Human cells for human proteins: Isotope labeling in mammalian cells in suspension for functional NMR studies

TL;DR

This paper introduces new protocols for isotope labeling in human cells, enabling NMR studies of difficult proteins like membrane proteins and nuclear receptors.

Contribution

A comprehensive suite of affordable and scalable protocols for isotope labeling in suspension HEK293 cells, including specific labeling strategies.

Findings

Uniform and specific isotope labeling, including ILV 13C-methyl labeling, is achievable in suspension HEK293 cells.

The protocols enable NMR studies of challenging human proteins such as membrane proteins and nuclear receptors.

Labeling is accomplished using a simple setup and cost-effective media based on labeled amino acids or microbial extracts.

Abstract

In biological and biomedical research, the focus progressively moves towards difficult human proteins, which often can only be expressed in higher eukaryotic cells. Nuclear magnetic resonance (NMR) could contribute significantly to the understanding of important proteins as it is one of the most information‐rich methods. However, to exploit the full potential of NMR, proteins must be isotope labeled. Although expression protocols in, for example, HEK293 cells are often established, isotope labeling is difficult and very expensive, and published protocols are mostly limited to adherent cells in plates. To resolve this disparity, we have developed a comprehensive suite of protocols for isotope labeling in HEK293 cells in suspension. We demonstrate uniform 15N and 13C labeling, as well as specific labeling, with special focus on methyl bearing amino acids, including the popular ILV 13C‐methyl labeling pattern. In addition, several innovative practices are introduced to deal with special cases. Labeling is achieved with a simple laboratory setup and affordable labeling media. These are based on either labeled amino acids, their precursors or amino extracts from microorganisms, like yeast, algae or bacteria. We demonstrate how these protocols enable NMR studies of proteins that are considered to be especially difficult to produce, for example, different human membrane proteins and nuclear receptors. We therefore expect that these new methods make many highly important proteins accessible to NMR studies and allow exploiting the high information content of this method for accelerating biological and pharmaceutical research.

Figures

Protein production in HEK293 cells. (a) Maintenance cycle of cells in suspension culture. Cells are kept in a shaker incubator with controlled temperature, humidity and CO2 levels. Every 2–3 days, cells are split to 0.3·106 cells/ml. Cell count and viability is monitored using a cell counter and a microscope. (b) Expression of proteins in HEK293 cells. The schematic workflow of protein expression is depicted for adherent cells (left) and stable cell lines in suspension culture (middle) and transiently transfected cells in suspension culture (lower right). After transfection, a stably transfected clone is selected by means of antibiotic resistance (2–3 weeks). For expansion, cells are seeded on progressively more plates (adherent culture) or adapted to growth in suspension (suspension culture, 2–3 weeks). Cells can be maintained in suspension for a period of a year and expanded whenever needed to larger volumes for expression (~1 week). For transient expression, typically a suspension adapted HEK293T cell line is expanded to the desired volume (~1 week) and transiently transfected for expression. Depending on the plasmid and the presence of a genomically integrated repressor element in the host cell line, induction may be required. Protein yields tend to be higher in stable cell lines, but transient expression can be accomplished in much shorter time.

Isotope labeling strategy for HEK293 cells. (a) Cells are expanded in unlabeled V3 medium until the desired volume is reached. Succeeding gentle centrifugation, cells are resuspended in V3 medium without yeast extract (V3⊖). In order to lower the amount of endogenous unlabeled amino acids, cells are incubated for 16 h in V3⊖. Subsequently, either labeled extract is added for uniform labeling or an amino acid mixture with the appropriate isotope labeling pattern is added to achieve amino acid type specific labeling. This is followed by induction of the protein production for stable cell lines. For transient expression, transfection with plasmid is performed during the incubation period. After 48–60 h of expression, cells can be harvested. (b) Bar graph displaying the quantity of isotope labeled amino acids and labeled yeast extract (YE) used individually for a desired labeling pattern (dark pink) and unlabeled amino acids and YE (light pink) which are added to V3⊖ to compensate for the lower amino acid content in V3⊖. For comparison the amino acid content of full V3 medium is shown in gray. Note that YE is missing in V3⊖ (light gray).

Approaches for uniform isotope labeling in mammalian cells in suspension. (a) Microscope images of fluorescent cells from full medium and labeling medium. (b) Yields of mEGFP in mg/L from different media formulations are shown: Full V3 medium (V3, red), V3⊖ with yeast extract added back (V3⊖ + YE, orange), V3⊖ with algal extract (Celtone) added (V3⊖ + AE, dark green), V3⊖ with delipidated (DL) algal extract added (V3⊖ + AEDL, bright green) and FreeStyle medium (blue). The yields of mEGFP are in the mg range because it was expressed as a fusion to β1AR and released by protease cleavage. (c–e) 2D TROSY‐HSQC spectra of u‐15N labeled mEGFP produced in stable HEK293 GnTI− cells either with (c) 15N labeled yeast extract or (d) delipidated 13C,15N‐labeled algal extract (ISOGRO, Sigma). All amino acid types were labeled, since identical signals are visible as reported from E. coli expression. Incorporation was 70%, as determined by mass spectrometry. (e and f) [15N,1H]‐ and [13C,1H]‐HSQC spectra of ILV‐labeled mEGFP produced with E. coli extracts with the same labeling pattern.

Amino acid type‐specific labeling using the 15N, 13C, and 2H isotope labeled amino acids indicated on the right in each spectrum. 2D [15N,1H] and [13C,1H] correlation spectra of mEGFP produced in stably transfected HEK293 GnTI− cells grown in suspension are shown. See text and Data S1 for exact experimental conditions.

Labeling of methyl groups of Val, Leu, and Ile using 13C and 2H isotope labeled precursors indicated on the right in each spectrum. 2D [13C,1H] correlation spectra of mEGFP produced in stably transfected HEK293 GnTI− cells are shown. (c) For Ile, only the uniformly 13C labeled precursor was available, the precursor in gray represents a proposed labeling pattern for similar spectra as in (a) and (b). The strong signal (marked with * in a) stems from detergent present in the buffer. See text and Data S1 for exact experimental conditions.

Special cases. (a) Methionine labeling in full media by enzymatic removal of intrinsic methionine from a medium with unknown composition, here HEK293 Freestyle medium (GIBCO). On the left, the reaction of the enzyme is shown (for reaction conditions see main text). Note that methanethiol is volatile and thus is continuously removed from the reaction resulting in complete turnover of methionine. In the middle, the methionine signal in the medium is shown at 0, 24 and 48 h after treatment with MGL. To monitor the reaction, the medium was spiked with 1 g/L 13C‐Met and 13C‐edited 1D 1H spectra were recorded. On the right, a 2D [13C,1H] HMQC spectrum of 13C‐Met labeled mEGFP is shown, produced from enzymatically methionine‐free HEK293 Freestyle medium (GIBCO). (b1) Isotope labeling of alanine with inhibition of alanine transaminase to suppress unwanted isotope dilution from pyruvate. On the left, the reaction of alanine transaminase is shown. Note that unlabeled pyruvate is present in rather high concentrations in the cell, thus driving the reaction towards the product alanine. Addition of l‐cycloserine (red) inhibits this reaction. On the right, a 2D [13C,1H]‐XL‐ALSOFAST‐HMQC spectrum of 2Hα,13Cβ‐Ala labeled mEGFP is shown, which was produced in presence of 50 mg/L of l‐cycloserine. (b2) Scrambling can also be observed for other amino acids, like in the example of valine labeling of mEGFP shown here. With the standard protocol, scrambling of the labeled valine precursor to leucine is apparent. Addition of an excess of unlabeled leucine (460 mg/L instead of 115 mg/L) changes the balance of the equilibrium reaction and suppresses scrambling. Note that the spectrum on the right hand side (same type as in b1) is plotted at very low contour levels, and it was recorded in detergent‐containing buffer, leading to the strong signal indicated by the dashed horizontal line. (c) When selectively 13C methyl labeled amino acids are not available, suppression of 1 J CC couplings in u‐13C labeled amino acids yields sharp signals. On the top left, a schematic representation of isoleucine is shown and on the right, the Mz‐excitation profile of an adapted BADCOP1 pulse is shown. The pulse was time‐reversed and the length adjusted in order to selectively invert β and γ1 nuclei, resulting in decoupling of the adjacent γ2 and δ1 nuclei, respectively. (BADCOP1_TR, pulse length: 1700 μs, offset: 16 ppm, simulated in Topspin shapetool). Typical chemical shifts of carbon nuclei are indicated with colored bands (BMRB; Ulrich et al., 2007). Below 2D [13C,1H]‐XL‐ALSOFAST‐HMQC spectra are plotted, showing the methyl region of 13C‐Ile labeled mEGFP, without (left) and with (right) application of a BADCOP1_TR 180° pulse. A slice through one representative signal is shown in the top right corner of each spectrum and illustrates the suppression of the homonuclear couplings as well as the concurrent ~2‐fold gain in sensitivity. Spectra were recorded on Bruker Avance IIIHD 800 MHz (a) and 900 MHz (b, c) spectrometers equipped with TCI CryoProbes.

