Rationally designed azobenzene photoswitches for efficient two-photon neuronal excitation

Manipulation of neuronal activity using two-photon excitation of azobenzene photoswitches with near-infrared light has been recently demonstrated, but their practical use in neuronal tissue to photostimulate individual neurons with three-dimensional precision has been hampered by firstly, the low efficacy and reliability of NIR-induced azobenzene photoisomerization compared to one-photon excitation, and secondly, the short cis state lifetime of the two-photon responsive azo switches. Here we report the rational design based on theoretical calculations and the synthesis of azobenzene photoswitches endowed with both high two-photon absorption cross section and slow thermal back-isomerization. These compounds provide optimized and sustained two-photon neuronal stimulation both in light-scattering brain tissue and in Caenorhabditis elegans nematodes, displaying photoresponse intensities that are comparable to those achieved under one-photon excitation. This finding opens the way to use both genetically targeted and pharmacologically selective azobenzene photoswitches to dissect intact neuronal circuits in three dimensions.

In the presence of water, we observed two sets of signals in the 13 C NMR corresponding to the glutamate moiety. This is due to an acid-base equilibrium of the amino and carboxylic groups.  Supplementary Figure 40 1 H and 13 C NMR spectra of compound 5a-1'.
-NHCO-  -NHCO- -NHCO- The difference in energy between the cis and trans isomers (E trans-cis ) and the minimum barrier height for the thermal cistrans isomerization are given (E ‡ cis-trans ). b The excitation energy (E exc ), the oscillator strength of the 1P absorption process (f), and the absorption crosssection of the 2P absorption process (σ 2 , in GM units) are given for the both isomers of each compound. c In all the cases, the lowest-energy barrier height for the thermal cistrans isomerization was found to correspond to an inversion mechanism. The difference in energy between the cis and trans isomers (E trans-cis ) and the barrier height for the thermal cistrans isomerization are given (E ‡ cis-trans ). b The excitation energy (E exc ), the oscillator strength of the 1P absorption process (f), and the absorption cross-section of the 2P absorption process (σ 2 , in GM units) are given for both isomers of each compound. c In all the cases, the lowest-energy barrier height for the thermal cistrans isomerization was found to correspond to an inversion mechanism. d Accurate description of the thermal isomerization barrier height would require inclusion of explicit water molecules in the calculation. e Calculation did not converge. f Calculation did not converge. From the σ 2 value computed in the gas phase and the equations given in reference 4 for the solvent dependence of σ 2 , an estimate of the 2P absorption cross-section in water was made (σ 2 = 100 GM).  Figure 10). Table 7. Primers used for Gibson assembly in the construction of pNMSB18.

Supplementary Methods
General procedure for the synthesis of azobenzene-based photoswitches: The preparation of ligands MAG 2P slow and MAG 2P_F slow was achieved via a multistep synthetic sequence (see Figure 2 in the manuscript). In both cases, we took the corresponding aminobenzoic acid (4a for MAG 2P slow and 4b for MAG 2P_F slow ) to form the azobenzene core (5a for MAG 2P slow and 5b for MAG 2P_F slow ), to which the different functional fragments of the target compounds were sequentially introduced: fully protected glutamate derivative 2 5 and maleimide acetic acid 3. 6 In addition, azobenzene model compounds Azo1' and Azo2' were prepared as references for the photochemical characterization of MAG 2P slow and MAG 2P_F slow , respectively.

