Monitoring Real-Time Metabolism of Living Cells by Fast Two

Feb 15, 2010 - Address: Istituto di Chimica Biomolecolare del CNR, Comprensorio Olivetti, Edificio A, Via Campi Flegrei 34, I-80078 Pozzuoli (Napoli),...
0 downloads 0 Views 857KB Size
Download PDF
Anal. Chem. 2010, 82, 2405–2411

Monitoring Real-Time Metabolism of Living Cells by Fast Two-Dimensional NMR Spectroscopy Andrea Motta,* Debora Paris, and Dominique Melck Istituto di Chimica Biomolecolare del Consiglio Nazionale delle Ricerche, I-80078 Pozzuoli (NA), Italy Living cell metabolism is often monitored by 1D NMR spectroscopy, but the spectral resolution and the short cell lifetime are certainly limiting aspects. 2D spectroscopy does yield higher resolution but is time-consuming since acquisition of the second dimension requires several minutes. However, after only few minutes, oxygen starvation changes cell metabolism, and long acquisition times may yield spectra that do not represent the cell physiological state. Accordingly, metabolic studies of cells require fast NMR data acquisition. Here, we have applied band-selective optimized flip-angle short-transient (SOFAST)-HMQC techniques to 15Nlabeled cells, showing for the first time that it is possible to obtain 2D 1H-15N correlation spectra of small metabolites directly in living cells, in a few seconds and with a high S/N ratio. SOFAST-HMQC spectra of 15N-labeled Thalassiosira rotula diatoms cells can be acquired in 10-15 s, and, as an application, we have detected a progressive variation of the amino acid content when diatoms are exposed to UV-B radiation, with no need of long analytical procedures to quantify the metabolic changes. We believe that fast acquisition techniques can easily be extended to other cell systems, foreseeing a wide application in the emerging fields of metabolomics and metabonomics, being able to picture the “instantaneous” in-cell metabolism. Nuclear magnetic resonance (NMR) is a well-established technique for monitoring metabolism in living cells. They are often investigated by one-dimensional (1D) NMR spectroscopy, therefore benefiting of real-time measurements since all spectral frequencies are excited by a single scan. However, 1D NMR lacks the resolution needed to cope with the degeneracy of the resonance frequency and a reasonable signal-to-noise (S/N) ratio because of the short acquisition time required for the short lifetime of samples. The lack of resolution can be circumvented by twodimensional (2D) spectroscopy that, compared with 1D, does yield higher resolution but is intrinsically time-consuming since data acquisition for the second dimension requires at least several minutes. The total experimental time will be the product of the number of scans Nscan required for a proper sampling of the indirect domain, and the single-scan duration (the repetition * Corresponding author. Address: Istituto di Chimica Biomolecolare del CNR, Comprensorio Olivetti, Edificio A, Via Campi Flegrei 34, I-80078 Pozzuoli (Napoli), Italy. E-mail: [email protected]. 10.1021/ac9026934  2010 American Chemical Society Published on Web 02/15/2010

time) Tscan, which includes the spin relaxation time necessary to restore the thermal equilibrium before the next additional measurement takes place. This recycle delay is, therefore, associated with the 1H spin-lattice relaxation time (T1), and, depending on its duration, acquisition times can be on the order of minutes, yielding total experimental times of hours. Cells are able to survive and stay suspended in the solvent medium for several hours, but after only a few minutes, oxygen starvation changes their metabolism and decreases the cytoplasmic pH.1 Therefore, long acquisition times may detect molecules originating from an “average” metabolism that does not correspond to a specific physiological state of the cell. For samples with short lifetime, data acquisition must be rapid, and fastacquisition 2D techniques, as those used to study the structure and dynamics of proteins in solution,2 are required. Two different strategies have been put forward for fast acquisition spectroscopy: the “single-scan” NMR3,4 and the band-selective optimized flipangle short-transient (SOFAST) heteronuclear multiple quantum correlation (HMQC).5,6 The single-scan approach is able to record any multidimensional NMR spectrum within a single repetition of the experiment, but with current spectrometer hardware, it typically lacks in sensitivity, resolution, and/or sufficient gradient strength over extended periods of time. Alternatively, the SOFAST method is able to drastically reduce Tscan by relaying on accelerated T1 of the spins of interest7 and on optimized flipangles (e.g., the Ernst angle8) to enhance the steady-state magnetization of the excited spins.9 Brutscher and co-workers have combined these features into single 2D and threedimensional NMR protocols,5,6,10,11 showing that it is possible to reduce Tscan down to e100 ms, obtaining 2D 1H-15N or 1 H-13C correlation spectra within seconds, and with high S/N ratio. (1) Serber, Z.; Selenko, P.; Ha¨nsel, R.; Reckel, S.; Lo ¨hr, F.; Ferrell, J. E., Jr.; Wagner, G.; Do ¨tsch, V. Nat. Protoc. 2006, 1, 2701–2709. j . J. Biomol. NMR 2003, 27, 101–113. (2) Freeman, R.; Kupcˇe, E (3) Frydman, L.; Lupulescu, A. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 15858– 15862. (4) Shrot, Y.; Frydman, L. J. Am. Chem. Soc. 2003, 125, 11385–11396. (5) Schanda, P.; Brutscher, B. J. Am. Chem. Soc. 2005, 127, 8014–8015. j .; Brutscher, B. J. Biomol. NMR 2005, 33, 199–211. (6) Schanda, P.; Kupcˇe, E (7) Pervushin, K.; Vo ¨geli, B.; Eletsky, A. J. Am. Chem. Soc. 2002, 124, 12898– 12902. (8) Ernst, R. R.; Bodenhausen, G.; Wokaun, A. Principles of Nuclear Magnetic Resonance in One and Two Dimensions; Oxford University Press: Oxford, 1987. (9) Ross, A.; Salzmann, M.; Senn, H. J. Biomol. NMR 1997, 10, 389–396. (10) Schanda, P.; Brutscher, B. J. Magn. Reson. 2006, 178, 334–339. (11) Schanda, P.; Van Melckbeke, H.; Brutscher, B. J. Am. Chem. Soc. 2006, 128, 9042–9043.

