Prototype Bayesian Meta-Learning for Few-Shot Image Classification.

Meta-learning aims to leverage prior knowledge from related tasks to enable a base learner to quickly adapt to new tasks with limited labeled samples. However, traditional meta-learning methods have limitations as they provide an optimal initialization for all new tasks, disregarding the inherent uncertainty induced by few-shot tasks and impeding task-specific self-adaptation initialization. In response to this challenge, this article proposes a novel probabilistic meta-learning approach called prototype Bayesian meta-learning (PBML). PBML focuses on meta-learning variational posteriors within a Bayesian framework, guided by prototype-conditioned prior information. Specifically, to capture model uncertainty, PBML treats both meta-and task-specific parameters as random variables and integrates their posterior estimates into hierarchical Bayesian modeling through variational inference (VI). During model inference, PBML employs Laplacian estimation to approximate the integral term over the likelihood loss, deriving a rigorous upper-bound for generalization errors. To enhance the model's expressiveness and enable task-specific adaptive initialization, PBML proposes a data-driven approach to model the task-specific variational posteriors. This is achieved by designing a generative model structure that incorporates prototype-conditioned task-dependent priors into the random generation of task-specific variational posteriors. Additionally, by performing latent embedding optimization, PBML decouples the gradient-based meta-learning from the high-dimensional variational parameter space. Experimental results on benchmark datasets for few-shot image classification illustrate that PBML attains state-of-the-art or competitive performance when compared to other related works. Versatility studies demonstrate the adaptability and applicability of PBML in addressing diverse and challenging few-shot tasks. Furthermore, ablation studies validate the performance gains attributed to the inference and model components.

Genetic Modulation of Biosynthetic Gas Vesicles for Ultrasound Imaging.

Gas vesicles (GVs) from microorganisms are genetically air-filled protein nanostructures, and serve as a new class of nanoscale contrast agents for ultrasound imaging. Recently, the genetically encoded GV gene clusters have been heterologously expressed in Escherichia coli, allowing these genetically engineered bacteria to be visualized in vivo in a real-time manner by ultrasound. However, most of the GV genes remained functionally uncharacterized, which makes it difficult to regulate and modify GVs for broad medical applications. Here, the impact of GV proteins on GV formation is systematically investigated. The results first uncovered that the deletions of GvpR or GvpU resulted in the formation of a larger proportion of small, biconical GVs compared to the full-length construct, and the deletion of GvpT resulted in a larger portion of large GVs. Meanwhile, the combination of gene deletions has resulted in several genotypes of ultrasmall GVs that span from 50 to 20 nm. Furthermore, the results showed that E. coli carrying the ΔGvpCRTU mutant can produce strong ultrasound contrast signals in mouse liver. In conclusion, the study provides new insights into the roles of GV proteins in GV formation and produce ultrasmall GVs with a wide range of in vivo research.

Biosynthetic Melanin/Ce6-Based Photothermal and Sonodynamic Therapies Significantly Improved the Anti-Tumor Efficacy.

Photothermal therapy (PTT) and sonodynamic therapy (SDT) are becoming promising therapeutic modalities against various tumors in recent years. However, the single therapeutic modality with SDT or PTT makes it difficult to achieve a satisfactory anti-tumor outcome due to their own inherent limitations, such as poor tissue penetration for the near-infrared (NIR) laser and the limited cytotoxic reactive oxygen species (ROS) generated from conventional sonosensitizers irradiated by ultrasound (US). Here, we successfully biosynthesized melanin with a controllable particle size with genetically engineered bacteria harboring a heat-inducible gene circuit. The biosynthetic melanin with 8 nm size and chlorin e6 (Ce6) was further encapsulated into liposomes and obtained SDT/PTT dual-functional liposomes (designated as MC@Lip). The resulting MC@Lip had an approximately 100 nm particle size, with 74.71% ± 0.54% of encapsulation efficiency for melanin and 94.52% ± 0.78% for Ce6. MC@Lip exhibited efficient 1O2 production and photothermal conversion capability upon receiving irradiation by US and NIR laser, producing significantly enhanced anti-tumor efficacy in vitro and in vivo. Especially, US and NIR laser irradiation of tumors received with MC@Lip lead to complete tumor regression in all tested tumor-bearing mice, indicating the great advantage of the combined use of SDT and PTT. More importantly, MC@Lip possessed good photoacoustic (PA) and fluorescence dual-modal imaging performance, making it possible to treat tumors under imaging guidance. Our study provides a novel approach to synthesize a melanin nanoparticle with controllable size and develops a promising combined SDT/PTT strategy to treat tumors.

