The Diagnostic And Prognostic Value of DYNLL1 And RAN In Lung Adenocarcinoma

Background: Lung adenocarcinoma (LUAD) is the most heterogeneous and aggressive among all NSCLC subtypes. This work aimed to determine the predictive ability of DYNLL1 and RAN in LUAD, as well as to explore the role of possible DYNLL1 and Ran molecular mechanisms on LUAD malignant behavior. Methods: Using data from the Cancer Genome Atlas and the Genotype-Tissue Expression, we compared the expression of DYNLL1 and its related gene RAN in LUAD and normal lung tissues. We also studied the prognostic value of DYNLL1 and RAN in terms of overall survival (OS) in LUAD, liver cancer and renal cancer patients. Using data from the online database STRING, we further investigated the protein–protein interaction (PPI), miRNA and transcription factor regulatory networks of DYNLL1 and Ran. Then, molecular docking was used to obtain accurate combination models between molecular compounds and DYNLL1 or Ran. We also studied whether DYNLL1 and RAN inuence the proliferation and migration of tumor cells with the stable knockdown LUAD cell lines. Results: The expression of DYNLL1 and RAN was signicantly upregulated in LUAD tissues compared to normal controls. And high expressions of DYNLL1 and RAN were associated with unfavorable overall survival in LUAD patients. We also found that knockdown of DYNLL1 and RAN inhibit the proliferation and migration of A549 tumor cells. Conclusions: Overall, our study provides evidence that DYNLL1 and RAN are essential for the progression of LUAD. And they may serve as potential tumor biomarkers for the diagnosis and prognosis of patients with LUAD.


Introduction
As the most common type of non-small-cell lung cancer (NSCLC), lung adenocarcinoma (LUAD) accounts for more than 50% of NSCLC [1]. The morphological and molecular classi cation of LUAD has undergone substantial changes in recent years, resulting in a classi cation scheme more in tune with patient prognosis and survival. Earlier studies revealed that activation of the EGFR, KRAS, and ALK genes de nes three different pathways that are responsible for a considerable fraction of LUAD [2]. Currently, some of the newer factors have found targetable drug molecules. Meanwhile, part of them have been included in current standard of clinical practice. This change in practice requires further development of detection methods [3]. Consequently, it is extremely important to identify molecular biomarkers that predict the diagnosis and prognosis of patients with LUAD, and potential new targets for molecular inhibition [4][5][6].
The LC8 family members of dynein light chain (DYNLL) were regarded as cargo adapters of molecular motors in some early studies [7][8][9]. And now, many studies have shown that they can act as hub proteins, interacting with short linear motifs located in intrinsically disordered protein fragments [10]. The LC8 family members of dynein light chain 1 (DYNLL1) is highly conserved among different species and widely expressed in many tissues [11]. It has been considered to be the main regulator of several cellular functions ranging from intracellular tra cking to apoptosis. For instance, the association of DYNLL1 and neuronal nitric oxide synthase can inhibit the pro-apoptotic function [12]. DYNLL1 also interacts and interferes with BimL which belongs to the pro-apoptotic Bcl-2 family [13]. Previous studies have suggested that overexpression and phosphorylation of DYNLL1 promote the growth of breast cancer cells. DYNLL1 phosphorylation by p21-activated kinase 1 has been shown to promote the survival and growth of oestrogen receptor (ER)-positive breast cancer cells [14,15]. On the other hand, the downregulation of DYNLL1 compromises the ER-nuclear accumulation and also its transactivation activity, suggesting that DYNLL1 has a potential chaperone-like activity in the nuclear translocation of ER.
These data de ne an important role of DYNLL1-ER interaction in supporting and strengthening ERinduced cellular responses in breast cancer cells [16].
Ras-related nuclear protein (Ran) is a small GTPase and a member of the Ras superfamily. As a molecular switch that modulates the GTP/GDP cycle, Ran functions with a variety of Ran-binding proteins to control a wide range of fundamental cellular functions, including mitotic spindle assembly, nucleocytoplasmic transport, nuclear pore complex formation and nuclear envelope, which nally in uence cell fate determination, including cell proliferation, differentiation, malignant transformation and cell death [17]. Overexpression of Ran GTPase has been observed in various cancer types, including colon, stomach, kidney, ovarian and pancreas cancer [18][19][20]. The upregulation of Ran expression may be a vital event in cell transformation and tumor progression [18]. Genetic instability is a well-known hallmark of cancer. Meanwhile, Ran GTPase plays a huge role in the mitotic instability of cancer cells [21]. Many studies have shown that the unusual localization of tumor suppressor proteins or oncogenes can be affected by Ran GTPase signaling in diverse types of cancer [22]. In detail, the localization of Ran-GTP is in the nucleus whereas Ran-GDP is restricted in the cytosol of cells. The GTP/GDP switch permits Ran to migrate across nuclear pore complexes and carry mRNA and protein along with it. Any instabilities in Ran expression has been shown to cause the abnormal transport of some tumor oncogenes such as NF-κB or Akt [23].
It has been reported that DYNLL1 is required for normal lung morphogenesis and ciliogenesis [24].
Meanwhile, RAN is associated with the occurrence and progression of many different types of cancer.
However, the role of DYNLL1 in LUAD and its relationship to RAN has not yet been explicated. In this study, we found that the expression level of DYNLL1 and RAN was higher in LUAD tissues than normal lung tissues. And the high expression of DYNLL1 and RAN was correlated with poor overall survival in lung adenocarcinoma, liver cancer and renal cancer patients. Knockdown of DYNLL1 and RAN inhibit the proliferation and migration of A549 tumor cells. Collectively, DYNLL1 and RAN may be powerful biomarkers to support the diagnosis and prognosis of lung adenocarcinoma as well as potential tumor treatment targets.

