1.

Introduction

Photoactive reporter genes allow optical imaging of in vivo biological activity, such as the spatial and temporal distribution of gene expression, ^{1, 2, 3, 4, 5, 6, 7} host-pathogen interactions, ^{8, 9, 10, 11, 12, 13} tumor growth and metastasis, ^{14, 15, 16, 17, 18, 19} and transplant studies.^{20, 21} While over 30 bioluminescence systems exist in nature, all have in common a luciferase-catalyzed oxidation of a luciferin substrate to produce light.²² Unlike fluorescence systems [i.e., green fluorescence protein (GFP)], bioluminescence imaging (BLI) requires no excitation source, which simplifies the optics involved in the imaging process, allows for detection of light emission from deeper tissues, and facilitates quantitative analysis. In addition, light emission from the chemiluminescent oxidation of the luciferin substrate emits broadband light that peaks at $563\mathrm{nm}$ ( $590\mathrm{nm}$ at body temperature of $37°\mathrm{C}$ ) but spans wavelengths from 500 to $700\mathrm{nm}$ , thus allowing deep tissue penetration (on the order of several centimeters), in particular of the red parts of the emission spectrum.^{23, 24} The main drawbacks to using bioluminescence for in vivo imaging are the extremely low intensity of light produced and the poor resolution due to tissue scattering.

Hyperosmotic agents such as glycerol, dimethyl sulfoxide (DMSO), Trazograph, polyethylene glycol, polypropylene glycol- (PPG) and polyethylene glycol- (PEG) based prepolymer mixture, and glucose have been shown to transiently clear tissue within $30\min$ of application in hamsters and rats^{25, 26, 27} and even in humans ^{28, 29, 30, 31} both in vitro and in vivo. Increases in transmission of up to 200% and decreases of the reduced scattering coefficient of a full order of magnitude were reported. Clearing is believed to be due to index matching of the extracellular fluid with the tissue solid components (mainly collagen) and dehydration due to osmotic pressure.

The objective of this study is to investigate if treating skin with a hyperosmotic agent (glycerol) in order to reduce scattering will improve the sensitivity and spatial resolution of bioluminescence imaging. This study consisted of four parts: in vitro measurement of tissue (i.e., mouse skin) optical properties, in vitro simulated BLI, in vivo macroscopic imaging, and in vivo simulated BLI.

2.

Materials and Methods

2.2.

In Vitro Simulated Bioluminescence Imaging

After determining the tissue optical properties, an in vitro simulated bioluminescence study was performed. The goal was to match the conditions encountered in BLI, specifically, intensity and wavelength. A tungsten white light source was bandpass filtered at 550, 600, and $650\mathrm{nm}$ [full width at half maximum $\left(\mathrm{FWHM}\right)=10\mathrm{nm}$ ], then coupled into a $200\text{-}\mathrm{\mu }\mathrm{m}$ diameter optical fiber with a $500\text{-}\mathrm{\mu }\mathrm{m}$ diameter spherical diffusing fiber tip at the distal end, which provided an isotropic light source. The three wavelengths were chosen because luciferase-catalyzed bioluminescence has its highest intensities in the wavelength range of 500 to $700\mathrm{nm}$ .²³ The shorter wavelength components of this spectrum are attenuated by hemoglobin absorption, so the chosen wavelengths of light provided a realistic simulation of actual luciferase bioluminescence. The diffusing fiber was taped to the bottom of a petri dish ( $100\mathrm{mm}$ diameter). A piece $(2\mathrm{cm}\times 2\mathrm{cm})$ of excised mouse skin was placed over the fiber tip on the bottom of the petri dish with the epidermal side facing the fiber and the subdermal side facing up. The sample was held in place by a metal retaining ring with an aperture large enough ( $1.5\mathrm{cm}$ diameter) such that it did not block any light. For all experiments, a constant volume of $25\mathrm{ml}$ of glycerol solution of 25%, 50%, or 100% concentration (v/v) was added to the petri dish such that a layer of glycerol, roughly 3-mm high was overlying the skin sample. The entire assembly of petri dish, fiber, skin sample, and glycerol was placed in the light-tight imaging chamber. The fiber was guided into the chamber via a double baffled black rubber seal to avoid stray light from entering the imaging chamber. Images were then taken with a liquid nitrogen-cooled, back-thinned, back-illuminated, deep depletion charge-coupled device (CCD) camera (Roper EEV $1300\times 1340$ ). Background subtraction was performed on all images as a means of noise reduction. Image integration times of $5\mathrm{s}$ were used and no on-chip binning was used. The setup is summarized in Fig. 1 . Intensity is reported as the integrated pixel intensity for a defined area of a CCD image at a specific time point relative to the first time point for each wavelength. Spot size is defined as the width of the section of the cross-sectional intensity profile where values are greater than a predefined threshold (100 photon counts).

