2.1.

FRET-Sensor TR23K

We use a genetically encoded caspase-3 FRET-sensor TR23K based on a pair of FPs—TagRFP as a donor and nonfluorescent chromoprotein KFP as an acceptor, which are connected by a flexible amino acid linker GTGGSGGDEVDGTGGSGDPPVAT^52,53 [Fig. 1(a)]. Normalized absorption and emission spectra are presented in Fig. 1(b). Spectral overlap integral for TagRFP-KFP is significantly higher than that for similar proteins (Table 1). Because KFP is almost nonfluorescent,⁶² there is no spectral crosstalk of donor and acceptor, and it is possible to detect FRET efficiency based on fluorescence lifetime of the donor protein—TagRFP. The DEVD tetrapeptide is embedded in the linker and specifically recognized by caspase-3 and cleaved, leading to the increase of TagRFP fluorescence intensity and lifetime.⁵² The use of the latter parameter is more convenient because it is independent of such factors as donor photobleaching and concentration of the sensor. Despite the tetrameric structure of TagRFP-23-KFP sensor, the DEVD site in the linker is located far enough from the barrels of TagRFP and KFP to be cleaved by caspase-3 without steric hindrances.^53,57

Fig. 1

(a) Schematic representation of FRET-sensor TR23K. (b) Normalized absorption spectra of TagRFP and KFP and fluorescence spectra for TagRFP. KFP is nonfluorescent.

Table 1

Spectral overlap integrals for FRET pair of FPs.

2.2.

Cell Culture, Transfection, and Treatment of Tumor Cells

Tumor cell lines of lung adenocarcinoma A549 and larynx carcinoma HEp-2 originating from the ATCC were obtained from the Russian Cell Collection (St. Petersburg, Russia). Cells were cultured in complete RPMI-1640 (Paneco, Russia) medium with 10% fetal bovine (Bioclot, SA), penicillin ($100\text{units}/\mathrm{ml}$), and streptomycin ($100\mathrm{mg}/\mathrm{ml}$) in 5% ${\mathrm{CO}}_2$ at 37°C in a humidified incubator.

The plasmid pTR23K contains encodes TagRFP and KFP in a single reading frame with a linker sequence encoding 23 amino acids.⁵¹

Transfection of A549 and HEp-2 cell line was carried out using the pLV lentivector system (Evrogen, Russia). The obtained pLVTR23K plasmid contains TagRFP-23-KFP gene with codon usage optimized for high expression in mammalian cells. The titer of viral particles was estimated at ${10}^5\mathrm{T}.\mathrm{U}/\mathrm{ml}$. One day before transfection, cells were plated in six-well plates at $1\times {10}^4$ in 2 ml of growth medium RPMI-1640 containing 5% fetal bovine serum so that cells would be 70% confluent the time of transduction. Near-confluent A549 cells were incubated with the threefold excess of viral particles for 24 h before being replenished with fresh medium. The cells were examined on an inverted fluorescence microscope 48 h post-transduction. The cells were grown in complete RPMI-1640 (Paneco, Russia) medium with 10% fetal bovine (Bioclot, SA) serum for 10 passages without losing their fluorescent properties.

For detection of activation of caspase-3 in vitro after treatment with anticancer drugs cells were seeded on a $\mu$-Dish 35 mm (ibidi GmbH, Germany) and incubated in 5% ${\mathrm{CO}}_2$ at 37°C in a humidified incubator. After 48 h of growing paclitaxel (Lance, Russia) were added in concentration of $30\mu \mathrm{g}/\mathrm{ml}$ to the A549-TR23K and incubated for 24 h. After 48 h of growing a combination of etoposide (Teva, Israel)/cisplatin (Teva, Israel) were added to the HEp-2-TR23K in concentration of $200/10\mu \mathrm{g}/\mathrm{ml}$ and incubated for 24 h. Posttreatment caspase-3 activation was visualized on time-resolved confocal fluorescence microscope.

2.3.

