|Home | About | Journals | Submit | Contact Us | Français|
13C-detected solid-state NMR experiments have substantially higher sensitivity than the corresponding 15N-detected experiments on stationary, aligned samples of isotopically labeled proteins. Several methods for tailoring the isotopic labeling are described that result in spatially isolated 13C sites so that dipole-dipole couplings among the 13C are minimized, thus eliminating the need for homonuclear 13C-13C decoupling in either indirect or direct dimensions of one- or multi-dimensional NMR experiments that employ 13C detection. The optimal percentage for random fractional 13C labeling is between 25% and 35%. Specifically labeled glycerol and glucose can be used at the carbon sources to tailor the isotopic labeling, and the choice depends on the resonances of interest for a particular study. For investigations of the protein backbone, growth of the bacteria on 2-13C-glucose containing media was found to be most effective.
The majority of aligned-sample solid-state NMR studies on proteins immobilized in supramolecular complexes, such as virus particles or membranes, have relied on 1H/15N double-resonance experiments (1). There are several advantages to this approach, including the relative ease and low cost of labeling all nitrogen sites in proteins obtained by expression in bacteria (2). Solid-state NMR experiments on uniformly 15N labeled proteins are straightforward because they have no nitrogen atoms directly bonded to other nitrogen atoms in either backbone or side chain sites. As a result, there is no need to implement homonuclear 15N decoupling at any stage in the pulse sequences, including during the direct acquisition of 15N signals, and the requisite heteronuclear decoupling is accomplished by irradiation of the 1H resonances. It is feasible to make accurate measurements of 1H chemical shift, 15N chemical shift, and 1H-15N heteronuclear dipolar coupling frequencies for individual sites, as well as to detect 1H-1H and weak 15N-15N homonuclear couplings using a wide variety of multidimensional NMR experiments.
However, there are two disadvantages to the 1H/15N double-resonance approach: it is restricted to the amide nitrogen sites in the polypeptide backbone and the few nitrogen-containing side chain sites, and the direct detection of 15N signals has low sensitivity because of its low gyromagnetic ratio. Both of these issues can be addressed by implementing 1H/13C/15N triple-resonance experiments on proteins labeled with both 13C and 15N. The sensitivity can be improved by detecting 13C signals because its gyromagnetic ratio is about 2.5 times larger than that of 15N, and there is the opportunity to obtain spectroscopic data from nearly all backbone and side chain sites. The design of triple-resonance experiments for stationary samples is considerably different than that for magic angle spinning (MAS) experiments. In this article we describe progress towards the development of isotopic labeling schemes compatible with direct detection of 13C signals in 1H/13C/15N triple-resonance experiments on stationary aligned samples (3 – 10) by examining a wide range of 13C labeling approaches (11) over a range of fractions of dilutions, ranging from 15% to 100% of all sites in the proteins, and 2-13C-glucose in addition to the complementarily labeled glycerols for labeling through metabolic pathways. Magnetically aligned filamentous bacteriophage particles are used as the samples to demonstrate the influence of the labeling patterns on solid-state NMR spectra.
In single-contact spin-lock cross-polarization experiments on single crystals of peptides and aligned samples of proteins, we have observed a four-fold improvement in the signal to noise ratio when 13C magnetization is detected compared to 15N magnetization for individual labeled sites in the same samples under equivalent experimental conditions (7). As is the case for 15N labeling of proteins expressed in bacteria, 100% uniform labeling of all carbons sites with 13C is straightforward when completely labeled carbon sources, such as 13C6 glucose or 13C3 glycerol, are used in the growth media. However, in protein samples where all of the carbon sites are labeled with 13C, the homonuclear dipole-dipole couplings among the dense network of bonded and nearby 13C nuclei present significant complications in solid-state NMR of stationary samples. Indeed, one of the principal advantages of high-speed magic angle spinning solid-state NMR experiments is that the homonuclear dipole-dipole couplings among the 13C sites are greatly attenuated. In contrast, in stationary samples, the 13C homonuclear couplings must be dealt with in order to obtain high-resolution spectra and to realize the sensitivity advantages of 13C-detection.
