Multiparametric Optimization of 31P NMR Spectroscopic Analysis of

May 5, 2010 - 31P NMR spectroscopy is known to be a fast and accurate method for analyzing phospholipid extracts from biological samples without prior...
1 downloads 10 Views 1MB Size
Download PDF
Anal. Chem. 2010, 82, 5433–5440

Multiparametric Optimization of 31P NMR Spectroscopic Analysis of Phospholipids in Crude Tissue Extracts. 1. Chemical Shift and Signal Separation Norbert W. Lutz* and Patrick J. Cozzone Centre de Re´sonance Magne´tique Biologique et Me´dicale, UMR CNRS 6612, Faculte´ de Me´decine de la Timone, Universite´ de la Me´diterrane´e, 13005 Marseille, France 31

P NMR spectroscopy is known to be a fast and accurate method for analyzing phospholipid extracts from biological samples without prior separation. However, the number of phospholipid classes and subclasses that can be quantitated separately in 31P NMR spectra of tissue extracts is critically dependent on a variety of experimental conditions. For solvent systems resulting in the formation of two phases, the effects of varying water and methanol content on chemical shift and line width of phospholipid signals have been previously determined. However, little attention has been paid to the influence that other extract components may exert on signal separation. We present, for the first time, a systematic and comprehensive study of 31P NMR chemical shift as a function of four experimental parameters: (i) extract concentration, (ii) concentration of chelating agent, (iii) pH value of the aqueous component of the solvent system, and (iv) temperature of the NMR measurement. This multiparametric study provides methodological guidelines for predictable and reproducible manipulation of 31P NMR spectra of brain phospholipids. It also provides a database for rational and efficient optimization of phospholipid spectra from other body tissues, cultured cells, and phospholipid-containing biofluids. 31

P NMR spectroscopy has been used for over 30 years to analyze phospholipid (PL) profiles in extracts of biological tissue.1 PLs constitute the matrix of human and animal cell membranes and provide numerous additional functions.2 Most importantly, membrane PLs are a rich source of cell signaling molecules,3 interact with membrane-embedded proteins (e.g., enzymes and transporters), and play an important role in cell recognition and cell-cell interaction. In the nervous systemsin particular, in the brainsPLs play a special role as crucial components of myelin. PLs are involved in both pathological * Author to whom correspondence should be addressed. Tel.: +33 491 324476. Fax: +33 491 256539. E-mail: [email protected]. (1) London, E.; Feigenson, G. W. J. Lipid Res. 1979, 20, 408–412. (2) Stryer, L. Biochemistry, 4th ed.; W. H. Freeman and Company: New York, 1995. (3) Tome, M. E.; Lutz, N. W.; Briehl, M. M. Biochim. Biophys. Acta 2003, 1642, 149–162. 10.1021/ac100514n  2010 American Chemical Society Published on Web 05/05/2010