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2405

Because of its adaptability to routine spectrometers, we have investigated the possibility of the use of the SOFAST-HMQC approach to explore cellular metabolism in 15N-labeled cells. We, here, report, for the first time, that the experiment allows acquisition of 2D 1H-15N correlation spectra of small metabolites directly in living cells in a few seconds, with a high S/N ratio, therefore affording a picture of the “instantaneous” incell metabolism. In particular, we have applied the SOFASTHMQC experiment to 15N-labeled diatoms cells, which are unicellular algae. They are at the base of the marine food web and are the major contributors to phytoplankton biomass worldwide. In response to favorable light and nutrient conditions, diatoms rapidly divide and form large blooms, and as blooms propagate, nutrients are depleted, growth ceases, and cells sink to the deep ocean. The sinking diatom blooms fuel the biological carbon pump and export carbon from the atmosphere to the deep ocean. Despite this, little is known about the molecular underpinnings of diatom biology. As a part of a long-running project, we have started a study of the metabolic profile of Thalassiosira rotula to understand how diatoms acquire nutrients, how they respond to stress, and how they activate chemical defense and chemical signaling that regulates algal bloom. Although useful information can be achieved by investigating the metabolic profile of polar and lipophilic extracts, in vivo studies of T. rotula cells in (artificial) seawater are expected to yield a more reliable understanding of the metabolic pathways. On the other hand, the presence of salt in the artificial seawater culture medium, used to suspend the cells in the NMR tube, will cause resonance broadening, and this, together with the degeneracy of the resonance frequency, will make 1D spectroscopy useless. T. rotula cells can easily be cultured on unlabeled and 15N-labeled media, and this warrants that a sufficient number of colonies can rapidly be obtained to test the potential application of the SOFAST-HMQC sequence to cells. The 2D correlation spectra obtained for T. rotula cells in 10-15 s with a high S/N ratio suggest that fast acquisition techniques introduced for proteins can be easily extended to other living cell systems, monitoring the metabolism under physiological or stressing conditions in the emerging fields of metabolomics and metabonomics.12,13 EXPERIMENTAL SECTION Cell Culturing. Axenic cultures of T. rotula cells were prepared as described.14 Briefly, diatoms were grown in Guillard’s (F/2) marine enrichment basal salt mixture powder medium, containing standard (35‰) or different salinities (20 and 45‰), and unlabeled or 15N-labeled NaNO3, on a 12 h light/12 h dark cycle, and a light intensity of 20.9 J mol-1 µm-2 s-1. Cells were kept in a 10 L carboy for 1 week and then harvested in the early stationary phase by centrifugation at 1200g in a swingout rotor. For UV irradiation,15 cells were harvested during (12) Nicholson, J. K.; Wilson, I. D. Nat. Rev. Drug Discov. 2003, 2, 668–677. (13) Beckonert, O.; Keun, H. C.; Ebbels, T. M. D.; Bundy, J.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. Nat. Protoc. 2007, 2, 2692–2703. (14) Miralto, A.; Barone, G.; Romano, G.; Poulet, S. A.; Ianora, A.; Russo, G. L.; Buttino, I.; Mazzarella, G.; Laabir, M.; Cabrini, M.; Giacobbe, M. G. Nature 1999, 402, 173–176. (15) Do ¨hler, G. Mar. Biol. 1984, 83, 247–253.