Choice of isotope labeling patterns based on coverage of the target protein (left) and quality of NMR spectra (right). On the left, ribbon representation of the GPCR β1AR of turkey (mgβ1AR) is shown with amino acid side chains acting as small light sources illuminating parts of the protein. The illuminated parts should approximately represent the regions of the protein on which the labeled nuclei can report, as their resonance is influenced by effects in their immediate surroundings. On the right, [13C,1H]‐XL‐ALSOFAST‐HMQC spectra of mgβ1AR are shown, with 13C methyl group labeling of Met, Leu, Ile, Val, Ala and ILV as indicated in the spectrum. All spectra were recorded at 298 K on Bruker Avance IIIHD 900 MHz and NEO 1.2 GHz (bottom spectrum) spectrometers equipped with TCI CryoProbes.

Examples of NMR studies enabled by mammalian isotope labeling. (a) Overlay of [13C,1H]‐XL‐ALSOFAST‐HMQC spectra of bovine rhodopsin in DM micelles (0.2%) in its dark (blue) and light‐activated state (red). NMR spectra of the stable dark state protein were recorded in 3 mm NMR tubes, which were protected from light by coverage in electric heat shrink. Light state spectra were obtained after 1 min illumination of the sample with a flash light of a mobile phone. The large chemical shift changes affecting essentially all resonances reflect the global conformational change of the receptor in response to the isomerization of 9‐cis to all‐trans retinal. (b) Overlay of [13C,1H]‐HMQC spectra of estrogen related receptor γ (ERRγ) in absence (black contour lines) and presence of different ligands (contour lines of different colors). These simple spectra show binding of all compounds except for the one giving rise to the blue spectrum on the left‐most spectrum. (c) Overlay of [13C,1H]‐XL‐ALSOFAST‐HMQC spectra of stabilized human β1 adrenergic receptor (hsβ1AR) in LMNG/CHS (0.01/0.001%) micelles without (blue) and with (green) the agonist isoprenaline bound. Also here, chemical shift changes reflect the change in equilibrium towards an active conformation upon agonist binding. (d) [13C,1H]‐XL‐ALSOFAST‐HMQC spectrum for human β2 adrenergic receptor (hsβ2AR) in MSPΔH5 nanodiscs (PO/PG) in its resting state (orange) and in presence of 1 mM Formoterol (violet). Spectra were recorded on Bruker Avance IIIHD 600 MHz (b) and 900 MHz (a, c, d) spectrometers equipped with TCI CryoProbes.

Full text

INTRODUCTION

1

Protein production in human cells for NMR studies

1.1

To understand human biology and disease, examining human proteins at the molecular level gives insights into their function and interactions, and ultimately can lead to ways of treating a disease using novel therapeutic molecules. Important insights at the molecular level can be obtained with several structural and biophysical techniques (Bai et al., 2015; Blundell et al., 2002; Boozer et al., 2006; Jelesarov & Bosshard, 1999; Kühlbrandt, 2014; Lutomski et al., 2022; Pellecchia et al., 2008). All have in common that the target biomolecule must be produced in isolated form. However, many human target proteins are difficult to produce in the preferred laboratory organisms for heterologous expression, such as E. coli bacteria. This is mainly due to the limited folding machinery of E. coli, which is not competent to fold more complicated human proteins. This applies mainly to large proteins consisting of several domains, to protein complexes, integral membrane proteins and proteins requiring post‐translational modifications. Therefore, for functional and structural studies by cryo‐electron microscopy or X‐ray diffraction, such proteins are produced in higher eukaryotic cells like insect and mammalian cells. Within the latter, human cells arguably represent the most fitting expression host and for many scientifically and medicinally highly important proteins, expression protocols for human cells are available (Assenberg et al., 2013; Fernández & Vega, 2013; Gräslund et al., 2008). The fundamental advantages of human cells as expression host are that proteins are correctly folded, assembled into larger complexes and post‐translationally modified in a proper way, as they are synthesized and matured by their native production machinery, which has co‐evolved with the target proteins.

Nuclear magnetic resonance (NMR) is presently the biophysical technique with the highest information content: molecular interactions, dynamics, function and three‐dimensional structure can be studied at the atomic level (Gronenborn & Polenova, 2022; Wüthrich, 2003). Importantly, the biomolecules under study do not need to be chemically altered and the molecules can be studied directly in near‐physiological solution conditions. The desired level of detail of NMR studies can be gradually adjusted, which reduces the complexity of the experiment and the analysis of NMR data. NMR is therefore highly popular for examining molecular interactions and for functional studies, as little assay development is required and experiments and analysis are comparatively straightforward. However, for resolving signal overlap in large biomolecules, the one requirement NMR has, is that the macromolecule under study is enriched with magnetically active isotopes, typically ^15^N and ^13^C. So far, protocols for stable isotope labeling are mainly available for E. coli cells and—with some limitations—also for insect cells in suspension culture (Franke et al., 2018; Gossert & Jahnke, 2012; Jin et al., 2023; Saxena et al., 2012; Skora et al., 2015; Strauss et al., 2003; Strauss et al., 2005). For human cells there are only few reported protocols, (Barbieri et al., 2016; Dutta et al., 2012; Egorova‐Zachernyuk et al., 2011; Hansen et al., 1992; Luchinat & Banci, 2017; Mallis et al., 2024; Rosati et al., 2024; Skelton et al., 2010; Subedi et al., 2024; Werner et al., 2008) most of which are based on culturing cells in plates or very small volumes (≤30 mL), such that only highly expressing proteins can be produced. Therefore, important classes of proteins may be excluded from unique insights that NMR in solution can provide and there is a mismatch between proteins of high interest in biomedical sciences and those amenable to NMR studies. Thus, methods for isotope labeling in human cells are highly sought to enable studying yet inaccessible aspects on highly important proteins.

Protein expression in human cell culture can be carried out in different ways (Figure 1), where the major discerning aspects are: (i) the way of transfection of cells with a gene of interest either transiently or by stable integration into the host's genome (Büssow, 2015), and (ii) the way of culturing cells either adherent on a petri dish or in suspension in a shake flask (Graham, 1987; Malm et al., 2020). (i) Human cells in culture allow transient or stable transfection with genes of interest, leading to protocols with either fast turnover or reproducibly high expression levels, respectively. This renders human cells a very attractive system compared to insect cells: Several mutants—for example, for resonance assignment purposes—can be produced in a matter of 1–2 weeks with transient transfection, while in insect cells the process of virus generation easily requires a month of time. For the central protein in a project, on the other hand, a stable cell line with well‐characterized expression characteristics can be kept and expressed whenever needed.

