Multiscale Photoacoustic Tomography of a Genetically Encoded Near‐Infrared FRET Biosensor

RESEARCH ARTICLE

www.advancedscience.com

Multiscale Photoacoustic Tomography of a Genetically

Encoded Near-Infrared FRET Biosensor

Lei Li, Hsun-Chia Hsu, Vladislav V. Verkhusha, Lihong V. Wang,*

and Daria M. Shcherbakova*

Photoacoustic tomography (PAT) with genetically encoded near-infrared

probes enables visualization of specific cell populations in vivo at high

resolution deeply in biological tissues. However, because of a lack of proper

probes, PAT of cellular dynamics remains unexplored. Here, the authors

report a near-infrared Forster resonance energy transfer (FRET) biosensor

based on a miRFP670-iRFP720 pair of the near-infrared fluorescent proteins,

which enables dynamic functional imaging of active biological processes in

deep tissues. By photoacoustically detecting the changes in the optical

absorption of the miRFP670 FRET-donor, they monitored cell apoptosis in

deep tissue at high spatiotemporal resolution using PAT. Specifically, they

detected apoptosis in single cells at a resolution of

≈

minamouseear

tumor, and in deep brain tumors (

>

3 mm beneath the scalp) of living mice at

a spatial resolution of

≈

150

m with a 20 Hz frame rate. These results open

the way for high-resolution photoacoustic imaging of dynamic biological

processes in deep tissues using NIR biosensors and PAT.

L. Li, H.-C. Hsu, L. V. Wang

Caltech Optical Imaging Laboratory, Andrew and Peggy Cherng

Department of Medical Engineering and Department of Electrical

Engineering

California Institute of Technology

Pasadena, CA 91125, USA

E-mail: lvw@caltech.edu

V. V. Verkhusha

Medicum, Faculty of Medicine

University of Helsinki

Helsinki 00290, Finland

V. V. Verkhusha, D. M. Shcherbakova

Department of Anatomy and Structural Biology and Gruss-Lipper

Biophotonics Center

Albert Einstein College of Medicine

Bronx, NY 10461, USA

E-mail: daria.shcherbakova@einsteinmed.org

V. V. Verkhusha

Science Center for Genetics and Life Sciences

Sirius University of Science and Technology

Sochi 354340, Russia

The ORCID identification number(s) for the author(s) of this article

can be found under https://doi.org/10.1002/advs.202102474

This is an open access article under the terms of the Creative Commons

Attribution License, which permits use, distribution and reproduction in

any medium, provided the original work is properly cited.

DOI: 10.1002/advs.202102474

1. 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 biolog-

ical tissues, providing PAT far better

spatial resolution than pure optical imag-

ing 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, deoxyribonu-

cleic acid/ribonucleic acid (DNA/RNA),

and melanin, and exogenous probes, such

as organic dyes, nanoparticles, and genet-

ically encoded chromophore-containing

proteins.

[4–14]

The use of genetically encoded bacterio-

phytochrome (BphP) based near-infrared

(NIR) fluorescent proteins (FPs) as contrast molecules in PAT

provides the most advanced technology for deep-tissue visualiza-

tion of molecules and cells at high spatial resolution in vivo.

[15–21]

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.

[22]

chromophores, they incorporate biliverdin, which is abundant in

eukaryotic cells as a product of heme metabolism.

[23]

While PAT

with BphP-based molecules is rapidly developing,

[17,19,24]

so far, it

has been applied to structural imaging, leaving dynamic molec-

ular processes largely unexplored.

Biological phenomena result from physico-chemical processes

of molecular binding, association, conformational change, and

catalysis.

[25]

To visualize dynamic processes, it is necessary to elu-

cidate the functional states of the constituent molecules at dif-

ferent time points. Förster resonance energy transfer (FRET),

uniquely sensitive to molecular conformation, association, and

separation in the 1–10 nm range, can resolve molecular interac-

tions and conformations.

[26]

Naturally, fluorescence microscopy

is suited for FRET imaging, which has provided valuable in-

formation for biomedical research. Multiphoton microscopy has

been widely used for FRET imaging with extended penetration

depth.

[27–32]

However, strong optical scattering in biological tis-

sue impedes high spatial resolution fluorescence imaging of

FRET at depths beyond 2 mm. Photoacoustic FRET imaging

of dyes

[33]

and FPs of green fluorescent protein (GFP) family-

based biosensors

[26]

was shown in model systems with purified

Adv. Sci.

2021

, 2102474

2102474 (1 of 8)

www.advancedsciencenews.com

www.advancedscience.com

Figure 1.