2406

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

exponential growth and exposed to UV-B radiation for 2 days (5 h per day) in a special quartz tube. A control suspension was put into a glass vessel (cells not irradiated with UV-B). The UV-B dose of 717 J m-2 day-1 was obtained by placing the lamp (Philips TL 40/12) 60 cm from the cultures.15 Prior to extraction of not irradiated cells, diatom cultures were allowed to settle overnight and the supernatant was gently removed by suction with a water pump. Extracts Manipulation. Combined extraction of polar and lipophilic metabolites from unlabeled and 15N-labeled diatoms cells was carried out using the methanol/chloroform procedure.16 Pelleted cells were resuspended in methanol (4 mL/g pellet)/water (0.85 mL/g pellet) and sonicated for 2 min. Then, 4 mL/g pellet of chloroform was added, and the homogenate was gently stirred and mixed on ice for 10 min using an orbital shaker (the solution must be monophasic). Another 4 mL/g pellet of chloroform and 4 mL/g pellet of water were then added, and the final mixture was shaken well and centrifuged at 12 000g for 15 min at 4 °C. This procedure separates a water/ methanol phase at the top (aqueous phase, with the polar metabolites), a phase of denatured proteins and cellular debris in the middle, and a chloroform phase at the bottom (lipid phase, with lipophilic compounds). The upper layer of each sample was transferred into glass vials, and after solvent removal under a stream of dry nitrogen, the sample was stored at -80 °C until required. For 1D and 2D NMR experiments, the polar extracts were resuspended in 700 µL of 1H2O-2H2O (90%-10%) and then transferred into an NMR tube. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on slab gel containing 12 and 15% acrylamide was performed using the standard procedure.17 Proteins were located on the gels using Comassie Brillant Blue staining. For 12% acrylamide, we used Phosphorylase b (97.4 kDa), bovine serum albumine (66.2 kDa), ovalbumin (45.0 kDa), carbonic anhydrase (31.0 kDa), trypsin inhibitor (21.5, kDa), and lysozyme (14.4 kDa), all from BIO-RAD. For 15% acrylamide, we used chymotrypsinogen A (24 kDa), cytochrome c (13 kDa), bovine pancreatic tripsin inhibitor (BPTI, 6.6 kDa), and insulin B-chain (3.5 kDa), all from Sigma. Size-exclusion chromatography was carried out at room temperature, using a 1.5 × 50 cm Sephadex G-50 Fine column and a flow rate of 0.2 mL/min. Separate chromatography experiments of standard amino acids were performed in 50 mM sodium phosphate, at pH 6.7, using a 55 µM concentration. Salmon calcitonin (3.4 kDa), bacitracin (1.4 kDa), and standard amino acids, all from Sigma, and sodium 3-(trimethylsilyl)-(2,2,3,32 H4)propionate (TSP, 172 Da), from Aldrich, were used as molecular mass standards. NMR Experiments. 1H NMR spectra were recorded at 600 MHz on a Bruker DRX-600 spectrometer, equipped with a TCI CryoProbe fitted with a gradient along the Z-axis. Spectra were referenced to internal TSP. Clean total correlation spectroscopy (TOCSY)18 spectra of cells and extracts were recorded using (16) Lindon, J. C.; Nicholson, J. K.; Holmes, E. Nat. Biotechnol. 2005, 23, 833– 838. (17) Sambrook, J.; Fritsch, E.; Maniatis, T. Molecular Cloning: A Laboratory Manual; Cold Spring Harbor Laboratory: Cold Spring Harbor, NY 1989. (18) Griesinger, C.; Otting, G.; Wu ¨ thrich, K.; Ernst, R. R. J. Am. Chem. Soc. 1998, 110, 7870–7872.

