In‐Cell Characterization of the Stable Tyrosyl Radical in E. coli Ribonucleotide Reductase Using Advanced EPR Spectroscopy

Abstract The E. coli ribonucleotide reductase (RNR), a paradigm for class Ia enzymes including human RNR, catalyzes the biosynthesis of DNA building blocks and requires a di‐iron tyrosyl radical (Y122 .) cofactor for activity. The knowledge on the in vitro Y122 . structure and its radical distribution within the β2 subunit has accumulated over the years; yet little information exists on the in vivo Y122 .. Here, we characterize this essential radical in whole cells. Multi‐frequency EPR and electron‐nuclear double resonance (ENDOR) demonstrate that the structure and electrostatic environment of Y122 . are identical under in vivo and in vitro conditions. Pulsed dipolar EPR experiments shed light on a distinct in vivo Y122 . per β2 distribution, supporting the key role of Y. concentrations in regulating RNR activity. Additionally, we spectroscopically verify the generation of an unnatural amino acid radical, F3Y122 ., in whole cells, providing a crucial step towards unique insights into the RNR catalysis under physiological conditions.


Wild-type protein expression, purification and preparation of EPR samples
E. coli BL21(DE3)-Gold (Invitrogen) were transformed with pTB-nrdB, which encodes for wt-β2, and plated on LB-agar plates with 100 µg/mL carbenicillin (Carb) at 37 °C. Positive clones were selected and a starter culture (5 mL) was grown overnight in LB-medium enriched with Carb at 37 °C until saturation. An intermediate culture (100 mL) was enriched with 1 mL of the starter culture, also grown overnight. The expression culture was grown with a 200-fold dilution of the intermediate culture in LB-medium containing Carb. At OD600 ~ 0.9, 1,10-Phenanthroline was added to a final concentration of 100 µM. After 20 min, protein expression was induced with 0.5 mM IPTG. Apo-b2 in E. coli was over-expressed because β2 endogenous expression level is far below the EPR detection limit [1] (see also Figure S2B). After 4 h of protein overproduction, the cells were harvested by gentle centrifugation. Although 7000 g for 20 min centrifugation was performed for E.coli whole-cell RNR experiments previously [1] , here the centrifugation is performed at 6000 g for 15 min at 4 °C in order to keep as many cells as possible intact and alive. A typical 2.5-3 g cell paste/L media was obtained, in great agreement with previously reported yield considering 1,10-Phenanthroline addition. [1] For in-vitro sample preparations, the apo-b2 construct was purified by anion-exchange chromatography and radical was generated as previously described. [2] Protein concentration was determined via UV-vis spectroscopy. 200 µL protein solution was transferred to EPR tubes and shock frozen in liquid N2. For incell sample preparations, the harvested cell pellet was washed several times and resuspended in 50 mM Tris pH 7.6 with 5% glycerol using 3 mL of buffer per g cell paste. 0.13 mM Fe II (NH4)2(SO4)2 (~5 equivalents of Fe II with respect to the estimated protein concentration) was added to this buffer. The added Fe concentration was only 13 % of that, which is shown to inhibit E. coli cell growth. [3] The suspension was allowed to sit on ice for 10 min and subsequently saturated with O2 gas on ice for 1 -2 min. 200 µL cell-suspension was directly transferred into EPR-tubes and frozen gently in an isopropanol rack at -80 °C to secure slow cooling and prevent cell damage. Spin concentrations of the samples are determined via 9.6 GHz cw-EPR measurements (see SI 2.3).