There are two approaches to dealing with the homonuclear 13C-13C dipole-dipole couplings in stationary aligned samples. The first is to apply multiple-pulse homonuclear decoupling sequences to the 13C nuclei (12). We have demonstrated the efficacy of this approach in the indirect dimensions of multidimensional experiments that enable 13C chemical shift frequencies to be measured; however, these experiments were performed with 15N-detection in the direct dimension (4). The detection of signals in the windows of multiple-pulse sequences rarely leads to optimal sensitivity because of the filtering limitations associated with the short periods of time available for sampling the signals. The second approach is to tailor the pattern of isotopic labeling so that the 13C labeled sites of interest are sufficiently isolated from other 13C nuclei to eliminate the need for homonuclear decoupling.
The isotopic labeling schemes described in this article generate samples with diluted, spatially isolated 13C sites so that dipole-dipole couplings among the 13C are minimized, thus eliminating the need for homonuclear 13C-13C decoupling in either indirect or direct dimensions of one- or multi-dimensional NMR experiments that employ 13C detection. The isotopic precursors utilized for tailored labeling of proteins expressed in bacteria range from specifically labeled two-, three-, or six-carbon molecules to random fractionally labeled growth media prepared from algae grown in the presence of a defined mixtures of 12C and 13C carbon dioxide. The coupling among 13C nuclei is avoided either as a result of the alternate site pattern of metabolic incorporation (13 – 15) or because of the low statistical probability of two 13C nuclei being bonded to each other. Here, results from uniformly 13C labeling and metabolically tailored 13Cα labeling based on [2-13C]-glucose, [2-13C]-glycerol, or [1, 3-13C]-glycerol are compared to fractional 13C labeling. Two-dimensional projections of triple-resonance solution-state NMR spectra are used to characterize the labeling patterns. The filamentous bacteriophages fd, which infects Escherichia coli, and Pf1, which infects Pseudomonas aeruginosa, are used for the experimental demonstrations. The structural forms of the major coat proteins are immobilized and aligned along with the virus particles by the magnetic field for solid-state NMR experiments, and the membrane-bound forms of the same proteins are solubilized in detergent micelles for solution NMR experiments. The bacteriophage samples provide direct insights into the spectroscopic effects of the 13C labeling schemes at different protein backbone sites in solid-state NMR experiments., They have been used in complementary magic angle spinning studies (16, 17).
Filamentous bacteriophage samples were obtained from bacterial cultures grown on algal-based media containing various fractions of 13C labeled nutrients, or on minimal media supplemented with [2-13C]-glucose, [2-13C]-glycerol, or [1,3-13C]-glycerol. In this article, our main focus is on the 1H-13Cα sites that are uniformly distributed throughout the protein backbone, separated from each other by three bonds, and have one directly bonded hydrogen (except for glycine and proline). At the intersection of peptide planes, the associated 1H chemical shift, 13C chemical shift, and 1H-13C dipolar couplings provide valuable structural constraints, and enable direct comparisons of sensitivity and resolution between 15N- and 13C- detected experiments.
In the absence of homonuclear decoupling, only proteins labeled such that the 13Cα sites are isotopically isolated from other 13C, i.e. bonded to 12Cβ and 12CO (Figure 1A), will yield solid-state NMR spectra with high resolution and sensitivity in stationary samples. The presence of 13C in either or both adjacent carbon sites (Figure 1B) will result in broadened signals due to the presence of unresolved 13C-13C dipolar couplings. The probability of having an isolated 13Cα site can be calculated from the individual probabilities of having 13Cα, 12Cβ, and 12CO present in the same polypeptide with random 13C labeling at various fractions. As shown in Figure 1D, the maximum probability occurs near a labeling ratio p =⅓. Because the maximum is broad and both natural abundance 13C and the enrichment of 13C in other proximate sites are potential complications, we prepared custom algal media with 13C percentages of 15%, 25%, 35%, and 45%, as marked by arrows in Figure 1D, in order to experimentally determine the optimal isotopic composition for solid-state NMR spectroscopy on stationary aligned samples of proteins.
