The in-cell NMR technique offers significant insights into the structure and function of heterologous proteins in the physiological intracellular environment at an atomic resolution. Escherichia coli (E. coli) is one of the most widely used host cells for heterologous protein expression in structural biological studies as well as for in-cell NMR studies to investigate fundamental structural characteristics and the physiochemistry of certain proteins and their intermolecular interactions under physiological conditions. However, in many cases, it is not easy to obtain well-resolved in-cell NMR spectra because the detectability and resolution of these spectra are significantly influenced by intracellular factors such as nonspecific intermolecular interactions. In this study, we re-examined the experimental parameters of E. coli in-cell NMR and found that the detectability and resolution of the NMR spectra clearly depended on the growth phase of the host cells. Furthermore, the detectability and resolution of the E. coli in-cell NMR spectra correlated with the soluble fraction amounts of the expressed target protein. These results indicate that the E. coli in-cell NMR spectrum of a target protein is a useful tool for monitoring the intracellular conditions of the host cell and for establishing the appropriate cultivation conditions for protein overexpression.
Solution NMR spectroscopy is nondestructive and is therefore widely used as a tool for the in vivo observation of metabolites and metal ions in living cells, for which it is ideal1. For example, in vivo quantitative, real-time monitoring of the dynamics of cellular sodium ions or ATP/ADP has been established by measuring the NMR spectra of 23Na or 31P nuclei, respectively1. Recently, in vivo NMR techniques have not only been applied to small biomolecules but have also been developed as in-cell NMR to study the structure and dynamics of heterologous proteins overexpressed in host cells.
In living cells, most proteins exert their biological function in an extremely crowded environment due to the high concentration of macromolecules (which can reach 400 g/L)2,3, and this can affect their conformation, stability, dynamics, and function4,5,6,7. The in-cell NMR technique can be used in the physiological intracellular environment to gain significant knowledge about the structural and functional mechanisms of heterologous proteins at an atomic resolution8,9,10,11,12,13,14.
Escherichia coli (E. coli) is one of the most widely used host cells for the overexpression of heterologous proteins, as well as for in-cell protein NMR studies to elucidate fundamental structural and physicochemical characteristics of proteins and their intermolecular interaction modes under physiological conditions8,9,10,11,12,13,14. However, in many cases it is not straightforward to record well-resolved E. coli in-cell NMR spectra for proteins as their detectability and resolution can be negatively influenced by many factors in the living host cells, such as nonspecific intermolecular associations. Conversely, this means that E. coli in-cell NMR spectra for proteins can sensitively reflect the intracellular physiological environment or phenomena involving protein dynamics.
Given that E. coli is the most widely utilized bioreactor for the overexpression of heterologous proteins, it is possible that the E. coli in-cell NMR spectrum of a recombinant protein would reflect the physiological conditions in a living host cell during the overexpression of that protein. In this study, we comprehensively re-examined the experimental parameters of E. coli in-cell NMR and investigated the types of information that would be reflected in the in-cell NMR spectrum of a recombinant protein.
Results and Discussions
The detectability and resolution of E. coli in-cell NMR spectra were significantly influenced by experimental parameters
It is widely accepted that the detectability and resolution of E. coli in-cell NMR spectra are influenced by several experimental parameters, such as the host cell cultivation conditions (e.g., the process or media composition of the subculture)8,15 and observation conditions (e.g., the final cell density in the NMR sample tube)16. In this study, we first re-examined the experimental parameters to elucidate the critical factors that could result in detectable and reproducible E. coli in-cell NMR measurements. Four experimental parameters of E. coli in-cell NMR were selected: the concentration of D2O in an NMR sample for 2H locking of the static magnetic field; the final cell density in the NMR sample tube; the final concentration of isopropyl β-d-1-thiogalactopyranoside (IPTG); and the cell density of the culture media (OD600 value for each stage of the growth curve) at the time of the addition of IPTG.
A previous report by Serber and colleagues indicated that a higher final cell density in an NMR sample tube gave broader, lower-resolution E. coli in-cell NMR spectra with lower sensitivity for the 2H lock signal15,16. The same tendency was observed in the present study, that is, a higher concentration of D2O and a lower cell density in the NMR sample can result in improved resolution of the E. coli in-cell NMR spectrum (Supplemental Fig. S1A,B). The detectability and resolution of an E. coli in-cell NMR spectrum do not depend on the IPTG dose (Supplemental Fig. S1C). The detectability and resolution of E. coli in-cell NMR spectra of GB1 drastically depended on the OD600 value at the time of the addition of IPTG, and this parameter was the most critical to improve in-cell NMR spectra (Supplemental Fig. S1).
