Combining Site-Directed Spin Labeling in Vivo and In-Cell EPR Distance Determination

Combining Site-Directed Spin Labeling in Vivo and In-Cell EPR Distance Determination. ChemRxiv. Preprint. Structural studies on proteins directly in their native environment are required for a comprehensive understanding of their function. Electron paramagnetic resonance (EPR) spectroscopy and in particular double electron-electron resonance (DEER) distance determination are suited to investigate spin-labeled proteins directly in the cell. The combination of intracellular bioorthogonal labeling with in-cell DEER measurements does not require additional purification or delivery steps of spin-labeled protein to the cells. In this study, we express eGFP in E.coli and use copper-catalyzed azide-alkyne cycloaddition (CuAAC) for the site-directed spin labeling of the protein in vivo, followed by in-cell EPR distance determination. Inter-spin distance measurements of spin-labeled eGFP agree with in vitro measurements and calculations based on the rotamer library of the spin label. Abstract Structural studies on proteins directly in their native environment are required for a comprehensive understanding of their function. Electron paramagnetic resonance (EPR) spectroscopy and in particular double electron-electron resonance (DEER) distance determination are suited to investigate spin-labeled proteins directly in the cell. The combination of intracellular bioorthogonal labeling with in-cell DEER measurements does not require additional purification or delivery steps of spin-labeled protein to the cells. In this study, we express eGFP in E.coli and use copper-catalyzed azide-alkyne cycloaddition (CuAAC) for the site-directed spin labeling of the protein in vivo , followed by in-cell EPR distance determination. Inter-spin distance measurements of spin-labeled eGFP agree with in vitro measurements and calculations based on the rotamer library of the spin label.


Introduction
The understanding of protein function and structure is crucially linked to the ability to study proteins in their native environment. Effects of not only molecular crowding but also post-translational modifications, presence of a variety of specific or non-specific interaction partners or chaperones have a great impact on proteins, yet are only incompletely understood through in vitro studies. 1 Facing the challenge of highly complex cellular compositions, in-cell electron paramagnetic resonance (EPR) spectroscopy in combination with site-directed spin labeling (SDSL) 2 has emerged as a valuable tool to provide structural information 3 as many cellular components are diamagnetic and therefore EPR-silent. Pulsed techniques such as double electron-electron resonance (DEER) provide access to long-range distance restraints in the nanometer range by measuring the dipole-dipole interactions between paramagnetic spin labels. 4,5 So far, in vivo DEER studies on spin-labeled proteins have either been conducted on outer membrane proteins upon cysteine-based spin labeling on the cellular surface [6][7][8] or relied on the delivery of spin-labeled protein into the cell, e.g.
Non-canonical amino acid (ncAA) incorporation and bioorthogonal in vivo spin labeling offer a more direct and elegant approach for in-cell EPR studies as they may combine expression, labeling and the EPR study of the protein of interest directly inside the same cell without additional delivery steps. While a number of studies on bioorthogonal spin labeling have been published in recent years, [19][20][21][22][23][24][25][26][27] the corresponding DEER measurements were limited to in vitro measurements.
Even in cases with confirmed in vivo labeling, purification and concentration of spin-labeled protein was required prior to a DEER measurement. The Steinhoff group has advanced nitroxide spin labeling via copper-catalyzed [3+2] azidealkyne cycloaddition (CuAAC) and reported labeling of eGFP at one site between the ncAA N-ε-propargyl-L-lysine (PrK) and a nitroxide spin label in E. coli. However, for DEER distance determination, the protein was purified and conventional cysteine labeling with MTSSL was used to introduce the second spin label. 20 In our previous work; 26 we have identified the ncAA para-ethynyl-phenylalanine (pENF) as a suitable choice for bioorthogonal spin labeling with CuAAC. pENF was incorporated into E. coli thioredoxin with high labeling yields and exhibited favorable linker properties for DEER distance determination in in vitro measurements. 26 Here, we extend this approach to bioorthogonal double spin labeling and DEER measurements directly in vivo ( Figure   1). We report the incorporation of pENF at two sites of eGFP via amber stop codon suppression in E. coli, develop conditions for CuAAC with an azide-bearing nitroxide spin label and combine it with direct in-cell DEER distance determination without any purification step. In-cell inter-spin distances of spin-labeled eGFP are comparable to in vitro measurements and calculated distances based on the rotamer library of the label. Figure 1 Schematic overview of in vivo spin labeling approach via copper-catalyzed [3+2] azide-alkyne cycloaddition followed by in-cell EPR distance determination The ncAA pENF is incorporated site-specifically into eGFP using amber stop codon suppression. The eGFP-expressing E. coli cells are subjected to CuAAC-based spin labeling and subsequent, DEER distance determination is performed directly inside the cell.