the time-proportional phase incrementation of the first pulse and incorporating the excitation sculpting sequence19 for water suppression. We used a double-pulsed field gradient echo, with a soft square pulse of 4 ms at the water resonance frequency, with the gradient pulses of 1 ms each in duration. In general, 256 equally spaced evolution-time period t1 values were acquired, averaging 2 (for diatoms) and 8 (for extracts) transients of 2048 points, with 6024 Hz of spectral width. Heteronuclear single quantum coherence (HSQC)-TOCSY experiment20 of extracts was acquired using the echo/antiechoTPPI gradient selection, with 256 evolution-time period t1 values, and averaging 4 transients of 2048 points over 6024 Hz of spectral width. Time-domain data matrices were all zerofilled to 4K in both dimensions, applying, prior to Fourier transformation, a Lorentz-Gauss window with different parameters for both t1 and t2 dimensions in all the experiments. The 1H-15N SOFAST-HMQC pulse sequence follows the scheme proposed.5 First, 1H pulses are applied band-selectively;7 second, the first 1H pulse has an adjustable flip angle R that allows further optimization of the sensitivity of the experiment for a chosen (short) scan time.9 In practice, the flip angle is chosen so as to ensure that part of the 1H magnetization is restored along the z-axis by the following 180° pulse; third, the small number of radio frequency pulses reduces signal loss due to pulse imperfections and B1 field inhomogeneities and limits the effects of sample and probe heating. The application of only two (band-selective) 1H pulses also ensures minimal perturbation of the unobserved proton spin polarization, thus optimally exploiting the spin-lattice relaxation enhancement effect. This can be achieved by the use of a polychromatic PC9 pulse shape for adjustable flip-angle band-selective excitation,21 which yields quite uniform excitation over the desired bandwidth for flip angles in the range 0° < R < 130°. As a refocusing pulse on the 1H channel, we tested the r-SNOB22 and RE-BURP23 profiles. Because of a signal increase of ca. 35%, we used RE-BURP instead of r-SNOB, confirming the finding of Schanda et al. for proteins.6 The acquisition parameters were as follows: R ) 120°, ∆ (1/2JHX) ) 5.4 ms, δ ) 1.8 ms, t1max ) 20 ms, t2max ) 40 ms, and trel ) 1 ms. Forty complex data points were acquired in the t1 dimension, adding 4 dummy scans (n ) 80 + 4). The bandselective 1H excitation (PC9, 3.0 ms) and refocusing (RE-BURP, 2.03 ms) pulses were centered at 8.0 ppm covering 4.0 ppm. 15 N was decoupled with GARP-4,24 with a 90° pulse length of 600 µs. 15N chemical shifts are relative to external 15NH4NO3 (5 M in 2 M HNO3). 15 N relaxation measurements were acquired implementing 2D standard proton-detected heteronuclear pulse sequences.25 T1 and T2 spectra were recorded with spectral widths of 6024 Hz sampled over 896 complex points in the ω2 (1H) dimension

and 1700 Hz over 64 complex points in the ω1 (15N) dimension, with 8 scans for each increment in the indirect dimension. The heteronuclear steady-state NOE spectra were acquired with a spectral width of 6024 Hz over 1024 complex points in the ω2 (1H) dimension and 1700 Hz over 96 complex points in the ω1 (15N) dimension. 15N decoupling during acquisition was achieved using a GARP-4 pulse sequence.24 The field strength of the CPMG refocusing train was ∼3.3 kHz and a 1.2 ms delay was used between the refocusing pulses.26,27 The effects of crossrelaxation between 1H-15N dipolar and 15N chemical shift anisotropy were removed applying 1H 180° pulses during relaxation delays.28 The relaxation delay for both T1 and T2 measurements was 1.5 s. T1 values were measured in a series of spectra with relaxation delays of 10, 20, 40, 80, 120, 180, 300, 500, and 800 ms. T2 measurements were taken with relaxation delays of 10, 20, 40, 60, 80, 100, 120, and 150 ms. To allow NOE evolution, 1H-15N steady-state NOE values were measured with two different data sets, one collected with no initial proton saturation and a second with initial proton saturation. The proton saturation period was 3 s. T1 and T2 spectra were processed using a sine bell apodization function shifted by 90° for both dimensions, zero-filling the matrices to 1024 × 128 real points. An automated baseline correction was applied in both dimensions, and a linear prediction of 64 points was applied in the ω1 (15N) dimension. T1 and T2 values were obtained by fitting peak intensities using single exponential decay, with the analysis of the uncertainties carried out in several different ways. The heteronuclear steadystate 15N-1H NOE values were obtained from the ratios of peak intensities in the saturated spectrum to those in the unsaturated spectrum. Errors were estimated by evaluating the standard deviation of the NOE.25 Model-free analyses29 were performed using the Modelfree software.30

Hwang, T.-L.; Shaka, A. J. J. Magn. Reson. 1995, 112, 275–279. Krishnamurthy, V. V. J. Magn. Reson. B 1995, 106, 170–177. j .; Freeman, R. J. Magn. Reson. A 1993, 102, 122–126. Kupcˇe, E j .; Boyd, J.; Campbell, I. D. J. Magn. Reson. B 1995, 106, 300– Kupcˇe, E 303. (23) Geen, H.; Freeman, R. J. Magn. Reson. A 1991, 93, 93–141. (24) Shaka, A. J.; Barker, P. B.; Freeman, R. J. Magn. Reson. 1985, 64, 547– 552. (25) Farrow, N. A.; Muhandiram, R.; Singer, A. U.; Pascal, S. M.; Kay, C. M.; Gish, G.; Shoelson, S. E.; Pawson, T.; Forman-Kay, J. D.; Kay, L. E. Biochemistry 1994, 33, 5984–6003.