Protein production in HEK293 cells. (a) Maintenance cycle of cells in suspension culture. Cells are kept in a shaker incubator with controlled temperature, humidity and CO2 levels. Every 2–3 days, cells are split to 0.3·106 cells/ml. Cell count and viability is monitored using a cell counter and a microscope. (b) Expression of proteins in HEK293 cells. The schematic workflow of protein expression is depicted for adherent cells (left) and stable cell lines in suspension culture (middle) and transiently transfected cells in suspension culture (lower right). After transfection, a stably transfected clone is selected by means of antibiotic resistance (2–3 weeks). For expansion, cells are seeded on progressively more plates (adherent culture) or adapted to growth in suspension (suspension culture, 2–3 weeks). Cells can be maintained in suspension for a period of a year and expanded whenever needed to larger volumes for expression (~1 week). For transient expression, typically a suspension adapted HEK293T cell line is expanded to the desired volume (~1 week) and transiently transfected for expression. Depending on the plasmid and the presence of a genomically integrated repressor element in the host cell line, induction may be required. Protein yields tend to be higher in stable cell lines, but transient expression can be accomplished in much shorter time.

(ii) In order to obtain large quantities of protein, human cells can be adapted to growth in suspension culture—notably an unnatural state for kidney cells—such that large numbers of cells can efficiently be cultivated (Figure 1b). The well‐established way of culturing HEK293 cells, however, is growing adherent cells as a monolayer in petri dishes (plates). Culture of adherent cells is relatively inexpensive: media can cost <20 EUR/L, and suitable incubators are found in essentially every cell biology laboratory. Several successful labeling experiments have been reported, based on growing adherent cells in plates of volumes of 5–10 mL (Barbieri et al., 2016; Dutta et al., 2012; Egorova‐Zachernyuk et al., 2011; Hansen et al., 1992; Luchinat & Banci, 2017; Rosati et al., 2024; Skelton et al., 2010; Werner et al., 2008) using homemade or commercial media for uniform ^15^N‐labeling, where it should be noted that the commercial medium is very costly (~6000 EUR/L). However, the difficult proteins we target here, are typically expressed at levels of 0.2–5 mg/L, such that production of sufficient amounts of protein for an NMR sample, would require growing up to a hundred plates with adherent cells grown in this expensive medium.

Since this is very impractical, culturing of cells in suspension is the preferred strategy for production of milligram amounts of protein (Williams et al., 2022; Yanaka et al., 2022). The major hurdles of suspension culture are, however, the high costs of media and incubators as well as the lack of proven protocols for isotope labeling in suspension culture. Suspension culture of HEK293 cells requires costly shaker incubators with controlled temperature, humidity and CO_2_ content, and expensive media (>200 EUR/L) specially designed for suspension culture at relatively high cell densities (>10^6^ cells/ml). With such high starting prices for regular culturing media, custom amino acid‐depleted media, as required for isotope labeling, often cost more than 2000 EUR/L, even without counting the cost of isotope labeled amino acids. Only recently, isotope labeling protocols for suspension culture have emerged, that are based on such custom dropout media where all amino acids or a subset of them were left away (Mallis et al., 2024; Subedi et al., 2024). In this way, the amino acids Ile, Leu, Val, Phe, and Tyr were successfully labeled in 30 mL cultures using ILVFY^−^ dropout media (Mallis et al., 2024). In an extensive study on scrambling of ^15^N labels, essentially all amino acids were labeled using a fully amino acid depleted medium, and high quality spectra were obtained, despite the relatively low estimated incorporation of 30%–52% (Subedi et al., 2024). It should also be noted that isotope labeling in relatively large volumes of suspension culture (1 L) has been reported for Chinese hamster ovary cells, where glycosylated immunoglobulin G1‐Fc was produced in uniform ^15^N labeled form. Here, a combination of isotope labeled algal amino acid extract and 16 labeled pure amino acids were used in the medium—rendering the approach rather expensive on the one hand, but on the other hand, glutamine was elegantly produced endogenously by co‐expressing glutamine synthase, thus significantly lowering the costs (Yanaka et al., 2022).

In summary, isotope labeling in human cells currently poses many practical and financial obstacles. We have addressed all these hurdles and have developed a comprehensive suite of protocols for suspension culture, which is both, enabling and cost‐effective at the same time. First, cells are grown in standard cell incubators, just with an added shaking plate. The growth medium, the second cost factor, is replaced by a formulation based on inexpensive media for adherent cells, supplemented with the key ingredients for stable growth in suspension. The final cost of a liter of labeling medium is that of the yeast extract‐depleted medium (<200 EUR/L) plus the cost of the isotope labeled amino acids, resulting in costs of 600 EUR/L for uniform ^15^N labeling, or <250 EUR/L for, for example, methionine ^13^C^ε^ labeling. Furthermore, we conserve the flexibility offered by mammalian expression systems allowing stable and transient expression.

In this study, we developed protocols for obtaining labeling patterns which proved to be the most impactful in previous NMR studies. NMR is mainly used for functional studies by characterizing molecular interactions, conformational changes and/or dynamics (Gronenborn & Polenova, 2022). In such studies, either uniformly or selectively ^15^N‐labeled proteins or proteins with ^13^C‐labeled methyl groups of the amino acids Ile, Leu, Val and Met were most often employed. We therefore developed protocols for all these important labeling patterns.

Our protocols have been tested with the named labeling patterns, but are surely not limited to these few amino acids. Except for heavily metabolized amino acids like Gln, Asp Glu, and Asn (side chain), most amino acids can be labeled with this method, as shown by a Subedi et al. (Subedi et al., 2024) Presently the limitation is mainly the availability of suitably isotope labeled amino acids or amino acid extracts. We thus expect that based on these methods, a large fraction of highly relevant proteins will newly become amenable to NMR studies and thus allow to address pressing questions in biological and medicinal sciences.

RESULTS

2

Cost‐effective suspension culture

2.1

The step towards suspension culture of mammalian cells represents a significant hurdle, as dedicated shaker‐incubators for mammalian cell culture with controlled humidity and CO_2_ levels are a large investment. Here, we used a basic incubator with active CO_2_ regulation and passive humidity control by means of a water bath in the base. Such incubators are easy to obtain as second‐hand models (Heraeus® and Binder were used here). The shaking function was added by using a shaking plate (Celltron®, Infors, Bottmingen, CH). While maintaining cells and during expression, it is important to monitor cell density and viability. For this task an economic miniaturized microscope with automated cell counting and viability determination software was used (CytoSMART Cell Counter). This entire setup should therefore be affordable for any laboratory.

Inexpensive media for cell growth in suspension are adaptable for stable isotope labeling

2.2

The highly optimized commercial media for suspension culture of HEK293 cells are costly (200–500 EUR/L) and custom amino acid‐depleted media are often prohibitively expensive (with quoted prices typically above 2000 EUR/L). This is a major issue when working with proteins with low expression levels like membrane proteins for instance. Media for suspension culture are inherently more expensive than media for adherent culture, because they must contain more nutrients to support growth to higher cell densities, and they also need to contain factors, which prevent cells from forming aggregates, thus keeping them in suspension—notably an unnatural state for HEK293 cells, without the native, multiple tight contacts between cells.