To confirm whether the NMR spectra of the target protein definitely originated within the living host cells, the NMR spectrum of E. coli suspension supernatant was measured immediately after the in-cell NMR measurement. No NMR signal was detected (Supplemental Fig. S2); thus, the improvement in the resolution of the E. coli in-cell NMR spectrum was not due to leakage of the target protein into the extracellular solvent in the NMR sample tube. In addition, we confirmed that in-cell NMR signals of target protein was not from the leakage expression of pET expression system (Supplemental Fig. S3).
The E. coli in-cell NMR spectrum for GB1 drastically depended on the stage of the growth curve
Because the OD600 value at the time of the addition of IPTG was a critical factor, in this section, the correlation between this and the detectability and resolution of the associated E. coli in-cell NMR spectrum was investigated in detail. Various E. coli cell suspension samples were obtained at all stages of the growth curve. Overexpression of the heterologous target protein was induced by IPTG at various OD600 values. In-cell NMR spectra for all samples were measured.
E. coli in-cell NMR signals were severely line-broadened when the OD600 value at the time of the addition of IPTG was 0.60, corresponding to the late lag to early log phase (Fig. 1 and Supplemental Fig. S4). The signals became narrower when the OD600 value was around 1.00, corresponding to the early to middle log phase, and the signal intensity decreased when the OD600 value was larger than 1.80, corresponding to the middle log to the stationary phase (Fig. 1). At the middle log to stationary phase, several unidentified strong signals appeared around 8 ppm (1H) and 115–120 ppm (15N) chemical shifts in the associated E. coli in-cell NMR spectra (Fig. 1). Thus, the detectability and resolution of the in-cell NMR spectra depended to a great extent on the stage of the growth curve (Fig. 1).
To confirm that the NMR spectra of the target protein definitely originated within the living host cells, the NMR spectra of E. coli lysate, in which overexpression of the target protein were induced by adding IPTG at various OD600 values in a similar manner as described above, were measured. As a result, signal line-width of all NMR spectra were remarkably improved (Supplemental Fig. S5). Its results support that the in-cell NMR spectra (Fig. 1), which shows host E. coli cell growth phase-dependent alterations of the detectability and resolution, were certainly originated in the living host cells.
Several other proteins, TTHA0227, TTHA0814, TTHA1718, hCaM and GFP, were tested to investigate whether the dependency of the E. coli in-cell NMR spectrum on the stage of the growth phase was not a GB1-specific phenomenon. The GB1, TTHA1718 and hCaM proteins are widely utilized for E. coli in-cell NMR experiments because they provide detectable E. coli in-cell NMR spectra15,16,17. However, there has been no report describing the in-cell NMR measurement of GFP, even though this is widely utilized as an analytical tool to visualize the intracellular localization of target proteins in a cell.
As shown in Fig. 2 and Supplemental Fig. S6, TTHA1718 showed similar tendencies to those of GB1, that is, substantial line-broadening at the late lag to early log phase, narrowing at the early to middle log phase, and a decrease at the middle log to stationary phase. Other TTHAs, hCaM and GFP did not show clear in-cell NMR signals (Supplemental Figs. S7–S11). For most samples, several unidentified strong signals appeared at similar positions at the middle log to stationary phase as with GB1. Indeed, superposition of the 2D 1H-15N HSQC spectra of GB1, hCaM, and GFP showed that the chemical shift values of these unidentified strong signals completely matched across the three spectra (Supplemental Fig. S12)18,19,20.
It was previously reported that bacterial cell lysate containing over-expressed heterologous proteins can be contaminated with water-soluble carbohydrates such as cyclic enterobacterial common antigen (ECACYC)21,22. Interestingly, the previous reports showed that the amount of the carbohydrate contaminants were increased depending on the cell culture conditions, such as length of cell culture period after induction of protein over-expression and composition of the culture media22. Our study demonstrated that the unidentified strong signals appeared depending on the length of cell culture period, more specifically the “before” induction of protein over-expression (Figs. 1, 2, and Supplemental Figs. S7–S10).