Materials and methods for the synthesis of azobenzene-based photoswitches:
Commercially available reagents were used as received. Solvents were dried by distillation over the appropriate drying agents. All reactions were monitored by analytical thin-layer chromatography (TLC) using silica gel 60 precoated aluminum plates (0.20 mm thickness). Flash column chromatography was performed using silica gel (230-400 mesh). 1

4-[(4-Acetylaminophenyl)azo]-3-fluorobenzoic acid, 5b-1' (Supplementary
To  Green and red fluorescent proteins were simultaneously excited at 488 nm for 343 ms, using bidirectional laser scanning at 400 Hz. Images were recorded with a resolution of 512x512, and with an imaging interval of 4 s. Green and red fluorescence were recorded with two different HyD detectors with a detection range from 500 to 550 nm and from 569 to 648 nm, respectively. Pinhole aperture was set at maximum (600 μm).
Whole field photostimulation flashes were fit to keep imaging interval, and periods lasted in total for 1 min. Photostimulation was done at 256x256 resolution with bidirectional laser scan. One-photon photostimulation was done at 405 nm (0.81 mW μm -2 ), and two-photon stimulation at 780 nm (2.8 mW μm -2 ). Back-photoisomerization was achieved at 514 nm (0.35 mW μm -2 ). Inter-stimulus imaging periods lasted 1.5 min. Intensity and duration of the photostimulation intervals were adjusted to obtain the optimal photoresponse and reproducibility. At the end of each experiment we reconfirmed that the neuron kept its healthy morphology.
Transgenesis was performed according to standard methods for microinjection. 11 To generate MSB104 strain a DNA mix containing 50 ng/ul pNMSB18, 1.5 ng/ul myo-2p:mCherry and 50 ng/ul Plus DNA ladder as carrier was injected into the gonad of GN692 young adult worms. The Primers used for Gibson assembly in the construction of pNMSB18 are shown in Supplementary

Imaging was performed 4 h after compound injection in TRN neurons co-expressing
GluK2-L439C-mCherry and GCaMP6s in a single focused plane. Neurons with healthy morphology and no signs of fluorescent aggregates were selected for photostimulation.
Green and red fluorescent proteins were simultaneously excited at 488 nm and 561 nm for 343 ms, using bidirectional laser scanning at 400 Hz. Images were recorded with a resolution of 512x512 and a digital zoom of 4, with an imaging interval of 660 ms.
Green and red fluorescence were recorded with two different HyD detectors with a detection range from 500 to 550 nm and from 569 to 648 nm, respectively. Pinhole aperture was set at ~500 μm.
Whole field photostimulation flashes were fit to keep imaging interval. Photostimulation was done at 256x256 resolution with bidirectional laser scan, with a digital zoom of 4.
One-photon photostimulation was done at 405 nm (15 μW μm -2 ), and two-photon stimulation at 780 nm (2.8 mW μm -2 ). Back-photoisomerization was achieved at 514 nm (1.2 μW μm -2 ). Intensity and duration of the photostimulation intervals were adjusted to obtain the optimal photoresponse and reproducibility. At the end of each experiment we reconfirmed that the neuron kept its healthy morphology.

Data analysis and statistics:
Amplitude of LiGluR photocurrents were analyzed using IgorPro (Wavemetrics). Displayed whole-cell current traces have been filtered using the infinite impulse response digital filter from IgorPro (low-pass filter with cutoff of 50 Hz).
The drift in current observed during light spectra recordings was corrected where appropriate with the IgorPro (WaveMetrics) software using a custom-made macro for drift correction.
1P and 2P calcium images were acquired with the Live Acquisition 2.1 software (Till Photonics) and stored by the Arivis Browser 2.5.5 (Arivis AG). These images were analyzed with ImageJ and the mean fluorescence value for each cell profile was calculated using the same software. The fluorescence signals were treated to obtain ∆F/F values according to: where F 0 is each cell's average signal for the experiment's baseline and F is the fluorescence signal upon stimulation. The resulting fluorescence ratios were analyzed in OriginLab. To obtain cell-averaged 1P action spectra, ∆F/F values were first normalized with respect to the maximum photoresponse obtained for each cell after perfusion of free glutamate at the end of the experiment. To obtain cell-averaged 2P action spectra, 2P ∆F/F responses were first normalized with respect to the 1P ∆F/F response at 405 nm for the same cell.