Cherenkov-excited Multi-Fluorophore Sensing in Tissue-Simulating Phantoms and In Vivo

Abstract

Cherenkov-excited molecular sensing was used to assess the potential for simultaneous quantitative sensing of two NIR fluorophores within tissue simulating phantoms through spectral separation of signals. Cherenkov emissions induced by external beam gamma photon irradiation to tissues/tissue simulating phantoms were detectable over the 500 nm to 900 nm wavelength range. The presence of blood was demonstrated to reduce the integrated intensity of detected Cherenkov emissions by near 50%, predominantly at wavelengths below 620 nm. The molecular dyes IRDye® 680RD and 800CW have excitation and emission spectra at longer wavelengths than the strongest blood absorption peaks, and also where the intensity of Cherenkov light is the lowest intensity, and so that the emission signal relative to background signal is maximized. Tissue phantoms composed of 1% intralipid and 1% blood were used to simulate human breast tissue, and vials containing fluorophore were embedded in the media, and irradiated with gamma photons for Cherenkov excitation. It was found that fluorescence emissions excited by the Cherenkov signal produced within the phantom could be detected at 5 mm depth into the medium within a 0.1-25 μM fluorophore concentration range. The detected fluorescence signals from these dyes showed linear relationships with radiation doses down to the cGy level. In vivo tests were only successful within the range near a μM, suggesting that these could be used for metabolic probes in vivo where the local concentrations are near this range.

1. Introduction

Molecular imaging in vivo requires both high resolution and multi-component information, ideally. Recent work in Cherenkov excited luminescence scanned imaging (CELSI) has shown potential for sensing and imaging molecular targets with very low volumetric sensitivity in tissue. The demonstration of the concepts of CELSI have shown in vivo and tissue phantom work with oxygen luminescence probes, however the developments around fluorescence reporters had previously been limited(1-3). This was extended into tissue-equivalent phantoms with two near-infrared (NIR) fluorescent reporters within the same volume. The latter issue is important, because many clinical and pre-clinical studies have shown the need to excite multiple fluorophores in the same tissue volume, for multi-parameter molecular imaging, either for better delineation of tumor volumes (4-7) or accurate quantification of molecular binding.(8-10) In this study, the Cherenkov-excited molecular sensing was extended to test the potential for quantifying two fluorescent tracers at the same time in tissue-equivalent media. Furthermore, these signals were assessed for linearity with concentration and applied radiotherapy dose.

Cherenkov light is produced when radiation in the MeV energy range travels through tissue, as illustrated in Figure 1(a), giving off a continuous spectrum of light that decays roughly as the inverse-square of the wavelength (1, 11, 12), as illustrated as the light blue line in Figure 1 (b). Thus, the apparent color of the Cherenkov emission is highly weighted in the ultraviolet and blue wavelengths when undistorted by the medium it is in. However, in mammalian tissue, due to the high absorption of short wavelengths by blood the signal is heavily attenuated and is largely seen as comprised of red and NIR wavelengths. Figure 1 (b) shows the absorption spectra of hemoglobin (weighted to 1% blood with oxygenated as HbO2 and deoxygenated as Hb-R) as well as water and lipids, and how the Cherenkov emission spectrum would be dominant in the wavelength region of 600nm – 900nm where there is a reduced absorption from either blood and water. However, at the local level, the Cherenkov light can be used to excite optical absorbing species such as fluorophores, and the resulting fluorescence emissions can exit the tissue as long as the emission peak has sufficient Stokes-shift to bring it out to the red-NIR diffusion transport window of tissue (2, 3, 13, 14).

Figure 1.

The generation of Cherenkov light from secondary electrons induced by MV x-ray irradiation of tissue is illustrated in (a), with the spectra of Cherenkov light superimposed on the absorbing species present in tissue in (b) showing the wavelength range were Cherenkov light emission dominates in vivo. The benefit of Cherenkov excitation by selective illumination at the point of irradiation is shown in (d) as compared to diffuse light excitation which decays exponentially with depth into the tissue (c). Finally, the entire concept of using selective illumination by MV x-ray beam to induce Cherenkov is shown (e), which subsequently is absorbed by local regions of fluorescence, and then the luminescence emission can be captured by a detector.