(26) Carr, H. Y.; Purcell, E. M. Phys. Rev. 1954, 94, 630–638. (27) Meiboom, S.; Gill, D. Rev. Sci. Instrum. 1958, 29, 688–691. (28) Palmer, A. G., 3rd; Skelton, N. J.; Chazin, W. J.; Wright, P. E.; Rance, M. Mol. Phys. 1992, 75, 699–711. (29) Lipari, G.; Szabo, A. J. Am. Chem. Soc. 1982, 104, 4546–4559. (30) Palmer, A. G., 3rd; Rance, M.; Wright, P. E. J. Am. Chem. Soc. 1991, 113, 4371–4380. (31) Kanaori, K.; Legerton, T. L.; Weiss, R. L.; Roberts, J. D. Biochemistry 1982, 21, 4916–4920. (32) Williams, S. P.; Haggie, P. M.; Brindle, K. M. Biophys. J. 1997, 72, 490– 498.

(19) (20) (21) (22)

RESULTS AND DISCUSSION In the cell, metabolites experience a viscosity of ca. 2 to 3 times that of water31,32 and interact with other components. As such, restriction of the rotational freedom may be predicted.31 However, their low molecular weight is likely to compensate for the viscosity effect, and an increase of the average effective T1 of in-cell metabolites can be expected. Therefore, a balance of intrinsic and extrinsic properties will affect metabolite relaxation. We first checked if high viscosity is a prerequisite for application of SOFAST-HMQC to low-molecular weight metabolites using a sample of 15N-labeled Ala-Leu dipeptide (5 mM, pH 6.9, 300 K) in the presence of SDS, with a viscosity of 9 relative to water (0.894 cP). Figure 1A reports the amide 1H-15N correlation peak of Leu, centered at 8.13 and 106.9 ppm, while the Ala amine protons

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2407

Figure 1. (A) 1H-15N SOFAST-HMQC spectrum of 15N-labeled AlaLeu dipeptide (5 mM, pH 6.9, 300 K) in the presence of SDS, with an acquisition time of 14 s. The peak, centered at 8.13 and 106.9 ppm, corresponds to Leu amide proton, while the Ala amine protons are observed at 7.65 ppm (1H) and 19.7 ppm (15N). The ∆ (1/2JHX) value was set to 5.4 ms since JHX ) 92.7 Hz; for the remaining acquisition parameters see the Experimental Section section. (B) Dependence of the cross-peak volume on the viscosity of the medium, relative to water.

are not visible in the depicted region but are at 7.65 ppm (1H) and 19.7 ppm (15N). The influence of the viscosity on the volume of the observed cross-peak was investigated by lowering the SDS concentration and, therefore, the relative viscosity from 9 to 1 (no SDS). In the 9-3 range, we observed that the crosspeak volume remained constant, to significantly decrease upon a reduction of the relative viscosity from 3 to 1 (Figure 1B). We estimated that in the absence of SDS (relative viscosity of 1) the cross-peak volume halves. Therefore, for a molecule as small as Ala-Leu, a viscosity of ca. 3 times that of water, which corresponds to the viscosity inside a living cell,31,32 maximizes the intensity of the 1H-15N SOFAST-HMQC peak. However, the efficient 1H-15N dipolar interaction is also important, since a well-defined cross peak, although with an intensity ca. 50% of the corresponding peak in SDS, is observed in the experiment without SDS (not shown). We also addressed the question whether the SOFAST-HMQC experiment works for single amino acids and, hence, in the presence of aminic solvent-exchangeable protons. Using a sample of 15N-labeled Leu (5 mM, 300 K) at four different pHs (1.4, 4.5, 7.1 and 9.2), with a ∆ (1/2JHX) value of 6.7 ms (JHX ) 74.6 Hz), we were able to clearly observe an intense 1H-15N correlation peak at pH 1.4 and 9.2, which decreased at pH 4.5 (65% intensity) and pH 7.1 (30% intensity). In the presence of SDS, with a relative viscosity of 3, we observed a substantial constant intensity, with ca. 10% reduction at pH 7.1. Such a behavior reproduces that reported in Figure 1B for the dipeptide Ala-Leu. The performance of SOFAST-HMQC was compared with other existing 1H-15N correlation experiments. We derived 1D 1 H-15N correlation spectra, with variable recycle delays, by setting t1 ) 0 in different 2D experiments of 15N-labeled AlaLeu dipeptide in the presence of SDS, with a viscosity of 9 relative to water. The S/N ratio vs the repetition time Tscan (comprising the duration of the pulse sequence, the acquisition time and the recycle delay) is reported in Figure 2 for the fast HMQC9 with flip angle R ) 120° (diamonds), fast HSQC33 (triangles), sensitivity-enhanced (se) water-flipback (wfb) HSQC34 (squares), and the SOFAST-HMQC with flip angle R ) 120° 2408

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

Figure 2. Sensitivity comparison of different 1H-15N correlation experiments: fast HMQC with flip angle 120° (diamonds); fast HSQC (triangles); sensitivity-enhanced water-flip-back HSQC (squares), and SOFAST-HMQC with flip angle 120° (circles). The S/N ratio (in arbitrary units), obtained for constant total experimental time in 1D (obtained by setting t1 ) 0 in all experiments) 1H-15N correlation spectra of 15N-labeled Ala-Leu dipeptide (5 mM, pH 6.9, 300 K) in the presence of SDS with a viscosity of 9 relative to water, is plotted vs the repetition time Tscan.