Instead of using expensive media that are designed for suspension culture, here, an economic medium for adherent culture with low nutrient content (derived from DMEM/Ham's F‐12; Dulbecco & Freeman, 1959; Ham, 1965; Yao & Asayama, 2017) was used as a base medium and only a few selected ingredients were added to allow cell growth in suspension. To increase the nutrient content, 4 g/L of glucose and 3 g/L of yeast extract were added. In order to maintain cells in suspension a proprietary mixture of lipids was added and albumin at 1 g/L was used to saturate surface receptors on the cells that would either attach to other cells or walls of the culture flask. Such a medium was jointly developed between Novartis AG and Bioconcept AG and resulted in the commercially available medium V3 (V3, Bioconcept AG) (BioConcept, n.d.). When compared to FreeStyle™ 293 medium (Gibco), V3 shows same doubling times of cells (20–25 h), even slightly higher transfection efficiency, but lower maximal cell density (3.5 vs. 5 × 10^6^ cells/ml). Since protein production is typically initiated at 1 × 10^6^ cells/ml, the maximally achievable cell density is not so relevant here, and protein yields are essentially the same in the two media. Only in the rare cases where expression proceeds for more than 5 days, nutrients become limiting. With a price point of <100 EUR/L at the time of writing, this brings the costs for propagating cells to an acceptable level for academic research.

This medium is an ideal platform for isotope labeling, since only the major source of amino acids, the yeast extract, needs to be replaced by an appropriate mixture of amino acids, defined by the desired isotope labeling pattern. Although this labeling medium is a custom medium, the price is affordable (<150 EUR/L, V3‐702‐I, Bioconcept AG), (BioConcept, n.d.) since the vendor only needs to exclude one ingredient. The disadvantage is that the medium still contains traces of amino acids from the underlying medium for adherent cell culture (typically tens of milligrams per amino acid, see Figure 2b). Furthermore, 5%–10% of fetal bovine serum (FBS) are added to the cell culture, which contain trace amounts of most amino acids. The unlabeled amino acids that are present in low amounts in media are, however, not so important, since in cell culture there is always a significant carry‐over of unlabeled amino acids from the growth culture, which need to be depleted during an incubation period prior to adding the labeled amino acid mixture. By adjusting the duration of the depletion period, also the small amount of unlabeled amino acids are sufficiently reduced, and incorporation levels of 70%–95% of the isotope can be achieved. This general approach therefore enables inexpensive isotope labeling with sufficient incorporation and high protein yields.

Isotope labeling strategy for HEK293 cells. (a) Cells are expanded in unlabeled V3 medium until the desired volume is reached. Succeeding gentle centrifugation, cells are resuspended in V3 medium without yeast extract (V3⊖). In order to lower the amount of endogenous unlabeled amino acids, cells are incubated for 16 h in V3⊖. Subsequently, either labeled extract is added for uniform labeling or an amino acid mixture with the appropriate isotope labeling pattern is added to achieve amino acid type specific labeling. This is followed by induction of the protein production for stable cell lines. For transient expression, transfection with plasmid is performed during the incubation period. After 48–60 h of expression, cells can be harvested. (b) Bar graph displaying the quantity of isotope labeled amino acids and labeled yeast extract (YE) used individually for a desired labeling pattern (dark pink) and unlabeled amino acids and YE (light pink) which are added to V3⊖ to compensate for the lower amino acid content in V3⊖. For comparison the amino acid content of full V3 medium is shown in gray. Note that YE is missing in V3⊖ (light gray).

Isotope labeling strategies

2.3

General isotope labeling strategy

2.3.1

The general isotope labeling strategy used here for HEK293 cells is the following: cells are maintained in unlabeled V3 medium and expanded to the desired volume for an expression experiment when needed. When a sufficient amount of culture is available (e.g., 1 L of 10^6^ cells/ml), cells are transferred via gentle centrifugation (maximally 300 g at 37°C) to V3 medium without yeast extract (V3^⊖^). Here, cells are incubated for 16 h in order to consume unlabeled amino acids remaining in the cells and in the base medium V3^⊖^. After this depletion time, an amino acid mixture with the appropriate isotope labeling pattern is added instead of 3 g of yeast extract, and protein production is induced. Depending on the desired isotope labeling pattern, the amino acid mixture can be either made up of pure amino acids or a labeled amino acid extract of microbial source, like yeast, algae or bacteria.

Uniform

15N labeling

2.3.2

The arguably most popular isotope labeling pattern for NMR studies is uniform ^15^N‐labeling. To this end, V3^⊖^ medium was supplemented with 5 g of u‐^15^N‐labeled yeast extract (autolysates, Cortecnet or Silantes). 5 g were determined to be the best compromise between high incorporation and low price. Except for the slightly higher amino acid content (5 g vs. 3 g of yeast extract), this medium exactly reproduces unlabeled V3 medium, in which the cells were grown, and thus protein yields and growth rates were indistinguishable from non‐labeled cultures (see Figure 3b). In Figure 3a, a spectrum of uniformly ^15^N‐labeled mEGFP (monomeric enhanced green fluorescent protein; Cormack et al., 1996; Zacharias et al., 2002) produced in this medium is shown. Overall, 70% incorporation was achieved, as determined by filtered and edited NMR spectra, and all amino acids were labeled, as evident from a comparison with deposited chemical shifts for a similar GFP construct (BMRB entry 5666).

Approaches for uniform isotope labeling in mammalian cells in suspension. (a) Microscope images of fluorescent cells from full medium and labeling medium. (b) Yields of mEGFP in mg/L from different media formulations are shown: Full V3 medium (V3, red), V3⊖ with yeast extract added back (V3⊖ + YE, orange), V3⊖ with algal extract (Celtone) added (V3⊖ + AE, dark green), V3⊖ with delipidated (DL) algal extract added (V3⊖ + AEDL, bright green) and FreeStyle medium (blue). The yields of mEGFP are in the mg range because it was expressed as a fusion to β1AR and released by protease cleavage. (c–e) 2D TROSY‐HSQC spectra of u‐15N labeled mEGFP produced in stable HEK293 GnTI− cells either with (c) 15N labeled yeast extract or (d) delipidated 13C,15N‐labeled algal extract (ISOGRO, Sigma). All amino acid types were labeled, since identical signals are visible as reported from E. coli expression. Incorporation was 70%, as determined by mass spectrometry. (e and f) [15N,1H]‐ and [13C,1H]‐HSQC spectra of ILV‐labeled mEGFP produced with E. coli extracts with the same labeling pattern.

This approach is very economic, since the isotope costs only amount to 5 g of labeled yeast extract. When compared to the only commercially available medium (BIOEXPRESS‐6000, CIL) with a price point >6000 EUR/L, the savings are more than 10‐fold. Further, the commercial medium is only available for adherent culture, such that there are additional savings in workload, as 1 L of easy‐to‐handle suspension culture corresponds to about one hundred plates of adherent culture. It should also be noted, that insect cells, the other popular eukaryotic expression system, have much higher requirements for amino acids and typically 10 g/L of labeled extracts are employed (Franke et al., 2018; Sitarska et al., 2015).

An alternative source of isotope labeled amino acids are algal extracts, which presently are commercially available with more diverse isotope labeling patterns than yeast extracts. We tested the products ISOGRO (Merck) and Celtone (CIL). Since algal extracts have a higher amino acid content than yeast extract (~60% vs. ~35%, respectively), only 4 g/L of algal extracts were employed. However, algal extracts severely impaired cell growth and resulted in nearly 90% reduced protein yields. Yields could be restored to about 50%–60% by de‐lipidation of the extract, and the same isotope incorporation level of 70% as for the yeast extract could be achieved (Figure 3c and Data S2: Protocol 4) (Barbieri et al., 2016). De‐lipidated algal extracts are therefore an alternative to yeast extracts, enabling more economic access to ^2^H and ^13^C isotope labeling. It should be noted here that 70% incorporation are sufficient for 2D spectroscopy and 3D NOESY experiments, but for 3D triple resonance experiments the efficiency of magnetization transfer will drop to 50% such that triple resonance experiments are only possible in selected cases (Sitarska et al., 2015).

