Radio Signals from Live Cells: The Coming of Age of In-Cell Solution NMR

Abstract

A detailed knowledge of the complex processes that make cells and organisms alive is fundamental in order to understand diseases and to develop novel drugs and therapeutic treatments. To this aim, biological macromolecules should ideally be characterized at atomic resolution directly within the cellular environment. Among the existing structural techniques, solution NMR stands out as the only one able to investigate at high resolution the structure and dynamic behavior of macromolecules directly in living cells. With the advent of more sensitive NMR hardware and new biotechnological tools, modern in-cell NMR approaches have been established since the early 2000s. At the coming of age of in-cell NMR, we provide a detailed overview of its developments and applications in the 20 years that followed its inception. We review the existing approaches for cell sample preparation and isotopic labeling, the application of in-cell NMR to important biological questions, and the development of NMR bioreactor devices, which greatly increase the lifetime of the cells allowing real-time monitoring of intracellular metabolites and proteins. Finally, we share our thoughts on the future perspectives of the in-cell NMR methodology.

Figure 1

Overview of the different approaches for in-cell NMR sample preparation. (a) Proteins (violet) can be directly expressed in the cells to be analyzed by NMR: bacteria and yeast cells (top, middle) are transformed with an inducible expression vector; insect cells (bottom left) are infected with a baculoviral vector; human cells (bottom right) are transfected with a constitutive expression vector. Protein expression is performed in isotopically enriched media. (b) Recombinantly expressed and purified proteins (teal) or nucleic acids (not shown) can be microinjected into X. laevis oocytes (top left) or delivered to human cells by using CPP-fused constructs (top right), by permeabilizing the cells with pore-forming toxins (middle) or by electroporating the cells (bottom). All cell types (but not proteins) are drawn approximately to scale; the oocyte is scaled down by 1:200.

Figure 2

Protein quinary interactions decoded by in-cell NMR. (a) 3D structure of three evolutionary divergent proteins: bacterial TTHA, human HAH1, and human SOD1 β-barrel; (b) the rotational mobility in bacterial cells of a set of surface mutations for the proteins shown in (a) depends both on the negative charge density (left) and on the exposed surface area of the hydrophobic side chains of valine (V), leucine (L), and isoleucine (I) residues (right): increasing charge density and decreasing hydrophobic surface result in faster rotational diffusion. Mobility^in cell is calculated as the ratio of in-cell peak heights and lysate peak heights. Line offset in the right panel is calculated from the regression lines (left panel) at density charge^density = 1. Reproduced with permission from Mu et al. Copyright 2017 Mu et al. (c) Comparison of models to describe in-cell NMR relaxation data: an apparent mass increase of the observed protein (left) results in poor agreement between R₁ and R₂, whereas a model where the observed protein is transiently interacting with cellular components of different sizes (right) fully reconciles the R₁ and R₂ data. Reproduced with permission from Leeb et al. Copyright 2020 American Chemical Society.

Figure 3

Maturation process of SOD1. (a) Scheme of SOD1 folding and maturation and molecular chaperone role of CCS. The preferred direction of each step is indicated by the size of the arrow. The effect of pathogenic mutations is shown with red arrows. Reproduced with permission from Luchinat et al. Copyright 2017 Luchinat et al. (b–e) In-cell ¹H–¹⁵N NMR spectra of WT (c–e) or mutant (b) SOD1 in human cells at different steps of the maturation (indicated by the corresponding drawing): (b) irreversibly misfolded mutant SOD1. Reproduced with permission from Luchinat et al. Copyright 2014 Nature Publishing Group; (c) apo-SOD1^SH, (d) E,Zn-SOD1^SH, and (e) Cu,Zn-SOD^SS WT SOD1. Reproduced with permission from Luchinat & Banci. Copyright 2018 American Chemical Society.

Figure 4

Conformational dynamics of IDPs probed by in-cell NMR. (a–c) The dynamics of α-syn in human-derived A2780 cells: (a) in-cell/in vitro relative signal intensity (I/I₀, top) and exchange contribution (R_ex, bottom) plotted for each residue of α-syn; (b) intramolecular PRE-derived distance profiles α-syn in buffer (gray) and in A2780 cells (red). Regions with decreased intensity, increased exchange, or increased compaction are indicated with arrows. (c) Intracellular α-syn is more compact (top) and interacts with cellular components at the N-terminus through hydrophobic residues and electrostatically at the C-terminus (bottom). Reproduced with permission from Theillet et al. Copyright 2016 Nature Publishing Group. (d, e) The dynamics and interactions of FG Nups in bacteria: (d) in-cell NMR relaxation and hetNOE measurements on two FG Nups regions, plotted for each residue, compared to in vitro data. The locations of the FG motifs are indicated with gray bars. (e) The location of FG Nups within the nuclear pore complex, and the main features of the FG-Repeat constructs analyzed by in-cell NMR. Reproduced with permission from Hough et al. Copyright 2015 Hough et al.