Fig. 1

Schematic of the experimental setup. A bioluminescence point source was simulated by using bandpass filtered light ( $550\mathrm{nm}$ , $600\mathrm{nm}$ , or $650\mathrm{nm}$ ) delivered via a $200\text{-}\mathrm{\mu }\mathrm{m}$ fiber to an isotropic diffuser tip. The same setup was used for in vitro (top left inset) and in vivo (bottom left inset) experiments.

Images were taken of the fiber alone, the fiber with overlying skin, and the fiber with overlying skin submerged in glycerol (time sequence). To minimize the effect of biological variation among skin samples, the experiment was designed to perform all three wavelengths in a single skin sample. This was accomplished by letting the camera controller take an image every $40\mathrm{s}$ for a duration of $5\mathrm{s}$ while rotating the three bandpass filters in front of the white light source every $40\mathrm{s}$ to provide time points for each wavelength in 2-min intervals.

3.

Results

Figures 2 show the increase in transmission [Fig. 2a] and decrease in diffuse reflectance [Fig. 2b], respectively, after treating native skin with 50% glycerol for $30\min$ (each line represents the average of three independent measurements). These figures also show that transmission and reflectance return to approximately their pretreatment values after rehydration with saline for $30\min$ . The reduced scattering coefficient $\left({\mu }_s^{\prime }\right)$ for skin [Fig. 3a ] decreases after $30\min$ of 50% glycerol exposure and returns to very near its untreated value after rehydration. The absorption coefficient $\left({\mu }_a\right)$ after $30\min$ of glycerol exposure is nearly the same as in the untreated state [Fig. 3b]. The limited thickness of the mouse skin samples $(\sim 220\mathrm{\mu }\mathrm{m})$ causes the inverse-adding doubling (IAD) program to return values that are not meaningful for weakly attenuated wavelengths in the cleared skin samples. Nevertheless, it can still be seen that ${\mathrm{\mu }}_s^{\prime }$ is reduced by approximately one order of magnitude for the wavelengths of interest.

Fig. 2

Total transmission (a) and reflectance (b) measurements of in vitro normal skin, glycerol treated skin (50%, $30\min$ ), and skin treated with glycerol for $30\min$ then placed in saline for $30\min$ . In the spectral range of interest (550 to $650\mathrm{nm}$ ), the total transmission increases (nearly two-fold) and the diffuse reflectance decreases (threefold to fourfold) due to glycerol application. Each line represents the average of three independent measurements $(n=3)$ .

Fig. 3

Reduced scattering coefficient (a) and absorption coefficient (b) of in vitro normal skin, glycerol treated skin (50%, $30\min$ ), and skin treated with glycerol for $30\min$ then placed in saline for $30\min$ . Each line represents the average of three independent measurements $(n=3)$ . Optical properties were calculated from transmission and reflectance measurements using an IAD program. In the spectral range of interest (550 to $650\mathrm{nm}$ ), the absorption coefficient is unaffected by the glycerol treatment, while the reduced scattering coefficient drops by nearly one order of magnitude.

Plots of intensity and spatial resolution for three wavelengths for the simulated bioluminescence experiment are shown in Figs. 4a and 4b . Both values decrease with increasing glycerol (50%) exposure. The maximum intensity decrease after $1800\mathrm{s}$ was 38.8% for 550-nm light, 48.4% for 600-nm light, and 42.6% for 650-nm light. Spot size decreased by 51.9% for 550-nm light, 60.0% for 600-nm light, and 68.0% for 650-nm light after $1800\mathrm{s}$ . The measured spot size was smaller for the shorter wavelength light at all time points.

