Control of the Two-photon Visual Process in ex vivo Retinas and in Living Mice

: Palcewska et al . first demonstrated near infrared (NIR) visual response in human volunteers upon two-photon absorption (TPA), in a seminal work of 2014, and assessed the process in terms of wavelength- and power-dependence on murine ex vivo retinas. In the present study, ex vivo electroretinography (ERG) is further developed to perform a complete characterization of the effect of NIR pulse duration, energy, and focal spot size on the response. The same set of measurements is successively tested on living mice. We discuss how the nonlinear intensity dependence of the photon absorption process is transferred to the amplitude of the visual response acquired by ERG. Finally, we show that the manipulation of the spectral phase of NIR pulses can be translated to predictable change in the two-photon induced response under physiological excitation conditions.


Introduction
The vision process relies on the photo-activation of rods and cones. Upon photon absorption, the retinal moiety of visual receptors changes its conformation, triggering a sequence of events culminating with the generation of an electric signal to the brain. [1][2][3] Human visual perception spans from 400 to 720 nm, [4] while the spectral sensitivity of mice is shifted to the ultraviolet. [5] Recently, we have studied the effect of spectral phase of green ultrafast pulses on the visual response of living mice. [6] In parallel, we have witnessed a series of experiments demonstrating approaches for widening the range of wavelengths that can be perceived by a species beyond the natural response of its photoreceptors. For instance, Ma et al. used lanthanide-doped nanoparticles in direct contact with the retina of living mice to upconvert NIR (980 nm) to green (535 nm) light. [7] An alternative approach to extend the range of perceived wavelengths, which does not require the use of exogenous agents, is based on the nonlinear absorption of NIR photons, usually generated by a short-pulse laser source. This has been demonstrated on ex vivo murine retina [4,8] and on human volunteers. [4,[9][10][11][12][13] In this work, we pursue this approach complementing ex vivo and in vivo measurements on mice, highlighting the dependence of the visual response on several excitation parameters (energy, size, and duration of the pulse), and we also show that the two-photon induced response can be modulated by the spectral phase of the light pulses under physiological excitation conditions. For all the experiments presented, the main observable is the electro-physiological signal emitted by the retina acquired by electroretinography (ERG). This signal shows two different characteristics: a hyperpolarization called A-wave, and a successive depolarization called B-wave ( Fig. 1.A, inset). The A-wave originates from the photoreceptors, emitted after the biological processes following the isomerization of rhodopsin, while the B-wave originates from bipolars cells, initiated by the hyperpolarization of the photo-receptors. [2] 1. Methods  The energy curves for SPA and TPA were fitted using the Naka-Rushton equation: [4,20,21] R max corresponds to the saturation level, E 1/2 is the required energy to generate a signal of amplitude while n relates to the slope in a logarithmic representation. The traces in all plots are normalized with respect to R max . For the energy dependence ( Fig. 2.A), the visual response at 960 nm is directly compared to that obtained by single-photon absorption (SPA) at 480 nm. The visible wavelength is obtained by doubling the frequency of the infrared pulse in a nonlinear crystal. We observed that substantially more energy is needed to measure a B-wave at 960 nm than at 480 nm (SPA:8.0 ± 3.1 pJ, TPA: 260 ± 34 nJ).
