2.1.

Near-Infrared Imaging Agents

Near-infrared dyes indocyanine green (ICG), IR-820, IR-806, and $3,3^{\prime }$ -diethylthiatricarbocyanine iodide (DTTCI) were purchased from Sigma-Aldrich (Saint Louis, Missouri) and used without further purification. Cypate, LS-276, LS-277, and LS-288 were synthesized as described previously,^{17, 18, 19} and their photophysical properties were described elsewhere.^{16, 20} Bovine serum albumin (BSA) was obtained from Sigma-Aldrich. Methanol, ethanol, acetone, dimethylsulfoxide (DMSO), methylene chloride (DCM), and chloroform were of spectroscopic grade quality and ultrapure water (Milli-Q, Millipore, Billerica, Massachusetts) was used throughout the study.

2.2.

In Vitro Spectroscopic Measurements

BSA $\left(50\mathrm{mg}\right)$ was dissolved in $1\mathrm{mL}$ of $0.01\text{-}\mathrm{M}$ phosphate-buffered saline (PBS, Sigma-Aldrich) to obtain physiologic-relevant concentration of albumin in human plasma. The stock solutions of the dyes $\left(50\text{to}100\mu \mathrm{M}\right)$ in DMSO were stored at $-20°\mathrm{C}$ in the dark. Aliquots were added to the appropriate solvents or BSA-PBS solution $\left(1\mathrm{mL}\right)$ and vortexed for $1\min$ . To prevent an inner filter effect, samples were further diluted with the appropriate organic solvent or PBS buffer to adjust the absorbance maximum between 0.1 and 0.3.

Absorbance spectra were recorded on a Beckman Coulter DU 640 spectrophotometer (Fullerton, California) and fluorescence steady-state spectra were recorded on a Fluorolog III fluorometer (Horiba Jobin Yvon, Edison, New Jersey) with $720\text{-}\mathrm{nm}$ excitation and $5\text{-}\mathrm{nm}$ band slits. All measurements were conducted at room temperature. The FLTs were measured using a time-correlated single-photon-counting (TCSPC) technique (Horiba) with $773\text{-}\mathrm{nm}$ excitation source NanoLed® (impulse repetition rate $1\mathrm{MHz}$ ) at $90\mathrm{deg}$ to the detector (Hamamatsu Photonics, Japan). For the FLT measurements, the absorbance of the working solutions was maintained below 0.15 at $773\text{-}\mathrm{nm}$ excitation wavelength by dissolving the stock solutions in the appropriate solvents. The FLT of cypate-agarose gel was recorded in solid state using a solid sample holder at about $45\mathrm{deg}$ to the path of light beam to avoid direct light reaching the detector. The angle was optimized with a blank agarose gel without cypate. The detector was set to $820\mathrm{nm}$ with a $20\text{-}\mathrm{nm}$ bandpass. The electrical signal was amplified by a TB-02 pulse amplifier (Horiba) fed to the constant fraction discriminator (CFD, Phillips, The Netherlands). The first detected photon was used as a start signal by a time-to-amplitude converter (TAC), and the excitation pulse triggered the stop signal. The multichannel analyzer (MCA) recorded repetitive start-stop signals from the TAC and generated a histogram of photons as a function of time-calibrated channels ( $6.88\mathrm{ps}$ /channel) until the peak signal reached 10,000 counts. The FLT was recorded on a $50\text{-}\mathrm{ns}$ scale. The instrument response function was obtained using a Rayleigh scatter of Ludox-40 (0.03% in MQ water; Sigma-Aldrich) in a quartz cuvette at $773\text{-}\mathrm{nm}$ emission. Decay analysis software (DAS6 v6.1, Horiba) was used for FLT calculations. The goodness of fit was judged by ${\chi }^2$ values, Durbin–Watson parameters, as well as visual observations of fitted line, residuals, and autocorrelation functions. Two-exponential analyses were normally used for data fitting, except where three-exponential decay equations were used to improve ${\chi }^2$ for analysis in BSA solutions. When the FLT components with insignificant contribution $(<3\%)$ were discarded, the decays were reported as single- or dual-component systems in the case of a two- or three-exponential fit. For solution systems with two emitting species, such as the BSA solutions, the FLTs were calculated by the following equations:

2.3.

