Multiscale photoacoustic tomography of a genetically encoded near-infrared FRET biosensor

Photoacoustic tomography with genetically encoded near-infrared probes enables visualization of specific cell populations in vivo at high resolution in deep tissues. Here, we report a near-infrared Forster resonance energy transfer biosensor based on a pair of the near-infrared fluorescent proteins, miRFP670-iRFP720, which enables dynamic photoacoustic imaging of active biological processes. Specifically, we detected apoptosis in single cells at a resolution of ~3 µm in a mouse ear and mouse brain tumors (>3 mm in depth) at a spatial resolution of ~150 µm. These results open the way for high-resolution photoacoustic imaging of dynamic biological processes in deep tissues.


INTRODUCTION
Photoacoustic (PA) tomography (PAT) breaks the optical diffusion limit by acoustically detecting optical absorption contrast. [1] Acoustic waves are orders of magnitude less scattered in biological tissues, providing PAT far better spatial resolution than pure optical imaging in deep tissues (>2 mm). [2,3] PAT is sensitive to the optical absorption of molecules. It is powerful at detecting both endogenous molecules, such as hemoglobin, cytochromes, DNA/RNA, and melanin, and exogenous probes, such as organic dyes, nanoparticles, and genetically encoded chromophore-containing proteins. [4][5][6][7][8][9][10][11][12][13][14][15][16][17][18] The use of genetically encoded bacteriophytochrome (BphP) based near-infrared (NIR) fluorescent proteins (FPs) as contrast molecules in PAT provides the most advanced technology for deep-tissue visualization of molecules and cells at high spatial resolution in vivo. [19][20][21][22] The BphP-based molecules with absorption peaks at 640-780 nm can be clearly distinguished from hemoglobin and provide the best sensitivity for in vivo visualization of cell populations. [23,24] As chromophores, they incorporate biliverdin, which is abundant in eukaryotic cells as a product of heme metabolism. [25] While PAT with BphP-based molecules is rapidly developing, [19,20] so far, it has been applied to structural imaging, leaving dynamic molecular processes largely unexplored.
Biological phenomena result from physico-chemical processes of molecular binding, association, conformational change, and catalysis. [26] To visualize dynamic processes, it is necessary to elucidate the functional states of the constituent molecules at different time points. Förster Resonance Energy Transfer (FRET), uniquely sensitive to molecular conformation, association, and separation in the 1-10 nm range, can resolve molecular interactions and conformations.
Naturally, fluorescence microscopy is suited for FRET imaging, which has provided valuable information for biomedical research. Multiphoton microscopy has been widely used for FRET imaging with extended penetration depth. [27][28][29][30] However, strong optical scattering in biological tissue impedes high spatial resolution fluorescence imaging of FRET at depths beyond 2 mm. Photoacoustic FRET imaging of dyes [31] and FPs of green fluorescent protein (GFP) family-based biosensors [32] was shown in model systems with purified molecules in vitro. However, the use of FRET sensors absorbing in blue/green light range seems to be incompatible with PAT imaging of deep tissues in vivo, because the strong optical absorption and scattering from endogenous molecules fundamentally limit blue/green light penetration. [33] Recently, following the development of spectrally distinct NIR FPs, such as those of (m)iRFP series of proteins, [34][35][36] genetically encoded NIR FRET biosensors became available. NIR FRET biosensors for GTPase, [36] protein kinases, [36,37] and calcium dynamics [38] allowed multiplexing with visible FPs for monitoring several processes in single cells. [39] NIR biosensors also can be combined with blue-light controlled optogenetic tools for cross-talk free all-optical control and readout. [39] NIR spectrum of these biosensors is obviously advantageous for their deep and sensitive imaging in living cells and live animals, because of deeper light penetration, less scattering and minimal autofluorescence. [40] In this work, we applied a genetically encoded NIR FRET biosensor as a probe in PAT. We performed multiscale imaging of biological processes based on FRET in vivo for the first time. We investigated two incarnations of PAT, photoacoustic microscopy (PAM) for single cells and photoacoustic computed tomography (PACT) for a deep-seated tumor in the brain, allowing imaging at different scales of spatial resolution and penetration depth.

