Five-dimensional two-photon volumetric microscopy of in-vivo dynamic activities using liquid lens remote focusing

. Multi-photon scanning microscopy provides a robust tool for optical sectioning, which can be used to capture fast biological events such as blood flow, mitochondrial activity, and neuronal action potentials. For many studies, it is important to visualize several different focal planes at a rate akin to the biological event frequency. Typically, a microscope is equipped with mechanical elements to move either the sample or the objective lens to capture volumetric information, but these strategies are limited due to their slow speeds or inertial artifacts. To overcome this problem, remote focusing methods have been developed to shift the focal plane axially without physical movement of the sample or the microscope. Among these methods is liquid lens technology, which adjusts the focus of the lens by changing the wettability of the liquid and hence its curvature. Liquid lenses are inexpensive active optical elements that have the potential for fast multi-photon volumetric imaging, hence a promising and accessible approach for the study of biological systems with complex dynamics. Although remote focusing using liquid lens technology can be used for volumetric point scanning multi-photon microscopy, optical aberrations and the effects of high energy laser pulses have been concerns in its implementation. In this paper, we characterize a liquid lens and validate its use in relevant biological applications. We measured optical aberrations that are caused by the liquid lens, and calculated its response time, defocus hysteresis, and thermal response to a pulsed laser. We applied this method of remote focusing for imaging and measurement of multiple in-vivo specimens, including mesenchymal stem cell dynamics, mouse tibialis anterior muscle mitochondrial electrical potential fluctuations, and mouse brain neural activity. Our system produces 5 dimensional (x,y,z,λ,t) data sets at the speed of 4.2 volumes per second over volumes as large as 160 x 160 x 35 μm. © 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

vascular flow patterns, or homing, rapid scanning of a relatively small volume is needed. In addition, for phenomenon such as neural firing, multiple events occur at different depths of the volume under study that need to be rapidly accessed [1,2]. In all of these situations, rapid lateral and axial scanning are needed to accurately capture biological dynamics. Lateral scanning that couples a galvo-driven slow axis with a fast axis using a resonant galvo or polygonal mirror can create a 2D scan field at 30-120 Hz or more depending on the field of view. Rapid axial scanning can reproducibly and precisely translate that to several volumes per second depending on the number of Z positions required.
Many point scanning microscopes generate an axial stack using mechanical displacement of the sample with a stepper motor [3], which are very precise and have several millimeters of travel, but are too slow for some live imaging applications. To rapidly scan during imaging, the objective or stage can be moved using a piezo mount, which works at several tens of hertz, but is limited by inertial forces that cause instability at higher speeds due to the weight of the objective lens. Another limitation of mechanical scanning is the addition of depthdependent optical aberrations, especially spherical aberrations [4]. Optical strategies that use simultaneous multi-focal imaging are also possible [5][6][7][8], but require highly specialized optical setups and often have optical performance trade-offs. Remote focusing is one of the simplest methods of rapidly modulating the focal plane using an active element remotely from the sample to adjust the axial focal plane. A number of technologies and active elements to use this process in point-scanning microscopy have been proposed [9], such as a voice coil motor [10], an acousto-optic modulator [11], a TAG lens [12], a deformable mirror [13,14], a thermal lens [15], an Alvarez lens [16], and a liquid lens [17][18][19][20][21][22][23]. Liquid lens technology is an inexpensive and stable remote focusing technology, and uses either hydraulic pressure [20][21][22] or the electrowetting phenomenon [17][18][19] to change the optical power of the lens. Liquid lens remote focusing is an easy to implement and affordable method for fast inertiafree volumetric imaging. However, high energy pulses required for multi-photon microscopy have drawn some concerns about long term use of liquid lenses with ultrafast lasers. In addition, adoption has been slow due to perceived challenges arising from suspected optical aberrations and difficult calibration of the liquid lens.
In this work we use an electrowetting liquid lens for remote focusing in multi-photon microscopy. Electrowetting occurs at the boundary of two immiscible liquids with different refractive indices that form an electrically adjustable curvature and hence an adjustable focusing power [17], with a settling time of a few milliseconds -a timescale useful for many biological phenomena. We measured the aberrations using a Shack-Hartmann wavefront sensor (SHWFS), and characterized hysteresis of defocus and several Zernike modes of the lens using a ramp signal with different input frequencies. We further measure the step response of the liquid lens. We calibrated our system using a test target made of fluorescent beads in a gel compared with mechanical Z-scanning, and were able to rapidly and reproducibly scan volumes. These measurements and analyses enabled us to measure and characterize the wavefront, in response to the input voltage. After the calibration step, we applied this method to multi-color volumetric imaging of live cells, blood flow in the skull of a live mouse, mouse tibialis anterior (TA) muscles, and neural activity in the mouse cerebral cortex.