Animal Treatment

All animal studies were performed in compliance with the Washington University School of Medicine Animal Studies Committee requirements for the humane care and use of laboratory animals in research. For in vivo NIR probe FLT measurement, male NCR nude mice were anesthetized and maintained with isoflurane gas (2% v/v) for fluorescent agent administration and imaging. Contrast agents dissolved in $100\mu \mathrm{L}$ of 20% DMSO were injected via lateral tail vein. Concentrations were based on absorbance at $780\mathrm{nm}$ , equivalent to $60\text{-}\mu \mathrm{M}$ cypate. For imaging of subcutaneously implanted cypate-loaded agarose gels, mice were anesthetized and maintained with isoflurane gas (2% v/v). A $5\text{-}\mathrm{mm}$ incision was made through the skin flank region and a pocket made in the subcutaneous tissue by blunt dissection. A $0.5\text{-}{\mathrm{mm}}^3$ slab ( $1\times 1\text{-}\mathrm{mm}$ square by $0.5\mathrm{mm}$ thick) of 1% agarose gel containing $60\text{-}\mu \mathrm{M}$ cypate was inserted into the pocket and the incision was closed with tissue adhesive. Gels were positioned such that the incision lay on the lateral aspect, with uninterrupted skin over the gel surface.

2.4.

Imaging Method and Analysis

In vivo mouse images were acquired with a time-domain diffuse optical imaging system (eXplore Optix, ART, Incorporated, Montreal, Canada). The system produced excitation light via a $780\text{-}\mathrm{nm}$ , $80\text{-}\mathrm{MHz}$ pulsed diode laser and captured emitted photons by fast photomultiplier tube with an $830\text{-}\mathrm{nm}$ telecentric emission filter. The detection system included a time-correlated single-photon counting device for in vivo fluorescence decay recording. The focal points of excitation and emission detection were separated by $3\mathrm{mm}$ to detect diffusely scattered fluorescence photons. Diffuse optical imaging improves depth resolution relative to reflectance imaging.²¹ FLT imaging in live mice was performed as reported previously.^{9, 10} Briefly, the animals were positioned on the heated imaging platform and a 2-D scanning region of interest (ROI) was selected by a top-view charge-coupled device (CCD) camera to include the area from the neck to the pelvis. The $780\text{-}\mathrm{nm}$ pulsed diode laser was set to $0.4\mu \mathrm{W}$ for excitation scans and adjusted for optimal signal strength, in the range $5\text{to}70\mu \mathrm{W}$ for fluorescence detection. ROI were raster scanned in $1.5\text{-}\mathrm{mm}$ increments for the controlled drug release (CDR) model and $3\text{-}\mathrm{mm}$ increments for the contrast agent imaging with $0.3\text{-}\mathrm{s}$ integration time per pixel. Fluorescence data were acquired $0.5\text{to}2.5\mathrm{h}$ after agent injection. Acquired images were analyzed with the OptiView software provided by ART. Data were reported as fluorescence intensity or FLT maps, with each pixel representing the integration or curve fitting of the acquired temporal point-spread function (TPSF), respectively, at each detection point. Pixels with less than $1\times {10}^5$ photon counts per second were not included in the FLT measurement due to insufficient signal. In vivo FLT maps are displayed after fitting a single exponential decay function for simplicity. To determine the best model for in vivo FLT results, FLT values from different solvent models were compared to the average FLT values in the liver region from in vivo imaging. The liver region was chosen for data sampling, as most agents accumulated in this organ nonspecifically relative to other tissues.

2.5.

Statistical Analysis

Column statistics were performed to assess the correlation of in vitro and in vivo FLT measurements. Percent differences were calculated for each in vitro model relative to the reference model to assess the ability of each solvent system to predict in vivo FLT behavior. The percent differences were compared to assess the agreement of each in vitro model in relation to the in vivo FLT measurements. All analyses were performed using Prism 4.0 (GraphPad Software, San Diego, California).

3.1.

Chemistry

NIR molecular probes synthesized in our laboratory and from commercial sources used in this study are shown in Fig. 1 . All compounds are based on heptamethine dyes that differ in the nature of the heterocyclic end groups or substituents at the mesoposition. The linear heptamethine dye cypate was prepared in high yield $(>60\%)$ from the reaction of glutaconaldehyde and the propanoic acid derivative of benzoindole, as reported previously.^{17, 22} The mesosubstituted dyes were prepared by a modified Suzuki coupling method.¹⁸ The purity of all prepared compounds was determined by HPLC and LC-MS analyses. The choice of the dyes was based on several criteria. First, the dyes were selected to represent a broad range of FLTs from $0.49\mathrm{ns}$ for IR-820 to $1.14\mathrm{ns}$ for DTTCI in albumin. Second, the probes were selected to represent different degrees of hydrophilicity determined by a number of sulfate groups in the molecule. For example, cypate with no sulfonate groups was sparingly insoluble in water, while LS-288 with four sulfonate groups was highly water soluble. Third, the synthesized dyes were designed to possess a reactive functionality (here, the carboxylic acid group) for labeling biomolecules such as peptides and proteins.