For pulse duration dependence ( Fig. 2.B), we applied a negative chirp using a multi-pass prism compressor (BOA -Swamp Optics), and measured the ERG as a function of the pulse duration of the infrared pulse. The B-wave amplitude decreases as the pulse duration increases at 960 nm. The same trace is plotted in Fig. 2.D as a function of intensity (blue triangles). For the beam diameter dependence ( Fig. 2.C), we focused the infrared light using a f = 36 mm lens, and measured the ERG as a function of the lens position. The size of the beam is measured using a beam profiler (Newport, LBP-1-USB).We observed that the B-wave decreases as the beam diameter increases, which is also expected from a multiphoton process depending non-linearly on the excitation intensity. The same trace is also reported in Fig. 2.D (black squares). Note that the beam diameter was kept below 250 um throughout the experiment to ensure that the density of the receptors in (1) The experimental set-up is reported in Fig. 1.A. The output from a 1 kHz Ti:Sapphire amplifier (Astrella, Coherent Inc.) is split into two pulse trains. The first one is frequency converted by a nonlinear optical parametric amplifier (TOPAS-White, Light Conversion) to produce 7 μJ, 80 fs pulses at 960 nm. An additional home-made OPA stage [14,15] pumped by the second pulse train is introduced to increase the output energy up to 40 μJ (120 fs pulse duration). The need of this amplification stage is mainly determined by the losses introduced by the gratings used for the 4ƒ zero-dispersion line housing a spatial light modulator (SLM-S640, by Jenoptik) at its Fourier plane, used for acting on the spectral phase of the 960 nm pulses. The beam exiting the 4fline is collimated and reduced in transverse size down to approximately 2 mm diameter. A fast silicon photodiode (SM05PD28) combined with a gated integrator (SR280 -Stanford Research Systems) monitors pulse energy shot-by-shot. A combination of a half wave plate and a polarizer set on a computer-controlled motorized rotational stage (PRM1/MZ8 -Thorlabs) is used to adjust pulse energy. A fast electronic shutter (Newport model 76992) allows to single out five consecutive pulses from the 1 kHz source to illuminate one eye of the mouse. The full cycle to the next irradiation lasts two seconds.
The in vivo / ex vivo system, in Fig. 1.B and Fig. 1.C, respectively, and the signal pre-amplifier (SR560 by Stanford Research System inc., working on battery power) are enclosed in a custommade Faraday cage to reduce the 50 Hz noise pickup. The analog signal is frequency filtered (between 0.3 and 3000 Hz), amplified by a factor 5000, and digitized by an analog-to digital converter (ADC -PCI-6220 by National Instruments) at 25 kHz sampling rate. During data acquisition, the ERG trace is sampled leaving a few milliseconds before and after each illumination event.
For in vivo measurements ( Fig. 1.C), the detection system and the measurement protocol are identical to those described in our previous work. [6] Briefly, 6-8 week-old C57BL/6 male mice are anesthesized using a mix of ketamine/xylazine as described in ref. [6] The beam illuminates the right eye, while the electro-physiological potentials are acquired from both eyes using contact lens electrodes (Ocuscience). The two electrodes are connected to the differential amplifier.
For ex vivo experiments ( Fig. 1.B), the eye is carefully pulled out of the orbit, cut from the optic nerve and placed in cold sterile Ames' medium (Sigma Aldrich) prepared according to manufacturer's instructions. The retina is dissected from the other eye tissues and placed, photoreceptors facing up, on a commercial holder (OcuScience) [16] at the sample position in Fig. 1.A. The signal is acquired via an electrode embedded in a bath of electrolyte solution (NaCl 140 mM, KCl 3.6 mM, MgCl 2 2.4 mM, CaCl 2 1.2 mM, HEPES 3.0 mM and EDTA 0.01 mM) in contact with the bottom of the retina (ganglion cell layer side). The other electrode is connected to the perfusion (Ames' medium heated at 37°C, oxygenated using a 95%/5% CO 2 / O 2 ), which flows at 50 mL/hour. The ex vivo signal is further filtered by a 50 Hz notch filter (2355-50-BNC by KR electronics, Inc). Note that the measurement performed requires only one retina, therefore the second dome on the holder is not used.

Visual Response ex vivo
For a precise study of TPA-induced vision, the ex vivo approach is ideal, as it presents the advantage of working without the lens (thickness: 1.6 mm [17] ) in the beam path. This helps reducing uncertainties in the intensity values, as in the near infrared the optical and spectroscopic characteristics of the mouse eye are not well established. [18,19] Aside from absorption, the lens could also affect pulse intensity at the retina by temporal pulse stretching. In our experiment, we successfully measured the B-wave amplitude as a function of three distinct excitation parameters: pulse energy, pulse duration, and beam diameter. The results are shown in Fig. 2. TPA: 34 nJ). We have conducted a simple quantitative comparison at the E 1/2 positions retrieved on both curves. The corresponding values for σ 1 I and σ 2 I 2 point to the fact that the pulses used in each experiment isomerize approximately the same number of molecules. Following Palczewska et al., [4] we can rule out a SPA contribution from the spectral tail of the 960 nm pulse.