Methods
Data mining. GEPIA2 (http://gepia2.cancer-pku.cn) [25], Xena Browser (http://xena.ucsc.edu/) [26] and The Human Protein Atlas (https://www.proteinatlas.org/), three online interactive web server for analyzing the RNA sequencing data of tumors and normal samples from the TCGA and GTEx projects, were used to analyze the expression pro les and prognostic value of DYNLL1 and RAN. The clinicopathological information, including age at initial pathologic diagnosis, smoking history, gender, pathologic stage, recurrence status, relapse-free survival (RFS) in months, overall survival (OS) status, and OS in months was downloaded from the TCGA database.

Cell line
The A549 human lung adenocarcinoma cell line (RRID: CVCL_0023) was were obtained from the American Type Culture Collection and maintained in Dulbecco's modi ed Eagle's medium (DMEM) basic medium supplemented with 10% fetal bovine serum and 1% antibiotics at 37 °C with 5% CO 2 . The protein bands were quanti ed relative to β-actin expression using ImageJ software (NIH, Bethesda, MD, USA, RRID:SCR_003070).

Cell proliferation analysis
Control and the DYNLL1 and RAN knockdown A549 cells (2 × 10 3 cells per well) were plated in 48-well plates and cultured for 72 h. Cell proliferation was determined using the colorimetric Cell Counting Kit-8 (CCK-8, Dojindo Laboratories, Kumamoto, Japan).

Cell migration analysis
A total of 5 × 10 5 cells were added in a volume of 100 μl (for ow conditions 340 μl) to the upper chamber of a 24-well transwell plate with a 8 μm pore in the insert (Corning International, Corning, NY). Lower wells contained 10% fetal bovine serum. The cells that migrated to the lower well after 6 hours were xed and stained with crystal violet and counted with microscope. Each experiment was performed in triplicate at least three times.
The diagnostic value of the expression levels of DYNLL1, RAN, NKX2 and NAPSA in LUAD was studied by analyzing the expression data from 515 LUADs and 347 normal tissues. Speci city and sensitivity were plotted on the x-and y-axes, respectively. The area under curve (AUC) was calculated to assess the ability of the expression levels of DYNLL1, RAN, NKX2 and NAPSA to predict the outcome of patients with LUAD.