and physiological states.4 Many diseases are associated with aberrations of the membrane PL composition and metabolism. Consequently, 31P NMR analysis of PL profiles has been suggested for differential diagnoses, e.g., for the characterization of tumors and other diseases,5-7 as research on PL metabolism in disease, has made significant progess over the past decade.8-10 31P NMR PL analysis is fast and accurate, and it can be applied to crude extracts without any need for prior PL separation.11 The number of known PL classes and subclasses that currently can be quantified varies between 5 and 20, depending on tissue type and experimental conditions. Although this technique is not appropriate for the identification of individual molecular PL species, it is more quantitative (provided saturation and polarization transfer effects are avoided5) and requires less sample preparation than more intricate methods, such as mass spectrometry. Nonetheless, the use of 31P NMR PL analysis for “phospholipidomics” is currently growing only at a slow pace. This is, in part, due to the sensitivity of this method to variations in several experimental parameters. In fact, the quality and reproducibility of 31P NMR PL spectra, as well as the number of PL classes and subclasses that can be quantified separately, vary considerably, as a function of optimization efforts. For the most frequently used PL solutions based on methanol-chloroform-water solvent systems, the effects of varying the water and methanol content on chemical shift and line width of PL signals have been previously determined.12,13 However, little (4) Aoki, J.; Inoue, A.; Okudaira, S. Biochim. Biophys. Acta 2008, 1781, 513– 518. (5) Solivera, J.; Cerdan, S.; Pascual, J. M.; Barrios, L.; Roda, J. M. NMR Biomed. 2009, 22, 663–674. (6) Merchant, T. E.; Minsky, B. D.; Lauwers, G. V.; Diamantis, P. M.; Haida, T.; Glonek, T. NMR Biomed. 1999, 12, 184–188. (7) Fuchs, B.; Schiller, J.; Wagner, U.; Hantzschel, H.; Arnold, K. Clin. Biochem. 2005, 38, 925–933. (8) Glunde, K.; Jacobs, M. A.; Bhujwalla, Z. M. Expert Rev. Mol. Diagn. 2006, 6, 821–829. (9) Beloueche-Babari, M.; Chung, Y. L.; Al-Saffar, N. M.; Falck-Miniotis, M.; Leach, M. O. Br. J. Cancer 2010, 102, 1–7. (10) Delikatny, E. J.; Cooper, W. A.; Brammah, S.; Sathasivam, N.; Rideout, D. C. Cancer Res. 2002, 62, 1394–1400. (11) Sotirhos, N.; Herslof, B.; Kenne, L. J. Lipid Res. 1986, 27, 386–392. (12) Edzes, H. T.; Teerlink, T.; van der Knaap, M. S.; Valk, J. Magn. Reson. Med. 1992, 26, 46–59. (13) Branca, M.; Culeddu, N.; Fruianu, M.; Serra, M. V. Anal. Biochem. 1995, 232, 1–6.

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

5433

Figure 1. (Top left panel) A one-phase system (left) was preferred over a two-phase system (right). The commonly used two-phase system hampers correct PL quantitation, because most of the upper phase is located outside the sensitive volume of the coil. (Top right panel) Complete range of the 31P NMR PL spectrum of rat brain. For better visibility of weak signals (PtdIP, PtdG), exponential line broadening (LB ) 3 Hz) was applied. In this representation, several PL signals are not well-resolved, notably in the PtdC and PtdE regions. For PL generating more than one 31 P NMR signal, observed nuclei are underlined (PtdIP, PtdIP, PtdIP2, PtdIP2). Currently unassigned signals are denoted by “Un” (where n ) 1, 2, ...). (Bottom left panel) PtdE and PtdS regions of the same spectrum. For better peak resolution, Lorentzian-Gaussian line shape transformation was applied (LB ) -1 Hz, GB ) 0.3). Because of these processing parameters, many very weak PL signals are difficult to detect. However, at least two peaks can be discerned for each PtdE, PtdEplasm, AAPtdE, and PtdS. For Figures S7-S12 in the Supporting Information, the more intense peak from each of these PL subgroups was chosen. (Bottom right panel) PtdC region obtained with the same processing parameters as the PtdE region. Several signals at the base of the dominating PtdC resonance were detected unambiguously, while they cannot be discerned in the upper spectrum generated with exponential line broadening. Besides the currently unassigned PtdC analog, PtdC1u, further minor resonances may be present upfield from PtdC.

attention has been paid to the influence that other extract components may exert on chemical shift and signal separation. Commonly used PL solutions that are based on methanol-chloroform-water systems result in two phases: one being predominantly nonpolar (chloroform/methanol), and the other predominantly polar (water/methanol). Although these systems may provide good 31P NMR signal resolution, they considerably complicate exact PL quantitation and may be a source of error,12 because complete phase separation only occurs over time. PLs concentrate in the nonpolar phase, the volume fraction of which would need to be accurately measured after phase separation to limit systematic quantitation errors. In addition, in many cases, the resolution that can be obtained from these two-phase systems is not optimal, as evidenced by the absence of fine structure from phosphatidylethanolamine analogs and phosphatidylserine.14 Our present study focuses on a novel, well-defined, and robust mixture forming a single homogeneous phase, thus entirely avoiding problems of phase separation (Figure 1, top left). This one-phase system permits more reliable and efficient PL quantitation than two-phase systems, and it provides high spectral resolution and sensitivity sufficient for quantitation of less-abundant PLs. We have occasionally used such mixtures in PL studies of extracts from cultured cells and (14) Estrada, R.; Stolowich, N.; Yappert, M. C. Anal. Biochem. 2008, 380, 41– 50.

