Abstract
Time is a critical dimension in chemistry and biology. Studying the composition of chemical systems as they evolve, including the effects of processing, use, and wear, provides access to the mechanisms underlying the transformations, either wanted or unwanted, as they occur. If monitoring of molecular events can occur in real-time, it can allow for adjusting the reaction conditions to maximize the wanted transformation and minimize the unwanted ones. In biological systems, which are constantly out of equilibrium, real-time monitoring provides precious insights into the kinetics of the underlying molecular events. Therefore, real-time approaches can radically change the approach to studying chemical and biological processes, paving the way to increased sustainability and a better understanding of living systems. However, selectively observing molecular events which take place within mixtures of high complexity is a challenging task. Commonly-used spectroscopies (UV-VIS, IR) are defeated by the fact that reactants, intermediates, and products have largely superimposable structures, which in turn are reflected in spectra that are difficult to disentangle, and the relation between the spectral response and concentrations must be identified (absorption coefficients). On the contrary, NMR is uniquely suited for characterizing the complex responses of these systems. It is intrinsically quantitative, as all the active nuclei provide a signal proportional to their concentration regardless of the chemical environment they are found in, and it is also perfectly selective as even the smallest structural perturbation is reflected in detectable shifts. Finally, its non-destructive nature makes it ideal for application to biological samples. Historically, however, NMR has not been included in online reaction monitoring applications because of its intrinsic low sensitivity, which in turn yields long measurement times, and the application to real-time monitoring of living cells is still in its infancy. In this project, we will develop new acquisition and processing schemes with the aim of reducing the time burden of the NMR experiments while preserving the level of attainable information, thereby pushing the usability of real-time NMR methodologies for real-time monitoring of chemical reactions and biological processes. In particular, we will focus on the development and implementation of blind-source-separation (BSS) methods and fast NMR experiments, and to apply them to a number of biotechnologically-relevant test cases including target-based drug screening in-cell, food processing, and biotransformations. We expect that the methods we propose will lead to a quick maturation of the application for real-time monitoring and that, in the longer run, they might become a standard for time-resolved NMR investigations.
Results achieved
The project "Time-resolved magnetic resonance to investigate dynamic events in biological systems and biotransformations" (TiReD) successfully achieved all three of its main objectives: the development of new data processing strategies, the creation of experimental and analytical techniques dedicated to the real-time monitoring of biological events, and their application to biologically relevant systems. Main Results by Task: • Demonstrated that standard augmentation strategies in Multivariate Curve Resolution (MCR) outperform tensor augmentation, reducing computational burden while preserving spectral information during noise filtering. • Released open-source software tools, including TrAGICo (extraction of trends, shifts, and intensities from NMR series), pyIHM (featuring a novel alignment algorithm for Indirect Hard Modeling), and t1t2une (automated optimization and configuration of NMR experiments). • Applied Fast Iterative Filtering (FIF) for baseline correction and developed a reference deconvolution technique with ridge regularization to improve resolution, signal-to-noise ratio, and linewidth in low-field dDNP experiments. Task 2: Machine Learning Models and Ligand Screening: • Developed machine learning models (PLSDA + SVM and Boruta + Random Forest Classifier) capable of accurately classifying intact-cell phenotypes from low-resolution in-cell ^1H NMR spectra. • Performed computational design and high-throughput screening of a targeted library of approximately 300 compounds directed toward non-canonical DNA structures (i-motifs and G-quadruplexes), using Principal Component Analysis (PCA) on ^1H NMR spectra to rapidly identify the most promising ligands. Task 3: Real-Time Monitoring of Biological Systems: • Bacterial Fermentation: Developed an in vitro real-time ^1H NMR monitoring protocol and MCR-ALS chemometric deconvolution to track the kinetic profiles of 11 key metabolites during lactic acid bacterial fermentation over a 24-hour period. • Hepatic Cell Metabolism: Monitored the metabolism of live human hepatocytes (HepG2) within an NMR bioreactor for 72 hours, highlighting the metabolic relationships among linoleic acid uptake, lipid accumulation, and glutathione depletion. • In-Cell ^19F NMR Drug Screening: Employed bioreactor-assisted in-cell ^19F NMR to investigate target engagement, membrane permeability rates, and quantitative intracellular binding affinities of novel fluorinated compounds against Carbonic Anhydrase (CA) isoforms in live human cells. Dissemination and Research Outputs: • Publications and Software: Produced 9 peer-reviewed scientific publications and released 3 open-source software repositories on GitHub (T1Ttune, pyIHM, and TrAGICo). • Compliance: Conducted all activities in full compliance with Open Access guidelines and the DNSH (Do No Significant Harm) principles.Project details
Unibo Team Leader: Elena Babini
Unibo involved Department/s:
Dipartimento di Scienze e Tecnologie Agro-Alimentari
Coordinator:
Università degli Studi di FIRENZE(Italy)
Total Unibo Contribution: Euro (EUR) 69.400,00
Project Duration in months: 24
Start Date:
28/09/2023
End Date:
28/02/2026
