Catálogo de publicaciones - libros

Brain infections are often acute and potentially life-threatening diseases of immunocompromised and healthy hosts. Rapid, accurate, and safe diagnostic methods are essential to initiate optimal treatment and reduce mortality from brain infections. We will focus here on focal brain infections as an example of the diagnostic potential and limitations of magnetic resonance imaging (MRI) and spectroscopy (MRS) methods. Particular attention will be paid to distinguishing between tumors and infective lesions. Different types of infections will be discussed.

Proton magnetic resonance spectroscopy (^1H MRS) provides detailed information on cellular metabolites in whole cells and tissues enabling a non-invasive assessment of the chemical composition of these samples and the changes that occur with a disease process. As such,^1H MRS can be used as an adjunct method for the pathological staging of human biopsy tissues [1-3]. Single pulse or one-dimensional (1D) proton spectra are rich with information, containing overlapping resonances from a number of mobile intracellular and extracellular metabolites and present in millimolar and submillimolar quantities. As a result of this overlap, information from metabolically important or potentially diagnostic resonances can be obscured, affecting both peak assignment and quantitative estimation of metabolite levels. One method of overcoming this problem has been the development and application of two-dimensional (2D) magnetic resonance (MR) spectroscopic methods to deliver additional chemical information than that available from ID spectra.

The pathology of human tissue can be determined by magnetic resonance spectroscopy (MRS) but it was a controversial field for over 20 years. Now MRS on human biopsies identifies disease processes, neoplastic status, and prognostic variables with high accuracy. The MRS databases were compared with careful and specialized histology [1,2]. The MR method is fast, accurate, and robust and for most organs complements routine histopathological diagnosis. For some organs such as the thyroid and esophagus it can provide information not available using a light microscope. However, rigorous procedures and quality control are required to ensure this level of accuracy and are the basis of this chapter.

The direct on-line ^1H NMR analysis of HPLC eluates from plant extracts using a flow probe offers the opportunity to gather structural information on individual plant metabolites without the often time-consuming preliminary purification of each metabolite. Two examples are discussed. Firstly, resolution of the major flavonoid constituents in “skullcap herb” in the absence of “standard compounds” can be achieved by a single pass experiment and interpretation of ID ^1H spectra. Secondly, the potential of the technique to allow a full structure elucidation of an unknown compound using a full range of 1D and homoand inverse detected heteronuclear 2D experiments will be examined using extracts of Aloe species from South Africa.

The design of drugs that modulate membrane bound receptors has always been of prime importance to the pharmaceutical industry. It is increasingly being recognized that the cellular membrane plays an important role in drug action and that understanding the manner in which drug molecules interact with lipid bilayers can enhance our abilities to design and develop improved medications. It has been proposed that lipophilic drugs reach their sites of action through cell membrane penetration and lateral diffusion within the membrane leaflet [1-4]. The protein supporting lipid membranes may be capable of a high degree of structural discrimination by positioning the ligand in a proper location, orientation, and active conformation for a productive interaction with the receptor. Therefore, acquiring detailed knowledge of drug and ligand physical properties that underlie their interaction profiles with the cellular membrane is of great interest in the discovery, design, and delivery of novel therapeutic agents. Information related to drug-membrane interactions, however, can rarely be derived directly from experiments with complex and labile natural membranes. Thus, various membrane models with different levels of complexity have been developed in order to exploit all the biophysical techniques available [5-7]. Whether it is in the form of liposomes, micelles, or other membrane mimetics, the interplay between the ligand and membranes is central to many aspects in pharmaceutical research.

NMR spectroscopy is the most important method for structure elucidation in solution. Once the atomic connectivity is established (constitution of the molecule) the proof of stereochemistry is the next step. NMR as an achiral method provides only relative stereochemical information unless chiral environments (solvents, interaction with shift reagents, chemical modification via chiral auxiliaries) are used. For determining the relative stereochemistry, the classical parameters are NOE effects and J coupling constants. Only recently, two different new sources of stereochemical information have been introduced for biomolecules: cross-correlated relaxation [1] andresidual dipolar couplings (RDCs) [2,3]. We will here consider the application of RDCs for small molecules such as drugs or drug-like molecules.