Up to this point, only commercially available extracts of isotope labeled amino acids were considered, due to the ease of accessibility and the relatively little work involved in the laboratory to prepare such media (Bligh & Dyer, 1959; White, 1988; White et al., 1979). Preparing isotope labeled amino acid extracts on site is laborious, but potentially allows accessing alternative labeling patterns, for example ILV‐labeling (Hansen et al., 1992; Linser et al., 2014). To this end, we grew E. coli cells in uniform ^15^N‐labeled form with selective methyl ^13^C‐labeling of Ile, Leu and Val methyl groups. In contrast to the original ILV‐labeling protocol (Rosen et al., 1996), the precursors α‐keto‐isovalerate and α‐keto‐butyrate were added from the beginning of the culture and at higher concentrations (180 and 160 mg/L, respectively), in order to fully label all E. coli cells. The lysed cells were digested with different proteases—enzymatic digestion is preferred over chemical hydrolysis, because in this way sensitive amino acids (Asn, Gln, Cys, Trp) are conserved in the process. After performing the same de‐lipidation protocol as for algal extracts, we were able to employ this E. coli extract in mammalian growth media (Figure 3e,f and Data S1: Protocol 3). The GFP protein produced in this way in mammalian cells, faithfully reproduced the labeling pattern from E. coli, albeit with lower incorporation (~66%), such that with this protocol we were capable of producing ILV‐^13^C‐methyl labeling in mammalian cells.

In summary, several alternatives for economic amino acid mixtures exist, which can be employed depending on the desired isotope labeling pattern. For uniform ^15^N labeling, yeast extract is preferred because nearly two‐fold higher yields can presently be obtained than with other extracts. For uniform ^13^C labeling, algal extracts are fundamentally more economic, since algae can grow on ^13^CO_2_ and don't require pricy ^13^C‐labeled glucose, and, finally, for more advanced labeling patterns E. coli is currently the best option.

Amino acid‐type specific labeling by replacement

2.3.3

With uniform labeling full coverage of a protein can be obtained, however, especially for larger proteins, spectra tend to become very crowded. In such cases, only labeling one or two amino acid types is a more promising alternative. The generalized approach for amino acid type specific labeling, is using V3^⊖^ medium and replacing the yeast extract with an amino acid mixture reflecting the desired isotope labeling pattern. Instead of yeast extract, the isotope labeled amino acid(s) and a mixture of the remaining amino acids is added, each at the corresponding amount present in 3 g of yeast extract (Table S1 and Figure S2b). The amount of the isotope labeled amino acid is adjusted such that it has a final concentration of at least 3‐fold the concentration of the unlabeled amino acid in the base medium before starvation.

This approach is exemplified with ^15^N‐valine labeling of mEGFP, where 150 mg/L of ^15^N‐Val was employed in the context of V3^⊖^ medium supplemented with the mixture of unlabeled amino acids shown in the Data S2 and Figure 2b. The resulting simplified spectrum is shown in Figure 4, where all valine resonances are visible. Interestingly, a small fraction of isotope scrambling is visible, indicating residual metabolic activity of mammalian cells. This indicates that potentially isotope labeling starting from simpler precursors may also be possible in these higher cells. This avenue was explored further below.

Amino acid type‐specific labeling using the 15N, 13C, and 2H isotope labeled amino acids indicated on the right in each spectrum. 2D [15N,1H] and [13C,1H] correlation spectra of mEGFP produced in stably transfected HEK293 GnTI− cells grown in suspension are shown. See text and Data S1 for exact experimental conditions.

For larger proteins however, [^13^C,^1^H] correlation spectra of methyl groups have proven to be the most sensitive approach (Rößler et al., 2020; Schütz & Sprangers, 2020; Tugarinov & Kay, 2004). Therefore, ^13^C‐labeling of methyl groups of Ile, Leu, Val and Met is highly important in this context. Here we demonstrate ^13^C‐methyl labeling of all these amino acids and additionally of alanine and phenylalanine. Alanine can be important as a more sensitive alternative to ^15^N‐labeling, as it can also be considered a probe of the protein backbone. The named amino acids can all be labeled with our general protocol. First, amino acid specific labeling of Val and Leu will be shown, and then Ala, Met and Ile will be used to illustrate special cases, like suppression of unwanted metabolism, labeling in full media, and spectroscopic techniques to obtain high resolution spectra even if selectively methyl‐labeled amino acids are not available.

Methyl group labeling of Val and Leu was achieved using 150 mg/L of ^13^C^γ^‐valine and 230 mg/L of ^13^C^δ^‐leucine (Tracer Technologies, Inc.), and ^13^C‐incorporation of 61% and 71% was achieved, respectively, as determined by mass spectrometry.

In E. coli for isotope labeling of Ile, Leu, and Val, an approach based on more economic precursors was established. We therefore sought a similar strategy for using amino acid precursors in mammalian cells, which would make methyl‐specific ^13^C labeling and side chain deuteration more accessible. In viability tests it was determined that for Val, Leu, and Ile, the precursors α‐keto‐isovalerate, α‐keto‐isocaproate, and α‐keto‐β‐methyl‐n‐valerate, can rescue the respective missing amino acid in HEK cells, and subsequently isotope labeling experiments were performed (Lichtenecker et al., 2004; Lichtenecker et al., 2013; Rosati et al., 2024; Rosen et al., 1996). Of these precursors, methyl labeled, otherwise deuterated variants are available for α‐keto‐isovalerate, and α‐keto‐isocaproate, thus allowing local deuteration in a relatively inexpensive manner. As shown in Figure 4, specific deuteration leads to markedly sharper signals for Val and Leu methyl groups (Dubey et al., 2021). However, here, some scrambling occurs: α‐keto‐isovalerate is also incorporated into leucine residues, albeit at a much lower degree than into valine (Figure 6b2). This can be prevented by adding an excess of unlabeled leucine, which shifts the metabolic equilibria towards valine. Employing 160 mg/L of ^2^H,^13^C α‐keto‐isovalerate (L03, MagLab) and 460 mg/L of unlabeled leucine resulted in essentially scrambling‐free spectra. This type of scrambling is known from E. coli and there, the causative enzyme can be inhibited by small amounts of leucine (20 mg/L). (Mas et al., 2013; Miyanoiri et al., 2013) Since in the presented labeling protocol leucine is already present at 115 mg/L, there rather seems to be an underlying equilibrium and the biochemical pathway is unclear. For labeling of leucine, 160 mg/L of ^2^H,^13^C α‐keto‐isocaproate (L05, MagLab) were employed and no significant scrambling was observed. The Ile precursor α‐keto‐β‐methyl‐n‐valerate was only available with a uniform ^13^C labeling pattern at the time of performing the experiments (F.‐X. Theillet, personal communication). Using 80 mg/L yielded a clean methyl spectrum, albeit with the additional complication of homonuclear ^13^C coupling (see Figure 6c for a remedy against peak splitting in this situation). Also here, V3^⊖^ medium was used as the base of the medium and unlabeled amino acids missing from the omitted yeast extracts were compensated using an amino acid mix. In Figure 5, spectra of ^13^C‐methyl labeled mEGFP are summarized.

Labeling of methyl groups of Val, Leu, and Ile using 13C and 2H isotope labeled precursors indicated on the right in each spectrum. 2D [13C,1H] correlation spectra of mEGFP produced in stably transfected HEK293 GnTI− cells are shown. (c) For Ile, only the uniformly 13C labeled precursor was available, the precursor in gray represents a proposed labeling pattern for similar spectra as in (a) and (b). The strong signal (marked with * in a) stems from detergent present in the buffer. See text and Data S1 for exact experimental conditions.

Special case I: Amino acid type specific labeling in full medium with amino acid in excess or by prior enzymatic depletion

2.3.4

In principle, isotope labeling can be achieved using full medium by adding the desired amino acid in 10‐fold excess over the amount of unlabeled amino acid in the medium. For some amino acid types, where the isotope labeled form is available at comparatively low cost, like ^13^C^ε^ methionine, this approach can be economic, since no custom medium is required. Addition of 500 mg/L of ^13^C^ε^ methionine to V3 medium prior to induction, yielded an isotope incorporation of 94%, as determined by mass spectrometry.