The isotope labeling patterns of the protein samples were analyzed using solution NMR spectra. In general, the proteins were uniformly labeled to the predicted extent in all carbon sites in both E. coli and P. aeruginosa when grown on the algal-based media. One-dimensional 1H-decoupled, 13C solution NMR spectra report on the 13C labeling at all sites, including the carbonyl and aromatic ring carbons that are not directly bonded to hydrogens. However, only a detailed analysis of individual sites reveals whether labeled sites are isotopically isolated. Since E. coli is the host for fd bacteriophage, its major coat protein represents the labeling that occurs in this bacterium. Solution NMR spectra of the membrane-bound form of fd coat protein in micelles are shown in Figure 2. The comparison of the spectrum obtained on a 25% uniformly labeled sample (Figure 2A) and that from a sample labeled from media containing 2-13C-glucose (Figure 2B) demonstrates that there is a greater extent of 13C labeling in the Cα region (45 ppm – 65 ppm) and reduced labeling in the aliphatic (0 ppm – 50 ppm), aromatic (110 ppm – 170 ppm), and carbonyl (165 ppm – 185 ppm) regions in the 2-13C-glucose labeled sample.
A combination of two-dimensional projections from three-dimensional heteronuclear solution NMR spectra is used to quantify the 13C enrichment at the Cα, Cβ, and CO sites of the proteins. The choice of experiments utilized for this purpose depends on the availability of resonance assignments as well as which sites are of interest. The labeling at Cα and CO sites can be assessed by comparison of peak intensities in 13C-edited 1H/15N correlation spectra of the various 13C labeled samples at equal concentrations, and this requires only backbone amide proton assignments (Figures 2C and and3).3). If 1H, 13C assignments are available, 1H/13C correlation spectra, e.g. two-dimensional 1H/13C HMQC or 1H/13C projections of three-dimensional HNCA, HNCB, and HNCO spectra can be used to elucidate the 13C labeling of the protein. The 13C labeling results for E.coli grown on [2-13C] glucose containing media are summarized in Figure 4 for the Cα and CO sites of fd coat protein. 13C is enriched in the Cα sites of all residues to level that is > 18%, except for leucine, and some approach 70%. The data summarized in Figure 4B show that the CO sites of isoleucine, leucine, proline, threonine, and tyrosine have enrichment levels > 18%, but that most amino acids undergo minimal labeling of their CO sites.
The coat protein of Pf1 bacteriophage reflects the metabolic pathways of its host organism P. aeruginosa, which are known to differ from those of E. coli (16, 17). The labeling patterns of proteins from P. aeruginosa were assessed using the same solution NMR approach used for E. coli. All of the proteins are 100% uniformly labeled with 15N in all nitrogen sites, and this provides a spectroscopic reference for the 13C resonance intensities. Column A in Figure 5 contains the 15N-edited 1H NMR spectra of all of the samples of the membrane-bound form of Pf1 coat protein in micelles that are directly comparable to the 13C NMR spectra in Column B. The corresponding one-dimensional solid-state 13C NMR spectra of the structural form of Pf1 coat protein in magnetically aligned bacteriophage particles are shown in Column C; these spectra are analyzed in three regions, with the broad band of resonances near 200 ppm from the carbonyl carbons, the intensity between 40 ppm and 80 ppm from the Cα sites, and the intensity between about 10 ppm and 40 ppm from other aliphatic carbons especially methyl groups. Although highly overlapped, the relative intensities of the signals provide a guide to the sensitivity in 13C-detected solid-state NMR experiments. Notably, the lowest signal to noise ratio is observed for the sample with the highest degree of 13C labeling (Figure 5D) because of the broadening effects of the homonuclear 13C dipolar couplings.
The one-dimensional solution 13C NMR spectra of the fractionally 13C labeled samples have peak intensities that are homogeneously scaled according to the 13C dilution ratio for all carbon regions compared to 100% uniform 13C labeling (Column B in Figure 5D- F ). Many signals of carbonyl, aromatic, and aliphatic carbons are resolved is the [45 %-13C] labeled protein, however, at 25% labeling the majority of signals are close to the level of noise under comparable experimental conditions, showing the direct effect of isotopic dilution when homonuclear dipolar couplings are not a factor. The 13C NMR spectra in Column B of Figure 5 indicate the differences in the labeling from media containing [2-13C]-glucose, [2-13C]-glycerol, and [1,3-13C]-glycerol. Compared to uniform labeling, the [2-13C]-glycerol sample has reduced intensity in the aliphatic, aromatic, and carbonyl regions of the spectrum, whereas the [1,3-13C]-glycerol sample shows a complementary labeling pattern with diminished intensity in the alpha carbon resonance region. The overall level of isotopic labeling in the [2-13C]-glycerol and [2-13C]-glucose samples is similar, however, the spectral patterns are different, for example the strong aromatic resonances in the spectrum of [2-13C]-glycerol are missing in the spectrum of the [2-13C]-glucose sample.