(2,3,5)F3Y122-β2 protein expression, purification and preparation of EPR samples
E. coli BL21(D3)-Gold cells (Invitrogen) were co-transformed with pBAD-nrdB122TAG and pEVOL-FnYRS-E3, and plated on LB-agar plates with 100 µg/mL carbenicillin (Carb) and 35 µg/mL chloramphenicol (Cm) at 37 °C. Positive clones were selected and a starter culture (5 mL) was grown overnight in LB-medium enriched with Carb and Cm at 37 °C until saturation. An intermediate culture (100 mL) was enriched with 1 mL of the starter culture, also grown overnight. The expression culture was grown with a 200-fold dilution of the intermediate culture in 2XYT medium containing Carb and Cm. At OD600 ~ 0.3, F3Y was added to a final concentration of 0.7 mM. After 30 min, 100 µM 1,10-phenanthroline was added to chelate iron. After further 30 min, protein expression was induced with 0.5% (w/v) L-arabinose. After 4 h protein overproduction, the cells were harvested by gentle centrifugation at 6000 g for 15 min at 4°C. For in-vitro sample preparations, the apo-(2,3,5)F3Y122 construct was purified by anion-exchange chromatography and radical was generated as previously described.
[4] Protein concentration was determined via UV-vis spectroscopy. 200 µL protein solution was transferred to EPR tubes and shock frozen in liquid N2. For in-cell sample preparations, the harvested cell pellet was washed several times and resuspended in 50 mM Tris pH 7.6 with 5% glycerol using 3 mL of buffer per g cell paste. 0.13 mM Fe II (NH4)2(SO4)2 (~5 equivalents of Fe II with respect to the estimated protein concentration) was added to this buffer. The BL21 cell-suspension was allowed to sit on ice for 5 min and afterwards saturated with O2 gas on ice for 1 -2 min. 200 µL of cell-suspension was directly transferred into EPR-tubes and frozen gently in an isopropanol rack at -80 °C to secure slow cooling and prevent cell damage.

Cell counting experiments
Prior to EPR experiments, four distinct cell suspensions were plated on LB-agar plates after 1:10 12 (1:10 11 for (2,3,5)F3Y122 construct) dilution to check the viability of the cells. The cell growth on plates with the E. coli cell suspensions that did not contain any overexpression plasmid were 3 -13 · 10 11 and 4 · 10 11 cells per mL cell suspension before and after iron addition, respectively. The cell growth on the plates with the E. coli cell suspension containing wt-b2 overexpression plasmid, which is used for EPR experiments, was 3 -60 · 10 11 cells per mL cell suspension, in great agreement with previously reported numbers. [1] The cell growth on the plates with the E. coli cell suspension containing (2,3,5)F3Y122 overexpression plasmid, which is used for EPR experiments, was at least 6 · 10 10 cells per mL cell suspension. These experiments proved that the cells used for EPR measurements were intact and alive. SI/4

Spectrometer specifications, sample concentrations and details of the EPR experiments
X-band: X-band cw-EPR measurements were carried out at T = 100 K using a Bruker EMX-Nano Benchtop spectrometer equipped with a continuous-flow nitrogen cryostat.
Q-band: Q-Band pulse EPR measurements were carried out at T = 10 K using a Bruker Elexsys E580 spectrometer equipped with a 150 W TWT amplifier, Bruker ER 5106QT-2 resonator, Bruker SpinJet AWG, Oxford Instruments CF935 continuous-flow helium cryostat and Oxford Instruments MercuryiTC temperature controller.
Orientation-selective DEER experiments were performed using the following dead-time free 4-pulse DEER pulse sequence: p/2obs-t1pobs-(t1+T)-ppump-(t2-T)-pobs-t2-echo with artifact-free 16-step phase cycling [5] and Gaussian pulses. [6] The frequency separation, Δf = fpump -fobs, was 84 MHz, and an overcoupled resonator with fpump set to the center of the resonator dip was used. The optimal ppulse lengths were determined using transient nutation experiments and were typically ~30 ns for the pump pulse and ~70 ns for the detection. DEER time traces were background-corrected by using an empirical second-order polynomial fitting, if not stated otherwise.
For orientation-averaging a recently reported procedure was used. [7] Each primary time trace was first normalized to the same signal intensity at zero time. Afterwards, these traces were normalized to the signal intensity at the pump position. Summation of the DEER traces led to the orientation-averaged time trace. Spin-concentrations: 240 µM (in vitro), 22 µM (in-cell).
ENDOR experiments were carried out at 10 K using a Bruker EN 5107D2 resonator and an AR 600 W radiofrequency (RF) amplifier (AR 600A225A). Orientation-selective 1 H Davies ENDOR spectra were recorded using the following microwave pulse sequence: p-Tp/2-t-p-t-echo. The RF pulse of variable frequency was applied during the time interval T and had a length of 17 µs. The first p pulse was a rectangular-shaped inversion pulse of 190 ns. Three consecutive measurements at field positions corresponding to g = 2.0094, 2.0059 and 2.0005 were performed for the in-cell and in vitro samples. For the sample of E. coli cells the ENDOR measurement was performed at g = 2.0094, with the microwave power optimized for a S = 5/2 species. Spin-concentrations: 18 µM (in vitro), 14 µM (incell).
W-band: W-band pulse EPR measurements were carried out at T = 20 K using the Bruker E680 spectrometer equipped with a Cryogenic Systems closed-cycle 6T magnet and a variable temperature insert (VTI) that allowed varying the temperature within the range 2 -300 K. Additional LakeShore 335 temperature controller equipped with a Cernox sensor was used to record the sample temperature. Due to a relatively high uncertainty in the external magnetic field values (B0), the field axis of the W-band EPR spectra was adjusted so that the Y122• spectral width matched the spin Hamiltonian parameters used for the X-and Q-band EPR/ENDOR simulations. This was achieved by compressing the x-axis of all field-swept W-band spectra by ~5%. Spin-concentrations: 18 µM (in vitro), 22 µM and 30 µM (in-cell).
Radical concentrations are detected via spin quantification experiments at X-band. An inherent error of at most 20 % should be taken into account. *: Samples were produced from different in vitro protein purification batches. a: Samples were produced from different cell growths, each starting from new LB-agar plates. All samples were prepared in the same way, except for the harvesting centrifugation step. In-cell 1 was centrifuged at 6000 g for 15 min at 4 °C (see SI 1.1), whereas in-cell 2,3 and 4 were centrifuged at 3000 g for 20 min at 4 °C.