5434

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

brain tissue;3,15,16 however, thus far, systematic and multiparametric studies of 31P NMR PL analysis for one-phase systems have been unavailable. To develop a methodology for the optimization of one-phase PL 31P NMR spectroscopy, we studied PL chemical shifts as a function of the following experimental parameters: (i) pH of the aqueous component of a well-defined methanol-chloroform-water solvent system; (ii) molar concentration of a chelating agent; (iii) tissue extract concentration (mg tissue per mL solvent); and (iv) measurement temperature. To the best of our knowledge, the combined effects of these four variables on 31P MRS chemical shift have not been investigated previously in tissue PL extracts. Based on our data, practical guidelines were derived that may serve to systematically and predictably optimize 31P NMR PL spectra of extracts from brain and other body tissues, cultured cells, and PL-containing biofluids. For the sake of clarity, the effects of all four parameters on PL line widths, as well as on the overall resolution of 31P NMR PL spectra, will be presented separately in Part 2 of this study.17 (15) Lutz, N. W.; Tome, M. E.; Aiken, N. R.; Briehl, M. M. NMR Biomed. 2002, 15, 356–366. (16) Viola, A.; Saywell, V.; Villard, L.; Cozzone, P. J.; Lutz, N. W. PLoS ONE 2007, 2, e157. (17) Lutz, N. W.; Cozzone, P. J. Anal. Chem. 2010, DOI: 10.1021/ac100515y.

EXPERIMENTAL SECTION Sample Preparation. Brains from a total of 10 Lewis rats were used. Animals were sacrificed by cervical dislocation. The brain was swiftly and entirely removed through an incision in the skull, freeze-clamped, and stored at -80 °C until extraction with methanol-chloroform-water (4:4:4 mL). For NMR spectroscopy, dried lipids were redissolved in a mixture of deuterochloroform, methanol, and an aqueous CDTA (trans-1,2-cyclohexyldiaminetetraacetic acid, cesium salt) solution (5:4:1). The amount of solvent was chosen to generate an extract concentration corresponding to 944 mg brain wet weight per mL solvent. After 31P NMR spectroscopy, the sample was diluted by a factor of 2, and again subjected to 31P NMR. Serial dilution was repeated until a final concentration of 118 mg brain wet weight per mL solvent was reached. Table S1 in the Supporting Information summarizes all of the samples used. All chemicals were purchased from Sigma-Aldrich (Saint Quentin, Fallavier, France), except for several PLs used for NMR signal assignment (Doosan Serdary Research, Toronto, Ontario, Canada). Further details of sample preparation are given in the Methods S1 section of the Supporting Information. 31 P NMR Spectroscopic Analysis. Spectra were acquired at 162 MHz (9.4 T) for 5 h from 277 K to 297 K, at 5 K intervals, on a model AVANCE 400 spectrometer (Bruker, Rheinstetten, Germany), with an acquisition time (AQ) of 3.165 s (16 000 data points), using an inverse-gated WALTZ16 proton decoupling device (using a power setting of 19 dB for a pulse width of 100 µs). Additional 31P NMR acquisition parameters included the following: sweep width, SW ) 25 ppm; pulse repetition time, TR ) 15 s; and pulse width (90°), PW ) 9.5 µs. Lorentzian/ Gaussian line shape transformation was applied to free induction decays before Fourier transform, phase, and baseline corrections. Spectra were referenced to the signal from the MDP solution mentioned previously (secondary external standard at 19.39 ppm). In the remainder of this study, observed 31 P nuclei are underlined in PL abbreviations where necessary, to avoid ambiguity. For instance, PtdIP refers to the 31P nucleus in the diesterified phosphate group of phosphatidylinositol 4-phosphate, while PtdIP refers to the monoesterified phosphate group of the same PL molecule. Further details on the NMR spectroscopic methods used are given in the Methods S1 section of the Supporting Information. THEORETICAL BASIS Distinguishing Phospholipid Groups and Subgroups by 31 P NMR Spectroscopy. PLs can be detected as a broad hump in in vivo 31P NMR spectra. Early work on dog brain from Chance’s laboratory demonstrated that up to 35% of an in vivo 31 P NMR signal may be due to unresolved PL signals, mostly stemming from partially mobile PL.18 However, to obtain detailed information on PL composition, PLs must be mobilized by appropriate extraction and sample preparation techniques. In this way, 31P NMR spectra yield signals that represent individual PL classes and subclasses. These phosphorylated compounds are characterized by distinct intramolecular environments close to the observed nucleus, i.e., at or near the (18) Cerdan, S.; Subramanian, V. H.; Hilberman, M.; Cone, J.; Egan, J.; Chance, B.; Williamson, J. R. Magn. Reson. Med. 1986, 3, 432–439.