Fig. 1

Chemical structures of NIR fluorescent dyes used in this study.

3.2.

Comparison of In Vitro and In Vivo Fluorescence Lifetimes of Molecular Probes

The spectral properties of the dyes were determined in solvents with different solvent polarity to explore the FLT behavior in different environments, as reported previously.¹⁶ In these solvents, all of the dyes absorbed and emitted light in the NIR region $\left(700\text{to}900\mathrm{nm}\right)$ . The dyes also exhibited relatively small bathochromic absorption and emission shifts $\left(5\text{to}30\mathrm{nm}\right)$ in high polarity relative to low polarity solvents, which is typical of heptamethine dyes.¹⁶ In contrast, the dynamic emission spectra show significant solvent-dependent changes in FLT values. The FLT changes correlated with the solvent polarity function, such as solvent orientation polarizability $\Delta f$ . In general, the FLT showed a decrease from low to high polarity solvents (from low to high values of $\Delta f$ ). The fluorescence decays in most solvents were monoexponential ( $>95\%$ fractional contribution from the major component constituted the two exponential analysis), suggesting the presence of one major form of the emitter in solution (Table 1 ). This indicates the homogeneity of the solutions, albeit with polarity-dependent FLTs in different solvents.

Table 1

FLTs (ns) of NIR molecular probes measured in various solvents and after intravenous administration in living mice. Data were fitted by a single exponential decay model. Standard deviations (sd) were less than 5% for in vitro measurements.

We also examined the FLT properties of the dyes in albumin, a major transport protein in blood. The average FLTs of single-exponential decay fits are shown in Table 1. According to molecular modeling studies,²⁰ albumin offers at least two binding pockets with significantly different polarities for the binding of heptamethine molecules. As a result, the molecular probes bound to albumin could exist in two different microenvironments that exhibit two distinct FLTs with corresponding fractional contributions. In contrast to the single exponential decays found in organic solutions, the fluorescence decays in albumin solutions show a two-exponential decay, revealing the presence of the two light-emitting states, except for DTTCI (Table 2 ).

Table 2

Measured two-component FLT ( τ , ns) of dyes from albumin solutions compared with FLTs in the liver region in mice after intravenous administration. Good agreement was observed between the two FLT components in vitro and in vivo. The in vitro and in vivo fractional composition of FLTs in the liver were significantly different for LS-276 and DTTCI, indicating the differences in protein-binding characteristics of the molecular probes. Standard deviations for in vitro measurements were less than 5%. In vitro τ1 (F1) is a fractional component (%); and NS is for not significant.

For in vivo FLT study, mice were imaged after intravenous injection of individual dyes. The resulting FLT maps of the dyes were relatively flat compared with the intensity images, indicating the poor dependence of FLT on the dye concentration (Fig. 2 ).

Fig. 2

(a) In vivo fluorescence intensity and (b) FLT ventral aspect image maps of mice injected with NIR fluorescent dyes $1\text{to}4\mathrm{h}$ postinjection. FLT maps were created from TPSF data acquired using the time-domain diffuse optical imaging system. FLT measurements were selected from the liver region (green rectangles) based on high fluorescence intensity and anatomical location for comparison to minimize discrepancies in organ distribution, as all dyes showed evidence of hepatobiliary elimination. Some of the dyes showed partial renal clearance with fluorescence signal from the bladder (indicated by abdominal ROI for LS-276). The collected signal was insufficient for accurate FLT measurement from the bladder for these specimens. Differences in the positions of the liver and kidney regions in different animals are due to the shifting of these internal organs in animals. (Color online only.)

