Lycopene is an important carotenoid pigment in red fruits and vegetables, especially in tomato. Although lycopene biosynthesis and catabolism have been found to be regulated by multiple factors including phytohormones, little is known about their regulatory mechanism. Cytokinins are crucial to various aspects of plant growth. Isopentenyltransferases (IPTs) catalyze the initial rate-limiting step of cytokinins biosynthesis, however, their roles in fruit ripening remain unclear. Here, the functions of SlIPT4, encoding an isopentenyltransferase, were characterized via RNAi-mediated gene silencing in tomato. As we expected, silencing of SlIPT4 expression resulted in accelerated leaf senescence. However, down-expression of SlIPT4 generated never-red orange fruits, corresponding with a dramatic reduction of lycopene. Among lycopene biosynthesis-related genes, the fact of remarkable decrease of ZISO transcript and upregulation of other genes, revealed that SlIPT4 regulates positively lycopene biosynthesis via directly affecting ZISO expression, and also supported the existence of regulatory loops in lycopene biosynthesis pathway. Meanwhile, the accumulation of abscisic acid (ABA) was reduced and the transcripts PSY1 were increased in SlIPT4-RNAi fruits, supporting the feedback regulation between ABA and lycopene biosynthesis. The study revealed the crucial roles of SlIPT4 in leaf senescence and the regulatory network of lycopene biosynthesis in tomato, providing a new light on the lycopene biosynthesis and fruit ripening.
Background: Isoorientin (ISO) is a glycosylated flavonoid with antitumor, anti-inflammatory, and antioxidant properties. However, its effects on bone metabolism remain largely unknown. Methods: In this study, we aimed to investigate the effects of ISO on receptor activator of nuclear factor-κB ligand (RANKL)-induced osteoclast formation in vitro and bone loss in post-ovariectomy (OVX) rats, as well as to elucidate the underlying mechanism. First, network pharmacology analysis indicated that MAPK1 and AKT1 may be potential therapeutic targets of ISO and that ISO has potential regulatory effects on the mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K)/protein kinase B (AKT) pathways, as well as oxidative stress. ISO was added to RAW264.7 cells stimulated by RANKL, and its effects on osteoclast differentiation were evaluated using tartrate‐resistant acid phosphatase (TRAP) staining, TRAP activity measurement, and F-actin ring analysis. Reactive oxygen species (ROS) production in osteoclasts was detected using a ROS assay kit. The effects of ISO on RANKL-triggered molecular cascade response were further investigated by Western blotting, quantitative real-time polymerase chain reaction, and immunofluorescence staining. In addition, the therapeutic effects of ISO were evaluated in vivo. Results: ISO inhibited osteoclastogenesis in a time- and concentration-dependent manner. Mechanistically, ISO downregulated the expression of the main transcription factor for osteoclast differentiation by inhibiting MAPK and PI3K/AKT1 signaling pathways. Moreover, ISO exhibited protective effects in OVX-induced bone loss rats. This was consistent with the results derived from network pharmacology. Conclusion: Our findings suggest a potential therapeutic utility of ISO in the management of osteoclast-associated bone diseases, including osteoporosis.
Intracranial aneurysms (IAs), a significant medical concern due to their prevalence and life-threatening nature, pose challenges regarding diagnosis owing to their diminutive and variable morphology. There are currently challenges surrounding automating the segmentation of IAs, which is essential for diagnostic precision. Existing deep learning methods in IAs segmentation tend to emphasize semantic features at the expense of detailed information, potentially compromising segmentation quality. Our research introduces the innovative Dual-Path Fusion Network (DPF-Net), an advanced deep learning architecture crafted to refine IAs segmentation by adeptly incorporating detailed information. DPF-Net, with its unique resolution-preserving detail branch, ensures minimal loss of detail during feature extraction, while its cross-fusion module effectively promotes the connection of semantic information and finer detail features, enhancing segmentation precision. The network also integrates a detail aggregation module for effective fusion of multi-scale detail features. A view fusion strategy is employed to address spatial disruptions in patch generation, thereby improving feature extraction efficiency. Evaluated on the CADA dataset, DPF-Net achieves a remarkable mean Dice similarity coefficient (DSC) of 0.8967, highlighting its potential in automated IAs diagnosis in clinical settings. Furthermore, DPF-Net’s outstanding performance on the BraTS 2020 MRI dataset for brain tumor segmentation with a mean DSC of 0.8535 further confirms its robustness and generalizability.
Data imbalance is often encountered in deep learning process and is harmful to model training. The imbalance of hard and easy samples in training datasets often occurs in the segmentation tasks from Contrast Tomography (CT) scans. However, due to the strong similarity between adjacent slices in volumes and different segmentation tasks (the same slice may be classified as a hard sample in liver segmentation task, but an easy sample in the kidney or spleen segmentation task), it is hard to solve this imbalance of training dataset using traditional methods. In this work, we use a pre-training strategy to distinguish hard and easy samples, and then increase the proportion of hard slices in training dataset, which could mitigate imbalance of hard samples and easy samples in training dataset, and enhance the contribution of hard samples in training process. Our experiments on liver, kidney and spleen segmentation show that increasing the ratio of hard samples in the training dataset could enhance the prediction ability of model by improving its ability to deal with hard samples. The main contribution of this work is the application of pre-training strategy, which enables us to select training samples online according to different tasks and to ease data imbalance in the training dataset.