FLIM-FRET Microscopy

FLIM-FRET of cells was performed using the time-resolved fluorescence confocal microscope MicroTime 200 (PicoQuant GmbH, Berlin, Germany). Fluorescence was excited by 532-nm picosecond diode laser PicoTA (PicoQuant GmbH, Berlin, Germany) (repetition rate and pulse-width are 80 MHz and 100 ps, respectively). The fluorescence signal was selected by a narrow bandpass filter (BP580/25, Chroma) and detected by a single-photon counting module (SPCM-AQRH, Perkin Elmer, Canada). Olympus (UPlanSApo $60\times /1.35$) has been used in all measurements. Cells were seeded on a $30\text{-}\mu$-Dish (ibidi GmbH, Germany), and after 24 h the intact cells were visualized before and after treatment with anticancer drugs. Fluorescence lifetime images were analyzed by SymphoTime analysis software (PicoQuant GmbH, Berlin, Germany). The amplitude weighted fluorescence lifetimes or mean fluorescence lifetimes (${\tau }_m=\frac{a_1{\tau }_1+a_2{\tau }_2}{a_1+a_2}$) were determined from the nonlinear least-square fitting of the decay curves. While single-exponential model was used to fit the decays of noninteracting donors, biexponential model was employed to fit the decay curves in case of interacting molecules as it is usually done in FLIM-FRET experiments.⁶³ The goodness of fit (chi-square) served as a guide for selection of the appropriate decay model.

2.4.

Animals

Female 12-weeks-old BALBc/nude mice (Animal Breeding Center, Pushchino, Russia) of 22 to 25 g weight were used for tumor cell inoculation. All experimental procedures were approved by the local ethical committee for animal experiments. Animals were treated humanely. All of the subcutaneous inoculations and tumor injections were administered with regard for alleviation of suffering. Mice were checked weekly for the development of tumors and for fluorescence signal. For fluorescence measurements, animals were anesthetized intramuscularly with a $50\text{-}\mu \mathrm{l}$ mixture of Zoletil 50 (Virbac, France) and Rometar 2% (Spofa Praha, Czech Republic) in equal proportion.

Cells for inoculation were harvested by trypsin treatment and washed in HEPES, $2\times {10}^6\text{cells}$ were suspended in 0.1 ml of Dulbecco’s phosphate buffer saline, inoculated dorsally in nude mice, and analyzed twice a week for fluorescence growing [Fig. 2(a)]. At indicated time points, the length, height, and width of tumors were measured by caliper. Tumor volumes ($V$) were calculated according to

Fig. 2

(a) 2-D planar fluorescence image of mice bearing tumors A549-TR23K before (left image) and after 7 days of Taxol treatment (right image) (iBox, ex. 502 to 547 nm, em. 580 to 640 nm); (b) FLIM measurements of intact anesthetized mouse using DCS-120 confocal scanner.

On the seventh day of growth, anticancer therapy was applied. For fluorescence measurements, animals were anesthetized intramuscularly with a $50\text{-}\mu \mathrm{l}$ mixture of Zoletil 50 (Virbac, France) and Rometar 2% (Spofa Praha, Czech Republic) in equal proportion.

All fluorescence measurements were performed using transdermal configuration for excitation and emission on a whole intact animal. Skin flap, glass window, or any clearance procedure was not used.

Taxol (paclitaxel Lance, Russia) was introduced i.v. at a dose of $15\mu \mathrm{g}/\mathrm{kg}$ for 5 days. Fluorescence was measured twice a week. Etoposide-Teva (Teva, The Netherlands) and cisplatin-Teva (Teva, Netherlands) were injected i.p. at a dose of 5 and $5\mu \mathrm{g}/\mathrm{kg}$ for 5 days i.v, respectively.

2.5.

Local Fluorescence Spectroscopy

The fluorescence spectra of the tumor xenografts were measured using a fiber optical spectrometer FS-003V spectrometer (CLUSTER Ltd., Russia). Fluorescence was excited by 532-nm diode-pumped solid-state laser in CW mode. The typical exposure times was 50 ms. A Y-shaped ring fiber bundle (seven fibers with a diameter of $100\mu \mathrm{m}$ and a numerical aperture of 0.22) was used to deliver the excitation laser radiation to the tissue (a central fiber) and to transmit the fluorescence radiation to the photodetector (six peripheral fibers). The laser power transmitted by the fiber was $\sim 5\mathrm{mW}$. The spatial resolution was 0.3 to 1 mm in the surface measurements. From three to five spectra were measured for each sample and processed using specialized software (LightView, CLUSTER Ltd., Russia).

2.6.

2-D Planar Imaging

2-D planar fluorescent images of anesthetized mouse were acquired in epifluorescence mode using the iBox imaging system (UVP) equipped with a 16-bit monochrome-cooled CCD camera (BioChemi HR Camera, UVP) and a 150-W halogen quartz lamp (BioLite, UVP). Animal plate was equipped by motorized platform for magnification and focusing. Measurements were performed using bandpass excitation (502 to 547 nm) and emission filters (570 to 640 nm). Images were obtained at various time points of tumor growth using vision works image acquisition, acquisition time was 60 s. Images were analyzed using vision works analysis software [Fig. 2(a)].