Fig. 4

Integrated photon intensity (photon counts) (a) and spot diameter (b) measured by the imaging system through skin as a function of time of exposure to glycerol. Both parameters decrease with time (and hence with decreasing the reduced scattering coefficient).

Figure 5 summarizes the spatial resolution data for the simulated bioluminescence study. Two trends are apparent: increasing spot size with decreasing glycerol concentration and increasing spot size with longer wavelength. Notably, 50% and 100% glycerol concentrations result in significantly decreased spot sizes (i.e., better spatial resolution) compared to native tissue $(p<0.001)$ while there is no statistically significant difference between native skin and 25% glycerol-treated skin.

Fig. 5

Effect of glycerol concentration on spot size reduction of skin in vitro for different wavelengths. Note that 25% glycerol is ineffective in clearing skin after $30\min$ and shows values that are not different from native skin. In all cases, the spot size for 650-nm light is larger than that for 550-nm light as expected. However, the change in spot size due to glycerol is most significant at the 650-nm wavelength.

The results of the in vivo macroscopic imaging study are shown as Fig. 6 . Figure 6a shows a section of mouse skin $30\min$ after injection of $0.2\mathrm{ml}$ of 100% glycerol. The same section of skin is shown $24\mathrm{h}$ later [Fig. 6b] when the clearing effects have reversed.

Fig. 6

Image of in vivo mouse skin $30\min$ after injection of $0.2\mathrm{ml}$ 100% glycerol (a) and $24\mathrm{h}$ after (b) glycerol application when clearing is reversed. The left top part of the image shows the area where glycerol was applied, the right bottom shows an area of normal skin. The luminescent spot (for $650\mathrm{nm}$ ) thresholded at photon counts $>100$ , generated by the $500\text{-}\mathrm{\mu }\mathrm{m}$ diameter diffusing fiber is shown in (a) in both the cleared and untreated portion of the skin. Note that the spot size reduced by a factor of approximately 2 in cleared skin.

Integrated intensity and spot diameter for the in vivo simulated BLI experiment are shown in Figs. 7a and 7b , respectively. Both plots show a decrease of intensity and spot diameter, respectively, as the tissue clears with time.

Fig. 7

Effect of glycerol application in vivo on integrated photon intensity (a) and spot diameter (b) measured by the imaging system through skin of a live mouse before, at $15\min$ , and at $30\min$ of glycerol application. Note that both parameters decrease with time (and hence with decreasing the reduced scattering coefficient) similar to what was seen in vitro.

4.

Discussion

In this study, we have investigated the use of glycerol as a tissue clearing agent in order to improve the sensitivity and spatial resolution of in vivo BLI. To address the feasibility of this approach, we conducted a set of experiments determining the optical properties of skin in vitro and designed a set of experiments to study the effects of skin clearing in experiments in which a diffusing fiber tip was used as a simulated and constant (point) source of bioluminescence emission.

The optical properties of mouse skin measured for the skin before treatment, after $30\min$ of glycerol (50%) exposure, and after rehydration with saline for $30\min$ (Figs. 2 and 3) were consistent with previous studies on rat skin and hamster skin.²⁶ Tissue optical clearing is believed to be due to a decrease in the reduced scattering coefficient by index matching of extracellular fluid and tissue solid components. Within the spectral range of interest (550 to $650\mathrm{nm}$ ) total (diffuse) transmission increased on average by a factor of 1.70 and diffuse reflectance decreased on average by a factor 3.37. After $30\min$ of saline rehydration, both transmission and reflectance values returned to within 10% of their untreated values. This was confirmed by the results of an IAD program, which calculated ${\mu }_s^{\prime }$ and ${\mu }_a$ from the spectrophotometer data. Over the spectral range of interest, the reduced scattering coefficient $\left({\mu }_s^{\prime }\right)$ decreased on average by approximately one order of magnitude, while the absorption coefficient $\left({\mu }_a\right)$ was nearly constant ( $<5\%$ change). The IAD model was unable to compute meaningful results for wavelengths $>450\mathrm{nm}$ in cleared skin due to a combination of low scattering combined with the relatively thin mouse skin thickness. Nevertheless, the reduction in diffuse reflectance, the increase in transmission, and the calculated optical properties in the range where the model did converge are in general agreement with data published on rat and hamster skin.^{25, 26, 27}

