BackgroundRadioresistance is the major cause of cancer treatment failure. Additionally, splicing dysregulation plays critical roles in tumorigenesis. However, the involvement of alternative splicing in resistance of cancer cells to radiotherapy remains elusive. We sought to investigate the key role of the splicing factor SRSF1 in the radioresistance in lung cancer.MethodsLung cancer cell lines, xenograft mice models, and RNA-seq were employed to study the detailed mechanisms of SRSF1 in lung cancer radioresistance. Clinical tumor tissues and TCGA dataset were utilized to determine the expression levels of distinct SRSF1-regulated splicing isoforms. KM-plotter was applied to analyze the survival of cancer patients with various levels of SRSF1-regulated splicing isoforms.FindingsSplicing factors were screened to identify their roles in radioresistance, and SRSF1 was found to be involved in radioresistance in cancer cells. The level of SRSF1 is elevated in irradiation treated lung cancer cells, whereas knockdown of SRSF1 sensitizes cancer cells to irradiation. Mechanistically, SRSF1 modulates various cancer-related splicing events, particularly the splicing of PTPMT1, a PTEN-like mitochondrial phosphatase. Reduced SRSF1 favors the production of short isoforms of PTPMT1 upon irradiation, which in turn promotes phosphorylation of AMPK, thereby inducing DNA double-strand break to sensitize cancer cells to irradiation. Additionally, the level of the short isoform of PTPMT1 is decreased in cancer samples, which is correlated to cancer patients' survival.ConclusionsOur study provides mechanistic analyses of aberrant splicing in radioresistance in lung cancer cells, and establishes SRSF1 as a potential therapeutic target for sensitization of patients to radiotherapy.

Alternative splicing is a critical process to generate protein diversity. However, whether and how alternative splicing regulates autophagy remains largely elusive. Here we systematically identify the splicing factor SRSF1 as an autophagy suppressor. Specifically, SRSF1 inhibits autophagosome formation by reducing the accumulation of LC3-II and numbers of autophagosomes in different cell lines. Mechanistically, SRSF1 promotes the splicing of the long isoform of Bcl-x that interacts with Beclin1, thereby dissociating the Beclin1-PIK3C3 complex. In addition, SRSF1 also directly interacts with PIK3C3 to disrupt the interaction between Beclin1 and PIK3C3. Consequently, the decrease of SRSF1 stabilizes the Beclin1 and PIK3C3 complex and activates autophagy. Interestingly, SRSF1 can be degraded by starvation- and oxidative stresses-induced autophagy through interacting with LC3-II, whereas reduced SRSF1 further promotes autophagy. This positive feedback is critical to inhibiting Gefitinib-resistant cancer cell progression both in vitro and in vivo. Consistently, the expression level of SRSF1 is inversely correlated to LC3 level in clinical cancer samples. Our study not only provides mechanistic insights of alternative splicing in autophagy regulation but also discovers a new regulatory role of SRSF1 in tumorigenesis, thereby offering a novel avenue for potential cancer therapeutics.

Colorectal cancer is the third most common type of cancer and the fourth leading cause of cancer-associated mortality worldwide. Serine/arginine-rich splicing factor 1 (SRSF1) is a well-characterized oncogenic factor that promotes tumorigenesis by controlling a number of alternative splicing events. However, there is limited network analysis, from a global aspect, to study the effect of SRSF1 on colorectal cancer. In the present study, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of available gene regulation data from The Cancer Genome Atlas database revealed the enriched functions and signaling pathways of SRSF1. Subsequently, Oncomine analysis was performed, which demonstrated that SRSF1 was upregulated in a number of types of colon cancer. From overlapping the analysis of 2,678 SRSF1-related genes and 3,625 colorectal cancer genes in GeneCards, 468 genes were identified as SRSF1-related colorectal cancer genes. The GO results revealed that these overlapped genes were primarily enriched in metabolic processes, response to DNA damage, regulation of the cell cycle and a number of additional biological processes. KEGG pathway analysis revealed that SRSF1-related colorectal cancer genes were associated with the cell cycle, deregulated signaling pathways associated with cancer progression and colorectal cancer signaling pathways. In addition, the Search Tool for the Retrieval of Interacting Genes/Proteins database and Cytoscape analysis demonstrated that 468 SRSF1-related colorectal cancer genes exhibit potential interaction networks in which these genes were enriched in DNA metabolic processes, cell cycle regulation and regulation of apoptosis. The results of the present study suggested that SRSF1 exhibited an increased degree of interaction with key molecules, including NUF2 NDC80 kinetochore complex component, kinesin family member 2C, structural maintenance of chromosomes 3, ATM serine/threonine kinase, BRCA1 DNA repair associated, protein kinase DNA-activated catalytic polypeptide, heat shock protein 90 alpha family class A member 1, ras homolog family member A, and phosphatase and tensin homolog. Collectively, the bioinformatics analysis of the present study indicated that SRSF1 may have key functions in the progression and development of colorectal cancer.

