati ed by the Ethics Committee of Shanghai Outdo Biotech Company. All participants offered written informed consent before surgery. The study conforms to the provisions of the Declaration of Helsinki. The data sets analyzed in this study are available on the public databases.

Consent for publication
Written informed consent for publication was obtained from all participants.

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
No potential con icts of interest were disclosed.