2.7.

FLIM-FRET of Tumor Xenografts

The fluorescence lifetime images of tumor xenografts were acquired on a macroscopic DCS-120 confocal scanning system (Becker & Hickl GmbH, Berlin, Germany) equipped with 532-nm picosecond diode laser PicoTA (PicoQuant GmbH, Berlin, Germany) (repetition rate and pulse-width are 80 MHz and 100 ps, respectively) and HPM-100-40 hybrid detector (Becker & Hickl GmbH, Berlin, Germany). Animals were put in a cassette on a mobile stage [Fig. 2(b)]. Fluorescence emission from a tumor was collected through the skin in the epi-illumination configuration.

The maximum diameter of the image area in the primary image plane of the scanner is about 15 mm. The size of the laser spot in the image plane is estimated to be $15\mu \mathrm{m}$. The average laser power incident on the sample is about $50\mu \mathrm{W}$. This resulted in energy density of $0.7\mu \mathrm{J}/{\mathrm{cm}}^2$, which is well below the damage threshold for biological samples. The numerical aperture of the detection beam is given by the beam diameter (about 3 mm) and the focal length of the scan lens (40 mm). The collection efficiency is thus considerably lower than in case of a microscope. However, macroscopic imaging can use much higher laser power, which compensates for low collection efficiency. To separate excitation and fluorescence signals long- and bandpass filters were used (HQ550LP and 580BP, Chroma). Image collection time for anesthetized mouse varied from 2 to 5 min depending on the level of fluorophores expression. Lifetime images have been analyzed with SPCImage 3.2 data analysis software (Becker & Hickl GmbH, Berlin, Germany). Standard deviations were calculated using SPCImage 3.2 data analysis software.

3.1.

FLIM of Transduced Tumor Live Cells In Vitro

Lentiviral transduction of A549 and HEp-2 cells with the TR23K caspase-3 sensor plasmid was accomplished with 70% to 90% efficiency.⁵³ This high efficiency allowed us to bypass the stages of selection and cloning while maintaining the initial heterogeneity of cell lines.⁶⁴ The cell lines obtained demonstrated a high level of expression of the sensor during the whole time of maintenance.

FLIM-FRET of transduced A549-TR23K and HEp-2-TR23K cells was performed using confocal time-resolved microscopy. Etoposide, cisplatin, and paclitaxel were used in the range of doses causing cytotoxicity in 30% to 90% of cells after 24 h of treatment (data not shown). Caspase-3 activity in cells was estimated by the shift in a lifetime distribution histogram from 1.8 to 1.9 ns (lifetime of intact sensor) up to the 2.5 ns (lifetime of cleaved sensor) as was demonstrated earlier.⁵³

Results on lifetime distribution analysis in HEp-2-TR23K differed from those in A549-TR23K. While HEp-2-TR23K have exhibited the sensitivity to cisplatin and etoposide, A549-TR23K proved to be sensitive only to paclitaxel. In this way, the cleavage of the caspase sensor caused by cisplatin and etoposide in the HEp-2-TR23K cells, and by paclitaxel in the A549-TR23K cells has been demonstrated in vitro [Figs. 3(a) and 3(b)], respectively. I was shown that these anticancer drugs could promote an activation of caspase-3 in doses and regimes suitable for in vivo study as it was shown earlier.^65–67 So it was expected that tumor xenografts based on TR23K-transduced cells would demonstrate the same specific responses to treatment with anticancer drugs described already.

Fig. 3

(a) FLIM of HEp-2-TR23K and corresponding lifetime distribution before (left image, solid line on the histogram) and after 24 h of treatment with $200\mu \mathrm{g}/\mathrm{ml}$ of etoposide and $10\mu \mathrm{g}/\mathrm{ml}$ of cisplatin (right image, dot line on the histogram); (b) FLIM of A549-TR23K and corresponding lifetime distribution before (left image, solid line on the histogram) and after 24 h of treatment with $30\mu \mathrm{g}/\mathrm{ml}$ of paclitaxel (right image, dash line on the histogram). Image size $80\times 80\mu \mathrm{m}$. (MicroTime 200, objective UPlanSApo $60\times /1.35$).

3.2.

