Development of 1h-nmr serum profiling methods for high-throughput metabolomics

• Autores: Rubén Barrilero Regadera
• Directores de la Tesis: Xavier Correig (dir. tes.)
• Lectura: En la Universitat Rovira i Virgili ( España ) en 2017
• Idioma: español
• Tribunal Calificador de la Tesis: Josep Ribalta Vives (presid.), Roger Mallol Parera (secret.), Francesco Savorani (voc.)
• Programa de doctorado: Programa de Doctorado en Tecnologías para Nanosistemas, Bioingeniería y Energía por la Universidad Rovira i Virgili
• Materias:
  • Matemáticas
    • Estadística
      • Análisis de datos
• Enlaces
  • Tesis en acceso abierto en: hdl.handle.net
• Resumen

  • The irruption of metabolomics is replacing the traditional approach of clinical diagnostics focused on single biomarkers, such as glucose or cholesterol, with the profiling of complex metabolite patterns reflecting metabolic activity in multiple biological pathways. Blood serum/plasma is one of the main biological matrices used in NMR-based metabolomics as its collection is minimally invasive, requires minimal sample manipulation and provides hundreds of metabolites encoding multisystemic biological information.

    High-throughput 1H-NMR profiling of serum/plasma allows a quantitative multi-compound analysis including lipoprotein classes and constituent lipids, albumin, and a large variety of low-molecular-weight metabolites (LMWM), including amino acids, creatinine, glycolysis-related metabolites, and ketone bodies, with a cost similar to standard lipids. The large physicochemical heterogeneity of these compounds requires the acquisition of three 1H-NMR measurements concerning the following molecular species: macromolecules, LMWM and lipids, where each measurement involves physical (sample extractions) and spectroscopic (editing NMR techniques) filters. However, molecular interactions and spectral complexity hamper a reliable metabolite profiling. Developing robust methods of metabolite profiling is therefore desirable to consolidate high-throughput 1H-NMR in the clinical practice.

    In our first work, we calibrated and evaluated regression models to estimate the concentration of the lipids used in the routine clinical practice (known as “lipid panel”). These lipids are still the main measurements and therapy targets of cardiovascular disease risk. Whereas most of the previous models have been calibrated using lipoprotein fractioning, our models were based on clinical enzymatic-colorimetric measurements in order to better reflect the clinical standards. Ultimately, these NMR-based regression models would lead to the incorporation of enzymatic-colorimetric lipids estimation in high-throughput 1H-NMR profiling of serum and plasma. To do so, we developed and validated 1H-NMR regression models of clinical measurements of total serum cholesterol and triglycerides, and cholesterol content of LDL, HDL and non-HDL particles, using 785 native serum/plasma samples comprising healthy subjects and several dyslipidaemias. Different combinations of 1D and 2D 1H-NMR experiments and chemometric techniques were evaluated, whereas plasma and serum were equally used from four different clinical centres. Our models performed similar to previous models based on small and more homogeneous cohorts, but the diverse matrix and physiological conditions of our samples made our models highly generalizable.

    In our second work, we addressed the quantitative issues affecting the “NMR-invisible” low-molecular-weight metabolites (LMWM) in 1H-NMR spectra of native serum. LMWM bind to proteins in native serum, consequently, their signals are totally or partially attenuated. These signal losses compromise absolute quantifications even if sophisticated signal deconvolution methods are used. In order to reduce protein binding effects on LMWM quantification, we developed a method to partially release bound LMWM from protein. Our method relies on promoting competition for ligand-binding sites of proteins by the addition of a small quantity of deuterated trimethylsilylpropanoic acid (TSP). In order to precisely quantify the extent of these interactions, we performed our quantifications using a multidimensional CPMG approach, which avoids the signal attenuations due to T2 relaxations inserted with 1D CPMG filters. The application of both strategies showed that TSP addition increases in approximately 30% the signal for clinically-relevant binding metabolites phenylalanine, leucine and isoleucine. Moreover, competitive binding strategies are fully compatible with high-throughput analysis.

    Finally, our third work addressed the quantitative profiling of serum lipids with 1H-NMR. Whereas 1H-NMR profiling of LMWM can be carried out with bioinformatics tools that allow automatic signal deconvolution based on specific metabolite signal patterns, similar solutions are not available for 1H-NMR profiling of lipids. In this context we present LipSpin, a freely-distributed software for the semiautomatic profiling of 1H-NMR spectra of lipids. Using a library of signal patterns based on mathematical and reference spectral models, a constrained lineshape fitting analysis provides the quantification of 15 different lipid-related variables about major lipid classes in serum (fatty acids, triglycerides, phospholipids and cholesterols). Lipid quantifications obtained with LipSpin agreed with those from conventional techniques and were applied to a dietary intervention study.