Cancer therapy is still a challenging issue. To address this, the combination of anticancer drugs with other therapeutic modalities, such as light-triggered therapies, has emerged as a promising approach, primarily when both active ingredients are provided within a single nanosystem. Herein, we describe the unprecedented preparation of tumor microenvironment (TME) responsive nanoparticles exclusively composed of a paclitaxel (PTX) prodrug and the photosensitizer pheophorbide A (PheoA), e.g., PheoA≅PTX₂S. This system aimed to achieve both the TME-triggered and controlled release of PTX and the synergistic/additive effect by PheoA-mediated photodynamic therapy. PheoA≅PTX₂S were produced in a simple one-pot process, exhibiting excellent reproducibility, stability, and the ability to load up to 100% PTX and 40% of PheoA. Exposure of PheoA≅PTX₂S nanoparticles to TME-mimicked environment provided fast disassembly compared to normal conditions, leading to PTX and PheoA release and consequently elevated cytotoxicity. Our data indicate that PheoA incorporation into nanoparticles prevents its aggregation, thus providing a greater extent of ROS and singlet oxygen production. Importantly, in SK-OV-3 cells, PheoA≅PTX₂S allowed a 30-fold PTX dose reduction and a 3-fold dose reduction of PheoA. Our data confirm that prodrug-based nanocarriers represent valuable and sustainable drug delivery systems, possibly reducing toxicity and expediting preclinical and clinical translation.

Artificial intelligence (AI) is being integrated into nearly every aspect of modern life, and machine learning (ML)–a subfield of AI–has the potential to accelerate the development of nanomedicines. Here, a machine learning workflow is presented to optimize the microfluidic‐based formulation of nanomedicines. A database of ≈200 unique nanomedicine formulations with over 550 total measurements is curated by producing liposomes, lipid nanoparticles, and poly(lactic‐ co ‐glycolic acid) nanoparticles, either empty or loaded with the model therapeutic agent, curcumin. Nanoparticle production parameters are systematically varied, and the resulting particles are characterized for their diameter, polydispersity index, and encapsulation efficiency. These data are used to train and validate 13 different ML models using open‐source libraries, with the task of returning the most accurate prediction of nanomedicine attributes. The most accurate models, based on random forest regression, are implemented to yield particles with user‐specified attributes. Finally, the proposed ML workflow, MicrofluidicML, is compared against generative large language models–OpenAI ChatGPT, Google's Gemini, and DeepSeek. MicrofluidicML provides a workflow where the researcher has complete governance and control of the input data with a relatively low computational overhead, and represents a step toward implementing a computationally lightweight ML framework to accelerate nanomedicine development.

Recombinant tissue-type plasminogen activator (tPA) is the only approved thrombolytic drug for acute ischemic stroke, a condition associated with severe disabilities and high mortality. However, when the blood-brain barrier (BBB) is damaged, tPA can exacerbate cerebral injury and increase the risk of hemorrhagic transformation, limiting its use to a small subset of patients. To address this challenge and minimize extravascular accumulation, we combined tPA with micrometer-sized particles (DPN). We then tested their safety and neuroprotective effects. After a 1 h transient occlusion of the middle cerebral artery, free tPA, tPA-DPN, or saline was administered to assess mice survival, neurological behavior, and infarcted area extent. Free-tPA exacerbated brain damage, resulting in a modest 10% survival rate at 24 h post intervention. Conversely, tPA-DPN displayed a far better prognosis, with a 75% survival rate comparable to that of saline. No statistical differences were documented between tPA-DPN and saline for the Activity Score and the Neurological Severity Score. tPA-DPN did not increase lesion volume or BBB permeability, unlike free-tPA, which led to an over 2-fold enlarged lesion volume and 50% higher BBB permeability. The safety profile of tPA-DPN is attributed to the robust conjugation of tPA onto DPN and the lack of DPN extravasation, resulting in negligible cerebrovascular damage of free tPA and glial and neuron impairment. The vascular confinement of tPA linked to microscopic particles reduces drug side effects and represents a valuable strategy for safe and effective tPA delivery, even in the postacute stroke phase.