SDS-PAGE gel electrophoresis and EPR analysis to determine possible protein leakage
In order to detect possible protein leakage that is more than the expected range for E. coli cells, experiments were performed with the supernatant of E. coli BL21(DE3)-Gold cell pellet collected after wt-β2 expression and Fe II and O2 treatment. Firstly, SDS-Page analysis was done by comparing the intensity of the collected supernatant with that of in vitro wt-β2 protein with known concentrations, namely 2 µM and 5 µM. The intensities of the loaded purified wt-β2 samples were clearly stronger than that of the collected supernatant (see Figure S1A). The comparison of the bands observed in SDS-PAGE gel revealed that the extracellular protein concentration of in-cell sample is < 2 µM. Secondly, the cw-EPR spectrum of the in-cell sample was compared to two supernatants collected from different incell samples. Comparison of the cw-EPR spectra demonstrated that the contribution of the extracellular proteins containing radicals is negligible. These results showed that the data collected and analyzed in the present study mostly belong to the radical that resides in wt-β2 in intact E. coli cells.

9.6 GHz EPR spectra of treated and untreated whole E. coli cells
Continuous wave (cw) EPR spectra of in-cell samples were recorded before and after treatment with Fe II and O2 ( Figure S2A). The cell pellet was resuspended in 50 mM Tris pH 7.6 with 5% glycerol using 3 mL of buffer per g cell paste. For the treated cells, the buffer contained ∼0.13 mM Fe II (NH4)2(SO4)2 additionally and the cell suspension was allowed to incubate on ice for 10 min and afterwards saturated with O2 gas on ice for 1 -2 min. In addition, cw-EPR spectra of E. coli cells without the wt-β2 expression plasmid pTB-nrdB were recorded in order to exclude any signal from the cells ( Figure S2B). EPR spectral comparison demonstrates that the treatment led to the generation of a radical species only in the presence of the wt-β2 expression plasmid pTB-nrdB, Fe II and O2. Figure S2. (A) Cw-EPR spectra of wt-β2 expressed in whole E. coli cells recorded before (orange) and after (black) Fe II and O2 treatment. Data were recorded at 9.6 GHz and 100 K with Bruker EMXnano. Experimental conditions: 31.6 mW power, 1.5 G modulation amplitude, 100 kHz modulation frequency, 5.12 ms as time constant, and 19.9 ms conversion time; 40 scans (orange) and 200 scans (black). (B) Cw-EPR spectra of E. coli cells grown without wt-b2 expression plasmid recorded at 9.6 GHz and 100 K. Before the EPR experiments, the cells were treated either with Fe II and O2 or only with O2. The same experimental conditions as in A); 30 scans (red and black). Signal arising from the resonator background is marked with an asterisk *. Table S3. 1 H Hyperfine coupling parameters used for spectral simulations of the Q-band ENDOR and multi-frequency EPR spectra of Y122• in whole E. coli cells, in combination with the g-tensor gx,y,z = 2.00915, 2.00460, 2.00225. [8] The Euler angles α, ß, and γ are defined within the EasySpin z,y',z'' convention. They refer to rotations from the g-tensor frame into the hyperfine tensor frames. Positive angles are clockwise rotations viewed along the rotation axis.

