Site Selective and Efficient Spin Labeling of Proteins with a Maleimide-Functionalized Trityl Radical for Pulsed Dipolar EPR Spectroscopy

Pulsed dipolar electron paramagnetic resonance spectroscopy (PDS) in combination with site-directed spin labeling (SDSL) of proteins and oligonucleotides is a powerful tool in structural biology. Instead of using the commonly employed gem-dimethyl-nitroxide labels, triarylmethyl (trityl) spin labels enable such studies at room temperature, within the cells and with single-frequency electron paramagnetic resonance (EPR) experiments. However, it has been repeatedly reported that labeling of proteins with trityl radicals led to low labeling efficiencies, unspecific labeling and label aggregation. Therefore, this work introduces the synthesis and characterization of a maleimide-functionalized trityl spin label and its corresponding labeling protocol for cysteine residues in proteins. The label is highly cysteine-selective, provides high labeling efficiencies and outperforms the previously employed methanethiosulfonate-functionalized trityl label. Finally, the new label is successfully tested in PDS measurements on a set of doubly labeled Yersinia outer protein O (YopO) mutants.


Dilution series
The following figures ( Figure    In the case of 5, the absorbance of the trityl is slightly increased in an acidic buffer system ( Figure S9a) as compared to alkaline conditions ( Figure S9b). For 6, the UV/Vis absorbance shows a strong decrease in the absorbance going from a 30 µM concentration to 20 µM ( Figure S10). This may be related to the experimental findings in section 6 below. For 9, the dilution series at pH 6.5 ( Figure S11a) and pH 7.0 ( Figure S11b) show no significant differences at a given concentration.

Calibration Curve for 9
For 9, the concentration dependent absorbance at 467 nm and 280 nm is plotted in Figure S12 with the resulting linear equations displayed in Table S1.  Table S1. Table S1. Linearized equations of the calibration curves shown in Figure S16. In order to evaluate the molar ratio between 9 and YopO in the labeling experiments, the absorbance peak of 9 at ~467 nm is used to determine the concentration of 9 in a labeled

Deconvolution of UV/Vis spectra
In order to evaluate the accuracy of the concentration determinations outlined above the where b denotes the scaling factor and c corresponds to an offset correction factor. These fits were then used to deconvolute the UV-vis spectra of labeled YopO mutants V599T9/N624T9 ( Figure S13b) and YopO S585T9/Q603T9 ( Figure S13c). Table S2 summarizes the quantification by both methods. Figure S13. UV/Vis spectra of 4.05 µM YopO (blue) and 10 µM TSL 9 (red) in labeling buffer (a). Recorded UV/Vis spectra after the labeling reaction and excess label removal (marine) and the corresponding deconvolution fit (red) of YopO V599T9/N624T9 (b) and YopO S585T9/Q603T9 (c). Table S2. Concentrations of YopO and 9 in the spectra of (Figure S13 b and c) determined using either the deconvolution or the maximum peak method. Comparing the resulting concentrations of YopO and 9 using either the deconvolution or the maximum peak value method, the deconvolution method gives slightly lower concentrations of YopO and slightly higher concentrations of 9. However, the deconvolution function is not able to correct for the bathochromic shift in the absorbance of 9, leading to unsatisfactory fits in the region above 300 nm. Overall, both methods give sufficiently accurate results for the determination of the concentrations of YopO and 9.

MALDI(+)-MS of YopO V599T9/N624T9 and S585T9/Q603T9
Both MALDI-MS analyses show masses of the respective non-, onefold-and twofold-labeled YopO mutants. This is inconsistent with the labeling degree distributions reported by ESI(+)-MS ( Figure 6 of the main text). Also, cw EPR and UV/Vis report a high labeling efficiency, which is not reflected in these MS results. However, the MALDI as well as the ESI sample preparation required acidic conditions (2% trifluoroacetic acid for MALDI and 0.1% formic acid for ESI(+)), which promote partial label detachment via retro-Michael reactions before and during the MS measurements. Several attempts to skip the acidic sample preparation failed.

Dimer Formation of 6 and Cleavage
Two separated 25 µM solutions of 6 and 9 in labeling buffer (20 mM PO i pH 6.8, 50 mM NaCl) were incubated for 16 h at 4 °C under the exclusion of light. The solutions were spun down to a final volume of 300 µL using VivaSpin 2/10k MWCO. From each solution a part was transferred into a 10 µL glass capillary and cw EPR spectra were recorded using a Bruker EMXnano spectrometer ( Figure S16). Then the sample of 6 was irradiated with UV light ( Figure S17).

Spin Count
The room temperature cw EPR spectra of both YopO mutants were used to derive the labeling efficiency in reference to the free TSL 9 buffer solution of known concentration (100 µM) ( Figure S18).  Figure S18. Room temperature cw EPR spectra obtained from YopO mutants (a) V599T9/N264T9 (109% labeling efficiency) and (b) S585T9/Q603T9 (95% labeling efficiency). The spin concentrations of the protein samples were determined in reference to the 100 µM TSL 9 solution in buffer (c).