phosphate moiety. Thus, in 31P NMR spectra, PLs are distinguishable based on (i) differences in the polar head, (ii) differences in the PL backbone, and (iii) differences in chemical bonds on glycerol C-1 and C-2. Further structural differences in fatty acid (FA) chains may cause peak splitting in certain cases, as discussed below. However, differences in unsaturation or peroxidation occurring at relatively distant sites of FA chains do not result in distinct 31P NMR resonances. Consequently, even the lowest achievable line width in our study (ca. 0.8 Hz) does not permit analysis of individual PL species, which is more appropriately done using chromatographic separation techniques coupled with mass spectrometry. Effects of the Intermolecular Environment on 31P NMR Chemical Shift of Phospholipids in Crude Extracts. Optimization of 31P NMR peak separation for PL analysis must take into account relative peak positions that are determined by chemical-shift differences. Chemical shifts are dependent on the chemical nature of the sample, as well as on experimental parameters (for instance, measurement temperature). For complex mixtures such as crude tissue extracts, it is virtually impossible to quantitatively characterize the effects of individual sample components on measured chemical shifts. Moreover, most NMR literature on PL involves bilayers in vesicles such as liposomes,19 or oil-in-water (O/W)-type micelles. Since, in our tissue extracts, the dominating solvent is nonpolar (chloroform), reverse micelles (water-in-oil) may need to be taken into account.20 Yet, very little is known about the processes governing PL chemical shift in impure reverse micelles. In addition, the precise intermolecular environment of PLs in tissue extracts that contain sizable amounts of nonphosphorylated lipids and water-soluble compounds such as CDTA complexes is unknown. These systems may possess characteristics of microemulsions,21 but structures are expected to vary as a function of experimental conditions. In this report, we do not attempt to present intramolecular or intermolecular PL structures and detailed explanations of their effects on NMR parameters, but rather pinpoint several essential factors that must be considered. First, the addition of chelating agent (CDTA) to extracts serves to “mask” divalent and, to a lesser degree, trivalent metal cations via the formation of stable cation-CDTA chelates. Unmasked paramagnetic cations would broaden PL 31P resonances, thus worsening PL signal resolution. Optimization of PL analysis by 31P NMR spectroscopy must take into account CDTAdependent effects on the 31P chemical shift. Second, tissue extract concentration affects both intermolecular aggregation among PL molecules and aggregation of PL molecules with nonphosphorylated lipids. Through this mechanism, the PL concentration may influence the chemical shifts of PL 31P resonances. From studies of reverse micelles of surfactants in organic solvents, it is known that there is a direct relationship between the water/surfactant molar ratio and the size of reverse micelles.22,23 Correspondingly, micelle formation and (19) Milburn, M. P.; Jeffrey, K. R. Biophys. J. 1987, 52, 791–799. (20) Kumar, V. V.; Manoharan, P. T.; Raghunathan, P. J. Biosci. 1982, 4, 449– 454. (21) Langevin, D. Ann. Rev. Phys. Chem. 1992, 43, 341–369. (22) Bru, R.; Sanchez-Ferrer, A.; Garcia-Carmona, F. Biochem. J. 1995, 310 (Pt 3), 721–739.