A variant of this approach, is using the enzyme methionine‐γ‐lyase (MGL) to degrade unlabeled methionine in the full medium—which in this case can be any medium formulation (Klopp et al., 2018). To this end, 2 mg/L MGL must be added to the medium, as well as the required co‐factor pyridoxalphosphate (0.1 mM) and, depending on the media, additional buffering might be required. Thanks to volatile products of the reaction, this enzymatic reaction runs to completeness and after 48 h at 37°C the medium is devoid of any unlabeled methionine. Before addition of isotope labeled methionine, the MGL enzyme must be inhibited with 1 mM propargylglycine, which is well‐tolerated by the cells. To demonstrate methionine labeling in other media with unknown composition, we performed a labeling experiment in HEK293 Freestyle medium (GIBCO Invitrogen), after treatment with MGL (Figure 6a).

Special cases. (a) Methionine labeling in full media by enzymatic removal of intrinsic methionine from a medium with unknown composition, here HEK293 Freestyle medium (GIBCO). On the left, the reaction of the enzyme is shown (for reaction conditions see main text). Note that methanethiol is volatile and thus is continuously removed from the reaction resulting in complete turnover of methionine. In the middle, the methionine signal in the medium is shown at 0, 24 and 48 h after treatment with MGL. To monitor the reaction, the medium was spiked with 1 g/L 13C‐Met and 13C‐edited 1D 1H spectra were recorded. On the right, a 2D [13C,1H] HMQC spectrum of 13C‐Met labeled mEGFP is shown, produced from enzymatically methionine‐free HEK293 Freestyle medium (GIBCO). (b1) Isotope labeling of alanine with inhibition of alanine transaminase to suppress unwanted isotope dilution from pyruvate. On the left, the reaction of alanine transaminase is shown. Note that unlabeled pyruvate is present in rather high concentrations in the cell, thus driving the reaction towards the product alanine. Addition of l‐cycloserine (red) inhibits this reaction. On the right, a 2D [13C,1H]‐XL‐ALSOFAST‐HMQC spectrum of 2Hα,13Cβ‐Ala labeled mEGFP is shown, which was produced in presence of 50 mg/L of l‐cycloserine. (b2) Scrambling can also be observed for other amino acids, like in the example of valine labeling of mEGFP shown here. With the standard protocol, scrambling of the labeled valine precursor to leucine is apparent. Addition of an excess of unlabeled leucine (460 mg/L instead of 115 mg/L) changes the balance of the equilibrium reaction and suppresses scrambling. Note that the spectrum on the right hand side (same type as in b1) is plotted at very low contour levels, and it was recorded in detergent‐containing buffer, leading to the strong signal indicated by the dashed horizontal line. (c) When selectively 13C methyl labeled amino acids are not available, suppression of 1 J CC couplings in u‐13C labeled amino acids yields sharp signals. On the top left, a schematic representation of isoleucine is shown and on the right, the Mz‐excitation profile of an adapted BADCOP1 pulse is shown. The pulse was time‐reversed and the length adjusted in order to selectively invert β and γ1 nuclei, resulting in decoupling of the adjacent γ2 and δ1 nuclei, respectively. (BADCOP1_TR, pulse length: 1700 μs, offset: 16 ppm, simulated in Topspin shapetool). Typical chemical shifts of carbon nuclei are indicated with colored bands (BMRB; Ulrich et al., 2007). Below 2D [13C,1H]‐XL‐ALSOFAST‐HMQC spectra are plotted, showing the methyl region of 13C‐Ile labeled mEGFP, without (left) and with (right) application of a BADCOP1_TR 180° pulse. A slice through one representative signal is shown in the top right corner of each spectrum and illustrates the suppression of the homonuclear couplings as well as the concurrent ~2‐fold gain in sensitivity. Spectra were recorded on Bruker Avance IIIHD 800 MHz (a) and 900 MHz (b, c) spectrometers equipped with TCI CryoProbes.

Special case II: Metabolism of human cells leading to isotope dilution

2.3.5

The amino acid metabolism of human cells has only low activity. One example shown above was the unwanted incorporation of α‐keto‐isovalerate into leucine, which could be controlled by shifting the metabolic equilibrium by adding an excess of unlabeled leucine. Another metabolic activity often observed in cell culture is the Cahill cycle (Felig, 1973), in which cells produce large amounts of unlabeled alanine from pyruvate, a highly abundant metabolite stemming from glycolysis (Figure 6b). In order to obtain sufficiently high incorporation of, for example, ^13^C‐Ala, the enzyme alanine‐transaminase, which is responsible for this reaction must be inhibited, which is easily achieved by adding 50 mg/L of l‐cycloserine to the medium (Skora et al., 2015; Wong et al., 1973). In our HEK293 cultures in V3 medium, we did not observe the detrimental levels of alanine production as for example previously in insect cells, but as a precaution, cycloserine was included in alanine labeling experiments, since it had no apparent negative effect on protein yields.

With this procedure a labeling efficiency of 87% could be achieved for ^2^H^α^, ^13^C^β^‐Ala, when employing 250 mg/L of labeled alanine (Figure 6b).

Special case III: Spectroscopic suppression

of

1J CC couplings for u‐ 13C‐labeled amino acids

2.3.6

For recording NMR spectra of methyl groups of proteins, amino acids with only the methyl carbon ^13^C‐labeled are highly preferred over uniform ^13^C‐labeled amino acids, because otherwise ^1^ J CC couplings from the neighboring ^13^C nucleus lead to splitting of the signals and thereby limit the resolution to 120–150 Hz in the carbon dimension. To avoid split signals, either the approach using selectively ^13^C‐methyl labeled E. coli extracts can be used or amino acids with this labeling pattern need to be purchased. Currently only selected amino acids like Ala, Met, Val and Leu are available with specific ^13^C‐labeling of methyl groups from vendors and some only at high expense (Dubey et al., 2021). In contrast all amino acids can be commercially acquired in uniformly ^13^C‐labeled form (u‐^13^C).

Here, modified spectroscopic techniques set the path for obtaining spectra with high resolution from uniformly ^13^C labeled amino acids. The recently introduced BADCOP pulses based on optimal control, are highly band selective and therefore allow differential treatment of methyl group carbons and their adjacent carbon (Coote et al., 2018). Time‐reversal of the BADCOP1 pulse and adjustment of its duration allowed selective decoupling of the γ2 and δ1 methyl groups in Ile from β and γ1 carbons, respectively, yielding high resolution spectra (Figure 6c). In this specific case, spectra were recorded at 900 MHz using a BADCOP1_TR pulse of 1700 μs duration and 16 ppm offset. The protein was isotope labeled employing 250 mg/L of ^13^C^6^ Ile and 73% incorporation was achieved.