Visual Response in-vivo
Ex vivo configuration allows for a more precise control of the parameters of the input beam and a better signal to noise ratio due to the absence of heart beat and respiration. However, the retinal pigment epithelium is detached from the retina during ex vivo experiments, which impairs renewal of the retinal, [16,28] , and reduces the signal over time. We therefore proceeded in measuring TPA-induced visual response in living mice. We first measured the power dependence of the ERG. The B-wave as a function of the energy for both visible (SPA) and infrared (TPA) femtosecond excitation is displayed in Fig. 3.
In agreement with the ex vivo results, the energy necessary to induce an electro-physiological response is of the order of the µJ, for a 120 fs pulse. In the experiment, the beam size at the retina needs to be carefully adjusted, in order to maximize the intensity for the TPA process. This was achieved using a telescope lens arrangement and adjusting the position of the second lens to optimize the ERG amplitude. We notice that the maximum B-wave amplitude in ex vivo retinas is smaller than its amplitude in living mice (34 µV, instead of 160 µV for the SPA case). It has been reported that this discrepancy can be addressed by a better sample preparation. [16] We also notice that the energy required to reach the E 1/2 (SPA: 2 · 10 -3 nJ and 10 -2 nJ, TPA: 0.180 µJ and 5 µJ for ex vivo and in-vivo respectively) increases by one order of magnitude as already observed by Vinberg et. al. [16] After determining the excitation parameters for obtaining TPA-induced visual response under physiological conditions in vivo, we investigated the sensitivity of the ERG signal to the temporal shape of the infrared femtosecond pulse. We relied on a phase-modulation scheme already applied to atomic and molecular systems. [29][30][31][32] The approach consists of scanning a π-phase step throughout the infrared pulse spectrum dispersed in the Fourier plane of a 4-f zero dispersion compressor [33] (see Fig.  1.A, PS) and measuring a signal related to the excitation of an electronic level (by absorption of fluorescence). Note that this optical scheme leaves the photon number (energy) and spectrum unchanged, while the temporal profile (and thus the intensity) of the pulse is modified.
We characterized the performance of our set-up for pulseshaping by performing the experiment on Rhodamine 6G in ethanol, using the dye fluorescence as an observable, right before and after the in-vivo measurements. The experiment is repeated on σ 1 I (s −1 ) σ 2 I 2 (s −1 )2 .6 ± 1.3 · 10 9 1.3 ± 0.7 · 10 10 the region probed was sufficiently homogeneous to perform a comparison. Looking at Fig. 2.A, one can note that the ERG signal does not increase linearly with energy for SPA, and quadratically for TPA. While the SPA follows the typical saturation trend (see inset) already reported [2,[22][23][24][25] which can be directly linked to the eye physiological response, the TPA trace requires further considerations. In the following, we rationalize it assuming that the ERG signal is determined by an interplay of two distinct contributions: i) a subpicosecond one, localized at the molecule site and associated with the photon absorption and the isomerization of rhodopsin, followed by ii) a much slower process governed by the physiological response and leading from the isomerized state to the generation of the B-wave . The physiological response is expected to be independent from the photo-excitation mechanism, but it might depend on the number of isomerized molecules. We can model the photo-molecular interaction (i) for SPA and TPA using the rate equations for a two-state model (not taking into account stimulated emission): For SPA, the number of excited molecules N 2 increases linearly with the energy. Assuming homogeneous molecule surface concentration at the retina, N 1 ∝ d 2 . Considering that , SPA should be insensitive to the beam diameter. As for all linear processes, we also expect no dependence on pulse duration.
For TPA, the number of excited molecules N 2 increases with the pulse energy at fixed pulse duration quadratically, and taking into account both that and N 1 ∝ d 2 , N 2 should scale as . Upon integration over the pulse duration, one can also see that . In agreement with our assumptions, we observe in Fig. 2.B that the dependence of the TPA signal on the pulse duration is approximated by . We can understand the good agreement with the rate equation of the ultrafast mechanism (i) as the parameter ∆t is significant only for the photo-dynamics at short times. On the other hand, the dependence of the ERG signal on the diameter (Fig.  2.C) behaves very differently from , indicating that the B-wave is modulated by the physiological response which is sensitive to the extent of the excited surface at the retina (see Fig. 2.D). [26] It is possible to compare the excitation rates for SPA and TPA (Equation 2 and Equation 3), using parameter values found in the literature. The SPA cross-section σ 1 for the cis-ground state can be derived from the molar extinction coefficient of rhodopsin at 500 nm (n = 40200 cm -1 M -1 ). [27] The TPA cross-section σ 2 was previously determined by Palczewska et al. [4] : σ 2 = 10 -50 cm 4 s. The full list of parameters used in the calculations are provided in Table 1. For both SPA and TPA, we used the E 1/2 value from the Naka-Rushton fit. The intensity is calculated as I = .