The decrease in light detected by the camera, seen after glycerol application (Fig. 4) in the in vitro simulated bioluminescence experiment was unexpected and in apparent conflict with the integrating sphere measurements, which showed a 1.7-fold increase in transmission. However, we speculate that this apparent discrepancy can be explained by the fact that fundamentally the two measurements are different. Traditionally, a double integrating sphere setup is used to measure total transmission and diffuse reflectance. In this setup, all of the light incident on the tissue is accounted for by one sphere or the other. More specifically, with regards to the transmission measurements, all transmitted light (i.e., total transmittance), irrespective of the angle with the tissue normal, is captured in this measurement. Hence, if the reduced scattering coefficient decreases as is the case in skin treated with glycerol, the total attenuation coefficient decreases, and hence, the overall transmission of light through the tissue increases. In contrast, the imaging geometry where the camera is positioned approximately $50\mathrm{cm}$ above the tissue sample and has an effective aperture of roughly $2.5\mathrm{cm}$ detects only light that is emitted within the acceptance cone angle of the camera $(\sim 7\mathrm{deg})$ . The decrease in light reaching the camera in cleared skin compared to native skin is believed to be due to the reduction of multiple scattering events. In the native skin, light emitted in a direction that would put it outside the camera’s acceptance angle can be multiply scattered and redirected to emit from the tissue in a direction within the camera’s acceptance angle. The net result of this is a decrease in light detected by the camera (causing a decrease of approximately 40%) when the reduced scattering coefficient of the sample decreases (by a factor of 10), even though the total transmission of light through the sample is increased. We hypothesize that this geometric effect is also responsible for the apparent conflict between our imaging data and data published by Vargas ²⁷ who showed increased fluorescence detected when skin was cleared and He, ³³ who showed increased signal detection in low light level imaging applications after tissue clearing. In both these studies, different photon collection geometries were used. Vargas used a multifiber probe consisting of $200\text{-}\mathrm{\mu }\mathrm{m}$ diameter optical fibers with a NA of 0.2 (acceptance angle $\sim 12\mathrm{deg}$ ) in contact with the tissue. In the study by He, the imaging geometry was not reported and apparently not considered in their Monte Carlo simulations. In addition, significantly thicker tissue samples were used, making a direct comparison difficult, if not impossible. Nevertheless, both these studies used geometries that are different from the typical BLI setup that employs a 1-in aperture roughly $50\mathrm{cm}$ away from the tissue surface. We conducted two experiments to verify our multiple scattering hypothesis. First, we imaged the fiber tip under a layer of scattering liquid consisting of polystyrene microspheres (diameter $0.954\mathrm{\mu }\mathrm{m}$ ) in decreasing concentrations (and hence decreasing scattering coefficients). The setup was similar to Fig. 1 but with the tissue layer replaced by scattering liquid. In this experiment (Fig. 8 ), we found that as ${\mu }_s^{\prime }$ is increased from 0 to $\sim 3{\mathrm{mm}}^{-1}$ , light reaching the camera increases. When ${\mu }_s^{\prime }$ was increased beyond $3{\mathrm{mm}}^{-1}$ , light reaching the camera decreases. The net effect of changing the reduced scattering coefficient $10{\mathrm{mm}}^{-1}$ (native skin) to $<1{\mathrm{mm}}^{-1}$ (cleared skin) is a reduction of light reaching the camera. This is consistent with our observation in native and clear skin. As we continue to increase ${\mu }_s^{\prime }$ , the slope of the curve changes and light levels reaching the camera start to decrease. The second experiment that was conducted to verify our hypothesis consisted of a Monte Carlo simulation. Using an isotropic light source underneath a slab geometry with the optical properties of normal and cleared skin, we determined the transmitted light as a function of radius and angle. Plots were generated of the light reaching the plane of the camera as function of radius. In this scenario (Fig. 8), we confirmed the data from the microsphere experiments and found that a combination of thin tissue layer, constant ${\mu }_a$ , and a reduction in ${\mu }_s^{\prime }$ from $10{\mathrm{mm}}^{-1}$ to less than $1{\mathrm{mm}}^{-1}$ leads to a decrease of light transmission within the acceptance angle of the camera, despite the fact that the total transmission increases. Although the decrease in intensity was not the desired effect, it may be a tradeoff for the increase in spatial resolution.