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

5435

micellar size in tissue extracts are dependent on PL concentration. Furthermore, micellar size determines the structure of water molecules within the micelle. As micellar size increases, the amount of “free” water (water molecules not bound to the PL polar heads) increases, relative to the amount of “bound” water,20 in analogy to data obtained from other surfactant-based systems.23 Also, the methanol distribution between three different pools (intramicellar; at the micellar interface; dissolved in chloroform) is concentration-dependent.21 As a result, varying intramicellar water/methanol ratios may affect 31P NMR chemical shift through the interaction of solvent molecules with the polar PL heads. Third, pH affects the PL 31P NMR chemical shift24 via three different mechanisms: (i) directly, via protonation of the PL phosphate moiety;1 (ii) indirectly, via protonation of compounds interacting with ions that potentially complex with the PL phosphate moiety (the latter effect most certainly applies to CDTA, because the stability constant of metal chelates is known to be pH dependent). In addition, (iii) indirect effects of pH on the structure of micelle-like aggregations cannot be excluded. Finally, the temperature of the sample during spectrum acquisition markedly affects the chemical shift of the PL 31P NMR signals,25 because a change in temperature may alter ion exchange and the structure and size of lipid aggregations.22 RESULTS AND DISCUSSION Peak Detection in Phospholipid 31P NMR Spectra of Brain Tissue Extracts. The number of peaks that could be consistently observed over the entire temperature range investigated varied as a function of experimental parameters. The maximum number was 36 for a 236 mg/mL brain extract at pH 7.9 and 200 mM CDTA (each of these peaks was observed for at least two different temperatures). This number was drastically decreased for spectra with major line broadening, because of either high extract or low CDTA concentrations, and/or because of suboptimal pH. Most PLs (see top right panel of Figure 1) yield one single 31P NMR resonance, representing either the respective phosphate moiety or two phosphate moieties with identical 31 P chemical shifts (cardiolipin, CL). In contrast, phosphatidylinositol monophosphates and diphosphates contribute two and three 31P NMR resonances, respectively. Furthermore, PtdS (phosphatidylserine), PtdE (phosphatidylethanolamine), PtdEplasm (ethanolamine plasmalogen, 1-alkenyl-2-acyl-sn-glycero-3-phosphoethanolamine), and AAPtdE (1-alkyl-2-acyl-snglycero-3-phosphoethanolamine) may each yield several partially resolved peaks (generally two to three), because of structural differences in FA chains, as indicated previously in the “Theoretical Basis” section (see bottom left of Figure 1). However, since AAPtdE occurs at a very low concentration, unambiguous detection of its peak splitting turned out to be difficult in some spectra. Further effects of the PL structure on the 31P NMR chemical shift are discussed in the Discussion S1 section in the Supporting Information. At high CDTA concentrations, the phosphatidylglycerol (PtdG) signal was virtu(23) Marhuenda-Egea, F. C.; Piera-Velazquez, S.; Cadenas, C.; Cadenas, E. Archaea 2002, 1, 105–111. (24) Sappey-Marinier, D.; Letoublon, R.; Delmau, J. J. Lipid Res. 1988, 29, 1237– 1243. (25) Metz, K. R.; Dunphy, L. K. J. Lipid Res. 1996, 37, 2251–2265.

