Cece, Enza (2023) Clinical-data driven design of nanoparticles. [Tesi di dottorato]

Anteprima

Testo Cece_Enza_36.pdf Download (7MB) | Anteprima

Abstract

Personalized medicine aims to identify treatments tailored for individual patients or groups of patients based on their unique features, including factors such as genetics, environmental influences, disease biomarkers and phenotype. This approach stands in contrast to the traditional “one-size” fits all strategy, which does not take in consideration the heterogeneity of the disease among different patients and does not consider the pathophysiological degree of the targeted disease. This leads to a poor clinical outcome, thus pushing personalized medicine into the clinical practice. In this regard, omics-based medicine (in terms of genomics, transcriptomics, proteomics, metabolomics, radiomics), next-generation sequencing, conventional tests (tests on body fluids, immunohistochemistry, flow cytometry, histopathology, and biopsy) and clinical imaging (Magnetic Resonance Imaging, Computed Tomography, Positron Emission Tomography, Ultrasound Imaging) are used to provide data about the pathophysiological state of the patient. Attempts in collecting and categorizing these data can be found in the creation of databases, like Cancer Imaging Archive, Cancer Genome, Cancer Proteome Atlas or cBioPortal for cancer genomics. In this scenario, Drug Delivery Systems (DDSs) and nanomedicine are establishing as a tool to contribute to the acceleration and translation of personalized medicine. DDSs can, actually, be designed and engineered with diverse physico-chemical properties, enabling precise diagnosis, targeted therapy at the tissue, cellular or molecular level and facilitating the monitoring of the treatment. The optimal design of a DDS relies upon a comprehensive analysis that considers the intended application (i.e., imaging, therapeutic, theranostics), addresses the targeted pathology and considers the active agents to be encapsulated. Furthermore, it considers factors such as biodegradability, biocompatibility, on- and off-target release kinetics and scalability production of the same. This approach enables the selection of materials and production processes best suited and tailored for the requirements of the DDS of interest. Even though a huge plethora of DDSs architecture is nowadays available, rational guiding principles to engineer their design features, through a disease and patient informed approach, are still missing. Thus, the adjustment of the physicochemical properties of the DDS, generally, relies on the evaluation of their in vitro and/or in vivo response for a specific clinical task of interest. This is a trial-and-error approach that results to be time consuming and, since it does not consider the whole framework shaped by the biological barriers encountered by injected DDSs, it drastically hinders their clinical translation. In average, a survey analysis of literature1 has found that less than 1% of the injected dose is capable of reaching the target site of interest. Indeed, biological barriers, encountered by DDSs intravenously injected, span across multiple scales and move from blood circulation to blood vessels’ margination and extravasation, diffusion in the Extracellular Matrix and uptake by the targeted cells. Each one of these barriers poses obstacles to the delivery and bio-distribution of DDSs at the target site. Transport conditions faced by DDSs are posed by the biological barriers and are extremely different depending on the degree of the disease and on the patient. Therefore, there is the necessity to transition towards the personalized design of the DDS and establish models and principles for an a-priori determination of the physico-chemical properties, best suited for the pathological condition of interest. As examples, Karageorgis et al.2 found that the accumulation of lipidic nanocapsules depends on the structural and permeability parameters of tumor microvasculature. Similarly, Sykes et al.3 histologically characterized tumors of different sizes in mice (ranging from low to high tumor volume due to disease progression), considering factors such as increased vascular density, cell density, and extracellular matrix content. They correlated these parameters with the tumor's capacity to accumulate PEGylated gold nanoparticles of various sizes. These examples demonstrate that it is possible to predict a-priori the accumulation of a DDS based on tumor characteristics and, consequently, adjust the properties of the DDS accordingly. However, the reported examples are focused on one type of tumor and one type of DDS. The paradigm shift of personalized nanomedicine needs to progress in two directions: a thorough characterization of the biological barriers and correspondent nano-bio interactions, as well as the development of tools through which a matching between the synthetic and biological identity of DDSs. This approach is expected to give the possibility to determine the optimal design of the DDS, so that it can achieve