(circles) experiments. As suggested, the se-wfb-HSQC and fast HMQC data were scaled by a factor of 2 to account for the intrinsic sensitivity gain obtained by phase-modulated quadrature detection and disregarded the differences in line width along the 15 N dimension between the various experiments, as they will only slightly affect the sensitivity comparison for small t1max values.5 For Tscan ca. 100 ms (vertical gray line in Figure 2), SOFAST-HMQC yields a sensitivity increase >300% with respect to standard techniques (se-wfb-HSQC, fast HMQC or fast HSQC), even for low-molecular weight metabolites: such a behavior compares well with that reported for proteins.5 Taken together, the above results indicate that the SOFASTHMQC experiment works also with low-molecular weight 15Nlabeled metabolites, with a higher S/N ratio with respect to other heteronuclear experiments. A question arises: Does the pulse sequence work for 15Nlabeled metabolites inside living cells? This was investigated on 15N labeled Thalassiosira rotula diatom cells suspended in artificial seawater as culture medium. Due to intracellular viscosity, a molecule in a cellular environment displays broad NMR line widths as a consequence of the reduced tumbling rate, and overlapped, poor quality spectra are the likely result. In our case, a further complication comes from the presence of high salt concentration in the seawater culture medium, used to suspend the cells in the NMR tube. The final result is that the 1D spectrum obtained for a 15N-labeled T. rotula sample containing ca. 50 million cells will show an unresolved “bumpy” distribution of the resonances (Figure 3, top trace). To account for the broad spectral appearance of metabolites inside T. rotula cells, we measured the 15N relaxation parameters (33) Mori, S.; Abeygunawardana, C.; Johnson, M. O.; van Zijl, P. C. M. J. Magn. Reson. B 1995, 108, 94–98. (34) Andersson, P.; Gsell, B.; Wipf, B.; Senn, H.; Otting, O. J. Biomol. NMR 1998, 11, 279–288.

Figure 3. 1H-15N correlation spectrum (central part) of a sample of 50 million 15N-labeled diatom cells (in seawater culture medium, 35‰ standard salinity, 300 K) recorded in 12 s. 1D traces correspond to the proton spectrum (top) and (left) to a column extracted along the 15 N dimension at the 1H frequency indicated by the dashed vertical line in the 2D spectrum.

T1 and T2 and the steady-state NOE using 2D standard protondetected heteronuclear NMR experiments (not shown).25 For T1 and T2, we obtained values in the ranges 480-680 ms and 110-125 ms, respectively, and derived, using the model-free analysis,29,30 a τc in the range 8.5-14.2 ns. The found relaxation times values recall those usually measured for small proteins in vitro, and this might reflect the increased intracellular (micro)viscosity as well as interactions with other components and/or to the presence of small amounts of paramagnetic ions in the cytoplasm.32 These results indicate that the cellular environment may be suitable for the application of fast 2D NMR spectroscopy to the observation of in-cell metabolites. The 1H-15N SOFAST-HMQC correlation spectrum of 50 million T. rotula cells is reported in Figure 3: it was directly acquired in the culture medium in an overall experimental time of 12 s. In such a short acquisition time, the NMR experiment certainly does not kill the cells, and in fact, the number of colonyforming units/OD is the same before and after the 12-s SOFASTHMQC experiment (data not shown). Furthermore, the high resolution observed in the 2D experiment can be appreciated from the left trace in Figure 3, extracted along the 15N dimension (vertical broken line in Figure 3). The robustness of in-cell SOFAST NMR spectroscopy was investigated by controlling several points.35 First, because of the high S/N ratio, we reduced the number of cells from 50 million down to 10 million, which, as shown in all the experiments below, appears to be sufficient for fast acquisition and high S/N spectra. Figure 4A reports the 1H-15N SOFAST-HMQC spectrum of a 10 million cell sample of 15N-labeled T. rotula, taken directly in the culture medium. It reproduces the spectral pattern of the more concentrated sample of Figure 3 and shows a high S/N ratio with well resolved resonances. Second, when dealing with living cells, it is important to consider that molecules outside the cell tumble faster and, therefore, exhibit sharper lines than internal metabolites in a more viscous environment. Consequently, a small fraction of extracellular molecules could contribute disproportionately to, or even dominate, the spectrum. This was (35) Serber, Z.; Keatinge-Clay, A. T.; Ledwidge, R.; Kelly, A. E.; Miller, S. M.; Do ¨tsch, V. J. Am. Chem. Soc. 2001, 123, 2446–2447.