5436

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

ally impossible to identify for all but the lowest temperatures (separately verified at 500 and 1000 mM CDTA); therefore, we did not attempt to evaluate this peak for CDTA concentrations above 200 mM. Unambiguous PtdG identification also turned out to be difficult at 50 mM CDTA (broad peaks). The origin of this particular behavior of PtdG may be related to the protonation and/ or complexation of the two vicinal hydroxyl groups in the glycerol moiety of the polar head. However, resolution of this issue must await further clarification. The size of a chemical-shift change induced by systematic variation of an experimental parameter, notably temperature, is dependent on the PL studied. Therefore, chemical-shift curves determined for different PL may cross for specific experimental conditions. This situation rendered an unambiguous identification of PLs difficult in cases where 31P NMR signals were particularly weak. On the other hand, our optimized protocol enabled the detection of minor PL signals in the first place and, therefore, provides a basis for future assignment of currently unidentified peaks. The Problem of Concurrent Effects of Experimental Conditions on 31P Chemical Shift. The capacity of 31P NMR spectroscopy to identify and quantitate individual PL classes and subclasses in one-phase extracts of brain tissue was determined to be critically dependent on experimental parameters such as the concentration of chelating agent, CDTA (Figure 2, left column), and sample temperature (Figure 2, right column). To illustrate the first example, both the top left and bottom left panels of Figure 2 present 31P NMR spectra for identical extract concentrations (236 mg/mL) and measurement temperatures (297 K), as well as virtually identical pH values in the aqueous component of the solvent (7.33 and 7.36, respectively). The sample with a very high CDTA concentration (1000 mM in the aqueous component of the solvent; see the bottom left panel of Figure 2) resulted in markedly narrower lines (1.5 Hz line width at half-height for phosphatidylcholine, PtdC) than did the sample with less CDTA (2.1 Hz, 200 mM; see the top left panel of Figure 2). Nevertheless, better spectral resolution in the PtdE/SM (sphingomyelin) region was achieved at the lower CDTA concentration (see the top left panel of Figure 2), because, at 1000 mM, the PtdEplasm signal was completely superimposed with the SM signal (see the bottom left panel of Figure 2). The second example compares spectra for identical extract concentrations (118 mg/mL), and for identical pH values (7.14) and CDTA concentrations (50 mM) in the aqueous component of the solvent. The spectrum acquired at low sample temperature (277 K, bottom right panel of Figure 2) showed markedly narrower lines (1.6 Hz line width at half-height for PtdC) than did the spectrum acquired at high temperature (297 K, top right panel of Figure 2, 2.3 Hz). Nonetheless, better spectral resolution in the PtdE/SM region was achieved at the higher temperature (top right panel of Figure 2), because, at 277 K, the PtdE signal was completely superimposed with the SM signal (see the bottom right panel of Figure 2). These examples illustrate the problems one may encounter due to concurrent effects of experimental conditions on chemical shift. A systematic study of LW effects is presented in Part 2 of this study.17

Figure 2. (Top left panel) PtdE region of a 31P NMR spectrum of a rat brain tissue extract. Extract concentration, 236 mg/mL; CDTA concentration and pH in the aqueous component of the solvent, 200 mM and 7.33, respectively; measurement temperature, 297 K. The PtdEplasm and SM signals are well-resolved. (Bottom left) PtdE region of a 31P NMR spectrum of a rat brain tissue extract. Extract concentration, 236 mg/mL; CDTA concentration and pH in the aqueous component of the solvent, 1000 mM and 7.36, respectively; measurement temperature, 297 K. The PtdEplasm and SM signals overlap entirely; they cannot be resolved despite reduced line width, compared to the top left spectrum. (Top right) PtdE region of a 31P NMR spectrum of a rat brain tissue extract. Extract concentration, 118 mg/mL; CDTA concentration and pH in the aqueous component of the solvent, 50 mM and 7.14, respectively; measurement temperature, 297 K. The PtdE and SM signals are well-resolved. (Bottom right) PtdE region of a 31P NMR spectrum of a rat brain tissue extract. Extract concentration, 118 mg/mL; CDTA concentration and pH in the aqueous component of the solvent, 50 mM and 7.14, respectively; measurement temperature, 277 K. The PtdE and SM signals overlap entirely; they cannot be resolved, despite reduced line width, compared to the top right spectrum.