PPI network analysis and Functional enrichment
Protein-protein (PPI) interactions network can visualize the patterns of molecular interactions and help to explain the mechanisms underlying phenotypes. PPI network analysis was performed using the online database STRING (https://string-db.org/) [27]. And the Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were analysed by STRING.
The iRegulon (http://apps.cytoscape.org/apps/iRegulon) plugin [30] in Cytoscape was performed to predict and analyze the interaction pairs of target genes-transcription factor (TF) in the PPI network. The parameters were set as follows: 0.01 act as the minimum identity among orthologous genes, and 0.001 serve as the maximum false discovery rate on motif similarity. The higher score of Normalized Enrichment Score (NES) in output results represented the more reliable results. The TF-target interaction pairs whose NES > 4 were selected for the network construction.

The molecular docking for DYNLL1 and RAN
The Connectivity map database (https://www.broadinstitute.org/conne ctivity-map-cmap), designed by Broad Institute of MIT and Harvard, was used to search the small molecular compounds which have potential roles with DYNLL1 and RAN [31,32]. To further understanding the docking mode of drug target, we obtained 3D structures of DYNLL1 and RAN from RCSB PDB (http://www.rcsb.org) and 3D structures of the small molecules which have potential roles with DYNLL1 and RAN from PubChem (https://pubchem.ncbi.nlm.nih.gov). Then we simulated a molecular docking using a web server Swissdock (http://www.swissdock.ch/docking), which was designed by Swiss Institute of Bioinformatics to predict the molecular interactions [33]. The 3D structures of DYNLL1 or RAN were uploaded as "Target selection" and those of small molecules as "Ligand selection" prior to a result report. Then, we used the USCF Chimera software (version 1.15, RRID:SCR_004097), a molecular modeling system, to calculate the possible binding modes and present an interactive visualization of data from Swissdock. The best docking simulation was ranked by the parameters of FullFitness and Estimated Energy.

Statistical Analysis
All statistical data were analyzed using SPSS 22.0 (SPSS, Chicago, IL, USA, RRID:SCR_002865) or GraphPad 7.0 (GraphPad Software, San Diego, CA, USA, RRID:SCR_002798). Chi square tests were used to analyze the associations between DYNLL1 and RAN expression levels and clinicopathological factors in the LUAD patients. Receiver operating characteristic (ROC) analysis were performed using SPSS 22.0 (Chicago, IL, USA). Diagnostic ability of the prediction model was evaluated, by calculating the area under a ROC curve. The ROC curve was drawn by plotting sensitivity against the false-positive rate. The cutoff value of DYNLL1 and RAN were calculated using the Youden index (speci city + sensitivity -1). Kaplan-Meier curves of OS and RFS were generated using GraphPad 7.0 (GraphPad Software, San Diego, CA, USA). Data were presented as mean ±SD. P values < 0.05 were considered statistically signi cant.