Lipid nanoparticles (LNP) have been extensively studied for their ability to encapsulate and protect RNA molecules from degradation. More recently, a few studies have begun to explore their applications as carriers for brain drug delivery via various administration routes. Nose-to-brain delivery represents a promising alternative to both invasive local injections and systemic administration, offering the possibility to bypass the blood-brain barrier and directly access the brain, achieve rapid absorption, reduce systemic exposure, and allow for ease of administration. In order to evaluate the viability of this alternative route, it is essential to acquire a better understanding of the intraneuronal mass transport of LNP, particularly in terms of how effectively and efficiently they deliver their payloads from the periphery to neuronal cell bodies. However, most previous studies have focused primarily on the delivery vector itself rather than on the fate of the transported cargo. In this study, we investigate the retrograde trafficking of nucleic acid-loaded LNP in primary cortical neurons, focusing on the transport of both the particle and the payload. Three distinct LNP were formulated to characterize different aspects of their interaction with the cells, with the major LNP player of this study containing a red-fluorescent Rhodamine B-tagged lipid and a green fluorescently FAM-tagged RNA. Flow cytometry was used to document LNP uptake by primary cortical neurons over time. Additionally, confocal microscopy was then used to investigate the colocalization of LNP and RNA after a conventional 2D culture treatment. As a final step, a compartmentalized chip that separates the somal and the axonal regions of cortical neurons was used to study the intraneuronal dynamics of LNP and their cargo. In this second setup, LNP were selectively administered at the axonal compartment, and the fluorescent signals from the vector (red) and the payload (green) were imaged through time-lapse microscopy. The progressive accumulation of RNA found at cellular bodies also in the absence of the red signal suggested an efficient retrograde transport of the LNP payload toward the soma. Comprehensively, this work demonstrates that primary cortical neurons are capable of efficiently uptaking LNP and of intracellularly transporting both LNP and their RNA cargo. Interestingly, a different colocalization trend (LNP-RNA) emerged depending on the followed setup. Localized axonal transfection appeared to favor dissociation of RNA from the LNP and subsequent accumulation at the soma. Overall, our work provides a fundamental in vitro proof of concept of the RNA delivery to the cellular bodies of primary cortical neurons via the retrograde transport of LNP vectors administered at the axonal termini. This finding, together with the image-analysis-based quantification of the RNA accumulation described in our work, paves the way for future studies aimed at designing lipid-based nanoparticles for RNA therapeutic delivery to the brain via peripheral administration.

Despite substantial improvements in breast cancer (BC) treatment there is still an urgent need to find alternative treatment options to improve the outcomes for patients with advanced-stage disease. Photodynamic therapy (PDT) is gaining a lot of attention as a BC therapeutic option because of its selectivity and low off-target effects. However, the hydrophobicity of photosensitizers (PSs) impairs their solubility and limits the circulation in the bloodstream, thus representing a major challenge. The use of polymeric nanoparticles (NPs) to encapsulate the PS may represent a valuable strategy to overcome these issues. Herein, we developed a novel biomimetic PDT nanoplatform (NPs) based on a polymeric core of poly(lactic-co-glycolic)acid (PLGA) loaded with the PS meso-tetraphenylchlorin disulfonate (TPCS_2a). TPCS_2a@NPs of 98.89 ± 18.56 nm with an encapsulation efficiency percentage (EE%) of 81.9 ± 7.92% were obtained and coated with mesenchymal stem cells-derived plasma membranes (mMSCs) (mMSC-TPCS_2a@NPs, size of 139.31 ± 12.94 nm). The mMSC coating armed NPs with biomimetic features to impart long circulation times and tumor-homing capabilities. In vitro, biomimetic mMSC-TPCS_2a@NPs showed a decrease in macrophage uptake of 54% to 70%, depending on the conditions applied, as compared to uncoated TPCS_2a@NPs. Both NP formulations efficiently accumulated in MCF7 and MDA-MB-231 BC cells, while the uptake was significantly lower in normal breast epithelial MCF10A cells with respect to tumor cells. Moreover, encapsulation of TPCS_2a in mMSC-TPCS_2a@NPs effectively prevents its aggregation, ensuring efficient singlet oxygen (¹O₂) production after red light irradiation, which resulted in a considerable in vitro anticancer effect in both BC cell monolayers (IC₅₀ < 0.15 µM) and three-dimensional spheroids.

Artificial intelligence (AI) is being integrated into nearly every aspect of modern life, and the collaboration between machine learning (ML) – a subfield of AI – and microfluidic fabrication techniques has the potential to accelerate the development of nanomedicines. Here, we present a machine learning workflow designed to optimize the microfluidic-based formulation of nanomedicines. A database of almost 200 unique nanomedicine formulations with over 550 total measurements was curated by producing liposomes, lipid nanoparticles, and poly(lactic-co- glycolic acid) (PLGA) nanoparticles, either empty or loaded with the model therapeutic agent curcumin (CURC), using a benchtop microfluidic system. Materials and flow parameters (input features), including the reagent concentrations, total flow rates, and aqueous:organic flow rate ratios were systematically varied, and the resulting particles were characterized for their hydrodynamic diameter (dH), polydispersity index (PdI), and encapsulation efficiency (EE) (output parameters). These data were used to train, test, and validate 13 different ML models, using a freely available and open-source librries, with the task of returning the most accurate prediction of the nanomedicine attributes – dH, PdI, EE. The most accurate ML models, based on random forest regression, were implemented to yield particles with user-specified attributes. Finally, the proposed ML workflow, dubbed MicrofluidicML, was compared against generative large language models – Open AI ChatGPT, Google’s Gemini, DeepSeek. MicrofluidicML provides a workflow where the researcher has complete governance and control of the input data, the computational overhead is much lower compared to the amount of data required to train a generative AI model, and the post-hoc SHAP analysis provides key information towards the critical factors affecting nanoparticle formation. The application of ML in the field of nanomedicine is inevitable, and MicrofluidicML represents a step towards implementing a computationally lightweight ML framework for accelerating nanomedicine development.