Light-Induced Triplet–Triplet Electron Resonance Spectroscopy

We present a new technique, light-induced triplet–triplet electron resonance spectroscopy (LITTER), which measures the dipolar interaction between two photoexcited triplet states, enabling both the distance and angular distributions between the two triplet moieties to be determined on a nanometer scale. This is demonstrated for a model bis-porphyrin peptide that renders dipolar traces with strong orientation selection effects. Using simulations and density functional theory calculations, we extract distance distributions and relative orientations of the porphyrin moieties, allowing the dominant conformation of the peptide in a frozen solution to be identified. LITTER removes the requirement of current light-induced electron spin resonance pulse dipolar spectroscopy techniques to have a permanent paramagnetic moiety, becoming more suitable for in-cell applications and facilitating access to distance determination in unmodified macromolecular systems containing photoexcitable moieties. LITTER also has the potential to enable direct comparison with Förster resonance energy transfer and combination with microscopy inside cells.

4 at a repetition rate of 20 Hz, and the microwave pulses were moved forward in time with respect to the fixed laser flashes. All spectra were acquired with one shot per point.
Time-resolved ESR (TRESR) was carried out in a critically coupled resonator using Laser 1 only, without field modulation or phase sensitive detection. A microwave attenuation of 33 dB and a time base of 4 ns were used. The signal was averaged between 1.4 and 2.4 µs after the laser flash (256 time points), around the intensity maximum of the time trace. Triplet state parameters were extracted via simulation of the spectrum using the Matlab® EasySpin routine (pepper function). 3 Electron spin-echo (ESE) experiments (i.e. field sweeps, T2 and DAF experiments) were performed in an over-coupled resonator using a standard Hahn echo sequence preceded by a laser flash (laser 1 -DAF -π/2 -τ -π -τ -echo), with a length of 16 ns for the π/2 pulse. Field swept electron spin-echo spectra were recorded using a DAF of 1066 ns, a τ value of 320 ns, a laser LITTER traces were acquired using the following pulse sequence: laser 1 -DAF -π/2 -τ -π -τ' -laser 2 -τ'' -echo, with a length of 16 ns for the π/2 pulse. In all cases a τ = τ' + τ'' = 2004 ns was used and the DAF was incremented from an initial value of 1958 ns in steps of 8 ns to reach a maximum increment of +2800 ns. A longer trace with a +5600 ns increment was recorded in order to determine the zero time of the experiment. The average energies per laser flash were set to 2 mJ (laser 1) and 2.5 mJ (laser 2), and the delay between flashes was tpp = 6.82 µs.
Measurements were carried out at three different values of the external magnetic field, corresponding to the Yand Zcanonical orientations of the ZFS tensor and a position in between Xand Y -(c.a. 335.1, 386.0 and 333.4 mT, respectively). The total accumulation times for these measurements were 20, 93 and 20 h, respectively. The accumulation time for the experiments in figures S2 and S3 was 2h per trace. Raw LITTER traces were phase-and background-corrected to obtain the form factors, and those were analyzed via Fourier Transform and Tikhonov Regularization using the Matlab® DeerAnalysis2018 routine to obtain the corresponding distance distributions. 4 The zero time used for the analysis (854 ns) was obtained from the full LITTER trace in Fig. 1

S2.1. Density Functional Theory calculations
Initial geometries for peptides [1] and [2] were built based on previously reported similar molecules, 2 using UCSF Chimera. 5 Geometry optimizations and spin density calculations were performed in vacuo using Gaussian® 09 (revision A.02). 6 Ground state geometry optimizations of peptides [1] and [2] were carried out in the singlet state, using the PBE1PBE functional and the 6-31g(d) basis set. Triplet state geometry optimizations of each of the TPP moieties from peptide [1] were subsequently performed using the functional B3LYP and the basis set SV(P). Electronic spin densities were obtained by single-point calculations on the TPP moieties previously optimized in the triplet state, using B3LYP and the basis set EPR-II. 7 Zero-Field Splitting (ZFS) tensor orientations for the two TPP moieties in peptide [1] in the triplet state were calculated using Orca (release version 4.2.0), 8 with the functional B3LYP and the basis set EPR-II. The spin-spin contribution to the ZFS was calculated using computed UNO (spin-unrestricted natural orbital) determinants. 9

S2.2. Orientation dependent simulations
Simulations of the LITTER data recorded on the Yand Zfield positions were performed using a modified version of the algorithm reported by Lovett et al., 10 to take into account the zero-field splitting and triplet populations. The spin system parameters were accounted for using EasySpin functions to simulate the spectra and resonant fields. 3 The input orientations were derived from the optimized structures of independently and simultaneously to the experimental datasets using an algorithm similar to that described by Marko et al. 11 The results of these fitting procedures are shown in Figure 4 (main text) and Figures S10 and S11 (SI).  and corresponding monoexponential fit (black) with a T2 value of (2.15 ± 0.05) µs. (b) Delay after flash experiment (red) and corresponding biexponential fit (black) with lifetimes of (0.34 ± 0.04) ms and (1.68 ± 0.03) ms for the two exponentials.