Each compound tested demonstrated relatively distinct FLTs ranging from $0.53\text{to}1.12\mathrm{ns}$ . The mean in vivo FLTs measured from the liver region for different dyes are shown in Table 1 and reflect the general trend that fluorescent probes with longer FLTs in vitro in any given solvent typically show longer FLT in vivo. An exception to the relatively flat FLT trend is the heterogeneous fluorescence intensity and FLT maps of LS-288 in mice (Fig. 3 ). The distinct short FLT in the bladder $\left(0.70\mathrm{ns}\right)$ relative to other parts of the body $\left(1.12\mathrm{ns}\right)$ was observed, indicating the water-rich environment in the urine (Table 3 ). The hydrophilic characteristics of the dye resulted in significant renal clearance relative to the more hydrophobic dyes used in this study.

Fig. 3

In vivo fluorescence intensity (a) ventral and (b) dorsal and FLT maps (c) ventral and (d) dorsal of nude mouse $2\mathrm{h}$ after intravenous injection of LS-288 solution. The FLT distribution is heterogeneous with distinctly different FLTs observed in the bladder and kidneys relative the rest of the body. Representative fluorescence TPSFs for the liver (1) and bladder (2) ROIs are shown in (e) and demonstrate the dominance of the long-FLT component in the liver and the short-FLT component in the bladder (multiexponential fitting data in Table 2). The mean detected FLT from each ROI is presented in (f), which compares favorably with the expected water-rich environment in urine. ROI 6 corresponds to the dorsal aspect of ROI 2 (bladder), indicating that the short FLT relative to ROI 7 is due to muscle tissue overlying ROI 2 in this position. A tomographic method could be used to isolate the contribution of adjacent tissues to the FLT.

Table 3

Measured FLTs ( τ , ns) of LS-288 in vitro and in vivo comparing the protein-bound fractions in albumin solution and in the liver ROI to that of relatively water-rich urine. Standard deviations were less than 5% for in vitro measurements.

Our results show that the dye FLTs in albumin solutions best approximates the in vivo range of FLT as measured from the liver region. A graphical box and whiskers plot of this analysis is shown in Fig. 4 . Two-exponential fitting of the acquired TPSFs for dyes in albumin solutions reliably predicted the range of FLTs observed in vivo, with the largest deviations from the mean (monoexponential fitting) in vivo measurements apparently due to differences in protein binding fractions (Table 2). FLT values for compounds in aqueous albumin solution showed the lowest bias relative to other solvents with an average difference of 0.7%.

Fig. 4

Box and whiskers plot of FLTs measured in various solvents and referenced to in vivo data from liver regions. The mean percent difference for fluorescence lifetimes for the dyes measured in albumin solution is close to 0 (0.7%), indicating good agreement between the two systems. The maximum difference for albumin was 28% with median difference of 2%. All of the other systems showed a median difference of 10% or greater. The largest mean differences between the in vivo and albumin models were observed for DTTCI (28%) and LS-276 (21%). These differences appear to be related to the protein binding fractions as shown in Table 2.

3.3.

Agarose Gel Implant Fluorescence Lifetime Imaging

Having demonstrated the potential to predict the in vivo behavior of NIR dyes in various mediums, we explored the use of FLT changes to monitor the release of encapsulated dyes in vivo. We used agarose gel implant containing cypate as a model of controlled drug release for this study. The site of surgery did not swell or show significant signs of inflammation during the imaging period, nor did the mice show signs of irritation or pain from the implant. High fluorescence intensity was measured from the area of the gel implant with significant fluorescence detected from the area surrounding the gel. We also detected fluorescence from the contralateral liver and kidney regions by $3\text{-}\mathrm{h}$ postimplant, demonstrating rapid cypate diffusion and absorption into systemic circulation [Fig. 5a ]. The measured FLT in the tissue surrounding the gel was substantially higher at several millimeters from the gel relative to the FLT of cypate in the agarose gel [Fig. 5b] with a near linear increase over distance from the gel center [Fig. 5c]. Because cypate is cleared from the body through the hepatobiliary pathway, it is also possible that the stomach and intestines contributed some of the detected fluorescence in the liver and kidney regions.

Fig. 5

(a) In vivo fluorescence intensity and (b) FLT maps of a mouse $3\mathrm{h}$ after subcutaneous implantation of cypate-loaded agarose gel (g). The FLT of cypate increased as it diffused through the subcutaneous space from the gel and was absorbed into the blood. Fluorescence can be seen in the right kidney (k) and the liver (l) regions, with distinctly different FLTs consistent with the dye environment. (b) A plot of FLT at distances from the agarose gel (black arrows) demonstrates the difference in cypate FLT as it diffused from the gel.