Results
The correlation between the expression of DYNLL1 and RAN in LUAD In order to nd potential tumor biomarkers for the diagnosis and prognosis of patients with LUAD as well as new factors involved in the development of LUAD, we analysis the top factors which in uence the relapse-free survival and overall survival in LUAD through the GEPIA2 database. And we found that DYNLL1 was among the top factors which signi cant in uence the overall survival time in LUAD.
Meanwhile, there were few researches focused on the role of DYNLL1 in cancer development, especially in lung cancer. Furthermore, the expression of DYNLL1 was signi cantly upregulated in LUAD samples than adjacent normal tissues (Fig. 1A). Further analysis showed that the expression level of RAN had positive correlation (r 2 = 0.248, P < 0.0001) with DYNLL1 (Fig. 1B). LUAD patients with higher expression levels of DYNLL1 always had higher expression levels of RAN. Meanwhile, the expression of RAN was also upregulated in LUAD samples (Fig. 1C). Heatmap also showed that DYNLL1 and RAN expression was signi cantly higher in LUAD tissues than in normal tissues (Fig. 1D).
Diagnostic value of DYNLL1 and RAN for LUAD patients Since the expression levels of DYNLL1 and RAN were upregulated in LUAD samples compared with controls, we next explored whether DYNLL1 and RAN may serve as potential diagnostic biomarkers in LUAD. ROC curve analysis was used to investigate the diagnostic value of DYNLL1 and RAN in distinguishing LUAD patients from normal conditions. The results showed that the AUC of DYNLL1 was 0.818 and the optimal cut-off value was 11.95, with a sensitivity of 76.5% and speci city of 71.1% (P < 0.0001) ( Fig. 2A). For RAN, the AUC was 0.994 and the cut-off value was 12.82, with a sensitivity and speci city of 95.3% and 92.9%, respectively (P < 0.0001) (Fig. 2B). As positive controls, we also analysed the ROC curves of two well-known diagnosis biomarkers, thyroid transcription factor 1 (TTF-1/NKX2-1) and napsin-A (NAPSA), in LUADs. The results indicated that the performance of the NKX2-1 [AUC, 0.570; 95% CI (con dence interval), 0.528-0.612] (Fig. 2C) and NAPSA (AUC, 0.515; 95% CI, 0.473-0.557) (Fig. 2D) were not satisfactory than DYNLL1 or RAN, which showed lower sensitivity and speci city than DYNLL1 and RAN. These results indicate that DYNLL1 and RAN have huge potential diagnose value for LUAD.
High DYNLL1 and RAN expressions were independent prognostic factors for poor OS and RFS in LUAD The associations between DYNLL1 and RAN expression and the demographic and clinicopathological parameters in patients with LUAD were summarized in Table 1. The results showed that the high DYNLL1 expression group had signi cantly higher mutation numbers (42/101, 41.6% vs. 35/129, 27.1%, P = 0.013) than the low DYNLL1 expression group. In addition, the high RAN expression group had a signi cantly higher proportions of male (103/196, 52.6% vs. 137/321, 42.7%, P = 0.035) and advanced neoplasm disease stage (53/193, 27.5% vs. 57/316, 18.0%, P = 0.011) compared with the low RAN expression group. In LUAD, the expression of DYNLL1 showed no differences among different pathological stages (Fig. 2E). Strikingly, RAN expression gradually enhanced with the increase of pathological stages (Fig. 2F). By conducting multivariate analysis, we found that, in addition to clinical stage, the expression levels of DYNLL1 and RAN were all independent prognostic factor for poor OS (HR = 0.694 and P = 0.014 for DYNLL1; HR = 0.575 and P < 0.001 for RAN) and RFS (HR = 0.543 and P = 0.002 for DYNLL1; HR = 0.521 and P = 0.001 for RAN) in LUAD patients (Table 2).  4. Current reformed smoker (for≤15 years); 5. Current reformed smoker (duration not speci ed).

High DYNLL1 and RAN expression was associated with unfavorable outcomes in LUAD patients
In order to reveal association between DYNLL1 and RAN gene expression levels and LUAD prognosis, we extracted the survival data in TCGA database. Next, we performed Kaplan-Meier survival curves and the cancer patients were divided into high/low expression group by using the best cutoff model. The results showed that patients with lower expression levels of DYNLL1 and RAN have better overall survival prognoses than those with higher expression levels of these two genes in LUAD (P = 0.0028, Hazard Ratio = 0.629 and P = 0.0005, Hazard Ratio = 0.600, respectively) ( Fig. 3A and Fig. 3B). Higher expression level of DYNLL1 also signi cantly associated with shorter relapse-free survival time (P = 0.0070, Hazard Ratio = 0.675) (Fig. 3C). Furthermore, the patients with both high expression of DYNLL1 and RAN showed the worst overall survival time (P < 0.0001, Hazard Ratio = 0.486) (Fig. 3D).
High DYNLL1 and RAN expression was associated with unfavorable outcomes in liver and renal cancer patients We also investigated whether DYNLL1 and RAN expression could in uence the patients' overall survival time in other 17 kinds of cancer types, including glioma, thyroid cancer, colorectal cancer, breast cancer and so on. Notably, results showed that the high expression of DYNLL1 and RAN had signi cantly worse OS compared to the low expression groups both in liver and renal cancer patients, respectively (P < 0.0001 for all, Hazard Ratio = 0.497, 0.506, 0.414, 0.506, respectively) (Fig. 4A, 4B, 4C and 4D). These results indicate that DYNLL1 and RAN participate in cancer development in a variety of cancer types.