Our NMR study showed that the 1H-15N chemical shifts of the unidentified strong signals were extremely identical to that of E. coli ECACYC (Figs. 1, 2, and Supplemental Figs. S7–S12). On the other hand, our further NMR experiments also demonstrated another possibility that the unidentified strong signals may be derived from fragments of cell wall peptidoglycan from the E. coli host cell (Supplemental Fig. S13).
These results indicate that the unidentified strong signals are not derived from the over-expressed heterologous proteins, but are derived from components of the host cells. Moreover, the results support the dependency of E. coli in-cell NMR spectrum on “stage of the growth curve” as a universal phenomenon regardless of the kind of over-expressed heterologous protein. In addition, they suggest that E. coli in-cell NMR spectra can sensitively reflect cell culture conditions during protein over-expression as further discussed below.
The E. coli in-cell NMR spectra depended on the yield of soluble and healthy protein
To identify the origin of the dependency of E. coli in-cell NMR spectra on the stage of the growth curve, we investigated the relationship between the spectra and the amount of soluble fraction of the overexpressed protein. As shown in Fig. 1 and Supplemental Figs. S14–S15, the soluble fraction of the overexpressed protein per medium reached its maximum when the protein overexpression was induced at OD600 = 1.20, and the best E. coli in-cell NMR spectrum (among those of OD600 = 0.6, 1.2, 1.8, and 2.4) was obtained at OD600 = 1.20 (Fig. 1). It is notable that the expression level of the insoluble fraction was slightly decreased when IPTG was added at OD600 = 1.20 (Fig. 1).
These results indicate that most efficient condition to yield soluble and healthy recombinant protein and give the best E. coli in-cell NMR spectra is to induce the overexpression of the target protein during the early to middle log phase. This suggests that the origin of the dependency of E. coli in-cell NMR spectra on the stage of the growth curve of the target protein arises from the yield of soluble and healthy recombinant proteins, making this a good tool for monitoring the best conditions for heterologous protein overexpression.
Two different expression systems, pET23,24,25 and pCold26, were tested to investigate whether E. coli in-cell NMR spectra were dependent on the expression system. The pCold expression system utilized cold-shock responses for heterologous protein overexpression and improved the expression level of soluble and healthy recombinant protein, suggesting that the intracellular physiological environment differed between the pET and pCold systems.
As shown in Fig. 3 and Supplemental Fig. S16, the pCold expression system showed much less dependency of the NMR spectrum on the growth curve than did the pET system (Fig. 1 and Supplemental Fig. S4). In the pCold system, the E. coli in-cell NMR spectra showed relatively higher resolution during the late lag to early log phase (Supplemental Figs. S4 and S16). This result is consistent with the difference in the potential soluble protein overexpression; that is, the pCold system is less sensitive to the induction timing of protein overexpression than the pET system.
The pCold system showed a similar tendency as the pET system, i.e., line-broadening at the late lag to early log phase, narrowing at the early to middle log phase, and a decrease at middle log to stationary phase (Figs. 1 and 3). Several unidentified strong signals also appeared at similar positions during the middle log to stationary phase, just as with the pET system. These results suggest that dependency of the host cell growth phase at the starting point of protein overexpression induction on the detectability and resolution of E. coli in-cell NMR spectra is not a phenomenon that is specific to the expression system.
The E. coli in-cell NMR spectra depend on the induction period
As described earlier, the detectability and resolution of the E. coli in-cell NMR spectra depended on the stage of the growth curve. Serber and colleagues reported that the detectability and resolution of the E. coli in-cell NMR spectra depended on the duration of the induction period15. Thus, results of our study demonstrated that the detectability and resolution of E. coli in-cell NMR spectra may depend on both the stage of the growth curve and the duration of the induction period. To evaluate the significance of these two factors, E. coli in-cell NMR spectra were measured at various durations of the induction period and at two different stages of the growth curve, namely OD600 values ~0.65 and 1.25, during the late lag to early log phase and the middle log phase, respectively. At both stages, OD600 ~ 0.65 and 1.25, the resolution was better for the E. coli in-cell NMR spectrum with the longer induction period (Fig. 4).