The extent of labeling of the Cα, Cβ, and CO sites differ only slightly from the expected average value for the various amino acids as indicated by the data in Figure 6A. In contrast to the even distributions of the random fractional labeling, the samples labeled by incorporation of [2-13C]-glucose have quite different levels of 13C enrichment in different sites (Figures 6B and 6C).
In the case of [2-13C]-glucose labeling in P.aeruginosa, at least 14 amino acids have isolated 13Cα labels. The enrichment ratio between the [2-13C]-glucose and a uniformly 13C labeled sample at the Cα position of glycine, leucine, glutamine, arginine, serine and tyrosine is less than 20 %. Glycine, serine and tyrosine are highly enriched in E. coli , however, due to the Entner-Doudoroff metabolic pathway in P.aeruginosa these amino acids are omitted in the 13C labeling . With the exception of glycine, leucine, glutamine, arginine, serine, and tyrosine, in total 14 of 20 amino acids in P.aeruginosa, and in E.coli all amino acids except from leucine provide isolated 13Cα labels (Figure 7) For [2-13C]-glycerol as labeling precursor in P.aeruginosa only leucine, glutamine, and arginine are not enriched above 20 % at the Cα site. Isoleucine and valine have significant 13C enrichment at the adjacent Cβ, and will be affected by 13C-13C dipolar coupling. A set of 15 of the 20 amino acids in proteins obtained from P.aeruginosa grown on [2-13C]-glycerol containing media is found to be suitable for 13C solid-state NMR experiments (Figure 4C).
The effects of tailoring the 13C labeling are readily observed in 13C NMR spectra. As noted above, because of the strong influence of the 13C-13C homonuclear dipole-dipole couplings there is not a simple relationship between the extent or type of 13C labeling and the resolution and sensitivity. The spectra of the random fractional 13C labeled samples have higher signal-to-noise ratios than that of a comparable uniformly 100% 13C labeled sample. The metabolic Cα labeling schemes based on [2-13C]-glucose and [2-13C]-glycerol show strong signals in the one-dimensional solid-state NMR spectra, compared to the broad, poorly resolved spectra from samples with uniform 100% 13C labeling. In contrast to the random fractionally labeled samples, the samples labeled with [2-13C]-glucose and [2-13C]-glycerol have relatively high overall labeling of the alpha carbons and significantly lower levels of labeling of carbonyl and aliphatic side chain carbons. The [1,3-13C]-glycerol sample has more extensive labeling of carbonyl and side chain methyl carbons, and the spectra have better resolution in these regions.
The effects of tailoring 13C labeling can also be observed in 15N NMR spectra. Although the effects are more subtle than those in the directly detected 13C NMR spectra, they are apparent in comparisons of 15N NMR spectra obtained with and without heteronuclear 13C decoupling. This is illustrated with the spectra in Figure 8. The spectra in the left column are all very similar; since they were obtained with 1H and 13C decoupling, the effects of the 13C-15N dipolar couplings are not seen. In contrast, the spectra in the right column were obtained on the same samples with only 1H decoupling. The broadening effects of nearby 13C labeled sites on the 15N amide backbone resonances can be observed to vary among the labeling schemes.