Choice of labeling pattern: Spectral quality and coverage

2.3.7

The methods developed here enable studies of biologically and pharmaceutically highly relevant proteins that were previously not amenable to NMR. In Figure 8, a number of examples from our projects are compiled, namely the GPCRs bovine rhodopsin, human β_1_ and β_2_ adrenergic receptors and the transcription factor ERRγ. For all proteins, expression in HEK cells was rather straightforward, as determined in small scale expression test. However, we first wanted to determine the most suitable isotope pattern for these relatively large and typically flexible proteins. To this end, we used the stabilized β_1_ adrenergic receptor from turkey (mgβ_1_AR; Warne et al., 2008) as a stepping stone to explore different isotope labeling patterns (^13^C^β^‐Ala, ^13^C^γ^‐Val,^13^C^δ^‐Leu, u‐^13^C‐Ile,^13^C^ε^‐Met). From the spectra shown in Figure 7, average signal intensities could be extracted, which established the following ranking for signal intensity: ^13^C^ε^‐Met >^13^C^δ1^‐Ile >^13^C^δ^‐Leu ≈^13^C^γ^‐Val >^13^C^β^‐Ala. Additionally evaluating signal overlap, it is evident that^13^C^ε^‐Met yields spectra of highest quality. When choosing the labeling pattern, however, coverage of important sites of a protein needs to be considered as well. In this respect, methionine yields the poorest coverage, as it is the least abundant of all methyl‐bearing amino acids. Depending on the amino acid composition of the target protein, or the site of interest, other amino acids might be more suitable. Another aspect regarding coverage, is whether side chain or backbone reporters are wanted. For the latter,^13^C^β^‐Ala is ideally suited, as the methyl group is firmly attached to the backbone and has no rotational freedom with respect to backbone atoms. In fact, ^13^C^β^‐Ala labeled receptor yielded high quality spectra as well, but overlap was already considerable. Modern algorithms capable of resolving overlap might therefore render Ala labeling an interesting alternative to amide labeling for monitoring backbone structure and dynamics.

Choice of isotope labeling patterns based on coverage of the target protein (left) and quality of NMR spectra (right). On the left, ribbon representation of the GPCR β1AR of turkey (mgβ1AR) is shown with amino acid side chains acting as small light sources illuminating parts of the protein. The illuminated parts should approximately represent the regions of the protein on which the labeled nuclei can report, as their resonance is influenced by effects in their immediate surroundings. On the right, [13C,1H]‐XL‐ALSOFAST‐HMQC spectra of mgβ1AR are shown, with 13C methyl group labeling of Met, Leu, Ile, Val, Ala and ILV as indicated in the spectrum. All spectra were recorded at 298 K on Bruker Avance IIIHD 900 MHz and NEO 1.2 GHz (bottom spectrum) spectrometers equipped with TCI CryoProbes.

For the receptors at hand, ^13^C‐methyl labeling of methionine seemed to be the most viable option, since methionine residues are quite evenly spread over the protein, and ^13^C‐labeled methyl groups of Val, Leu and Ile yielded spectra with strong overlap of exchange broadened signals, emphasizing the challenges that these highly flexible family of receptors pose to NMR studies.

Enabling aspects of isotope labeling in mammalian cells

2.3.8

For all the following examples we used ^13^C^ε^‐Met labeling, because of the superior spectral quality. First, we present bovine rhodopsin, for which expression in HEK293 cells was previously established for the constitutively active variant (N2C, M257Y, D282C) (Chatterjee et al., 2015; Deupi et al., 2012; Kubatova et al., 2020; Tsai et al., 2018; Weiss et al., 1994), and^13^C^ε^‐Met isotope labeling was thus straightforward. Commercially available 9‐cis‐retinal was used as surrogate for the natural ligand 11‐cis‐retinal. The obtained protein preparation in DM micelles was tested by UV–visible spectroscopy to test for light contamination and to assess the quality of the sample. Functional integrity of the produced protein was demonstrated by recording spectra of the dark state of rhodopsin and of the same protein after illumination. Essentially all of the visible resonances of the receptor shift upon illumination, reflecting the overall conformational change of rhodopsin, when changing from an inactive state to an activated state. Rhodopsin is exemplary for a multitude of proteins, which require a higher expression host, and where production in mammalian cells is already established. Thanks to the above presented methods such proteins are now readily amenable to NMR studies.

Pharmaceutically highly relevant targets like GPCRs can be studied by NMR (Eddy et al., 2018; Isogai et al., 2016; Kofuku et al., 2012; Shimada et al., 2019). However, establishing a single receptor for NMR characterization is already a magnificent task and therefore even large NMR laboratories typically work on just one specific type of receptor. Since HEK293 cells are highly competent at expressing receptors, we were able to not only produce the above‐mentioned rhodopsin in labeled form, but also the pharmacologically relevant human β_1_ and β_2_ adrenergic receptors, which are the target of some of the most prescribed drugs.

For producing human β_1_ and β_2_ adrenergic receptors (hsβ_1_AR and hsβ_2_AR, respectively) we therefore focused on^13^C^ε^‐Met labeling. For hsβ_1_AR we transferred the mutations found to stabilize mgβ_1_AR (Isogai et al., 2016; Rößler et al., 2025) to the human counterpart and were able to obtain NMR spectra of higher quality than with wild type protein. The construct can be activated by agonists and binds to G‐protein surrogates (Rößler et al., 2025). Our laboratory was able to produce these different receptors with relative ease. We attribute this to expression in mammalian cells. Since currently HEK293 cells are probably the most competent expression host for human proteins, the first step in protein production—expression of correctly folded protein—is solved, and one can focus on optimization of subsequent steps.

A further example of application of labeling in mammalian cells enabling medicinal research is the estrogen‐related receptor γ (ERRγ) (Wang et al., 2006): This difficult‐to‐produce protein was labeled in mammalian cells and could be used to test ligands discovered in biochemical assays for binding. In Figure 8 a number of successful ligand binding experiments are shown, where the ligands induced chemical shift perturbations, which are indicative of binding.

Examples of NMR studies enabled by mammalian isotope labeling. (a) Overlay of [13C,1H]‐XL‐ALSOFAST‐HMQC spectra of bovine rhodopsin in DM micelles (0.2%) in its dark (blue) and light‐activated state (red). NMR spectra of the stable dark state protein were recorded in 3 mm NMR tubes, which were protected from light by coverage in electric heat shrink. Light state spectra were obtained after 1 min illumination of the sample with a flash light of a mobile phone. The large chemical shift changes affecting essentially all resonances reflect the global conformational change of the receptor in response to the isomerization of 9‐cis to all‐trans retinal. (b) Overlay of [13C,1H]‐HMQC spectra of estrogen related receptor γ (ERRγ) in absence (black contour lines) and presence of different ligands (contour lines of different colors). These simple spectra show binding of all compounds except for the one giving rise to the blue spectrum on the left‐most spectrum. (c) Overlay of [13C,1H]‐XL‐ALSOFAST‐HMQC spectra of stabilized human β1 adrenergic receptor (hsβ1AR) in LMNG/CHS (0.01/0.001%) micelles without (blue) and with (green) the agonist isoprenaline bound. Also here, chemical shift changes reflect the change in equilibrium towards an active conformation upon agonist binding. (d) [13C,1H]‐XL‐ALSOFAST‐HMQC spectrum for human β2 adrenergic receptor (hsβ2AR) in MSPΔH5 nanodiscs (PO/PG) in its resting state (orange) and in presence of 1 mM Formoterol (violet). Spectra were recorded on Bruker Avance IIIHD 600 MHz (b) and 900 MHz (a, c, d) spectrometers equipped with TCI CryoProbes.

In summary, we show that with the ability of performing isotope labeling in mammalian cells, a large hurdle towards NMR studies of difficult‐to‐produce proteins is removed. Sufficient amounts of protein can be produced and isotope incorporation is high enough for functional studies and for characterizing molecular interactions at the atomic level.

METHODS

3

For better readability, the protocols for isotope labeling can be found in a separate file in the Data S2. Here, purification protocols for individual proteins are given.