METHODS
To explore NIR FRET with PAT, we developed an improved FRET biosensor for caspase-3 based on miRFP670 donor and iRFP720 acceptor separated by the caspase-3 cleavage site (Fig. 1A). We chose to work with the caspase-3 biosensor, because this type of biosensor has been extensively characterized and provides robust responses in single cells. Thus, it perfectly suits the need to test the performance of the novel technology. The optical absorption spectra of miRFP670 and iRFP720 are red-shifted relative to hemoglobin (Fig. 1B). The donor miRFP670 [35] is spectrally similar to the previously developed dimeric iRFP670 [34] that was successfully used in PAT. The miRFP670-iRFP720 FRET pair is characterized by a substantial spectral overlap between the donor fluorescence and acceptor excitation spectra (Fig. 1C). The miRFP670-iRFP720 caspase-3 biosensor provided the highest response to cleavage in the suspension of mammalian cells. When excited at 610 nm, the fluorescence changes in the donor channel were almost 2-fold at the peak emission wavelength (Fig. 1D). This is higher than that of the previously reported caspase-3 biosensor containing a monomeric miRFP720 acceptor and having the reverse orientation of the donor and the acceptor in a fusion. According to the spectra (Fig. 1D) obtained for the HeLa cells stably expressing the miRFP670-iRFP720 biosensor, the donor/FRET ratio was 49%, compared to 34% reported for the biosensor with the miRFP720 [36]. Upon excitation, for the miRFP670-iRFP720 caspase-3 biosensor, the generated PA signal consists of the following Eq. (1): Here, is the PA signal generated by the donor miRFP670 in the presence of the acceptor iRFP720. is larger than , which is the PA signal generated by the donor in the absence of FRET, because the energy absorbed by the donor is transferred to the acceptor due to FRET [31]. is the direct PA signal generated by the acceptor iRFP720. In this work, we used 610-nm light for FRET imaging. Although 610 nm is away from the donor excitation peak at 643 nm (Fig. 1B), a use of 610 nm allows minimizing background by reducing the direct absorption from the acceptor (iRFP720). is the PA signal from endogenous molecules at 610 nm. Once caspase-3 cleaves the -DEVD-linker, miRFP670 and iRFP720 become separated from each other, resulting in no FRET. The PA signals can be expressed as: By detecting the PA signal difference between and , we can photoacoustically monitor changes in FRET. To obtain miRFP670-iRFP720 caspase-3 biosensor, iRFP720 acceptor was inserted, and the fluorescent proteins miRFP670 and iRFP720 were swapped. The linker between two proteins is GGDEVDGPVAT and the plasmid was obtained from pEGFP-N1 plasmid. First, we tested the miRFP670-iRFP720 biosensor, which was stably expressed in HeLa cells, using fluorescence microscopy. Upon addition of staurosporine (STS) to the cells, caspase-3 activation resulted in biosensor responses (Fig.  2). The monitored fluorescence intensity ratio between the FRET (605/30 nm excitation and 725/40 nm emission filters) and the donor (605/30 nm excitation and 667/30 nm emission filters) channels decreased to the minimum of ~40% in 30 min for individual cells and in less than 2 h for a population ( Fig. 2A, B). Fluorescence images (Fig. 2C) illustrate the dynamics in individual cells at subcellular resolution.
As shown in Figure 3A, a pulsed Nd:YAG laser pumps a dye laser to provide 610-nm light for PA excitation. The illumination beam is focused by an objective through a ring-transducer into the mouse ear from the top. The resultant PA waves are detected by a ring-shape ultrasonic transducer. Volumetric imaging is acquired by two-dimensional raster scanning of the mouse ear. The PAM system provides a spatial resolution of ~3-4 µm with a penetration depth of 1 mm. The PACT system used in this study is shown in Fig. 3B. A lab-made optical parametric oscillator (OPO) laser, pumped by an Nd:YAG laser with a 10-Hz pulse-repetition rate, is used to output 610 nm for PA excitation. The laser beam is first homogenized by an optical diffuser (EDC-5, RPC Photonics) and then illuminated the mouse brain from the top. The maximum light fluence on the skin of the animal is ∼10 mJ cm -2 , which is well below the American National Standards Institute safety limit. The PA signals are detected by a full-ring ultrasonic transducer array (Imasonic). A lab-made 512channel preamplifier was directly connected to the ultrasonic transducer array housing. The pre-amplified photoacoustic signals were digitized using a 512-channel data acquisition system. The digitized radio frequency data were first stored in the onboard buffer, then transferred to a computer. The digitized raw data were fed into a half-time dual-speed-of-sound universal back-projection algorithm for image reconstruction.