Two-dimensional 1H-13C PISEMA spectra of aligned bacteriophage samples prepared with the various 13C labeling schemes are compared in Figure 9. In the 100% uniformly 13C labeled sample, the strong network of 13C-13C homonuclear couplings at all sites interferes with the experiment and, with the exception of the methyl carbon region between 10 ppm and 40 ppm, there is essentially no intensity observable in the displayed spectral region, which encompasses all of the aliphatic carbon sites in the protein. In contrast, all of the samples with tailored 13C labeling yield resolved spectra. The spectra from Pf1 coat protein obtained from bacteria grown on media containing [2-13C]-glycerol or [1,3-13C]-glycerol are complementary, due to the specific metabolic labeling pattern of the glycerol precursors. There is notably more intensity from aliphatic side chain carbons in the spectrum obtained from the [1,3-13C]-glycerol due to its labeling pattern. Although the total amount of 13C in the protein is reduced, there is a net gain in signal-to-noise ratio in all of the one-dimensional spectra. Although less than half of the carbon sites are labeled in the case of [45 %-13C] phage, the observed signal-to-noise ratio is about twice that observed for the 100% labeled phage. As expected, the signal-to-noise ratio decreases with decreasing 13C content, however, even [15 %-13C] phage gives spectra with better signal-to-noise ratios than those from a 100% uniformly 13C labeled sample. The gain in signal-to-noise ratio is greater in the case of proteins obtained from media containing specifically labeled glucose or glycerol.
The analysis of the two-dimensional solid-state NMR spectra reveals different line shape behavior for the 13C chemical shift and the 1H-13C dipolar coupling dimensions. For all of the samples the full width at half height is approximately 250 Hz to 300 Hz in the 13C chemical shift dimension, except with [45 %-13C] labeling where the samples have somewhat broader lines of 400 Hz. For the 1H-13C dipolar coupling dimension, the different dilutions of the 13C affect the line width of the two-dimensional spectra. For the spectra of the random fractional labeled samples the full width at half height decreases with increasing 13C dilution from 1300 Hz for [45 %-13C] phage to 700 Hz for [15 %-13C] phage. For the metabolically tailored 13C labeling schemes the full width at half height of the two dimensional spectra is between 1350 Hz and 900 Hz, with the [1,3-13C]-glycerol and [2-13C]-glucose labeled samples providing spectra with somewhat better resolution than those labeled with [2-13C]-glycerol.
Although there is a reduction in signal-to-noise with decreasing 12C to 13C isotope ratio. However, this is compensated by improvement in the line shape in the 1H-13C dipolar coupling dimension whereas the 13C chemical shift dimension is hardly affected. For the two different Cα labeling precursors, [2-13C]-glycerol and [2-13C]-glucose, the later shows a better line shape in both dimensions of the two dimensional spectra. The [1,3-13C]-glycerol sample yields slightly better line shapes for the Cα sites.
Uniform isotopic labeling of proteins has been an integral part of the experimental design from the beginning of the field (18, 19), and 100% uniform labeling with 13C and 15N is widely used in triple-resonance solution NMR experiments as well as MAS solid-state NMR experiments. Because of the natural separation of nitrogen sites in proteins, there is no need to vary the extent of 15N labeling because of through-space of through-bond couplings. However, all carbons are bonded to at least one other carbon and in most cases two other carbons, resulting in dense networks of homonuclear scalar and/or dipole-dipole couplings that can interfere with both solution NMR and solid-state NMR experiments. The two basic strategies for 13C labeling that evaluated in the context of aligned-sample solid-state NMR experiments with the data in Figures 8 and and99 were previously used for solution NMR and magic angle spinning solid-state NMR. In solution NMR, the preparation of protein samples that were uniformly fractionally 13C labeled enabled J couplings to be used to assist in making resonance assignments (20) and to simplify relaxation pathways (21). Biosynthetic incorporation of specifically 13C labeled metabolic precursors, including glycerol (22), glucose (23), and pyruvate (24), were also used to facilitate 13C relaxation studies and triple-resonance experiments solution NMR. Increased resolution in magic angle spinning solid-state NMR experiments has resulted primarily from the use of samples with complementary labeling patterns obtained from growth on media containing [1-13C]-glycerol and [2,3-13C]-glycerol (13 – 15).