Optical setup
Our microscope (Fig. 1) uses a Ti:Sapphire laser (Coherent Chameleon Ultra II) with a wavelength range of 680nm -1080nm, and intensity modulation using a Pockels cell (Conoptics). A second femtosecond fiber laser (Calmar Cazadero) source was used for testing the thermal reponse of the lens. The fiber laser source produces 1550 nm, 370 fs pulses which were frequency doubled to produce 775 nm. A pinhole was used at the focus of our beam expansion telescope to improve the uniformity and Gaussian properties of the beam. We limited the be liquid lens ( , where M is the magnification from the sample plane to the image plane right after the liquid lens, n is the refractive index of the immersion medium, f L3 and f LL are the focal lengths of lens L 2 and the liquid lens respectively. We calculate a theoretical 36.8 μm total focal length displacement given the optical power range of the liquid lens. This range was selected to match the 2D scanning rate of our system (110 frames per second) to provide a continuously scanned volume for a reasonable number of volumes per second (4.2 volumes/s, 26 frames per volume) for our biological imaging targets with Nyquist sampling. For other applications such as brain neural activity measurement in multiple cortical layers, this range could be altered to be as large as ~100 µm by changing the focal length of L 3 and magnification of the system to provide appropriate axial scanning.

Preparation of the fluorescent beads in a polyacrylamide gel
To produce a volume of point sources fixed in space we made a poly-acrylamide gel mixed with 200 nm fluorescent Tetraspec beads (Invitrogen T7280). The gel was made by combining 45 μl Tetraspec beads (diluted 1:50 in ultrapure water), 45 μl 30% Acrylamide (Bio-rad 1610156), 2 μl Ammonium persulfate, and 0.2 μl TEMED. The mixture was transferred to the well of a coverglass-bottom petri dish and imaged after the gel solidified.

Preparation of the cell sample
Mesenchymal stem cells (MSCs), isolated from a wild-type mouse, were grown on a 35mm Petri dish in growth medium (α-MEM, 10% fetal bovine serum, Penicillin Streptomycin, Lglutamine). MSCs were stained with lipophilic membrane dye DiD according to the manufacturer's protocol (Invitrogen) 30-40 minutes before imaging. Cells were transferred to a hemocytometer and imaged in suspension, with a heating strip and an infrared thermometer to maintain the temperature of the cells at 37° C.

Preparation of the tibialis anterior sample
The tibialis anterior (TA) muscle was extracted from a transgenic mouse with C57Bl/6 background ubiquitously expressing mitochondrial Dendra2 green/red photoswitchable monomeric fluorescent protein (Jackson Laboratory, #018385) [26], 1 hour before imaging. We permeabilized the muscle in buffer 1 containing ( ) dye was also added to the buffer to stain the nuclei. Following incubation with TMRE and NucBlue, the TA was rinsed in buffer 2 for 15 minutes. The TA was then removed and fixed to a dissection-gel petri dish at proximal and distal tendons. For imaging, the TA was submerged in buffer 2 and the following substrates were added to stimulate mitochondrial respiration: glutamate (10 mM), malate (5 mM), ADP (2.5 mM). To increase the rate of fluctuations, FCCP (1µM) was added.

Intravital imaging
For skull imaging, one hour before surgery a dose of Meloxicam (1mg/kg) was injected subcutaneously to the mouse. The mouse was initially anesthetized using 4% isoflurane (100 ml/min oxygen flow), and restrained using a 3D printed stereotaxic holder. The rate of isoflurane was then reduced to 1.5% during imaging. A 20 μL dose of 70 kDa rhodamine-B dextran (Nanocs) was administered through retro-orbital injection, to visualize the vasculature. Five minutes before making an incision, 50 μL of 0.25% bupivacaine was locally applied as analgesia. An incision was made on the scalp from between the eyes toward both the ears to make a flap. The periosteum layer was removed, and the area of imaging was cleaned using a cotton swab; immediately sterile phosphate buffered saline (PBS) was applied to the incision site. The animal was placed under the microscope objective and sterile PBS was added to fill the gap between the skull and the objective lens. After the imaging session, the animal was euthanized using CO 2 and cervical dislocation. For mouse neural activity experiments, four weeks prior to imaging, the mouse was transduced with a viral construct encoding the red fluorescent calcium sensor jRGECO in neurons, via direct cortical injection through a cranial window that was surgically placed above the barrel cortex (3mm diameter, 120um thickness). First, the animal was anesthetized using isoflurane (4% induction and 1.6% maintained). A craniotomy was performed above the barrel cortex (AP: −2, ML: −5 from bregma), and a 1 µl dose of pAAV.Syn.NES-jRGECO1a.WPRE.SV40 [27] virus was injected (addgene: #100854; Titer: 4.1 x 10 13 genome copies/mL), at a rate of 0.2 µl/min using a glass micropipette (Eppendorf Celltram Oil pump). A coverslip glass was positioned and cemented (UV cured dental cement, Henry Schein) on top of the craniotomy after infusion of 1% agarose on the surface of the exposed dura. Before imaging, the animal received one dose of a cocktail of ketamine/xylazine (156.25 mg/kg ketamine, 6.25 mg/kg xylazine) as anesthesia and placed on a 3D printed stereotaxic holder. Air blow stimulation were given through a PVC tube oriented towards the contralateral whiskers to the craniotomy.
All animal procedures and experiments were approved by the UGA Institutional Animal Care and Use Committee (IACUC).