E. coli in-cell NMR spectra with the same finishing point for the induction but with different starting points and induction periods were compared. The spectrum associated with induction during the late lag to early log phase (OD600 value ~0.65) with a 4.5-h induction period had the same finishing point of induction as those induced in the middle log (OD600 value ~1.25) and stationary (OD600 value ~2.6) phases with 3.0- and 1.5-h induction periods, respectively (Fig. 4 and Supplemental Fig. S17). The detectability and resolution of the E. coli in-cell NMR spectra clearly depended on the starting point of induction, and the best point for this was the middle log phase (OD600 value ~1.25), as well as described in the Results section. These results demonstrate that the detectability and resolution of E. coli in-cell NMR spectra depended on both the stage of the growth curve and the duration of the induction period. Strictly, it was the starting point of induction, not the duration of induction, that was the critical determinant of the detectability and resolution of E. coli in-cell NMR spectrum.
The results of this study showed that the starting point of induction is the critical determinant of the detectability and resolution of E. coli in-cell NMR spectra. In addition, we have proposed that the origin of the dependency of these spectra on the stage of the growth curve of the target protein arises from the yield of soluble and healthy recombinant proteins. However, the physiological meaning of this is not clear.
To establish the physiological meaning of the starting point of induction, we analyzed the E. coli growth curves for the cases where IPTG was added when OD600 was ~0.65 or 1.25. In these cases, the OD600 values finally reached 1.5 and 2.6 at the stationary phase, respectively. The OD600 value 2.6 was identical to the case of no induction; that is, when the starting point of induction was the middle log phase, the E. coli cell was able to grow healthily similar to the case where there was no induction. These results suggest that the detectability and resolution of E. coli in-cell NMR spectrum can provide information about the intra-cellular physiological conditions of the E. coli host cells. It should be noted that each growth curve stage is from the mixture of all phases in the life cycle of the E. coli host cell, as the duplication time of the cell ~20 min is significantly shorter than the time range of both the growth curve ~6 hours and the induction time of the protein overexpression ~2–6 hours. Therefore, any in-cell NMR spectra represent the mixture of all phases in the life cycle of the E. coli host cell.
In this study, we showed that the detectability and resolution of E. coli in-cell NMR spectra depended on the growth curve stage of the host cells, and in particular on the starting point of the induction of the target protein overexpression. The spectra correlated with the soluble expression level of the target protein, and the origin of the dependency of the spectra on the stage of the growth curve of the target protein was expected to arise from the yield of soluble and healthy recombinant proteins. When the starting point of induction was the middle log phase, the E. coli cell was able to grow healthily, similar to the no induction case, and this gave the best spectrum. Thus, the E. coli in-cell NMR spectrum can provide information about the intracellular physiological conditions in the E. coli host cells as ensemble average, and could be utilized to monitor soluble and healthy protein production.
All chemicals were purchased from Wako Chemicals (Osaka, Japan) and Nacalai Tesque, Inc.(Kyoto, Japan) unless specifically noted, except for the isotope-enriched amino acids, which were purchased from Cambridge Isotope Laboratories.
The pET expression plasmid containing cDNA coding GB1 (the T2Q mutant, pI = 4.5, whose amino acid sequence is MQYKLILNGKTLKGETTTEAV DAATAEKVFKQYANDNGVDGEWTYDDATKTFTVTE), thermus thermophilus HB8 (TTHA1718) and its homologues (TTHA0227, TTHA0814), human calmodulin (hCaM), and green fluorescent protein (GFP) were individually incorporated into the E. coli BL21(DE3) strain (New England Biolabs, MA, USA). A fresh single colony of the BL21(DE3) transformants was picked and cultivated in 15 mL LB medium in a 50 mL Falcon tube at 37 °C overnight (ca. 12 h). The whole cell suspension was then centrifuged at 2,000 g for 10 min at room temperature, and its supernatant was eliminated. The cell pellets were re-suspended in 150 mL of fresh M9 minimal media, which contained 15NH4Cl (1 g/L) and isotopically unlabeled d-glucose (4 g/L) as the sole nitrogen and carbon sources, respectively; this was aseptically prepared in a 300-mL baffled Erlenmeyer flask and cultured at 37 °C with horizontally rotational shaking at 200 rpm. When the cell density (OD600; optical density at 600 nm) values reached the desired value, IPTG was added into the medium to achieve a final concentration of 1 mM, and cell cultivation was continued for a further period (typically 3 h). After the cell cultivation finished, the cell suspension was centrifuged at 2,000 g for 10 min at room temperature, and the required amount of cell pellets (typically ca. 0.55 g wet cell weight) was gently re-suspended with ~500 μL of fresh M9 minimal media containing 10% [v/v] D2O. The cell suspension was then gently transferred into a 5-mm diameter glass tube.