Figure 5

Protein-observed intracellular ligand screening approaches. (a) HTS for compounds that disrupt protein–protein interactions (PPIs) using in-cell NMR combines the advantages of in vitro and in vivo studies by providing residue-specific information in a physiologically relevant environment. Spectra from samples treated with mixtures through the matrix approach are analyzed by SVD. Adapted with permission from DeMott et al. Copyright 2018 American Chemical Society. (b) ¹H–¹⁵N in-cell NMR spectra of FKBP in bacteria in complex with unlabeled FRB in the absence (black) and presence (red) of rapamycin. (c) Surface residues of FKBP involved in the interaction with rapamycin; FRB is shown in blue. Reproduced with permission from Xie et al. Copyright 2009 American Chemical Society. (d) ¹H–¹⁵N in-cell NMR spectra of CA II in human cells in the absence (black) and presence (red) of acetazolamide (AAZ); (e) 3D view of AAZ bound to the catalytic zinc ion in the active site of CA II (PDB: 3HS4). (f) Imino region of the 1D ¹H NMR spectra of CA II in human cells in the absence of ligands (black) and treated with AAZ (red), MZA (magenta), and other ligands (blue). Reproduced with permission from Luchinat et al. Copyright 2020 Luchinat et al.

Figure 6

Structure and interactions of nucleic acids revealed by in-cell NMR. (a) ¹H–¹³C NMR spectra of isotopically labeled DNA (top) and RNA (bottom) hairpins recorded in vitro (left) and in X. laevis oocytes 5 h after a microinjection. The hairpin structures are shown in the left panels. Adapted with permission from Hänsel et al. Copyright 2009 American Chemical Society. (b) Different topologies of telomeric RNA G-quadruplex observed by ¹⁹F NMR experiments in vitro (top) and in X. laevis oocytes (bottom). The ¹⁹F signals corresponding to the G-quadruplex dimer (orange) and two-subunits stacked G-quadruplex (violet) are color-coded according to the structures (right). Adapted with permission from Bao et al. Copyright 2017 Bao et al. (c) Interaction of a DNA hairpin (MH-DNA) with netropsin in human cells: (top left) flow cytometry analysis after electroporation; viable MH-DNA containing cells are shown in red. (bottom left) Localization MH-DNA (green) inside the cell nucleus (blue). (right) Deconvoluted imino region of MH-DNA 1D ¹H NMR spectra obtained alone and in the presence of netropsin in vitro (green and blue) and in human cells (red). Reproduced with permission from Krafcikova et al. Copyright 2019 American Chemical Society.

Figure 7

Schematics of wide-bore (a–c) and modern narrow-bore (d–g) NMR bioreactors for mammalian cells. (a) Perfusion bioreactor by Foxall et al.; cells are embedded in low-gelling agarose. (b) Stirred bioreactor by Freyer et al.; spheroids are kept in suspension by mechanical stirring. (c) Hollow-fiber reactor by Gillies et al.; cells growing in suspension are directly inoculated, whereas cells growing in adhesion are embedded in collagen beads prior to inoculation. (d) Perfusion bioreactor by Kubo et al.; cells are embedded in Mebiol gel. (e) bioreactor by Breindel et al.; cells embedded in low-gelling agarose are perfused with a horizontal drip irrigation system. (f, g) Two alternative bioreactor setups based on a sealed flow unit; (f) cells are kept as a suspension and the nutrients diffuse through a coaxial microdialysis membrane (Cerofolini et al.); (g) cells are embedded in low-gelling agarose and perfused as in (a, d) (Luchinat et al.). Schematics were redrawn to scale based on the technical details and illustrations reported in the works cited above.

Figure 8

Application of bioreactors to real-time in-cell NMR. (a) Analysis of the cellular metabolic state: ³¹P NMR spectra of HeLa cells in the absence (top) and in the presence (bottom) of a flow of nutrients. (inset) Monitoring the ratio between P_β-ATP and inorganic phosphate (P_i) as a function of time reveals a decrease of cellular metabolic activity upon treatment with cytotoxic compounds. Reprinted from Carvalho et al., with permission from Elsevier. (b) Changes of protein quinary structure upon antibiotic binding to the ribosome: (top) overlay of the in-cell ¹H–¹⁵N NMR spectra of thioredoxin (Trx) in E. coli cells before (red) and after (blue) treatment with tetracycline; changes in Trx peak intensities as a function of time (bottom left); possible mechanism of how an antibiotic binding to the ribosome induces Trx-mRNA interactions (bottom right). Reprinted with permission from Breindel et al. Copyright 2017 American Chemical Society. (c) Changes in protein and glutathione redox state: real-time in-cell NMR spectra of Trx and glutathione in HeLa cells after treatment with ATG, a thioredoxin reductase inhibitor (black), and the oxidant tert-butyl hydroperoxide (TBH, green). Reprinted with permission from Mochizuki et al. Copyright 2018 American Chemical Society. (d) Monitoring protein–ligand interactions in real time: reconstructed ¹H NMR spectra of carbonic anhydrase II (CA II) in HEK293T cells before (black) and after treatment with acetazolamide (AAZ, red) or methazolamide (MZA, magenta); concentration profiles of ligand-bound CA II as a function of time (bottom); example of a time series of raw in-cell ¹H NMR spectra (right). Reprinted with permission from Luchinat et al. Copyright 2020 American Chemical Society.