Recently, deep reinforcement learning, associated with medical big data generated and collected from medical Internet of ings, is prospective for computer-aided diagnosis and therapy. In this paper, we focus on the application value of the second-generation sequencing technology in the diagnosis and treatment of pulmonary infectious diseases with the aid of the deep reinforcement learning. Specifically, the rapid, comprehensive, and accurate identification of pathogens is a prerequisite for clinicians to choose timely and targeted treatment. us, in this work, we present representative deep reinforcement learning methods that are potential to identify pathogens for lung infection treatment. After that, current status of pathogenic diagnosis of pulmonary infectious diseases and their main characteristics are summarized. Furthermore, we analyze the common types of secondgeneration sequencing technology, which can be used to diagnose lung infection as well. Finally, we point out the challenges and possible future research directions in integrating deep reinforcement learning with second-generation sequencing technology to diagnose and treat lung infection, which is prospective to accelerate the evolution of smart healthcare with medical Internet of ings and big data.

PurposeTo evaluate the feasibility of using a simplified non-coplanar volumetric modulated arc therapy (NC-VMAT) and investigate its dosimetric advantages compared with intensity modulated radiation therapy (IMRT) and coplanar volumetric modulated arc therapy (C-VMAT) for hippocampal-avoidance whole brain radiation therapy (HA-WBRT).MethodsTen patients with brain metastase (BM) were included for HA-WBRT. Three treatment plans were generated for each case using IMRT, C-VMAT, and NC-VMAT, respectively.ResultsThe dosimetric results of the three techniques complied roughly with the RTOG 0933 criteria. After dose normalization, the V30Gy of whole brain planned target volume (WB-PTV) in all the plans was controlled at 95%. Homogeneity index (HI) of WB-PTV was significantly reduced in NC-VMAT (0.249 ± 0.017) over IMRT (0.265 ± 0.020, p=0.005) and C-VMAT (0.261 ± 0.014, p=0.020). In terms of conformity index (CI), NC-VMAT could provide a value of 0.821 ± 0.010, which was significantly superior to IMRT (0.788 ± 0.019, p<0.001). According to D2% of WB-PTV, NC-VMAT could provide a value of 35.62 ± 0.37Gy, significantly superior to IMRT (36.43 ± 0.65Gy, p<0.001). According to D50% of WB-PTV, NC-VMAT can achieve the lowest value of 33.18 ± 0.29Gy, significantly different from IMRT (33.47 ± 0.43, p=0.034) and C-VMAT (33.58 ± 0.37, p=0.006). Regarding D2%, D98%, and Dmean of hippocampus, NC-VMAT could control them at 15.57 ± 0.18Gy, 8.37 ± 0.26Gy and 11.71 ± 0.48Gy, respectively. D2% and Dmean of hippocampus for NC-VMAT was significantly lower than IMRT (D2%: 16.07 ± 0.29Gy, p=0.001 Dmean: 12.18 ± 0.33Gy, p<0.001) and C-VMAT (D2%: 15.92 ± 0.37Gy, p=0.009 Dmean: 12.21 ± 0.54Gy, p<0.001). For other organs-at-risk (OARs), according to D2% of the right optic nerves and the right lenses, NC-VMAT had the lowest values of 31.86 ± 1.11Gy and 7.15 ± 0.31Gy, respectively, which were statistically different from the other two techniques. For other organs including eyes and optic chiasm, NC-VMAT could achieve the lowest doses, different from IMRT statistically.ConclusionThe dosimetry of the three techniques for HA-WBRT could roughly comply with the proposals from RTOG 0933. After dose normalization (D95%=30Gy), NC-VMAT could significantly improve dose homogeneity and reduce the D50% in the brain. Besides, it can reduce the D2% of the hippocampus, optic nerves, and lens. With this approach, an efficient and straightforward plan was accomplished.