Fractionation of the cell lysate into soluble and insoluble components (the supernatant and precipitate, respectively) was performed as follows. First, 5 mL of cell suspension was centrifuged at 2,000 g for 10 min at room temperature, and the cell pellet was washed gently once using phosphate buffered saline (PBS). The cell pellet was gently re-suspended using 1 mL of fresh PBS, and the cells were disrupted with sonication. Following centrifugation at 17,000 g for 10 min at 4 °C, 75 μL of the supernatant was retrieved and was mixed well with 25 μL of 4 × loading buffer for SDS–PAGE and 1 μL of 2-mercaptoethanol. The residual pellet (cell debris) was re-suspended with 600 μL of PBS and was mixed well with 200 μL of 4 × loading buffer for SDS–PAGE and 8 μL of 2-mercaptoethanol. SDS–PAGE was applied to 10 μL aliquots of these soluble and insoluble fractions; the SDS–PAGE gel was then stained using Coomassie Brilliant Blue (CBB) and the area of the band of the target protein was quantified by using ImageJ software.
All of the NMR spectra were recorded on either a Bruker Ascend 500 MHz or an AVANCE 600 MHz equipped with a cryogenic BBO or QCI-P probe, respectively, at a sample temperature of 25 °C. Two-dimensional (2D) 1H-15N BEST-TROSY NMR spectra, with 1024 data points in the F2 (1H) dimension and 96 increments with 16 ppm and 36 ppm spectral width in the F2 and F1 (15N) dimensions, respectively, and glycine-selective 1H-15N HSQC with MUSIC27,28 were recorded with 32 scans. The NMR data were processed and analyzed using programs NMRPipe29 and Sparky (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco), respectively.
Chary, K. V. R. & Govil, G. NMR in Biol System From Molecules to Humans (Eds.: R. Kaptein), Supringer, Berlin, pp. 423–52 (2008).
Zimmerman, S. B. & Trach, S. O. Estimation of macromolecule concentrations and excluded volume effects for the cytoplasm of Escherichia coli. J. Mol. Biol. 222, 599–620 (1991).
Luby-Phelps, K. Cytoarchitecture and physical properties of cytoplasm: volume, viscosity, diffusion, intracellular surface area. Int. Rev. Cytol. 192, 189–221 (2000).
Hatters, D. M., Minton, A. P. & Howlett, G. J. Macromolecular crowding accelerates amyloid formation by human apolipoprotein C-II. J. Biol. Chem. 277, 7824–30 (2002).
Stagg, L., Zhang, S. Q., Cheung, M. S. & Wittung-Stafshede, P. Molecular crowding enhances native structure and stability of alpha/beta protein flavodoxin. Proc. Natl. Acad. Sci. USA 104, 18976–81 (2007).
Schlesinger, A. P., Wang, Y., Tadeo, X., Millet, O. & Pielak, G. J. Macromolecular Crowding Fails To Fold a Globular Protein in Cells. J. Am. Chem. Soc. 133, 8082–5 (2011).
Wang, Y., Sarkar, M., Smith, A. E., Krois, A. S. & Pielak, G. J. Macromolecular Crowding and Protein Stability. J. Am. Chem. Soc. 134, 16614–18 (2012).
Reckel, S., Hansel, R., Löhr, F. & Dӧtsch, V. In-cell NMR spectroscopy. Prog. Nucl. Magn. Reson. Spectrosc. 51, 91–101 (2007).
Pielak, G. J. et al. Protein nuclear magnetic resonance under physiological conditions. Biochemistry 48, 226–34 (2009).
Maldonado, A. Y., Burz, D. S. & Shekhtman, A. In-cell NMR spectroscopy. Prog. Nucl. Magn. Reson. Spectrosc. 59, 197–212 (2011).
Mercatelli, E., Barbieri, L., Luchinat, E. & Banci, L. Direct structural evidence of protein redox regulation obtained by in-cell NMR. Biochim. Biophys. Acta 1863, 198–204 (2016).