Figure 4. 1H-15N SOFAST-HMQC spectrum of 15N-labeled T. rotula in varying conditions: (A) in vivo spectrum of 10 million cells directly in the culture medium, 35 ‰ standard salinity, acquired in 12 s; (B) supernatant of the sample used in (A) after removal of all cells by centrifugation and filtration (vertical scale × 8); (C) pellet after resuspension in fresh culture medium; (D) polar extract obtained with the methanol/chloroform protocol (see text). Peaks are labeled with the single-letter code for amino acids; the asterisk marks a yet unidentified peak.

investigated after removal of the cells from the sample by centrifugation and filtration and analyzing the supernatant. It contained no detectable extracellular metabolites as its corresponding SOFAST-HMQC spectrum (Figure 4B, vertical scale × 8) showed no signals, therefore ruling out in Figure 4A any interference from extracellular metabolites. This was confirmed by the following step. The pellet separated from the supernatant was resuspended in fresh standard culture medium giving a spectrum (Figure 4C) identical to that observed when in vivo (spectrum Figure 4A). It is concluded that the cross peaks we observed in the SOFAST-HMQC experiments of Figures 3 and 4A stem from molecules within the cell and that the amount of the released molecules, if present, are beyond detection. When investigating intracellular 15N-labeled metabolites in vivo by NMR, care must be taken to avoid detection of resonances originating from proteins within the cell, which might become labeled because of the unspecific labeling process. This was examined by analyzing the polar extracts of the diatom cells using the methanol/chloroform protocol. The used procedure separates the polar metabolites in the water/ methanol phase at the top, a phase of denatured proteins and cellular debris in the middle, and a chloroform phase at the bottom, with lipophilic compounds.16 As a proof to rule out the presence of signals originating from polypeptides/proteins in the above SOFAST-HMQC spectra, we carried out SDSPAGE gels of the polar extracts obtained from 10 and 50 million cells. Figure 5 reports a 12% acrylamide gel (5A) and a 15% acrylamide gel (5B). In both, the absence of bands in lanes 1 and 2 (reporting a 10 million cell extract ran in duplicate) and lanes 3 and 4 (50 million cell extract ran in duplicate) confirmed the total absence of polypeptides/proteins down to a molecular weight of 3 kDa. For lower molecular weight, we resorted to size-exclusion chromatography under the experimental conditions used for NMR analysis. At pH 6.9, all the molecules present in the polar extract Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2409

Figure 5. SDS polyacrylamide gel electrophoresis of 15N-labeled T. rotula polar extracts: (A) 12% acrylamide, and (B) 15% acrylamide. In both, lane S reports prestained protein standards with molecular weight indicated on the left side; lanes 1 and 2, 10 million cells ran in duplicate; lanes 3 and 4, 50 million cells ran in duplicate. Comassie Brillant Blue staining was used to visualize proteins. (See the Experimental Section for protein standards).

eluted with an apparent molecular mass comparable to that of TSP (172 Da). The experiments described above confirm that the cross peaks we observed are associated with metabolites within the cells and that the presence of polypeptides/proteins in the spectra can be safely excluded. In fact, the SOFAST-HMQC spectrum of the polar extract (Figure 4D) compares well with the in vivo (Figure 4A) and the resuspended pellet (Figure 4C) data, showing only small differences in chemical shift, possibly reflecting differences in salt composition of the in vitro NMR buffer and the cytoplasm. Identification of the cross peaks was achieved upon a careful titration of the extracts with standard amino acids. However, because of the variability of the 1H and 15N chemical shifts, the assignment was confirmed by an HSQC-TOCSY experiment, which unambiguously linked the 1H-15N fragment to the amino acid chain. The cross peaks, labeled with the one-letter code in the extract spectrum (Figure 4D) and in the in vivo spectrum of 10 million cells (Figure 4A), show a striking correspondence. It is important to notice that the spectral position of free amino acids corresponds to that found within the cell; a similar behavior is observed for proteins inside and outside the cell.35 However, as for proteins, the great advantage of the observation of in-cell metabolites by fast NMR spectroscopy does not lie in the structural investigation but on the possibility to examine the behavior of metabolites directly in the cellular compartments and follow their fate upon a change of the physiological state of the cell as well as in the possible interaction with unlabeled/labeled proteins. A crucial point for the application of SOFAST-HMQC analysis of different cell types relies upon the minimal metabolite concentration that can be detected. By integrating the extract spectra and referring to the standard TSP of known concentration (100 µM), the analysis of ten spectra estimated for Phe an average concentration of 0.29 ± 0.09 µM per 106 cells, which appears to be a respectable low limit for NMR-based metabolomics. As an application, we followed the effects of UV-B radiation on T. rotula grown in different salinities. It has been reported that the UV-B radiation, which is known to damage biological systems, results in a stress situation that alters the total amino 2410