Residual dipolar couplings (RDCs) provide important constraints for the determination and refinement of protein NMR structures. Based on echo-anti-echo manipulation or IPAP principle, a suite of sensitivity enhanced experiments are described for measuring backbone ^1H^N−^15N, ^15N−^13C′, ^1H^N−^13C′, ^13C′−^13C^α, ^13C^α−^1H^α, ^15N( i )−^13C^α( i ), ^1H^N( i )−^13^α( i ), ^15N( i )−^13C^α( i −1), and ^1H^N( i )−^13C^α( i − 1) dipolar couplings in proteins. The accuracy of the measured couplings can be assessed by comparing the experimentally obtained values with those predicted based on high resolution structures. Even forverysmall RDCs, such as the ^15N( i )−^13C^α( i − 1) couplings that are smaller than 0.3 Hz, a correlation coefficient of 0.83 is obtained, attesting to the accuracy of couplings obtained with these sensitivity-enhanced IPAP experiments. We also present a novel application for the use of RDCs. Under certain conditions, the folded state of aprotein comprises detectable, conformational sub-states. Such sub-states at local sites, so-called melting hot spots, are characterized by re-orienting bond vectors. Determination of RDCs allows for efficient and easy detection of such hot spots.

The binding of IgE to its high-affinity receptor, FcεRI, is fundamental to allergic disease. Molecules that block this interaction could therefore act as useful therapeutics for the treatment of asthma, allergic rhinitis, and other forms of atopy. To this end, binding selections using the extracellular portion of the α-chain of FcεRI (FcεRIα) and polyvalent peptide-phage libraries have yielded two distinct classes of peptide ligands that antagonize IgE binding to its receptor and prevent downstream IgE-mediated signaling events in basophils [1,2]. NMR spectroscopy has been used to characterize the structures of these peptide antagonists and their modes of binding to FcεRIα.

The advent of indirect detection has revolutionized the world of NMR spectroscopy in all areas of chemistry. for the purpose of assembling fragments of moieties as defined by NMR resonances into meaningful structural fragments or even complete structures, HMBC [1] is one of the most powerful tools available to the modern NMR spectroscopist. In situations of low sensitivity brought about most commonly by the availability of vanishingly small amounts of sample to analyze, direct observation of ^13C resonances can be difficult or impossible. Accordingly, the sensitivity afforded by indirect detection with HMBC and variants provides a tool that has the potential to access the key and difficult quaternary ^13C atoms while at the same time assembling the puzzle pieces into a coherent structure.