Modification of PEG reduces the immunogenicity of biosynthetic gas vesicles.

Nanobubbles have received great attention in ultrasound molecular imaging due to their capability to pass through the vasculature and reach extravascular tissues. Recently, gas vesicles (GVs) from archaea have been reported as acoustic contrast agents, showing great potential for ultrasound molecular imaging. However, the immunogenicity and biosafety of GVs has not yet been investigated. In this study, we examined the immune responses and biosafety of biosynthetic GVs and polyethylene glycol (PEG)-modified GVs (PEG-GVs) in vivo and in vitro. Our findings suggest that the plain GVs showed significantly stronger immunogenic response than PEG-GVs. Less macrophage clearance rate of the RES and longer circulation time were also found for PEG-GVs, thereby producing the better contrast imaging effect in vivo. Thus, our study demonstrated the PEG modification of biosynthetic GVs from Halobacterium NRC-1 is helpful for the future application of GVs in molecular imaging and treatment.

Biosynthesis of Melanin Nanoparticles for Photoacoustic Imaging Guided Photothermal Therapy.

Photothermal therapy (PTT) has attracted considerable attention in recent years due to their unique advantages in minimal invasiveness and spatiotemporal selectivity. However, the fabrication procedures of PTT agents frequently require complex chemical and/or physical methods that involves harsh and environmentally hazardous conditions. Here, a genetically engineered bacterium is developed to synthesize melanin nanoparticles under mild and environmentally friendly conditions. The biosynthetic melanin nanoparticles exhibit excellent biocompatibility, good stability, and negligible toxicity. In addition, the biosynthetic melanin nanoparticles have strong absorption at near-infrared (NIR) region and higher photothermal conversion efficiency (48.9%) than chemically synthesized melanin-like polydopamine nanoparticles under an 808 nm laser irradiation. Moreover, the results show that the biosynthetic melanin nanoparticles have excellent photoacoustic imaging (PAI) performance and can be used for PAI guided PTT in vivo. In conclusion, the study provides an alternative approach to synthesize PTT agents with broad application potential in the diagnosis and treatment of cancer.

Forkhead box family transcription factors as versatile regulators for cellular reprogramming to pluripotency.

Forkhead box (Fox) transcription factors play important roles in mammalian development and disease. However, their function in mouse somatic cell reprogramming remains unclear. Here, we report that FoxD subfamily and FoxG1 accelerate induced pluripotent stem cells (iPSCs) generation from mouse fibroblasts as early as day4 while FoxA and FoxO subfamily impede this process obviously. More importantly, FoxD3, FoxD4 and FoxG1 can replace Oct4 respectively and generate iPSCs with germline transmission together with Sox2 and Klf4. On the contrary, FoxO6 almost totally blocks reprogramming through inhibiting cell proliferation, suppressing the expression of pluripotent genes and hindering the process of mesenchymal to epithelial transition (MET). Thus, our study uncovers unexpected roles of Fox transcription factors in reprogramming and offers new insights into cell fate transition.

Induction of Pluripotent Stem Cells from Mouse Embryonic Fibroblasts by Jdp2-Jhdm1b-Mkk6-Glis1-Nanog-Essrb-Sall4.

Reprogramming somatic cells to pluripotency by Oct4, Sox2, Klf4, and Myc represent a paradigm for cell fate determination. Here, we report a combination of Jdp2, Jhdm1b, Mkk6, Glis1, Nanog, Essrb, and Sall4 (7F) that reprogram mouse embryonic fibroblasts or MEFs to chimera competent induced pluripotent stem cells (iPSCs) efficiently. RNA sequencing (RNA-seq) and ATAC-seq reveal distinct mechanisms for 7F induction of pluripotency. Dropout experiments further reveal a highly cooperative process among 7F to dynamically close and open chromatin loci that encode a network of transcription factors to mediate reprogramming. These results establish an alternative paradigm for reprogramming that may be useful for analyzing cell fate control.