Protein-protein interaction network analysis
In order to study how DYNLL1 and RAN were involved in the development of LUAD, STRING was performed to construct the protein-protein interactions between DYNLL1 and RAN. In the PPI network, DYNLL1 and RAN had close connections through RANBP1/2, RANGAP1, NUTF2 and XPO1/5 (Fig. 5). Furthermore, we did functional analysis based on the PPI network. The top processes and pathways were showed in Table 3. The analysis revealed that DYNLL1 and RAN may interact through intracellular protein transport pathways and participate the microtubule-based process and mitotic cell cycle process. MicroRNAs (miRNAs) play critical roles in tumor progression. So, we analysed the miRNAs which have potential roles to regulate the expression of DYNLL1 and RAN. As shown in Fig. 6B-6C, DYNLL1 interacts with 10 miRNAs (hsa-miR-92a-3p, for example), RAN interacts with 97 miRNAs (has-miR-197-3p, for example).
Identi cation of candidate drugs target to DYNLL1 and RAN The next question was to identify candidate drugs that were available to target DYNLL1 and RAN. Based on Connectivity map database, we screened the small molecular compounds which have potential roles with DYNLL1 and RAN. We found that ampiroxicam, 5-iodotubercidin, and pidorubicine had the highest possibility to interact with DYNLL1. And SCH-79797, GBR-12935, and ouabain had the highest possibility to interact with RAN. Then, the compounds were used to construct the molecular docking model with DYNLL1 and RAN, respectively. The Swissdock was used to generated a series of docking modes and USCF Chimera software was used to present an interactive visualization of data from Swissdock. The docking mode of the complex were shown in Fig. 7.

Knockdown of DYNLL1 and RAN inhibit the proliferation and migration of A549 tumor cells
To further understand the role of DYNLL1 and RAN in the development of LUAD, we generated stable DYNLL1 and RAN knockdown cells by transfection with DYNLL1 and RAN -speci c short hairpin RNAs (shRNAs), respectively. The knockdown e ciency of DYNLL1 in A549 cells were con rmed by both qRT-PCR and Western blotting (Fig. 8A-8B). And DYNLL1-shRNA-1 and DYNLL1-shRNA-5 showed higher knockdown e ciency (Fig. 8A-8B). Moreover, we investigated the effect of DYNLL1 knockdown on A549 cell proliferation. The results showed that knockdown of DYNLL1 signi cantly inhibited the proliferation of A549 tumor cells (Fig. 8C). Meanwhile, the knockdown e ciency of RAN in A549 cells were showed in Fig. 8D-8E and RAN-shRNA-3 and RAN-shRNA-5 showed higher knockdown e ciency. Furthermore, the knockdown of RAN also restrained the proliferation of A549 tumor cells (Fig. 8F). We next explored whether DYNLL1 and RAN in uence the migration of A549 tumor cells. And the results showed that both knockdown of DYNLL1 and RAN in A549 cells suppressed the migration ability of tumor cells (Fig. 8G-8H). These results indicated that DYNLL1 and RAN may promote the proliferation and migration of tumor cells in the development of LUAD.