Characterization of a NIR miRFP670-iRFP720 FRET caspase-3 biosensor and PA imaging setup. A) Schematic design of FRET miRFP670-

iRFP720 caspase-3 biosensor that is activated during apoptosis. Activated caspase-3 cleaves the -DEVD- linker and releases miRFP670 and iRFP720

from each other, thus decreasing FRET. PA: photoacoustic signal. FR: fluorescence. B) Overlay of molar extinction spectra of oxyhemoglobin (HbO2),

deoxyhemoglobin (HbR), miRFP670, and iRFP720. C) Excitation and emission spectra for miRFP670 and iRFP720. Dashed area indicates spectral overlap

between the emission spectrum of miRFP670 (donor) and excitation spectrum of iRFP720 (acceptor). D) Spectral changes of miRFP670-iRFP720 caspase-

3 biosensor measured in the suspension of HeLa cells stably expressing biosensor before (black line) and 3 h after (red line) staurosporine (STS)-ind

uced

apoptosis. The spectra were normalized to the fluorescence intensity of iRFP720 excited at 700 nm, which does not change during apoptosis.

molecules in vitro. However, the use of FRET sensors absorbing

in the blue/green light range seems to be incompatible with PAT

imaging of deep tissues in vivo, because the strong optical ab-

sorption and scattering from endogenous molecules fundamen-

tally limit blue/green light penetration.

[34]

Recently, following the development of spectrally distinct NIR

FPs, such as those of (m)iRFP series of proteins,

[35–39]

genet-

ically encoded NIR FRET biosensors became available. NIR

FRET biosensors for GTPase,

[37]

protein kinases,

[37,40]

and cal-

ciumdynamics

[41]

allowedmultiplexingwithvisibleFPsformon-

itoring several processes in single cells.

[42]

NIR biosensors also

can be combined with blue-light controlled optogenetic tools

for cross-talk free all-optical control and readout.

[42]

NIR spec-

trum of these biosensors is obviously advantageous for their

deep and sensitive imaging in living cells and live animals, be-

cause of deeper light penetration, less scattering, and minimal

autofluorescence.

[43]

In this work, we applied a genetically encoded NIR FRET

biosensor as a probe in PAT. We performed multiscale imag-

ing of biological processes based on FRET in vivo for the first

time. We investigated two incarnations of PAT, photoacoustic mi-

croscopy (PAM)

[44]

for a shallow tumor in a mouse ear at optical

resolution and photoacoustic computed tomography (PACT)

[9]

for a deep-seated tumor in the brain at a resolution limited by

acoustic diffraction, allowing imaging at different scales of spa-

tial resolution and penetration depth. We demonstrated that the

FRET-based caspase-3 biosensor enabled visualization of drug-

induced apoptosis in single cells at optical resolution using both

PAM (3

m) and fluorescence microscopy (resolution limited by

diffraction,

≈

350 nm). We also monitored apoptosis in brain tu-

mors (

3 mm beneath the scalp) in vivo at 150-

m resolution

using PACT.

2. Results

2.1. Characterization of a NIR FRET Biosensor

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 (

Figure 1A

). We

chose to work with the caspase-3 biosensor, because this type of

biosensors has been extensively characterized and provides ro-

bust responses in single cells. Thus, it perfectly suits the need

to test the performance of the novel technology. The optical ab-

sorption spectra of miRFP670 and iRFP720 are red-shifted rela-

tive to hemoglobin (Figure 1B). The donor miRFP670

[36]

is spec-

trally similar to the previously developed dimeric iRFP670

[35]

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 (Figure 1C).

The miRFP670-iRFP720 caspase-3 biosensor provided the high-

est response to cleavage in the suspension of mammalian cells.

When excited at 610 nm, the fluorescence changes in the donor

channel were almost twofold at the peak emission wavelength

Adv. Sci.

2021

, 2102474

2102474 (2 of 8)

www.advancedsciencenews.com

www.advancedscience.com

Figure 2.

Fluorescence imaging of caspase-3 activity in HeLa cells stably expressing miRFP670-iRFP720 caspase-3 biosensor during STS-induced apop-

tosis. A) Kinetics of the fluorescence intensity ratio (FRET/miRFP670) for individual cells undergoing apoptosis. Each line represents one single c

ell.

The images were taken every 10 min. B) Fluorescence intensity ratio (FRET/miRFP670) before and after STS-induced apoptosis.

30 cells, error bar,

s.d. ***,

0.001, calculated using a paired Student’s

-test. C) Representative fluorescence images in FRET (red) and miRFP670 (blue) channels at

selected time points. Scale bars, 10

(Figure 1D). This is higher than that of the previously reported

caspase-3 biosensor containing a monomeric miRFP720 accep-

tor and having the reverse orientation of the donor and the

acceptor in a fusion.

[37]

According to the spectra (Figure 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.

[37]

Upon excitation, for the miRFP670-iRFP720 caspase-3 biosen-

sor, the generated PA signal consists of the following Equa-

tion (1):

(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.

[33]

is the direct PA signal gener-

ated by the acceptor iRFP720. In this work, we used 610-nm light

for FRET imaging. Although 610 nm is away from the donor ex-

citation peak at 643 nm (Figure 1B), a use of 610 nm allows for

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, re-

sulting in no FRET. The PA signals can be expressed as:

(2)

By detecting the PA signal difference between

and

we can photoacoustically monitor changes in FRET.