SUPPORTING INFORMATION
SI/9

Estimation of in-cell spin concentration
The continuous wave (cw) EPR signal intensity reports on the spin concentration in a sample. [9] Therefore, we performed cw-EPR experiments with in-cell and in vitro wt-β2 samples to estimate in-cell spin concentration. The protein and radical content of the in vitro sample were determined as 200 ± 50 µM and 1.2 Y•/protein dimer via UV-vis spectroscopy prior to EPR experiments. Based on these data, we calculated the Y122• concentration as 240 ± 60 µM in the in vitro sample. Subsequently, intensities of EPR spectra recorded with five in-cell samples prepared from different growths and/or distinct overexpression levels were compared to that of the in vitro sample. This comparison demonstrated that the in-cell radical concentration was ca. 18 µM on average with a maximum variation between batches of ± 6 µM, approximately an order of magnitude lower than that of the 240 µM in vitro sample (blue trace in Figure  S3). It should be noted that 18 µM is the average bulk concentration in the EPR tube which contains 200 µL of the in-cell sample. In order to estimate the spin concentration within the cells, we calculated the intracellular aqueous volume of BL21(DE3) cells in the EPR tube either as 127 or as 167 µL based on two different methods. [10] As the protein leakage out of the cells was less than 2 µM (see SI 2.1.), the spin concentration within the cells was estimated approximately as 25 ± 4 µM. We note that the highest spin concentration we detected in our in-cell samples with different growth conditions was approximately 22 µM (darkest grey line in Figure S3). This sample has been used for multi-frequency EPR and DEER experiments (named in-cell 1). According to the calculation mentioned above, the intracellular Y122• concentrations of in-cell 1 is 32 ± 6 µM. The sample used for ENDOR experiments (named in-cell 3) showed an intracellular Y122• concentration of 20 ± 4 µM. SI/10

Hyperfine coupling pattern of Y122• shown along with its 94 GHz EPR spectra
The 94 GHz EPR spectrum of Y122• is dominated by g-and hyperfine (hf) anisotropy. The rhombic g-tensor causes spectral splitting (first level in spectral splitting scheme shown below). The spectrum is further split at the three principal g-value positions due to anisotropic hf couplings from magnetically coupled nuclei as listed in Table S3.   Table S1. (B) Direct comparison of two in-cell EPR spectra recorded at 94 GHz without background correction. Position of the gx component is displayed with a grey vertical line. Spectral feature originating from Mn 2+ in cells, which overlaps with the gx region of Y122•, is marked with an asterisk, * (see Figure S6). Difference in the EPR line shapes of the two in-cell samples observed in the gx region is due to: i) a small difference in the microwave power for the two measurements, which strongly affected the Mn 2+ intensity; ii) distinct radical to Mn 2+ concentration ratios. An unidentified background signal observed both in vitro and in-cell and unrelated to Y122• is marked with #. SI/11

Detection and evidence of Mn 2+ species in the cellular environment
In order to display the contribution of the Mn 2+ species to the spectra of in-cell sample, 9 Hahn-echo and 3 refocused echo (pump pulse off) experiments at 34 GHz were recorded with three different samples; 1) in vitro sample (240 µM Y122• radical concentration), 2) incell sample 1 (22 µM Y122• bulk radical concentration), and 3) E. coli cells lacking the β2 overexpression plasmid (0 µM Y122• radical concentration). They were either optimized to detect Y122• or Mn 2+ at pump and/or detect frequencies of the DEER experiments. The optimization for the Mn 2+ detection was performed by lowering the microwave power and shot repetition time. The six characteristic hyperfine lines of Mn 2+ species are only visible in the in-cell and E. coli cells samples (see Figure S5). These experiments clearly demonstrate the huge contribution of Mn 2+ species to the detected DEER signal at observer frequency. For the quantification of the Y122• and Mn 2+ contributions to the refocused echo with the pump pulse on, see SI sections 2.9 and 2.10. Figure S5. Pulse EPR spectra of Y122• in in-cell (left) and in vitro (middle) samples compared to E. coli cells without overexpression plasmid (right). The detection is either optimized for Y• (red) or for Mn 2+ (green). Furthermore, results of the optimization for Y• at the detect frequency (blue) are shown along with the refocused echo field-swept spectra (black). The EPR spectra were recorded at 10K and normalized to the same video gain and number of scans. Experimental conditions: tπ = 28 ns for the measurements at the pump frequency and tπ = 70 ns for the detect frequency; t1 = 300 ns (green with 200 ns), t2 = 2.1 µs (refocused echo); srt = 4 ms (green with 2 ms), shots per point = 100, microwave power = 20 mW (green with 3.2 mW).