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

Figure 6. Comparison of the 1H-15N SOFAST-HMQC spectra of 15 N-labeled T. rotula grown under 35‰ standard salinity as control (A) and exposed to UV-B irradiation (B). Acquisition time was 12 s; irradiation was 717 J m-2 d-1 (weighted) for 5 h per day over period of 2 d. Peaks are labeled with the single-letter code for amino acids.

acid content of UV-B exposed diatoms.15 Figure 6 depicts the SOFAST-HMQC spectra of T. rotula grown at 35‰ standard salinity (spectrum Figure 6A) and that of diatom cells grown at the same salinity and exposed to (weighted) 717 J m-2 d-1 (spectrum Figure 6B). A comparison of the spectra of cells grown at 20 and 45‰ salinity with those obtained at standard salinity (35‰) indicated no significant variation of the cross peaks’ volume and the spectral patterns (not shown), confirming that salinity variation does not noticeably alter the physiological state the cells. Therefore, the different salinity appears to be a relatively minimal stress for the investigated diatoms, confirming the reported data.15 On the contrary, the effects of UV-B radiation are clearly visible in spectrum Figure 6B, since, compared to Figure 6A, we observed a reduction of the Asp, Asn, and Gln peaks’ volume and an increase of Thr, Tyr, Val, Ile, Leu, and Lys peaks’ volume, while Arg and Ala remained unchanged. These results show that the SOFAST-HMQC experiment can fruitfully be used to follow alteration of diatom cells metabolism for a picture of the effects of UV-B radiation as well as of other stressing factors. Our preliminary data confirm those reported,15 but the spectra were obtained without any chemical manipulation of the samples, therefore eliminating all drawbacks due to long analytical procedures. CONCLUSIONS We have reported that 2D 1H-15N correlation spectra of 15Nlabeled metabolites can be directly recorded in living cells using the SOFAST-HMQC pulse sequence. To the best of our knowledge, this is the first time that high-quality 2D correlation spectra of metabolites have been directly recorded in living cells within 10-15 s of experimental time and high S/N. We are aware that snapshot analysis of metabolism in live cells requires not only a short acquisition time but also fast manipulation of the cells in preparing the NMR sample and fast optimization of the spectrometer setup. The preparation of any cell line for NMR studies follows standard procedures,

often lasting not less than 10-15 min. To shorten this time period, we always prepared two sequential samples: the first, to set up the spectrometer, took a total time (centrifuge and suspend the cell pellet, loading the tube, tune, shim, etc.) of ca. 10 min, and a second one for acquisition, therefore without the need of spectrometer setup, which required a total time (centrifuge and suspend the cell pellet, loading the tube, and acquire) of ca. 3 to 4 min. Certainly, particularly fast metabolic turnovers may be altered even in 3 to 4 min due to partial oxygen deprivation, nutrient depletion, or waste buildup. Nevertheless, considering that our spectra were acquired in the culture medium, we are confident that the total time was short enough to prevent a drastic effect on the overall metabolic response. We have also shown that the spectra stem from internal metabolites only, without any contribution from extracellular molecules and/or labeled proteins eventually obtained from the unspecific labeling process used to culture 15N-labeled cells. Identification of amino acid cross peaks in the cellular extracts of T. rotula showed a striking correspondence with the in vivo spectral positions. However, the observation of in-cell metabolites by fast NMR spectroscopy does not involve the structural investigation but the possibility to observe the fate of metabolites upon a change of the physiological state of the cell as well as their molecular interactions directly in the cellular compartments. As shown for UV-B irradiated T. rotula cells, the reported approach allows real-time NMR studies of metabolic processes

in living cells. Examination of the effects of UV-B radiation on T. rotula by SOFAST-HMQC experiments has immediately detected a variation of the amino acid content, monitoring, without any chemical manipulation of the samples, the alteration of diatom cells metabolism upon UV-B irradiation. Obviously, it is desirable to extend the investigation to eukaryotic cell systems, and potential applications include in-cell investigation under physiological or stressing conditions, highthroughput characterization of cell lines by NMR, and induction of metabolic changes by potential drugs, as well as investigation of the primary nitrogen metabolism in plant cells. In general, application in the fields of metabolomics and metabonomics can be predicted, and many of the above applications are in progress in our laboratory. ACKNOWLEDGMENT We thank Rainer Ku¨mmerle (Bruker Biospin AG, Fa¨llanden, Switzerland) for critical reading of the manuscript, Giuliana d’Ippolito (ICB-CNR, Pozzuoli, Italy) for careful preparation of the 15 N-labeled sample of diatom cells, Giuseppina Andreotti (ICBCNR, Pozzuoli, Italy) for the SDS-PAGE gels of the diatom cells extracts, and Emilio P. Castelluccio (ICB-CNR, Pozzuoli, Italy) for the computers’ maintenance. Received for review November 25, 2009. Accepted February 4, 2010. AC9026934

Analytical Chemistry, Vol. 82, No. 6, March 15, 2010

2411