The twin challenges of biomolecular NMR spectroscopy are sensitivity and resolution. Nonuniform sampling, in which data collection is tailored to meet the requirements of sensitivity or resolution, enables new approaches to meet these challenges. We describe the use of nonuniform sampling in biomolecular NMR and the use of maximum entropy reconstruction to process the data. Data collected using nonuniform sampling cannot be processed using conventional methods such as the discrete Fourier transform or linear prediction, because these methods require data sampled at uniform intervals. A two- to three-fold reduction in acquisition time can routinely be realized for each indirect time dimension that employs nonuniform sampling, making this approach a valuable complement to cryogenic probes and high-field magnets for the most demanding biomolecular NMR applications. The approach is especially well suited for high-throughput applications of NMR spectroscopy in drug discovery. The development of Fourier transform (FT) NMR spectroscopy by Ernst and Anderson [1] in the late 1960s set the stage for the subsequent development of multidimensional experiments [2-A]. By resolving individual nuclear coherences, these experiments enable systematic determination of the structure of biological macromolecules in solution. However, the need to avoid spectral aliasing when using the discrete FT (DFT) to compute frequency spectra from the time domain data imposes substantial data collection requirements; it is not unusual for each additional dimension to increase the amount of data that must be collected, and the time required to complete the experiment, by two orders of magnitude. The Nyquist theorem places a lower bound on the sampling rate, or time between samples (the dwell time), needed to avoid aliasing. This rate is inversely proportional to the width of the spectrum. Frequency resolution comparable to the natural line widths of the resonances is needed to exploit the dispersion of resonances at a given magnetic field, and high-resolution requires data collected at long acquisition times, in essence because the closer two resonances are in frequency, the longer it takes for a detectable difference in their signals to evolve. Since the DFT requires samples collected at uniformly spaced intervals, acquiring data at long acquisition times requires collecting samples at all of the intervening multiples of the dwell time. The data collection requirements increase with increasing magnetic field, because the increased dispersion of resonances necessitates shorter dwell times. In the 1980s, linear prediction (LP) extrapolation of NMR data beyond the sampled interval was introduced [5,6], and it remains in common use today. LP extrapolation dramatically reduces the need for aggressive apodization when working with short data records, providing improvements in both sensitivity and resolution in favorable situations. However, LP has drawbacks; it has recently been shown that LP extrapolation generates falsepositive peaks when applied to noisy data or when used to extrapolate too far beyond the measured interval [7]. A more subtle but insidious defect is that LP extrapolation can cause slight frequency shifts of real signals in these situations. These shifts are not apparent on casual inspection but have potentially serious consequences, such as improper resonance assignments. Furthermore, the extent to which NMR data can be extrapolated using LP depends mainly on the signal-to-noise ( S/N ) ratio, and consequently additional LP extrapolation is not a general solution to increased data requirements imposed by the increased spectral dispersion at higher magnetic fields. Conventional approaches employing the DFT, including LP extrapolation, require acquisition times that can be prohibitively long, especially for biomolecules that are fleetingly stable or for data collected at very high magnetic field. Thus, there is considerable impetus for developing alternative methods that can produce highresolution multidimensional spectra in less time. A wide variety of different approaches are actively being developed. One approach involves multiplexing, resulting in coherence frequencies that encode more than one resonance frequency in a single dimension. These include “reduced dimensionality” experiments such as G-matrix FT (GFT) [8,9] and Hadamard encoding [10]. A drawback of these approaches is that they require special radiofrequency pulse sequences as well as additional post-processing steps to disentangle the multiplexed frequencies. A different approach to multiplexing has been developed by Frydman and colleagues [11], which utilizes orthogonal magnetic field gradients to partition the sample into “subsamples,” each reporting a different part of the spin response. This approach is not practical when sensitivity is already a limiting factor for measuring the response of the entire sample, but as advances in probe technology continue to improve sensitivity and gradient strength (which determines the resolution), it will likely find increasing application in biomolecular NMR. Parametric methods that model the NMR signal as a sum of exponentially decaying sinusoids, such as LP singular value decomposition [12], Hankel SVD [13], filter diagonalization method (FDM) [14], maximum likelihood [15], and Bayesian [16] methods are in principle capable of producing high-resolution spectra from very short data records. These methods often exhibit “spontaneous splitting,” resulting in multiple sinusoids for a single resonance when applied to noisy or nonideal (non-Lorentzian) data, complicating the parametric interpretation. Indeed the regularized resolvent transform [17], in which an FDM-like approach is used to extrapolate the signal, rather than compute a parametric decomposition, was developed to avoid this problem. Another class of methods utilizes symmetries or redundancy in multidimensional spectra to avoid acquiring all combinations of the evolution times for the different dimensions. An approach inspired by the methods used in medicine to reconstruct images in computerized axial tomography exploits the relationship between line projections in the frequency domain and linear cross sections in the time domain. Called back-projection reconstruction [18], this approach can, in favorable circumstances recover a complete two-dimensional spectrum from a small number of cross sections through the full time domain data matrix. However, it has been pointed out that in the general case n cross sections are needed to unambiguously resolve n peaks in multiple dimensions [19], and for complex biomolecules, this may not result in significant time savings. Another approach, called three-way decomposition [20], exploits the fact that line shapes in multiple dimensional spectra can be decomposed into products of one-dimensional line shapes. Back-projection reconstruction and three-way decomposition are special cases of nonuniform sampling in the time domain. In principle, any method that approaches spectrum reconstruction as an inverse problem, that is, computes spectra and inverts the spectrum to obtain “mock” data that are then compared with the empirical data for consistency, can be used with nonuniformly sampled time domain data. A very general method that makes no assumptions regarding the nature of the signals, and imposes no particular constraints on the subset of the time domain data matrix that must be collected, is maximum entropy (MaxEnt) reconstruction [21]. It has been shown recently to be particularly robust, yielding fewer false positives and more accurate frequencies than LP extrapolation [7]. It is also versatile: it can be used to construct spectra from the same data utilized by backprojection reconstruction, or can be used in combination with other methods, such as GFT (Jim Sun and Gerhard Wagner, personal communication). In principle, it can be used with any pulse sequence, and the computed spectra are compatible with conventional analysis tools developed for Fourier spectra. In this chapter, we describe the use of nonuniform sampling and MaxEnt reconstruction in multidimensional NMR experiments and its application to biomolecules. In practice, factors of two to three savings in data acquisition time per indirect time dimension are readily achieved. The robustness and versatility of MaxEnt reconstruction make it a powerful tool for improving sensitivity and resolution, and saving time in biomolecular NMR experiments.