Protein purification

3.1

Purification of hsβ_1_AR, hsβ_2_AR and mgβ_1_AR: The frozen cell pellet was resuspended in solubilization buffer (20 mM HEPES pH 7.5, 300 mM NaCl, 1 mM EDTA, 10% glycerol, 1 mM PMSF and 1% LMNG w/v, hsβ_1_AR, 2% DM w/v, mgβ_1_AR or 1% DDM w/v, hsβ_2_AR) and treated for 20 s with a TURRAX IKA T18 disperser equipped with a S18N‐19G dispersing tool. The homogenized solution was stirred at 4°C for 1 h and subsequently cleared by centrifugation at 185,000 g for 1 h. The supernatant was filtered using a 0.45 μm nitrocellulose membrane (Merck, MF‐Millipore) and subsequently loaded onto a 5 mL StrepTrap HP column (GE). The column was washed with 10 column volumes of washing buffer (20 mM HEPES pH 7.5, 300 mM NaCl, 0.01% LMNG (w/v) and 0.001% CHS (w/v), or 0.1% DDM (w/v) or 0.2% DM (w/v)) and afterwards protein was eluted with 4 column volumes of elution buffer (20 mM HEPES pH 7.5, 300 mM NaCl, 2.5 mM desthiobiotin, 0.01% LMNG (w/v) and 0.001% CHS (w/v), or 0.1% DDM (w/v) or 0.2% DM (w/v)). 3C HRV protease and 1 mM DTT were added to the eluted protein and incubated overnight at 4°C to cleave off the mEGFP‐tag. Afterwards, the protein was concentrated in a 50 kDa concentrator and further purified on a Superdex 200 column (GE) for size‐exclusion chromatography in SEC buffer (20 mM HEPES pH 7.5, 300 mM NaCl, 0.01% LMNG (w/v) and 0.001% CHS (w/v), or 0.1% DDM (w/v) or 0.2% DM (w/v)). For hsβ_1_AR and mgβ_1_AR, the receptor‐containing fractions were pooled and concentrated in a 50 kDa concentrator (Amicon Ultra, Merck MF‐Millipore) to a concentration above 50 μM and subsequently diluted with D_2_O and SEC buffer to 50 μM with 10% D_2_O. These samples were aliquoted in 100 μL fractions and flash frozen in liquid nitrogen. Aliquots were stored at −80°C prior to use. For hsβ_2_AR the SEC eluate was only concentrated to 700 μL and further transferred into nanodiscs. To this end 100 mM stock solutions of POPC and POPG powders in 200 mM sodium cholate were prepared by repeated shock‐freezing of the suspension in liquid nitrogen and subsequent thawing at RT, finally resulting in a clear solution. To 700 μL of hsβ_2_AR solution, 60 μL POPC and 40 μL POPG stock solution were added and the sample was incubated for 5 min on ice. Subsequently, the sample was supplemented with MSP1D1∆H5 and the volume was increased to 1 mL with SEC buffer to achieve final concentrations of 10 mM lipids (with a 3:2 ratio of POPC:POPG), 20 mM sodium cholate and 200 μM MSP1D1∆H5. This mixture was incubated overnight at 4°C. 800 mg Bio‐Beads (SM‐2 resin, Bio‐Rad) were added in 4 portions to the solution. After every addition the mixture was stored for 1.5 h on ice and mixed from time to time by inversion of the reaction tube. Bio‐Beads were separated from the solution in Mobicols, using filters with 35 μm pore size (MoBiTec), by centrifugation for 1 min at 1500×g. Removal of tags was achieved by the addition of 50 μL 3C HRV protease (in‐house production, His‐tagged, 1 mg) as well as 1 mM DTT and incubation overnight. Subsequently, the mixture was purified by SEC on a Superdex 200 Increase 10/300 column (GE) with SEC buffer (20 mM HEPES pH 7.5, 150 mM NaCl). Fractions with GPCR‐containing nanodiscs were pooled and concentrated in 50 kDa Amicon‐Ultra 0.5 mL centrifugal filters (Merck Millipore) to reach a final concentration of ∼60 μM. The sample was stored at 4°C until further usage.

Bovine rhodopsin and MSP1D1∆H5 were purified according to established protocols (Hagn et al., 2013; Tsai et al., 2018).

Mass spectrometry

3.2

The labeling efficiency of methyl group labeled samples was assessed by mass spectrometry at the Functional Genomics Center Zurich (FGCZ). Purified mEGFP obtained in isotope labeled fashion from HEK293 cultures was buffer exchanged to remove traces of detergent using a 3 kDa concentrator (Merck) and 10 μg of mEGFP was subsequently hydrolysed under 6 M HCl with 0.1% w/v phenol steam at 110°C for 24 h before being dried by SpeedVac. (Crabb et al., 1997) The free amino acids were derivatized (Cohen, 2003) using the AccQ‐Tag kit (Waters) following the recommendations of the supplier and using an internal standard (MSK‐A2, Cambridge Isotopes Laboratories). A Waters UPLC H‐Class Plus system with Acquity UPLC Quadrupole Solvent Manager and Sample Manager were used for the analysis. Detection was facilitated using a Waters QDa single quadrupole mass detector in positive mode operated by one MS and followed by selected ion monitoring for every individual amino acid in its dedicated retention window. The labeling efficiency was calculated from the ratio of the quantity of labeled amino acid, as identified by increased molecular weight, and total quantity of the given amino acid type in the analyzed sample. The data is shown in Table S1.

Isotope incorporation of uniform labeled samples was determined by integration and comparison of suitable regions in ^15^N/^13^C‐edited and ^15^N/^13^C‐filtered 1D ^1^H NMR spectra. For the pulse sequence see Data S1 (Methods).

DISCUSSION AND CONCLUSION

4

We present here a suite of protocols for isotope labeling in mammalian cells, which enables studying many aspects of difficult proteins by NMR.

The hurdles to establish a eukaryotic expression system in an NMR laboratory were high, regarding both cost and work. We therefore shed costs in every step and simplified protocols considerably, resulting in a simple setup (shaking plate for suspension culture in regular incubator) and affordable media, where costs were lowered by an order of magnitude.

We established a wide range of isotope labeling patterns suitable for studying for instance small proteins or intrinsically disordered regions of larger proteins (uniform ^15^N/^13^C labeling); for investigating individual aspects of proteins using simplified spectra (amino acid‐type selective ^2^H/^15^N labeling); and for characterizing large proteins (selective ^13^C/^2^H methyl group labeling). The label incorporation that can be achieved with these cost‐efficient protocols is in the order of 70–90% overall, where the labeling degree within individual amino acids is, however, typically 99%, such that for example in site specific deuteration the neighboring hydrogens will all be ^2^H labeled.

Thus, the methods presented here, enable a wide range of labeling patterns with stable isotopes in HEK293 cells, which arguably represent the most fitting expression system for human proteins. We demonstrate how these protocols enable NMR studies of proteins that are considered to be especially difficult to produce, for example different human membrane proteins and nuclear receptors.

We therefore hope that these methods signify a breakthrough allowing to fully exploit the high information content of NMR for biological and biomedical research.

AUTHOR CONTRIBUTIONS

Philip Rößler: Conceptualization; writing – original draft; methodology; writing – review and editing; formal analysis; investigation; visualization. Marco M. Ruckstuhl: Investigation; methodology; visualization; formal analysis. Arnelle Löbbert: Investigation; methodology; visualization; formal analysis. Timo Stühlinger: Investigation; methodology; formal analysis; visualization. Lucia R. Franchini: Investigation; methodology. Ching‐Ju Tsai: Conceptualization; supervision. Roman Lichtenecker: Investigation; methodology. Binesh Shrestha: Conceptualization. Simon H. Rüdisser: Investigation; methodology; supervision; formal analysis; software. Robert Konrat: Conceptualization; supervision; resources. Gebhard F. X. Schertler: Conceptualization; supervision; resources. Alvar D. Gossert: Conceptualization; funding acquisition; writing – original draft; writing – review and editing; visualization; methodology; supervision; resources; formal analysis.

CONFLICT OF INTEREST STATEMENT

Robert Konrat and Roman Lichtenecker are shareholders in MAG‐LAB, Vienna, Austria. Roman Lichtenecker is partly employed by the same company.

Supporting information

DATA S1. Supplementary Methods.

DATA S2. Supplementary Protocol.