Discussion
In this study, we identi ed the aberrantly expressed DYNLL1 and RAN involved in LUAD through the comparison of RNA expression pro les in cancerous tissues with that of normal lung tissues based on validation from TCGA and GTEx datasets. Additionally, we discovered that the expression of DYNLL1 and RAN signi cantly predicted the overall survival time in patients with LUAD. And knockdown of DYNLL1 and RAN inhibit the proliferation and migration of A549 tumor cells.
Previous studies have suggested that DYNLL1 is a physiologic substrate of p21-activated kinase 1 (Pak1). And the DYNLL1-Pak1 interaction plays a critical role in breast cancer cell survival [14].
Additionally, as an inhibitor of DNA end resection, DYNLL1 also signi cantly in uences genomic stability and response to chemotherapy-induced DNA damaging. The loss of DYNLL1, induced DNA end resection and restored homologous recombination (HR), was shown to contributed to the cisplatin and olaparib resistance in BRCA1-de cient high-grade serous ovarian carcinoma (HGSOC) cells [34]. Another study showed that DYNLL1 could act as a protein hub for the 53BP1 oligomerisation and its recruitment to DSBs [35]. These studies led us to highlighting a novel role of DYNLL1 in cancer progression rather than a motor that transports cargo along microtubules.
Ran GTPase plays important roles in cell cycle progression through the regulation of the microtubule polymerization, mitotic spindle formation, and cytoskeleton organization during mitosis. Consequently, dysregulation of Ran could therefore lead to the microtubule dysfunction, aneuploidy and disruption in cell migration [36,37]. Some study has shown that high expression of Ran GTPase promotes the invasion and metastasis in human clear cell renal cell carcinoma [38]. Moreover, Ran stimulates membrane targeting and stabilization of RhoA which enhances ovarian cancer cell invasion [39]. It is also found that the overexpression of Ran is correlated with metastasis potential in pancreatic cancers.
Additionally, the knockdown of RAN signi cantly inhibited the capability of pancreatic cancer cells to metastasize to the liver [40]. Consistently, overexpression of RAN leads to the activation of PI3K/AKT signaling and promotes invasive ability in non-small cell lung cancer cells [41]. Strikingly, Ran expression is essential for mitosis of cancer cells but not normal cells [19,42]. It has been shown that Ran expression is more important to tumor cells with K-Ras activating mutations than the K-Ras wild-type normal cells [43]. In another study, cancer cells with hyperactivation of the Ras/MEK/ERK and PI3K/Akt/mTORC1 pathways are more dependent on Ran expression than their wild-type counterparts [44]. On the other hand, Ran GTPase has also been implicated in drug resistance and cancer therapeutic development. The inactivation of Ran GTPase can signi cantly increase the sensitivity of breast cancer cells to ge tinib [45]. Furthermore, functional depletion of Ran GTPase suppresses the proliferation and migration of both glioblastoma and breast cancer cells [46]. It was also showed that the inhibition of Ran in SCID mice can induced ovarian cancer regression [47].
In our study, we found that DYNLL1 and RAN were co-expressed in LUAD patients. And high expression of DYNLL1 and RAN associated with unfavorable overall survival in LUAD patients. Notably, DYNLL1 was most well-regarded as cargo adapters of molecular motors. Meanwhile, Ran GTPase could also function as transporter. The co-expression of the two proteins may strengthen the transportation and hyperactivation of oncogenes. Combined the protein-protein interaction and functional enrichment analysis, the top pathways that DYNLL1 and Ran participate in including intracellular protein transport, cellular protein localization, microtubule-based process and mitotic cell cycle process. We proposed that DYNLL1 and Ran may promote the LUAD development through the in uence of oncoprotein transportation or directly regulates cell mitosis.
In target genes-miRNA regulatory networks, RAN interacts with more miRNAs than DYNLL1. On the other hand, in target genes-TF regulatory network, DYNLL1 seems to be regulated by more transcription factors that of RAN. We also investigated the candidate drugs which could target DYNLL1 and RAN.
Ampiroxicam is a prodrug of piroxicam which is an anti-in ammatory agent [48]. Pidorubicine, also known as epirubicin, is a 4'-epi-isomer of doxorubicin. And pidorubicine has shown good antitumor effects by interference with the synthesis and function of DNA [49]. The two drugs could be good candidates to interact with DYNLL1 in the treatment of LUAD. Ouabain is a cardiotonic steroid hormone which has a role as some ATPase inhibitor. Whether it could be a candidate to interact with Ran in the treatment of LUAD need more studies. The datasets used and/or analyzed during the present study are available from the corresponding author on reasonable request.     The association between DYNLL1 and RAN expression and survival in Liver and Renal Cancer patients.  Protein-protein interaction analysis of DYNLL1 and RAN with STRING. Each node represents a different gene. Each line represents a connection between two different genes.