the desired in vivo behaviour and, consequently, the desired clinical outcome. While the need to characterize biological barriers is evident, it becomes equally apparent that a standardized and uniform methodology for achieving this goal is lacking. Clinical parameters are being extracted through a wide range of techniques, resulting in non-uniformity. On the other hand, clinical imaging is emerging as a pivotal tool for tumor characterization. Nonetheless, the acquisition of high-quality, artifact-free clinical images remains a formidable challenge, especially when dealing with early cancer lesions. In such cases, image reconstruction remains a crucial and critical step. To match the synthetic and biological identity of DDSs is necessary to have a rational workflow that systematically incorporates each clinical tumour parameter into the design of the DDS. To translate this approach in a clinical and practical context, gathering preliminary data about nanoparticles from in vitro studies and in vivo studies, in vitro ECM penetration studies, open-source datasets, literature studies is the first step to perform. In a manner similar to the collection of clinical data on tumors, information about nanoparticles has been systematically gathered and organized within various databases. As examples, caNanoLab is a data sharing portal designed to facilitate information sharing across the international biomedical nanotechnology research community to expedite and validate the use of nanotechnology in biomedicine; nano (a Nature Portfolio Solution) retrieves detailed information on properties, applications, toxicity and preparation methods of thousands of nanomaterials and devices; PubVINAS is a friendly online nanomodeling tool based on big data curation of nano-biological activities and nanostructure annotations; Nanoparticle Information Library (NIL) helps occupational health professionals, industrial users, worker groups, and researchers organize and share information on nanomaterials, including their health and safety-associated properties; compendium for Biomaterial Transcriptomics (cBiT) collects transcriptional profiles of cells after biomaterial exposure (knee and dental implants) and, finally, Nanomaterial-Biological Interactions (NBI) knowledgebase gives information on the toxicity on zebrafish of nanomaterials. These databases collectively serve to facilitate research, collaboration, and informed decision-making in the field of nanotechnology and nano-bio interactions. However, they are associated with some challenges that make it difficult to access the information stored within them. These datasets are characterized by a high volume of non-organized data and the complexity and heterogeneity of the data collection methods, making it difficult to compare the data among them. What is currently lacking is the integration of data from clinical tumor characterization and nanomedicine, as well as the establishment of meaningful relationships between these datasets. In this context, Machine Learning (ML) and Artificial Intelligence (AI) are increasingly establishing as powerful tools to find hidden relationships in complex datasets characterized by big data. ML/AI tools have already found their application also in nanomedicine; in particular, they have been used to adjust the production process parameters to achieve desired physicochemical properties of the nanoparticles or of the microparticles. On the other hand, they have been used for the prediction of nanoparticles’ cytotoxicity, the cell uptake efficiency, depending on the material, size, charge and concentration of the nanoparticles, the prediction of the composition of the protein corona and, consequently, the cellular recognition mediated by it. Moreover, ML/AI has been also used to define Quantitative Structure Activity Relationships (QSARs), revealing non-evident links between structural properties of nanoparticles and their biological activity in vitro. As example, Mirkin et al. 4 correlated through a XGBoost Model (a ML algorithm) the in vitro immune response of macrophages with the structural properties of spherical nucleic acids (SNAs), by analysing a library of 1000 SNAs derived from the combination of 11 different design parameters. They selected the design parameters able to impact the most on the levels of immune activity. However, each ML/AI model has been implemented using single experimental datasets, analyzing individual DDS architectures and specific tumor scenarios, which limits its ability to capture the complexity of all the different architectures available and tumor types under treatment. As there is a need to fine-tune the properties of DDSs to achieve a specific synthetic identity matching with the desired biological identity, microfluidics stands out as a reliable tool for production. It offers precise control over the final architecture and functionality of DDSs at the molecular level, resulting in narrow size distributions and very low batch-to-batch variability.

Downloads

Actions (login required)

Modifica documento