RESULTS
Next, we visualized the FRET sensor expressing cells in the same conditions using optical-resolution PAM (Fig. 3A).
Under 610-nm illumination, PAM [41][42][43] revealed the HeLa cells at a spatial resolution of ~3-4 µm. After baseline imaging, STS was added to the cell culture media to induced apoptosis. The kinetics of the PA signals for individual cells (~30 min) and the population (less than 2 h) were similar to those observed in fluorescence measurements (Fig. 4A). After 120 min, the PA signals decreased to ~70% of the baseline level (Fig. 4B). PAM images also visualize the signal changes in individual HeLa cells during the apoptosis at subcellular resolution (Fig. 4C). To monitor the FRET processes in deep tissue, we imaged deep-seated tumors in the brain using PACT [7,44] (Fig. 5A).
A tumor was induced in the mouse brain by the injection of 1×10 6 HeLa cells expressing miRFP670-iRFP720 caspase-3 biosensor. Three weeks after injection, the mouse was imaged by PACT. During the in vivo experiments, the mouse was mounted onto a holder with the water bag placed on top, and ultrasound gel applied between the scalp and the water bag for ultrasonic coupling. To illuminate the whole brain, a broad laser beam set at 610 nm was used. Throughout the experiments, the scalp was not removed. We first obtained the baseline image of the tumor (3 mm beneath the scalp), we then injected STS (10 µL) in DMSO into the tumor to induce cell apoptosis. The mouse brain was monitored using PACT for ~3 hours. PACT image of a mouse brain 2-h post-injection is shown in Fig. 5B, where the tumor was highlighted by computing the difference from the baseline image. A threshold level of four times the noise level, estimated as the standard deviation of the background signal outside the imaged region, was applied. We selected the three regions for monitoring of the signal changes: the tumor region, the contralateral region, and the biggest vessel in the brain (Fig. 5B). PA signals of the tumor region decreased obviously, while PA signals from non-tumor regions had no significant changes, which indicates the caspase-3 activities inside tumor cells after the STS injection (Fig. 5C, D).

DISCUSSION
Here, we demonstrated multiscale PA imaging of FRET biosensors in living cells and live animals for the first time, to the best of our knowledge. We observed that fluorescence imaging of the same NIR biosensor in similar conditions correlated well with the PA imaging. Specifically, the kinetics of the biosensor responses in individual cells, the variability between cells, and responses in large cell populations were similar. Further, the kinetics of biosensor responses is similar for cell populations observed in vivo, showing the PA signals decreased to a plateau in 120-150 min after the STS stimulation. We explored the possibility of applying PAT at different resolutions and penetration depths in vivo. In both PAM and PACT, we distinguished vasculature based on hemoglobin contrast from tumor cells expressing the NIR FRET biosensor. PAM has monitored the FRET process at single-cell resolution within 1-mm in depth. PACT has visualized the FRET dynamic in deep seated tumors (>3 mm in depth) at 150-µm spatial resolution. For fluorescence imaging, it is currently not possible to obtain anatomical and functional details at such resolution beyond 1 mm in depth.