Influence of Temperature on 31P NMR Chemical Shift. Correct identification and accurate quantitation of PL based on 31 P NMR spectra of tissue extracts require that individual PL signals be sufficiently intense and well-separated from each other. While both the signal-to-noise ratio and the peak separation are affected by line widths (see Part 2 of this study17), absolute and relative peak positions on the chemicalshift scale are also crucial. For this reason, we determined 31P NMR chemical shifts for mono- and diesterified phosphate moieties of PLs, as a function of extract and CDTA concentrations, sample temperature, and pH. Most 31P NMR resonances detected in tissue PL extracts stem from diesterified phosphate moieties; these will be presented first. Within the temperature range measured, PtdC chemical shifts increased, as the temperature increased, in a virtually linear fashion (Figure 3, top left panel, r ) 1.00). Increases ranged between 0.003 and 0.004 ppm/K, and they were not correlated with the pH, extract concentration, or CDTA concentration used (see Figure 3 and Figures S1-S6 in the Supporting Information). Chemical shifts of PLs other than PtdC usually followed the trends shown by PtdC; however, in some cases, they had a tendency to level off at higher temperatures (see Figure 3 and Figures S1-S6 in the Supporting Information, top left). These measurement points often were somewhat better fitted by logarithmic curves or secondorder polynomials than they were by straight lines. On the basis of linear fits, the temperature sensitivity of chemical shifts varied as a function of the PL investigated, notably of their head groups.

PtdC analogs that differed from PtdC only by the double bonds in fatty acid chains (1-alkyl-2-acyl-sn-glycero-3-phosphocholine, AAPtdC, or 1-alkenyl-2-acyl-sn-glycero-3-phosphocholine, PtdCplasm) exhibited virtually the same slope as did PtdC (range given above). In contrast, PtdE and its analogs (AAPtdE and PtdEplasm) had a slope that was approximately twice as steep, in the range of 0.007-0.008 ppm/K (see Table S2 in the Supporting Information for the selected values obtained for all major PL peaks). SM, which is characterized by a choline moiety in the polar head, but also by a backbone different from that of the PtdC analogs, had slopes in the range of 0.004-0.005 ppm/K, i.e., somewhat higher than the values measured for the other cholinecontaining PLs. Although the temperature dependence of PtdG resembled that of PtdC, the steepest slope (ca. 0.01 ppm/K) was measured for phosphatidylserine (PtdS), which is the only PL measured that contains an amino acid in the polar head. Wherever sufficiently resolved, two peaks for each PtdE, AAPtdE, PtdEplasm, and PtdS signal were evaluated separately. PL with polyol moieties (phosphatidylinositol, PtdI; PtdIP; phosphatidylinositol4,5-diphosphate, PtdIP2) consistently exhibited a moderate temperature dependence (ca. 0.006 ppm/K) for the 31P chemical shifts of their diesterified phosphate groups. This sharply contrasts with the behavior of monoesterified phosphate groups. Their 31P NMR chemical shifts were virtually invariant to temperature changes, or slightly decreased with increasing temperature for PtdIP, PtdIP2, and phosphatidic acid (PtdA). Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

5437

Figure 3. 31P NMR chemical shifts of PL peaks from rat brain extracts, as a function of measurement temperature (individual series measured for five different temperatures). Only those peaks that were clearly detectable for at least two temperatures have been included. In this figure, as well as Figures S1-S12 in the Supporting Information, shortened abbreviations have been used due to space restrictions: PC ) PtdC (phosphatidylcholine); PCpl ) PtdCplasm (phosphatidylcholine plasmalogen); AAPC ) AAPtdC (alkyl-acyl-phosphatidylcholine); PIP2 ) PtdIP2 (phosphatidylinositol diphosphate); PI ) PtdI (phosphatidylinositol); PS ) PtdS (phosphatidylserine); PE ) PtdE (phosphatidylethanolamine); PEpl ) PtdEplasm (phosphatidylethanolamine plasmalogen); AAPE ) AAPtdE (alkyl-acyl-phosphatidylethanolamine); PA ) PtdA (phosphatidic acid); PG ) PtdG (phosphatidylglycerol). Currently unassigned peaks are denoted by “Un” (where n ) 1-23). Peaks that are not currently identified peaks but most probably represent analogs of phosphatidylcholine are denoted by “PC1u” and “PC2u. Less-confident assignments are followed by question marks (e.g. “lyPC?” indicates a possible lyso-phosphatidylcholine peak). Conditions: extract concentration, 236 mg brain tissue per mL solvent; CDTA concentration, 50 mM; pH 7.14 in the aqueous component of the extract solvent used.