Characterization of the liquid lens
We performed wavefront measurement to specifically characterize the optical aberrations produced by the liquid lens. For this purpose the liquid lens was directly conjugated to the SHWFS without passing through the rest of the system to eliminate system aberration contributions to the measurement. The acquired wavefronts from the SHWFS were reconstructed using the methods described in [28]. For conversion of the extracted defocus term to optical power in diopter unit we use the following formalities. The defocus term in radial coordinates has the following form [29]: where c 3 is the coefficient of the Zernike mode. We can write a general term for the focal distance f of a spherical surface [30] Because the simplify it to and therefore We applie to the liquid l 3(a)-(c) we s decreases. Th aberration is negligible (<0 consistent res defocus (Fig.  3 ith high peak s effect, we us nergy, and mea MHz, 10 MH uce 3.3 nJ, 10 n r respectively. ted peak pow low 60°C are s e high energy p shone on the li vals of 5 minut triangle signa tization noise n the hysteresis vefront measurem respectively. D sh and descending hy duced by the liqu he thermal respon e power, respectiv s was exposed to very 15 minutes energy is the sed a fiber laser asured the tem Hz, and 5 MH nJ, and 20 nJ p We observed wers (Fig. 3(f)suitable for op pulses on the l iquid lens for tes ( Fig. 3(i) The secon mitochondria dynamic elect [35] (mito-De 585/40 nm fil Fig. 6(a). We ( Fig. 6(b) id lens volume n R-GECO cal the barrel co s of air towards a suitable lo ure over an ar ormed using Im matic segment tensity over t ntensity thresho nd dilation (3 s separate reg pute intensity ver the volume tain the calcium lls might be di as one by the so rons from dif ( Fig. 8(c) Fig. 8. Volumetric imaging of calcium activity in the mouse barrel cortex. (A) shows a preprocessed volumetric frame from the timelapse. The volume dimensions are 80 µm x 80 µm x 25 µm (x,y,z). After 3D registration several ROIs were located in the data set using mean volumetric fluorescence intensity, shown in (B). From each ROI, the intensity fluctuation is derived and background normalized (F0) to obtain the DF/F0 calcium activity fluctuations, shown in (C). The red ticks (STIM) indicate the timing of an air blow stimulation on the contralateral whiskers to the recording site. 1050 nm excitation light was used for imaging. Visualization 4 is available in the supplemental information.

Conclusions
In conclusion, we presented characterization and application of a liquid lens for fast volumetric imaging of in vivo samples. One main concern for using the liquid lens is the potential for optical aberrations due to the curvature produced by the electrical potential. Here we used a wavefront sensor for characterizing wavefront error produced by the liquid lens. We show that other than the desired defocus, astigmatism is the main aberration error in the system; which was minimized by fine alignment of the liquid lens, resulting in a minimal magnitude. Another major concern with using liquid lens technology for remote focusing of femtosecond pulsed lasers is the effect of high powered pulses on the performance and stability of the lens. We therefore measured the temperature of the body of the device and its hysteresis response after 2 hours of continuous exposure and did not observe a significant increase in the temperature, alteration of focusing speed, or alteration of focusing accuracy. We also measured the full range step response of the device and a settling response time of 92 ms. To speed up the axial scanning and avoid over/undershoots, we used a triangular waveform and an electrical low pass filter which eliminates the high frequency digitization effect. We then calibrated the remote focusing method versus a precise mechanical scanning method, and calculated the ratio of voltage to focal shift. And finally, we applied this method to several relevant in-vivo dynamically varying samples such as cells, muscle mitochondrial activity, intravital imaging of blood flow in the skull, and neural firing in the barrel cortex. Liquid lens technology has recently attracted attention for applications such as headmountable brain neural activity measurement [39], or endoscopic imaging by providing both lateral [40] and axial [41,42] scanning. Due to their high potential for low cost and stability, many different methods of producing liquid lenses have been proposed. All of these methods and devices require careful characterization to generate high quality reproducible results. The characterization and calibration methods performed in this paper can be applied to those and other similar applications.

Funding
National Science Foundation (1706916); National Institutes of Health (R21EB027802); Center for Regenerative Engineering and Medicine (REM); Soft Bones Foundation.