FLIM and Steady-State Measurements of Tumor Xenografts In Vivo

Tumor xenografts of A549-TR23K and HEp-2-TR23K were based on the same cells expressing the caspase-3 sensor. Control tumor xenografts A549-TagRFP and HEp-2-TagRFP were based on cells expressing TagRFP only, thus imitating the ideal situation of total cleavage of the sensor under similar physiological conditions. Fluorescence measurements were performed after establishing of xenografts, which took about 2 weeks. It should be noted that xenografts based on A549 cells were of highly fluorescence and characterized by nodular and expansive tumor growth. The other xenograft model using HEp-2 cells exhibited weak fluorescence and characterized by both surface growth and tissue penetration.

The monitoring of fluorescence of tumor xenografts has been done by local fluorescence spectroscopy and 2-D planar imaging by iBOX as well as lifetime imaging by confocal laser scanner DCS-120. In control experiments, it has been shown that the change in fluorescence intensity was proportional to increasing tumor volume of A549-TagRFP and HEp-2-TagRFP xenografts during the initial stage of growth (data not shown). For xenografts expressing the TR23K biosensor, the relation to the increase of tumor volume was not reproduced in a consistent manner [Fig. 4(b)]. Spectra of xenografts expressing the TR23K sensor with the maximum of fluorescence at 585 nm did not show a significant shift compared to control xenografts expressing only TagRFP with the maximum of fluorescence at 595 nm [Fig. 4(c)].

Fig. 4.

Fluorescence of tumor xenografts A549-TR23K on 8th, 16th, 25th day of tumor growth: (a) pseudocolor mapping of the intensity of tumor fluorescence (iBox, ex 502 to 547 nm, em 580 to 640 nm). (b) Fluorescence of the tumor xenograft A549-TR23K (white) and the tumor size (gray) normalized to the 25th day of growth on 8th, 16th, 25th day of tumor growth. ROI of tumor fluorescence were depicted manually using Image J (NIH). The same ROI were moved to the nonfluorescent skin region. Fluorescence contrast was calculated as a relationship of integral fluorescence of tumor to integral fluorescence of skin. Normalized index was calculated as a ratio of fluorescence contrast on indicated dates to the fluorescence contrast at the last date of measurements (25th day). (c) Spectra of tumor xenografts A549-TR23K and A549-TagRFP (FS-003V spectrometer, ex. 532 nm).

Figure 5 demonstrates the diversity of tumor-to-skin fluorescence intensity ratio obtained by planar imaging and lifetimes of xenografts expressing the TR23K sensor, only TagRFP, and without any FPs. The fluorescence intensity distribution was more diverse compared to lifetime readouts.

Fig. 5

Distribution of mean fluorescence intensity versus lifetimes for tumor xenografts at different stages of growth (from 8 to 32 days), $\tau$, mean lifetime ($n=6$ to 10) with calculated custom standard deviation.

It should be noted that xenografts based on A549 cells were of highly fluorescence and characterized by nodular and expansive tumor growth. The other xenograft model using HEp-2 cells exhibited weak fluorescence and characterized by both surface growth and tissue penetration. Tumor to skin ration of fluorescence in A549-based tumors and HEp-2-based tumors differed three to five times according to Fig. 5.

The relationship among fluorescence intensity, lifetime, and tumor volume of xenografts expressing control TagRFP is presented in Fig. 6. The most significant feature is that the lifetime of tumor increases on initial stage of growing and approaching the value of 2400 to 2500 ps which corresponds to the lifetime of TagRFP in vitro.

Fig. 6

Relationship between tumor volume and average fluorescence intensity and lifetime for A549-TagRFP.

3.3.

Lifetime Imaging in Xenografts Based on the HEp-2 Tumor Cells

Here, we presented the analysis of FLIM and lifetime distribution in one set of experiments. Four animals were monitored and treated simultaneously from the 0 to 32nd days of xenografts stabilization and growing. The mean fluorescence lifetime was determined from one/two exponential fit of the decay curve.