Simulation of cw EPR Spectra
The cw EPR spectra of YopO V599T9/N624T9 (Table S3, entry a; Figure 6 in the main text) and S585T9/Q603T9 (Table S3, entry b; Figure S8b) were simulated using the "chili" routine of EasySpin 1 taking into account g-anisotropy and a rotational correlation time . The hyperfine coupling tensor was assumed to be isotropic. The spectrum of free label 9 (Table S3, entry c; Figure S6b) was simulated using the "garlic" routine of EasySpin. 1 All simulation parameters are summarized in Table S3

Relaxation Time Measurements
Relaxation times T 1 and T m were measured with the Inversion Recovery (IR) pulse sequence ( Figure S19a) and via a two-pulse Electron Spin Echo Envelope Modulation (2pESEEM) experiment ( Figure S19b). The pulse sequences were applied at the maximum of the field sweep spectrum and the temperature was set to 50 K, 60 K, 70 K and 80 K. Both IR and 2pESEEM experiments included phase cycling, two steps for 2pESEEM and four steps for IR. 4 All parameters of the IR and 2PESEEM pulse sequences are given in Table S4.

S15
The T 1 relaxation times were extracted by multiplying the recorded IR curves by -1 and fitting a single exponential decay (y = a • exp(-x/T 1 ) + c). 12 The T m relaxation times were obtained by fitting a stretched exponential decay (y = a • exp((-x/T m ) c ) + d) to the echo decay curves acquired by the 2pESEEM experiment. 4 The traces recorded from double mutant YopO V599T9/N624T9 are displayed in Figure S20 and the fit values for T 1 and T m are summarized in Table S5.

DQC
The six-pulse DQC sequence ( Figure S21) was applied at the magnetic field position which yielded the maximal intensity in the field-swept EPR spectrum. The phase of the microwave radiation was adjusted such that the intensity of the DQC echo was maximal in the real channel. Pulse lengths and interpulse delays are given in Table S6. The shot repetition time (SRT) was set to 15.3 ms. 5 A 64-step phase cycle was applied to remove undesired echoes and thus extract the pure double quantum coherence pathway contributions. 6,7 In order to remove deuterium ESEEM from the dipolar traces, a modulation averaging procedure was applied ( 1 and  2 in 8 steps of 16 ns). 2  Table S6. Pulse sequence settings for DQC. Figure S21. Schematic representation of the DQC pulse sequence. The employed sequence was adapted from literature procedures. 6

SIFTER
The SIFTER sequence ( Figure S22) in conjunction with a 16-step phase cycle 8 was applied at the magnetic field position yielding the highest signal amplitude in the field sweep spectra.
Modulation averaging ( 1 and  2 in 8 steps of 16 ns) was applied to remove deuterium ESEEM from the time traces. All pulse lengths, interpulse delay times and further parameters are given in Table S7. Table S7. Pulse sequence settings for SIFTER. Figure S22. Schematic representation of the SIFTER pulse sequence. The employed sequence was adapted from literature procedures. 9

PELDOR
For the PELDOR experiment ( Figure S23) 9 on 9 the settings in Table S8 were used. The length of the pump pulse (π) B was determined by a transient nutation experiment. The pump pulse (π) B was set to the maximum of the field sweep spectrum and the observer pulses were applied at a frequency offset of -15 MHz relative to the pump frequency. Regarding the suppression of deuterium ESEEM, an 8-step modulation averaging procedure was applied with a time increment of 16 ns. Additionally, a two-step phase cycle was used in order to remove undesired echoes and to correct for receiver baseline offsets.  For the PELDOR experiment on R1-labeled YopO, the pump pulse was applied at the magnetic field position, which yields the maximal signal amplitude. The detection sequence was applied at a frequency offset of -100 MHz with respect to the pump frequency. The other parameters were set as given in Table S9. The optimal length of the (π) B pump pulse was determined by a transient nutation experiment. As mentioned above, a modulation averaging procedure and a two-step phase cycle was used to average out deuterium ESEEM and to remove unwanted echoes as well as baseline offsets.

Original PDS Time Traces, Background Removal and Validation
All PDS data was analyzed using the DeerAnalysis 2018 package for MATLAB. 10 In PDS, the resulting time trace is a convolution of the wanted dipolar interaction between the pair of spin labels within one protein molecule (intramolecular) and a background contribution between spins located on different macromolecules (intermolecular). The intramolecular dipolar interaction can be extracted by different procedures depending on the respective experiment: for PELDOR spectroscopy, the background is usually fitted directly to the time trace assuming a three-dimensional distribution of background nano-objects. 11,12 For the singlefrequency experiments DQC and SIFTER, however, such an analytical treatment of the background is not applicable. 12,13 In this case, experimental background data obtained by performing DQC/SIFTER measurements on labeled single cysteine mutants have been used. [14][15][16] Then, an 8 th order polynomial was fit to the thus obtained time traces ( Figure S24) quantifying the experimental background. The DQC and SIFTER time traces were then

Signal to Noise Ratio (SNR) determination
The quality of the recorded dipolar traces can be estimated by the signal-to-noise ratio (SNR) defined as where λ is the modulation depth of the dipolar trace, t is the acquisition time of the respective experiment, and σ N is the standard deviation of the noise of the trace. In order to deconvolve the noise from the wanted signal, the signal has been approximated by a polynomial fit (polynomial of second to 5 th order). Subtracting this fit from the measured traces yields the pure noise contributions. The SNR has been calculated from the raw data prior to background-correction using the software SnrCalculator. 17 The thus obtained SNR values of all dipolar traces shown either in the main text or the supporting information are compiled in Table S10.  Table S11.  After plasmid amplification, the mutagenesis was confirmed via Sanger sequencing ( Figure S28).