S3.2. Optimisation of experimental conditions for LITTER
The effects of laser flash energy in signal intensity and modulation depth were studied (Fig. S4).
Higher laser 1 energies improved signal intensity ( Fig. S4 (a)), in consistence with the formation of more triplet spins before the Hahn echo sequence, but the signal was nearly saturated at 2 mJ. However, they did not cause any effective change in modulation depth (Fig. S4 (b)) as expected, therefore rendering better modulation-to-noise ratios (MNR) at higher energies. Higher laser 2 energies clearly improved modulation depth (Fig. S4 (d)), as expected from more triplet spins being formed during the Hahn echo sequence, but had a negative effect in signal intensity (Fig. S4 (c)). Two different factors seem to be contributing to the observed decrease in signal intensity at high laser 2 energies. First, the shorter T2 due to the formation of more second triplets by laser 2 10 on molecules where a first triplet had already been formed by laser 1 (Fig. S5). Second, the incomplete decay of the photoexcited triplets back to the ground state due to the TPP triplet lifetime at cryogenic temperatures being comparable to the short repetition time (50 ms) used in the LITTER experiment, which becomes more important when larger numbers of triplet states are formed by larger laser energies. That would also be consistent with the nonlinearity in the laser 1 energy dependence of the signal intensity ( Fig. S4 (a)). The sample was aerated in an attempt to increase signal intensity by reducing triplet lifetime. The signal was indeed slightly increased for very short τ values, but the improvement was lost for the τ ~ 2 µs of the LITTER experiment, due to the reduction of T2 caused by the presence of paramagnetic oxygen.  The importance of a good overlap volume between the two laser beams for modulation depth was studied by changing the sample position and concentration (Fig. S6). In the optimized setup, the ~ 5 mm height sample is centered in the resonator and laser 1 is focused in a ~ 1 mm spot on the center of the sample. The sample above the volume excited by laser 1 absorbs light from laser 2 but does not contribute to the signal, reducing the number of laser 2 photons that reach the overlap volume (therefore it will be called "dead volume"). By pushing the sample tube further down into the resonator, the dead volume is reduced, leading to an increase in the number of laser 2 photons 12 that reach the overlap volume and a corresponding increase in modulation depth (Fig. S6 (b)), similar to effect of increasing laser 2 energy (Fig. S4 (d)). The decrease in signal intensity could be explained by a fraction of the laser 1 photons missing the sample, above the top.
Doubling the concentration of [1] from 40 µM to 80 µM in the optimized setup also doubled signal intensity because of twice as many molecules being photoexcited by laser 1 (Fig. S6 (c)). However, this also intensified the attenuation of laser 2 by the dead volume, leading to a significant decrease in modulation depth.  porphyrin has been omitted for clarity. The bonds defining the three dihedral angles studied are indicated as "A" (blue, rotation around the C α -C amide bond of the Ala, corresponding to the conversion between the bent and extended conformers), "B" (green, rotation around the C phenyl -15 C amide bond) and "C" (red, rotation around the C porphyrin -C phenyl bond). (b) Energy profiles of the rotation around the abovementioned bonds, calculated using Gaussian 09 (PBE1PBE, 6-31g(d)).

The two energy minima in the profile for dihedral A (blue) corresponding to the [1] bent and [1]
extended conformers are indicated, with an energy barrier of interconversion of 11 kJ/mol. The absolute minimum of each energy profile has been assigned an energy of 0 kJ/mol.

S3.4. Modulation depth analysis
Assuming that all molecules have two porphyrin moieties that can be excited, the modulation depth observed will be a function of the porphyrin excitation efficiencies by laser 1 (e1) and laser 2 (e2), Porphyrin excitation efficiencies e1 and e2 can be estimated from the absorbance of the sample at the wavelength of irradiation, the laser pulse energies used and the geometry of the experiment. In our case we have estimated that the pulse energies for both laser 1 and laser 2 are high enough to optically saturate the overlap volume, making e1 = e2 = 1. This is consistent with the marginal change in echo intensity observed when increasing the energy of laser 1 from 2 mJ to 3 mJ ( Fig.   S4 (a)). Using the reported value of φT = 0.8 for free-base TPP, 12 we predict a modulation depth Δ = 0.16, which is slightly larger than our measured values. This difference likely originates from experimental factors such as the background correction used in the data analysis, as the background decay used impacts the measured modulation depth, the fact that porphyrin excitation efficiencies may not be exactly 1, or the fact that the second laser may not excite completely the portion of the sample excited by the first laser. Furthermore, the calculation above assumes that all molecules have two moieties that can be excited. This assumption may not be experimentally appropriate as sample bleaching can occur and was observed during our experiments as a decrease in both echo intensity and modulation depth over time.