The tailored labeling from the labeled glycerol improves the resolution that is possible with magic angle spinning of 100% uniformly 13C labeled samples. In contrast, in aligned-sample solid-state NMR, in order to obtain high-resolution spectra and realize the sensitivity gains feasible with 13C-detection, the homonuclear dipole-dipole couplings have to be dealt with either by multiple-pulse homonuclear decoupling or isotopic labeling. As described here, an approach applicable to proteins obtained by expression in bacteria is to tailor the 13C labeling to label sites of interest, for example the alpha carbons, while adjacent carbon sites are unlabeled. In this continuation of our development of this approach, several complementary labeling schemes were tested for two different bacteria. The 13C dilution from random fractional labeling media had predictable outcomes for both tested expression systems. However, the implementation of metabolic precursors for labeling depends on the organism’s specific metabolic pathways, and requires experimental verification. The choice of labeling strategy is an integral part of the experimental design, which contrasts with the situation for most solution NMR and magic angle spinning solid-state NMR studies that can be accomplished with only one or a few labeled sample prepared according to the previously demonstrated labeling schemes.
For aligned-sample solid-state NMR of proteins, the optimal percentage for uniform fractional 13C labeling is between 25% and 35%. Specifically labeled glycerol and glucose can be used at the carbon sources to tailor the isotopic labeling, and the choice depends on the resonances of interest for a particular study. For investigations of the protein backbone, 2-13C-glucose was found to be most effective. In the case of P. aeruginosa there are 14 of the 20 amino acids with isolated Cα position in the protein backbone and for E. coli. 19 amino acids have 13Cα sites.
The 100% Uniformly 13C and 15N labelled samples were obtained in the conventional way by using a minimal salts media with 15N labelled ammonium sulfate and 13C6 labeled glucose as the sole nitrogen and carbon sources. Uniform fractional 13C labeling was performed by growing bacteria on standard BioExpress Cell Growth Media prepared with the stated percentages of 13C. The metabolically 13C labelled samples were obtained by supplementing the using [2-13C]-glycerol, [2-13C]-glycerol, [1,3-13C]-glycerol, or as the sole carbon source. All of the isotopically labelled compounds and media described in this article are from Cambridge Isotope Laboratories (www.isotope.com).
The 50 mg/ml solutions of aligned bacteriophage particles were prepared as described previously (23). The coat proteins subunits were separated from the phage particles and solubilized in SDS for the solution NMR analysis (24).
The solid-state NMR experiments were performed on a Varian Inova spectrometer with 1H, 13C, and 15N frequencies of 500.125 MHz, 125.76 MHz, and 50.68 MHz respectively. A home-built triple resonance probe with a single 5 mm solenoid coil was used and the RF power levels were adjusted to generated 50 kHz RF fields on all three channels. PISEMA was utilized for the separated local field experiments because its cycle time was short enough to accommodate the large frequencies of 1H-13C dipolar couplings. The 1H carrier frequency was set to the resonance of water at 4.7 ppm; the 13C and 15N carrier frequencies were set to 100 ppm on their respective scales. Solid samples of adamantane and ammonium sulfate served as external chemical shift references for 13C and 15N, respectively. The two-dimensional solid-state NMR spectra were acquired with 64 points in t1 and 512 complex points in t2. The experimental data were zero filled in t1 to 2K and in t2 to 4K data points and multiplied by a sine bell window function before Fourier transformation in each dimension. The recycle delay was 6 s.
The solution NMR experiments were performed on a Bruker Avance 800 MHz spectrometer with 1H, 13C, and 15N frequencies of 800.034 MHz, 201.203 MHz, and 81.076 MHz respectively. For these experiments, the coat proteins of the bacteriophages were solubilized in SDS micelles at a concentration of 1 mM. Resonance assignments of Cα, Cβ or CO sites for the solution-state NMR spectra were obtained from three-dimensional HNCA, HNCB, and HNCO experiments on a 100% uniformly 13C, 15N labelled sample. Resonance assignments of Hα for the 1H,13C-HSQC spectra were obtained from a three-dimensional 15N-edited NOESY-HSQC experiment with a mixing time of 120 ms. Cβ and CO resonance measurements NMR spectra were obtained from three-dimensional HNCA, HNCB, and HNCO experiments. 13C chemical shift was referenced indirectly to the 1H chemical shift of DSS.
We thank Christopher Grant, Chin Wu, and Xuemei Huang for helpful discussions and assistance with the instrumentation. This research was supported by grants from the National Institutes of Health, and utilized the Biomedical Technology Resource for NMR Molecular Imaging of Proteins, which is supported by grant P41EB002031. F.V.F. was supported by an EMBO postdoctoral fellowship (ALTF 214-2007).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.