Mean lifetime of nontreated xenograft HEp-2-TR23K on 17th day of growth was of $1679\pm 99\mathrm{ps}$ [Fig. 7, images (a) with corresponding histograms (a′)]. Lifetime distribution in treated HEp-2-TR23K xenograft changed to $1957\pm 110\mathrm{ps}$ on the 7th day after chemotherapy’s start (23rd day of tumor growth) [Fig. 7, image (b) with corresponding histogram (b′)]. Lifetime of another nontreated tumor xenograft within the same terms of growth (23rd day of tumor growth) was estimated as $1830\pm 114\mathrm{ps}$ [Fig. 7, images (c) with corresponding histograms (c′)]. We observed that the stage of tumor growth correlated with an increase in lifetime. In control experiment with HEp-2-TagRFP xenograft that simulates totally split sensor, the mean lifetime was $2212\pm 59\rho \mathrm{s}$ within the same terms. Remarkable heterogeneity of TagRFP lifetime distribution in this control xenograft was noticed, most probably due to the different depth and specificity of tumor invasion as presented in the corresponding image [Fig. 7, image (d) with corresponding histogram (d′)].

Fig. 7

FLIM-FRET of tumor xenografts based on the larynx carcinoma HEp2 (a–d, upper images), pseudocolor scale bar (e) and corresponding lifetime distribution histograms: (a) HEp2-TR23K before treatment with cisplatin and etoposide with corresponding histogram (a′) (before treatment); (b) HEp2-TR23K on the 7th day of treatment with cisplatin and etoposide with corresponding histogram (b′) (cisplatin and etoposide 7 days); (c) HEp2-TR23K control with correspondence histogram (c′) (control 7 days); (d) HEp2-TagRFP with corresponding histogram (d′) (control TagRFP); (e) pseudocolor scale bar from 1000 to 2500 ps.

3.4.

Lifetime Imaging in Xenografts Based on the A549 Tumor Cells

The mean lifetime in highly fluorescent A549-TR23K sensor-expressing xenografts starts from $1672\pm 174\mathrm{ps}$ on the 16th day [Fig. 8, images (a) with correspondent histogram (a′)]. On the 7th day after chemotherapy the distribution became bimodal, the component of $1700\pm 106\rho \mathrm{s}$ retained [Fig. 8, image (b) with corresponding peak (b′)] and a peak of $2100\pm 216\mathrm{ps}$ appeared [Fig. 8, image (b), area (b′), and histogram (b′′)]. The decay curve of A549-TR23K xenografts after treatment from the region of interest (ROI) was fitted with two exponential model (${\chi }^2=1.1$) (Fig. 9).

Fig. 8

FLIM-FRET of tumor xenografts based on the lung adenocarcinoma A549 (a–c, upper images) and corresponding lifetime distribution (bottom images): (a) A549-TR23K before treatment with Taxol (16th day of growth) with the corresponding histogram (a′ (before treatment); (b) A549-TR23K on the 7th day of treatment with Taxol (23rd day of growth) with the corresponding two peak histogram, short-lived (b′ and long-lived (b′′) (Taxol 7 days). The areas of short-lived (b′) and long-lived (b′′) components of cleaved sensor are bounded by white line at the image (b); (c) A549-TR23K control, (23rd day of growth) with the corresponding histogram (c′ (control 7 days); (d) A549-TagRFP, (23rd day of growth) with corresponding histogram (d′ (control TagRFP); (e) pseudocolor scale bar from 1000 to 2500 ps.

Fig. 9

Time trace in A549-TR23K xenograft after treatment with Taxol calculated from the ROI (red) of image from Fig. 8(b), IRF-instrument response function. $\text{Binning}\times 10$, ${\chi }^2=1.1$.

The fluorescence of nontreated A549-TR23K xenograft did not change significantly within a week, it was measured as $1604\pm 213\mathrm{ps}$ on 23rd day [Fig. 8 image (c) with corresponding histogram (c′] under the same conditions of animal maintenance. The mean lifetime of control xenograft A549-TagRFP, imitating total cleavage of the sensor, was $2400\pm 49\mathrm{ps}$ [Fig. 8, image (d) with corresponding histogram (d′] and is very narrow. The difference of lifetime distribution of the same FP TagRFP and sensor TR23K in xenografts bearing A549 and xenografts bearing HEp-2 could be connected with individual characteristics of cell lines and features of growth of xenografts. We presumed that every animal bore only one tumor.

The study of the tumor models allowed us to determine that the anticancer treatment led to the shifts of mean lifetime distribution in both types of xenografts. In case of A549-TR23K, the bimodal distribution lifetime signal [Fig. 8, image (b) with corresponding histogram] supported the idea of heterogeneous responses of a tumor to chemotherapy.^68,69 In the case of HEp-2-TR23K, the shift was not so pronounced and with a single mode which coincided with a more homogeneous response of the tumor to chemotherapy.