Inomata, K., Kamoshida, H., Ikari, M., Ito, Y. & Kigawa, T. Impact of cellular health conditions on the protein folding state in mammalian cells. Chem. Commun. 53, 11245–8 (2017).
Tanaka, T. et al. High-resolution protein 3D structure determination in living eukaryotic cells. Angew. Chem. Int. Ed. 58, 7284–8 (2019).
Nishida N., Ito, Y. & Shimada, I. In situ structural biology using in-cell NMR. Biochim. Biophys. Acta Gen. Subj. in press (2019).
Serber, Z., Ledwidge, R., Miller, S. M. & Dӧtsch, V. Evaluation of Parameters Critical to Observing Proteins Inside Living Escherichia coli by In-Cell NMR Spectroscopy. J. Am. Chem. Soc. 123, 8895–901 (2001).
Serber, Z. et al. High-Resolution Macromolecular NMR Spectroscopy Inside Living Cells. J. Am. Chem. Soc. 123, 2446–7 (2001).
Sakakibara, D. et al. Protein structure determination in living cells by in-cell NMR spectroscopy. Nature 458, 102–5 (2009).
Ikeya, T. et al. NMR protein structure determination in living E. coli cells using nonlinear sampling. Nat. Protoc. 5, 1051–60 (2010).
Hamatsu, J. et al. High-resolution heteronuclear multidimensional NMR of proteins in living insect cells using a baculovirus protein expression system. J. Am. Chem. Soc. 135, 1688–91 (2013).
Xu, G. et al. Strategies for protein NMR in Escherichia coli. Biochemistry 53, 1971–81 (2014).
Erbel, P. J. et al. Identification and biosynthesis of cyclic enterobacterial common antigen in Escherichia coli. J. Bacteriol. 185, 1995–2004 (2003).
Erbel, P. J. et al. Cyclic enterobacterial common antigen: potential contaminant of bacterially expressed protein preparations. J. Biomol. NMR 29, 199–204 (2004).
Studier, F. W. & Moffatt, B. A. Use of bacteriophage T7 RNA polymerase to direct selective high-level expression of cloned genes. J. Mol. Biol. 189, 113–30 (1986).
Rosenberg, A. H. & Studier, F. W. T7 RNA polymerase can direct expression of influenza virus cap-binding protein (PB2) in Escherichia coli. Gene 59, 191–200 (1987).
Studier, F. W., Rosenberg, A. H., Dunn, J. J. & Dubendorff, J. W. Use of T7 RNA polymerase to direct expression of cloned genes. Methods Enzymol. 185, 60–89 (1990).
Qing, G. et al. Cold-shock induced high-yield protein production in Escherichia coli. Nat. Biotechnol. 22, 877–82 (2004).
Schubert, M., Smalla, M., Schmieder, P. & Oschkinat, H. MUSIC in triple-resonance experiments: amino acid type-selective 1H-15N correlations. J. Magn. Reson. 141, 34–4 (1999).
Schubert, M., Oschkinat, H. & Schmieder, P. MUSIC, selective pulses, and tuned delays: amino acid type-selective 1H-15N correlations, II. J. Magn. Reson. 148, 61–72 (2001).
Delaglio, F. et al. NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR 6, 277–93 (1995).
We would like to express our gratitude to Prof. Conggang Li (Wuhan Institute of Physics and Mathematics) for kind instruction about the amino acid sequence of GB1 protein and Misaki Nakamura (Yokohama National University) for gel quantification. This work was financially supported in part by KAKENHI (JP19H04856, JP17H05873, JP17H03641, and 26840050), AMED-CREST (JP19gm0910005), and BINDS for T.S., T.F., C.K. from MEXT and AMED, Japan, and performed under the Collaborative Research Program of the Institute for Protein Research, Osaka University (CR-19-02).
The authors declare no competing interests.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Sugiki, T., Yamaguchi, Y., Fujiwara, T. et al. In-cell NMR as a sensitive tool to monitor physiological condition of Escherichia coli. Sci Rep 10, 2466 (2020). https://doi.org/10.1038/s41598-020-59076-2
Journal of Magnetic Resonance (2021)
Analytical Chemistry (2021)
Proceedings of the National Academy of Sciences (2020)