Fig. 8

Photon intensity detected by the imaging system as a function of the reduced scattering coefficient. Experimentally $(▴)$ , a serial dilution of polystyrene microspheres was made and simulated bioluminescence from the diffusing fiber tip was measured. Simulation results $(•)$ were obtained from Monte Carlo simulations with the appropriate geometry to simulate light collection within the cone angle and aperture of the detector.

Glycerol application caused a decrease in spot size [Fig. 7a] ranging from 51.9% for $550\mathrm{nm}$ light to 68.0% for $650\mathrm{nm}$ light. The difference in resolution improvement between the wavelengths used is believed to be due to absorption (and hence the shorter mean free path) of shorter wavelength light by hemoglobin. Higher glycerol concentrations were also shown to be more effective at improving spatial resolution at $30\min$ postinjection (Fig. 5). Notably, 100% and 50% glycerol are effective in reducing the imaged spot size (i.e., increasing spatial resolution), while 25% glycerol was not able to cause changes in spot size that were significantly different from native tissue. This is due to the higher osmotic pressure of the higher concentration of agent.^{26, 27}

The in vivo macroscopic imaging study (Fig. 6) showed that, in accordance with previous studies, gross subdermal structures normally obscured by turbid tissue can be made visible by treatment with hyperosmotic agent. It was also shown that the clearing effects reversed 12 to $24\mathrm{h}$ after injection.

Penetration of optical clearing agents through intact skin (in vivo) is minimal and extremely slow. This has been attributed to the fact that these agents are hydrophilic and penetrate the lipophilic stratum corneum poorly. In this study, we circumvented this problem by injecting the clearing agent directly in the skin (intradermally). Subdermal injections had no noticeable clearing effects. Given the relatively small thickness of the tissue $(\approx 220\mathrm{\mu }\mathrm{m})$ , it is difficult to inject the necessary amount $\left(0.2\mathrm{ml}\right)$ of glycerol. It was found that relatively high glycerol concentrations (50 to 100%) were necessary to achieve clearing within the desired time frame (15 to $30\min$ ). At this concentration, the solution has a high viscosity, further increasing the difficulty of injection.

Promising recent work by Khan, who have explored a number of different skin clearing agents, has shown that a combination of Food and Drug Administration–approved lipophilic PPG based polymers and hydrophilic PEG is an effective, topically applied skin clearing agent. Although we have not tried this agent in mice, in vitro and in vivo results in humans suggest that it does not have the problems associated with glycerol.²⁸

The results of the in vivo simulated bioluminescence experiment [Figs. 7a and 7b] were consistent with the trends seen in the in vitro experiments (Figs. 4 and 5). A 41.8% decrease in integrated intensity and a 38.5% decrease in spot size were seen $30\min$ postinjection. Although we limited our tissue clearing to the clearing of skin for practical reasons (it is difficult to obtain adequate diffusion of the agent in thicker layers, especially in vivo), BLI is particularly useful for the imaging of processes that occur several millimeters or even centimeters below the tissue surface. For those cases, clearing only the skin (200 to $300\mathrm{\mu }\mathrm{m}$ in thickness in a typical mouse) will have a significantly reduced effect. Nevertheless, it is conceivable to extend this approach to clearing deeper tissues layers, in which case, spatial resolution is significantly improved and evidence suggests that the photon flux may in fact be improved,³³ depending on the imaging geometry.

5.

Conclusions

In summary, we have shown that applying a hyperosmotic clearing agent to skin of small rodents has the potential to improve spatial resolution of BLI owing to a reduction in the reduced scattering coefficient in skin by one order of magnitude. In contrast, reducing the scattering coefficient reduces the amount of light reaching the camera due to a reduction in the amount of multiple scattered light that reaches the camera aperture and thus reducing sensitivity of the method. Tissue thickness and imaging geometry appear to be the critical factors that determine the effect of tissue clearing on spatial resolution and sensitivity.