Subtraction of the Mn 2+ spectral features present in the EPR spectra of in-cell samples. Details of orientationselective 1 H ENDOR spectra recorded at 34 GHz
In order to demonstrate that the Mn 2+ species present in E. coli cells is the reason behind the marginal differences between the in vivo and in vitro EPR line shapes (especially at frequencies higher than 9.6 GHz), we recorded EPR spectra of E. coli cells without Y122• at 34 and 94 GHz. The cells either lacked the overexpression plasmid or were harvested prior to the radical generation. Subsequently, we used these spectra for background correction of the in-cell data. The difference spectra are nearly identical to those recorded in vitro (orange vs green in Figure S6). These results clearly demonstrate that the only difference observed is due to the Mn 2+ species inherently present in the cells.
In addition, the magnetic field positions corresponding to gxy, gy and gyz molecular orientations used for orientation-selective 1 H Davies ENDOR measurements at 34 GHz are shown with vertical gray lines in Figure S6 (left).
For the W-band in-cell spectrum two types of background measurements had to be performed. The unidentified background peak labeled with #, as well the sharp hyperfine structure lines originating from the MS = -1/2 ↔ +1/2 transition of Mn 2+ were both detected for the E. coli cells harvested prior to the radical generation (solid magenta trace, Figure S6 right). However, the Mn 2+ EPR intensity was strongly dependent on the microwave power. Since it was impossible to perfectly reproduce the same conditions for the background measurement, this background trace could only be used to isolate (dashed pink trace) and subsequently subtract the background peak labeled with #. To obtain a Mn 2+ background, the in-cell spectrum was remeasured with a short shot repetition time, reduced by the factor of 30 (dark purple trace). This strongly suppressed the slowly relaxing Y122• contribution, while reproducing the faster-relaxing Mn 2+ background under the same conditions (microwave power and signal phase) as used for the in-cell Y122• measurement. This trace was subsequently used to suppress the Mn 2+ spectral features in the in-cell Y122• spectrum. SI/13

Details of the Y122• DEER measurements and analysis
The DEER time trace, V(t), is a product of intermolecular contribution B(t) and intramolecular interaction F(t). [12] In order to separate F(t), which is the first step in analysis, V(t) is divided by B(t), which is commonly known as background function. Along with the background-corrected data, Fourier transform (FT) of F(t) is also shown below. In an ideal case, in which all the orientations are excited (no orientation selectivity), FT results in a spectrum called Pake pattern. DEER time traces of rigid spin pairs that are strongly correlated depend on the relative orientation of these spins, known as orientation selectivity. As our DEER experiments did not result in ideal Pake patterns, we recorded orientation-selective DEER experiments shown below in Figure S7. Additionally, two-pulse echo decay measurements were performed ( Figure S8).
Parameters chosen for background correction procedure might result in unreal distance distribution peaks. DeerAnalysis2019 offers a validation tool to estimate the errors in the determination of the mean distances and their distributions. This tool calculates a distance distribution and carries out statistical analysis for a given set of parameters. Here, we calculated confidence intervals for in-cell and in vitro DEER traces by varying the starting values of the background fit by ± 50% in 10 steps and the background dimensionality in 10 steps by ± 0.5 with respect to the value chosen for the data analysis (shaded areas in Fig. S7). The detected mean distances with incell and in vitro samples were both validated.  SI/16

Origin of the second distance observed with in-cell sample
The analysis of the orientation-averaged in-cell DEER trace resulted in a second distance distribution peak that was validated by the validation tool of DeerAnalysis. The contribution of this distance, and thus the number of spin pairs (if real) resulting in this peak, is tiny as assessed by comparison of peak intensities of the distance distributions. We evaluated three possibilities as the origin of this peak, and subsequently suggest that it is an orientation-selection artifact. The three possibilities are;