Results and Discussion
We decided on eGFP as a model protein for our study since in vivo CuAAC-based spin labeling on this protein has previously been shown 20 and the fluorescence properties of eGFP facilitate in-cell monitoring of the protein. We chose position Y39TAG in the β2-β3 loop region and position L221TAG in the β11-strand of the barrel as labeling sites. 28 We co-transformed plasmids pBAD_GFP_Y39/L221-TAG and pEVOL_pCNF (encoding a polyspecific Methanocaldococcus jannaschii tRNA Tyr (CUA)/tyrosyl-tRNA-synthetase (YRS) pair evolved for the genetic encoding of para-cyano-L-phenylalanine, pCNF) 29, 30 into BL21-Gold (DE3) E. coli, induced the culture and purified Y39/L221pENF eGFP by Ni-NTA chromatography via its C-terminal His6-tag. ESI-MS data confirmed the correct incorporation of pENF at two positions ( Figure S1). In vitro labeling reactions were performed as previously described 26 with 1 mM CuSO4, 3 mM 2-(4-((bis((1-(tert-butyl)-1H-1,2,3-triazol-4-yl)methyl)amino)methyl)-1H-1,2,3-triazol-1-yl) acetic acid (BTTAA), 1 mM sodium ascorbate (NaAsc) and 1 mM 3-(azidomethyl)-2,2,5,5-tetramethyl-1-pyrrolidinyloxyl (az-proxyl) and yielded Y39/L221pENF-L eGFP with a labeling efficiency of approximately 70 % based on the ratio of spin concentration to protein concentration. Incorporation of pENF and labeling did not significantly affect the function of eGFP as in vitro excitation and emission fluorescence spectra of Y39/L221pENF before and after labeling did not differ from eGFP wildtype spectra ( Figure S2). In vitro DEER measurements of spin-labeled Y39/L221pENF-L eGFP resulted in a narrow distance distribution (HWHM: 0.5 nm) with a mean peak at 2.3 nm ( Figure S3) and agree well with the calculated distances based on the crystal structure of eGFP (PDB structure 4EUL) and our previously published rotamer library 26 of the pENF-L spin label. In case of in vivo CuAAC, a major concern is copper-mediated cytotoxicity. Copper ions are linked to the formation of reactive oxygen species (ROS) 31 and are known to impair cytoplasmic proteins with Fe-S clusters due to their thiophilic tendency. 32 However, it has been shown that this toxicity is greatly reduced by adding chelators to the copper species with BTTAA being a ligand that shows especially promising properties for in vivo CuAAC applications. 33 To investigate the effects of in vivo CuAAC on E. coli, we monitored the bacterial growth immediately after CuAAC treatment as well as the long-term growth behavior with a plate sensitivity assay ( Figure 2). Cells were treated with the labeling reagents copper(II)sulfate, BTTAA and sodium ascorbate (in a 1:3:1 ratio), diluted, and the optical density of the cells at 600 nm for in-cell EPR distance determination were frozen in liquid nitrogen as part of the DEER sample preparation. We therefore expect a similar level of integrity as seen during FACS measurements in the respective DEER samples. For a qualitative description of the nitroxide reduction, the amplitude of the centerline from averaged CW spectra was plotted against the time. The signal intensity at t = 60 min was set to 100 %.
Next, we were interested in studying possible nitroxide reduction inside E. coli ( Figure 3). Proxyl-based spin labels as employed in this study are prone to fast biological reduction and it has been shown that they can be rendered EPRinactive on a minute timescale. 34  We limited the labeling times for the in vivo labeling approach to 40 minutes at room temperature, followed by additional washing steps to remove unbound nitroxide spin labels. In total, approximately 60 minutes passed between the addition of the nitroxide spin labeling reagent to E. coli cells and the start of the EPR measurement. Time-dependent EPR spectra are shown in Figure 3 A. For a qualitative description of the signal reduction, we plotted the amplitude of the nitroxide center field peak against the time (Figure 3 B). Upon performing the CuAAC-based spin labeling procedure, E. coli cells expressing Y39/L221pENF eGFP show a typical nitroxide EPR spectrum that prevails up to 120 minutes after the addition of azido-proxyl spin labeling reagent.
In principle, EPR signals in E. coli can stem from several possible species. Besides spin-labeled Y39/L221pENF-L eGFP, free spin label from the labeling reaction, as well as labeled pENF, both in free form and esterified to tRNA Tyr (CUA) might contribute to the signal. To limit off-target labeling, protein-expressing cells were transferred to fresh LB medium without pENF and protein expression was prolonged for 90 minutes prior to CuAAC labeling. In addition, approximately 9 % of E. coli proteins are terminated with an amber stop codon (TAG) 36 and pENF might also eventually be incorporated into these proteins. To investigate potential sources of signal, we co-transformed E. coli cells with plasmids for eGFP wild-type and the tRNA Tyr (CUA)/YRS pair and expressed eGFP wild-type in the presence of pENF in the medium. In this scenario, the EPR signal after CuAAC labeling procedure can only arise from free or tRNA-esterified spin-labeled pENF or from labeled pENF incorporated at amber sites of off-target proteins. However, we did only observe neglectable signal intensities in this sample (Figure 3 A, black). In addition, we tested our washing protocol after CuAAC-based spin labeling with E. coli cells treated with CuAAC labeling reagents and azido-proxyl in the absence of pENF and found that free spin-label is effectively removed from the bacterial solution ( Figure S7). We, therefore, assumed that the main contribution to the signal stems from spin-labeled Y39/L221pENF-L eGFP. Moreover, the spectral shape of the positive sample was similar to in vitro EPR spectra of in vitro labeled Y39/L221pENF-L eGFP ( Figure S8).
Echo-detected field sweeps of E. coli samples after CuAAC spin labeling contained spectral contributions from Cu(II) species as a result of the incomplete removal of the catalyst after labeling ( Figure S9). The phase memory time of the spins was reduced to 0.66 µs for cellular samples compared to 2.44 µs from the in vitro experiment in deuterated aqueous solution ( Figure S10). In conclusion, we have advanced CuAAC-based spin labeling of alkyne-bearing proteins and direct DEER distance measurements to in vivo environments. Our labeling conditions do not exhibit copper-mediated cytotoxicity in E. coli.
We observe moderate nitroxide reduction during labeling, allowing for double protein labeling with sufficient yields for subsequent DEER measurements. Moreover, we observe minimal background signal from nitroxides not attached to our target protein positions. This enables DEER studies that deliver valuable information on the structure of the labeled protein, and argue for a similar conformation that eGFP adopts in the E. coli cytoplasm and in vitro.
Taken together, our approach combines natural translation, folding, and processing of a target protein with bioorthogonal double labeling and direct DEER distance measurements directly in the natural environment of a bacterial cell. It overcomes the necessity for introducing spin-labeled proteins into cells, e.g. via electroporation or hypo-osmotic shock and thus represents a new access point to in-cell EPR studies of protein structure and function.