Note that this simplified model assumes a square pulse. The rates found for both conditions are very similar: σ 1 I = 2.6 ±1.3 · 10 9 s -1 and σ 2 I 2 = 1.3 ± 0.7 · 10 10 s -1 for SPA and TPA respectively. We assumed an uncertainty of 5 fs on the pulse duration, 5 µm on the diameter and used the error given by the fitting procedure for the energy (SPA: 3.1 pJ and (2) (3) (4) sitive to the pulse duration, beam size and energy. Interestingly, the dependence on pulse duration and energy follows the trend set by the I 2 dependence of TPA, while the behavior for diam-two different animals. An ERG is measured every 2 seconds upon 5 pulse irradiation at 1 kHz. Every 2 ERGs, the π-step is spectrally shifted by a few nm (2.3 and 1.5 for mouse 1 and 2 respectively, corresponding to 10 and 15 pixels on the SLM). The scan is repeated during 45 min, 3 times per mouse (with one anaesthesia every 45 min).The same experiment on rhodamine 6G in ethanol is performed using a two-lens scheme (ƒ = 3.5 cm), a filter (BG40, Thorlabs), and a photomultiplier tube (Hamamatsu H7826). Fig. 4.A shows the experimental laser spectrum centered at 960 nm (dots). In Fig. 4.B we show the calculated temporal profile assuming a positive chirp of 1200 fs 2 /rad (black spectral phase trace in Fig. 4.A) to account for the fact that the pulse autocorrelation indicates a pulse of approximately 120 fs for a spectrum which is capable of supporting a Fourier-transform pulse duration of 27 fs. The blue trace in Fig. 4.B corresponds to the calculated temporal profile when the π-step is set at the center of the pulse spectrum (960 nm, blue spectral phase trace in Fig. 4.A). This condition corresponds to the largest modulation we can expect from a signal scaling with I 2 , as shown in the continuous trace in Fig. 4.C, calculated by integrating the square of the temporal dependence (Fig. 4.B) at each spectral position. This is exactly what is observed for the rhodamine 6G trace in Fig. 4.D (green circles), where the minimum of the fluorescence as a function of the π-step position occurs around this position. We then repeated the procedure in living mice monitoring the B-wave amplitude as a reporter for phase-modulation. The results obtained independently for two mice are shown in Fig. 4.D (red triangles). We observed that the B-wave amplitude obtained from living mice and the fluorescence of Rhodamine 6G upon TPA have the same trend: a decrease when the π -step is located at the center of the spectrum. To highlight the good agreement between the in vivo experiment and the theoretical behavior, in Fig.  4.C, we overlapped the experimental trace obtained on mouse 1 with the calculated spectral response.

Conclusion
We performed a comprehensive study of the visual process generated by TPA. The ERG response upon infrared excitation is sen-  Fig. 4. π-step experimental results and simulations. A: Measured spectrum (red dots), spectral phase representations of a positive linear chirp of 1200 fs2/rad with (blue) and without (black) π-step at 960 nm. B: Calculated temporal intensity as a function of time for chirped pulses with a π-step at 960 nm (blue line) and without π-step (black line). C: Calculated amplitude modulation for different π-step positions (red lines). Normalized B-wave amplitude of mouse 1 (red dots). D: B-wave amplitude obtained on two different mice (red triangles) and by fluorescence on Rhodamine 6G in ethanol (green circle) as a function of the π-phase step spectral position.
eter is different, pointing towards the interplay of a short-term photodynamic process (absorption by the rhodopsin and photoisomerization) and a longer term physiological modulation of the visual response. We demonstrated the acquisition of ERG signals by TPA on living mice. In this context, we showed it is possible to perform spectral-phase shaping of the infrared beam to control the amplitude of the generated B-wave.

License and Terms
This is an Open Access article under the terms of the Creative Commons Attribution License CC BY 4.0. The material may not be used for commercial purposes. The license is subject to the CHIMIA terms and conditions: (https://chimia.ch/chimia/about).
The definitive version of this article is the electronic one that can be found at https://doi.org/10.2533/chimia.2022.570