1.
A distinct Y122•-Y122• distance: It cannot be a Y122•-Y122• distance arising from conformationally distinct Y122• pairs because we only detected one Y122• conformation. As explained in SI 2.4, any conformational change of this radical regardless of its contribution to the spectrum would manifest itself in the spectral line shape of 94 GHz EPR spectrum as observed previously. [14] 2. Mn 2+ -Mn 2+ distance: As reported previously, Mn 2+ can occupy the iron site in wt-b2 in vitro. [15] Considering the high Mn 2+ content within the cells, we investigated a possible Mn 2+ -Mn 2+ distance. In this regard, we performed Mn 2+ DEER experiments with E. coli cells that either contain the apo wt-b2 protein or lack the overexpression plasmid. As shown below in Figure S9, none of the experiments showed any dipolar modulations, and thus did not result in any distances. These data displayed that the detected distances throughout this work do not belong to a Mn 2+ -Mn 2+ pair, and are not affected by the presence of Mn 2+ in the cells; however, presence of Mn 2+ causes a reduction in modulation depth and in signal-to-noise ratio. Figure S9. DEER measurements of E. coli cells that contain the wt-b2 overexpression plasmid (green) or lack it (black) at 34 GHz and 10 K. The fits that are shown with dashed lines are based on a homogenous, three-dimensional background function.

Orientation selection artifact:
The observed second distance resembles the one extracted by DeerAnalysis when the distance vector rY122-Y122 is parallel to the magnetic field, with the nII component dominating the dipolar spectrum (DEER time traces named '3' in Figure S7). For orientation-averaging, each primary time trace was first normalized to the same signal intensity at zero time. Afterwards, these traces were normalized to the signal intensity at the pump position. As the normalization of individual in-cell DEER traces was not ideal because of the poor signal-to-noise ratios (SNRs) and huge Mn 2+ contribution at pump and detect positions, the distance resulting from the nII component might be over-pronounced in the orientation-averaged trace leading to the observed second peak. Indeed, during our experiments we realized that an improved SNR of the in-cell DEER trace recorded at D1 position led to a reduction of this second peak intensity in distance distribution analysis. SI/17

DEER data of three distinct in-cell samples
Three in-cell samples from distinct cell growths were prepared as explained in SI 1.1. Additionally, two in vitro samples were prepared to mimic in-cell samples having almost same Y122• fractions (named mimic 1 and mimic 2). The details of spin concentrations of all samples are given in Table S2. Q-band field-swept EPR spectra of these samples recorded via refocused spin echo with the pump pulse applied at the primary echo position (normalized to the Mn 2+ intensity), primary DEER time traces and their corresponding form factors are shown below in Figure S10. Note that even though background slopes of in-cell 2 and 3 are slightly steeper than that of incell 1, they do not exceed the slope of the 70 and 150 µM in-vitro samples shown in Figure 5B. SI/18

Calibration curve for determining the expected modulation depths of in-cell samples
The DEER modulation depth is reduced by the presence of Mn 2+ EPR signals overlapping with that of the radical. To create a calibration curve, in vitro samples containing different amounts of Y122• and Mn 2+ were prepared (at most 150 µM and 280 µM, respectively; see Table S2).
The relative contribution of the Y122• signal for each sample was estimated via the refocused echo signal intensity recorded at the observe frequency. The primary echo intensity fractions of Mn 2+ and Y122• do not properly represent the corresponding values contributing to the DEER signal due to a shorter Tm time of Y122• (the radical's relative contribution to the refocused echo intensity is reduced compared to that of Mn 2+ ). Furthermore, we found that switching on the p pump pulse optimized for an S = 1/2 species induces significant changes in the detected Mn 2+ line shape (see Figure S11). The importance of the pump pulse influence on the Gd 3+ refocused echo signal has been previously reported in the literature. [16] Thus, field-swept EPR spectra of several in-cell samples and various in vitro b2/Mn 2+ mixtures were recorded by integrating the refocused echo with the pump pulse applied at zero dipolar time, with the same t1 and t2 delays as used in the DEER measurements. The detection window, as in DEER, was placed symmetrically around the maximum of the refocused echo of the S = 1/2 species. In the next step, each field sweep was scaled to that of E. coli cells containing only Mn 2+ and no Y122• radicals ( Fig. 5A and Fig. S10). Subtracting