Chemicals
All chemicals were obtained from Carl Roth or Sigma Aldrich unless stated otherwise. BTTAA were purchased from Jena Bioscience and pENF from Achemblock. Azido-proxyl was synthesized according to published procedure. 1 (2000 g, 2 min 4°C, 4x 1 mL washing solution). After discarding the supernatant, cell pellets were either subjected to room temperature cw EPR spectroscopy or shock-frozen in liquid nitrogen for DEER distance determination. On average, 60 minutes passed between exposure of the nitroxide labeling reagent to the cells and shock-freezing of the sample or start of the cw EPR experiment, respectively.

X-band cw EPR
Room-temperature cw EPR spectra were recorded at 20°C and with an X-band spectrometer (EMX-Nano, Bruker with a cylindric cavity mode TM1110). Typically, 40 µL sample volume was filled into a In vivo measurements were either averaged over 5 scans for nitroxide reduction kinetics or consecutive scans were added up until the highest possible signal-to-noise ratio was achieved ( Figure   S8). All recorded spectra were processed using MatLab2018A (the MatWorks, Inc.).

Q-band pulsed EPR
Pulsed EPR experiments were performed at Q-band (34 GHz) frequency at 50 K with a shot repetition time of 4 ms to avoid nitroxide saturation. Echo signals were detected with an integrator gate width corresponding to the respective π pulse length and a video-bandwidth of 20 MHz.

Phase memory relaxation
The phase memory time Tm was determined by increasing the interpulse delay of a Hahn echo sequence (starting with τ = 800 ns) and extracting the time point at which the signal intensity decreased to 1/e of the initial intensity at t = 0 µs.

Figure S6
Flow cytometry analysis to determine the copper toxicity. Events were pre-gated based on the forward and sideward scattering (top row). Bottom row plots GFP fluorescence intensity against PI intensity.

Figure S7
Removal of excess spin-label after CuAAC labeling reaction. E. coli cells expressing Y39/L221pENF eGFP (green) or WT eGFP in the absence of pENF (black) were subjected to the CuAAC labeling reaction and washing procedure. CW EPR spectra were recorded 60 minutes after nitroxide exposure to the cells and averaged over 5 scans to improve the signal-to-noise ratio. The black spectrum indicates that the unbound spin-label reagent az-proxyl is almost completely removed after the labeling and subsequent washing steps. .

Figure S8
Comparison of Y39/L221pENF-L eGFP in vitro (orange) and in vivo (green). Cw EPR spectra were recorded in X-band at room temperature and accumulated over 10 minutes (in vitro) or 50 minutes (in vivo).