(A minor trend toward negative slopes is observed; see Figures S1-S6 in the Supporting Information.) The unassigned peaks, U3 and U4, exhibited a temperature dependence similar to that of PA, whereas the U5 chemical shift slightly increased with temperature. Differences in temperature sensitivity of chemical shifts between PLs resulted in alterations in relative peak positions over a given temperature range. Thus, several signals that are relatively distant from each other at one temperature, approach or even overlap at a different temperature (see Figure 3; also see Figures S1-S6 in the Supporting Information, notably, the far-right, topcenter, and bottom-left diagrams). Even the order in which peaks occur in the spectrum changed for some signals, as depicted by crossing lines for SM and PtdE (notably, the far-right diagrams in Figures S1 and S3-S6 in the Supporting Information). The crossing temperature depended on the extract and CDTA concentrations, as well as on pH. In Figures S1, S5, and S6 in the Supporting Information, the lowest crossing temperature was ca. 277 K, and the highest was ∼287 K. For particular combinations 5438

Analytical Chemistry, Vol. 82, No. 13, July 1, 2010

of CDTA concentration and pH, the temperature sensitivity of chemical shifts for most PLs was determined to be unexpectedly low for an extract concentration of 236 mg/mL. This is consistent with the anomalous concentration dependence described in the following paragraph. Influence of Extract Concentration on 31P NMR Chemical Shift. Generally, PtdC chemical shifts decreased moderately, as the extract concentration increased, in an approximately linear manner, although there was an anomaly at 236 mg/mL for several PLs (see Figure S7 in the Supporting Information). For this concentration, chemical shifts were smaller for temperatures higher than (and greater for temperatures lower than) the values expected for a near-linear relationship, as seen for pH 8.0 (50 and 200 mM CDTA) and pH 7.4 (1000 mM CDTA). Nevertheless, for all combinations of pH and CDTA concentration, plots of chemicalshift decreases for different temperatures resulted in almostequidistant lines (see Figure S7 in the Supporting Information), which reflects the linear temperature dependence described in the previous paragraph. The dependence of chemical shift on

extract concentration was less pronounced as the sample temperature decreased and the CDTA concentration increased. Thus, at 297 K and 50 mM CDTA (pH 7.4), the PtdC chemical shift decreased by 0.027 ppm as the extract concentration increased 8-fold (r ) 0.98), but it decreased by only 0.002 (1000 mM CDTA, pH 7.4) or 0.003 ppm (200 mM CDTA, pH 8.0) at 277 K. The extract concentration dependence of chemical shifts for all major PLs generally followed the pattern observed for PtdC (see Figures S7 and S8 in the Supporting Information). However, there only was a minor trend toward decreasing chemical shift for PtdI, even at 50 mM CDTA (see Figure S8A in the Supporting Information). CL, which is a PL with four fatty acyl (FA) chains rather than two, exhibited an unusual decrease in chemical shift at the highest extract concentration (pH 7.4, 50 mM CDTA; see Figure S8D in the Supporting Information). The behavior of PtdIP2 represents a clear exception to the rule that PL chemical shifts decrease as the extract concentration increases. In fact, PtdIP2 chemical shifts generally increased with, or were largely independent (for 1000 mM CDTA) of, extract concentration (see Figure S8A in the Supporting Information), again with the characteristic anomaly previously described for an extract concentration of 236 mg/mL (see Figure S7 in the Supporting Information). For the 31P NMR resonances from monoesterified phosphate groups (PtdA, PtdIP, and PtdIP2), decreases in the chemical shift, as a function of extract concentration, were more pronounced (see Figures S8D and S8E in the Supporting Information) than those described previously for diesterified phosphate. While chemical-shift variations for the latter never exceeded 0.05 ppm over the entire concentration range (for any combination of pH and CDTA concentration), variations in the former attained values up to 0.3 ppm in the case of PtdIP2 and were most pronounced at low CDTA concentrations (Figures S8D and S8E in the Supporting Information). For 50 mM CDTA, PtdIP2 chemical shifts initially showed a modest increase between 118 mg/mL and 236 mg/mL before consistently decreasing with extract concentration. For PtdA, chemical-shift differences were highest at 200 mM CDTA, whereas for 50 mM CDTA, the concentration dependence was almost plateaulike for extract concentrations of