Background correction details
The DEER time traces were background corrected using second-order polynomial fitting (poly2) or homogeneous three-dimensional background correction (hom3). Both background functions resulted in almost identical distances and distributions (± 0.01 nm). However, the detected modulation depths were marginally different. We carefully compared the dipolar spectra resulting from distinct background functions for all the recorded traces. Poly2 consistently resulted in more reliable Pake patterns for all samples under investigation (see Figure S12A as an example). There are three possible reasons why poly2 was better for our samples: 1) Low modulation depth and SNR. This could be the reason for in-cell traces but not for the in vitro ones because the in vitro DEER traces display high SNRs. 2) A possible non-homogenous environment of the spins under investigation. Some examples of this kind are membrane proteins and/or pronounced aggregation of biomolecules. [17] A previous in vitro DEER study performed with E. coli RNR samples having different Y122• concentrations showed that b2 dimers are homogenously distributed even at concentrations as high as 2.3 mM. [18] This concentration is ten times higher than that of our in vitro sample with the highest Y122• concentration investigated here. Therefore, we concluded that aggregation of RNR b2 dimers is not the reason in our case.
3) The dipolar evolution time (spacing between the first Hahn echo and third detection pulse) of the DEER experiments is not long enough. A proper fitting can only be achieved when the dipolar evolution time in pulse sequence is significantly longer than the time required for dipolar modulations to fully decay. Unless the evolution time is set to extremely long values, the dipolar modulation observed with most of the RNR samples does not fully decay due to strong correlation between spins. The analysis of the long DEER traces shown in Figure S12B with both background functions resulted in highly similar Fourier transformed data; however, the hom3 has a deeper hole in the centre of the Pake pattern. This occurs when 'part of the biradical contribution is attributed to background' (Jeschke, G., DeerAnalysis User Manual2013, https://epr.ethz.ch/software.html). Although we prolonged the dipolar evolution time as long as possible, dipolar modulations did not fully decay (lower trace in Figure S12B). We concluded that being unable to record traces whose dipolar modulations fully decay is the reason why poly2 resulted in more reliable Pake patterns in RNR samples. Additionally, poly2 has been the method of choice to analyze in vitro DEER data recorded with RNRs previously. [11a, 14, 19] In light of these results, we decided to employ the calibration curve obtained with poly2 in the main text. Note that the conclusion of presence of b2 dimers carrying only one Y122• in the cells was also reached with the hom3 analysis, although poly2 and hom3 resulted in different modulation depth parameters, and thus slightly different calibration curves.   S14. Cw-and derivative pulse EPR spectra of F3Y122• measured at 9.6 and 34 GHz. In-cell (black) and in vitro (blue) data are shown along with corresponding simulations (red dashed lines) and the supernatant spectrum (green). Simulation parameters reported in literature [4a, 20] are changed minimally during spectral simulations (see Table S4). Asterisks denote couplings seemingly absent in the experimental spectra due to a background signal. 9.6 GHz data were recorded at the Bruker EMXnano with 2.5 mW power (31.6 mW power for green), 1.5 G modulation amplitude, 100 kHz modulation frequency, 5.12 ms as time constant, and 19.9 ms conversion time at 100 K, 50 scans (black), 20 scans (blue), 100 scans (green). 34 GHz data were recorded with Bruker ELEXSYS-II E560 equipped with the 5106QT-2 resonator at 10 K. Experimental conditions: π = 32 ns, τ = 300 ns, srt = 4 ms, spp = 10 and 20 mW power, 1 scan.  [4a, 20] were changed in order to achieve the best fit. Best agreement with the experimental data was achieved with gx,y,z = 2.0085(5), 2.0045(2), 2.0022 (3). Hyperfine parameters used for simulating the 9.6 GHz and 34 GHz data are given in regular and italic font, respectively. The small differences between the two sets are likely due to the presence of the second conformation. Euler angles α, ß, and γ are defined within the EasySpin z,y',z'' convention. They refer to rotations from the g-tensor frame into the hyperfine tensor frames. Positive angles are clockwise rotations viewed along the rotation axis. The hf coupling uncertainty was at most 20%.