2.2. Visualization of the Caspase-3 Activity in Live Cells

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 acti-

vation resulted in biosensor responses (

Figure 2

). The monitored

fluorescence intensity ratio between the FRET (605/30 nm excita-

tion 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 (Figure 2A,B). Fluorescence im-

ages (Figure 2C,D) illustrate the dynamics in individual cells at

subcellular resolution.

Next, we visualized the FRET sensor expressing cells in the

same conditions using optical-resolution PAM (

Figure 3A

). Un-

der 610-nm illumination, PAM

[44]

revealed the HeLa cells at a

spatial resolution of

≈

3–4

m. After baseline imaging, STS was

added to the cell culture media to induce apoptosis. The kinetics

of the PA signals for individual cells (

≈

30 min) and the popu-

lation (less than 2 h) were similar to those observed in fluores-

cence measurements (Figure 3B). After 120 min, the PA signals

decreased to

≈

70% of the baseline level (Figure 3C). PAM images

also visualize the signal changes in individual HeLa cells during

the apoptosis at subcellular resolution (Figure 3D).

2.3. Multiscale PAT of Caspase-3 Activities In Vivo

We then carried out PA FRET imaging in vivo using both PAM

and PACT. We first imaged the caspase-3 activities in a mouse ear

tumor using PAM at a single-cell resolution. A xenograft tumor

was induced in the mouse ear by the injection of 1

HeLa

Adv. Sci.

2021

, 2102474

2102474 (3 of 8)

www.advancedsciencenews.com

www.advancedscience.com

Figure 3.

PAM of caspase-3 activity in HeLa cells stably expressing miRFP670-iRFP720 caspase-3 biosensor during STS-induced apoptosis. A) Setup of

optical-resolution PAM. B) PA amplitude changes for individual cells undergoing apoptosis. Each line represents one single cell. The images were ta

ken

every 10 min. C) PA amplitude before and after STS-induced apoptosis.

40 cells, error bar, s.d. ***,

0.001, calculated using a paired Student’s

-test. D) Representative PAM images of cells at selected time points. Scale bars, 10

cells expressing miRFP670-iRFP720 caspase-3 biosensor. Three

days after injection, the mouse ear was imaged by PAM with 610-

nm illumination at a spatial resolution of

≈

3–4

m. The tumor

and the ear vasculature (vessels are shown in gray and the tumor

is shown in color) were clearly resolved, as shown in

Figure 4A

After baseline scanning, STS was injected into the tumor subcu-

taneously. We then monitored the PA signal changes in the tu-

mor area. An obvious PA signal decrease was observed after the

injection, as shown in Figure 4B, indicating the caspase-3 activity

during cell apoptosis.

To monitor the FRET processes in deep tissue, we imaged

deep-seated tumors in the brain using PACT

[9]

(

Figure 5A

). A

tumor was induced in the mouse brain by the injection of 1

HeLa cells expressing miRFP670-iRFP720 caspase-3 biosen-

sor. 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 ap-

plied 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. The optical fluence on the scalp was 10 mJ cm

−

Throughout the experiments, the scalp was not removed. In the

experimental group, we first obtained the baseline image of the

tumor (3 mm beneath the scalp), we then injected STS (10

L) in

DMSO into the tumor to induce cell apoptosis. The mouse brain

was monitored using PACT for

≈

3 h. PACT image of a mouse

brain 2-h post-injection is shown in Figure 5B, where the tumor

was highlighted by computing the difference from the baseline

image. A threshold level of four times the noise level, estimated

asthestandarddeviationofthebackgroundsignaloutsidetheim-

aged region, was applied. We selected the three regions for mon-

itoring of the signal changes: the tumor region, the contralateral

region, and the biggest vessel in the brain (Figure 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

(Figure 5C,D). In the control group, we injected DMSO (10

into the tumor and monitored the PA signal changes for

≈

post-injection. Signals from three similar regions (as labeled in

Figure 5B) were analyzed and are plotted in Figure 5E. No sig-

nificant changes were observed either inside the tumor or in the

normal tissue (Figure 5F).

3. Discussion

Here, we demonstrated multiscale PA imaging of FRET biosen-

sors in living cells and live animals for the first time, to the best

of our knowledge. We observed that fluorescence imaging of the

sameNIRbiosensorinsimilarconditionscorrelatedwellwiththe

PA imaging. Specifically, the kinetics of the biosensor responses

in individual cells, the variability between cells, and responses in

large cell populations (Figures 2A and 3B) 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 (Figures 4B and 5C).

We explored the possibility of applying PAT at different reso-

lutions and penetration depths in vivo. In both PAM and PACT,

wedistinguishedvasculaturebasedonhemoglobincontrastfrom

tumor cells expressing the NIR FRET biosensor. PAM has mon-

itored the FRET process at single-cell resolution within 1-mm in

depth. PACT has visualized the FRET dynamic in deep seated

Adv. Sci.

2021

, 2102474

2102474 (4 of 8)