The ability to spatially shape radiation beams delivered by a linear accelerator (LINAC), combined with highly sensitive optical detection instrumentation, allows the ability to adaptively image/measure Cherenkov stimulated light emissions with highly resolved spatial information from the tissue (15, 16). Figure 1 (c) and (d) shows a comparison of the light excitation in tissue from diffuse light © versus linac delivered x-rays, emitting Cherenkov light selectively into the medium. The diffuse light applied is attenuated exponentially with depth, while the Cherenkov light can be selectively placed anywhere by the Linac x-ray beam. The MV x-ray radiation travels deep into tissue spinning off secondary electrons (Figure 1(a)), and these secondary electrons each give off 100-1000 optical photons by the Cherenkov effect. Taken together the selective delivery of Cherenkov combined with localization of molecular probes, as illustrated in figure 1(e) can be used to sample signals from a subsurface location, such as a lymph node for metastatic burden estimation, or to areas for radiation treatment. In summary, the benefit of Cherenkov excited luminescence sensing are:

1. Cherenkov light is broadband but in vivo 600-900nm light has appreciable penetration.

2. This light is generated at the site of x-ray irradiation dose deposition by secondary electrons, which can be throughout tissue for high MV x-ray linac delivery

3. The selective excitation can then lead to capture of the emission, with knowledge of where the signal came from within the tissue.

Thus, as illustrated in previous studies (1, 3), the concept of Cherenkov-excited luminescence detection can work reasonably well, depending upon the probe and the tissue geometry. Using spectrometer readout to filter the fluorescence emission from the Cherenkov light (excitation signal) is the topic examined in this study. This spectral decoupling of the output signals works well, as long as an estimate of contributing spectral bases is available to fit to. In this study, the challenge of simultaneously detecting two molecular species from Cherenkov excitation was examined in realistic tissue simulating environments.

2. Materials & Methods

2.1. Experimental set up and LINAC beam delivery

The experimental set up was largely set by the linear accelerator (Varian Clinic 2100C, Varian Medical Systems, Palo Alto, USA) used for MeV energy delivery, in a treatment room of the Norris Cotton Cancer Center, Lebanon, NH. For all experiments, a 10 x 10 mm² beam was used at 6 MV photon energy. The LINAC was set up to deliver radiation in 3.25 μs bursts of radiation at 360 Hz repetition rate, with a dose rate of 600 Monitor Units per minute, roughly equivalent to 600 cGy/min.

A single fiber bundle was used for light collection, and delivery to a spectrometer (Acton Insight, Princeton Instruments, Acton, USA) through a 500 nm longpass filter (FEL500, Thorlabs Inc.). A fast-gated intensified CCD camera (ICCD, PI-MAX3, Princeton Instruments, USA) was used at the spectrometer output for multispectral detection. The spectrometer and ICCD camera were placed just outside the radiotherapy room, so that no shielding from the radiation generated by the LINAC was necessary. The single fiber bundle (CeramOptec, Germany) was 15 meters in length, and housed 19, 200-μm diameter silica fibers that conducted emitted light from the treatment region (i.e. tissue-simulating phantom surface, or animal skin) to the spectrometer. The LINAC trigger SNYC output was used to trigger acquisitions by the camera at minimum trigger delay equal to the camera insertion delay of 27 ns and radiation pulse gate width of 3.25 μs. The ICCD was cooled to a control point of −25 °C and the spectrometer grating used in all experiments was 300 lines/mm, with center wavelength set to 700 nm.

2.2. Liquid Phantoms & Fluorescent Inclusions

Liquid tissue simulating ‘background’-simulating phantoms were prepared to contain 1% v/v concentration of Intralipid (Fresenius Kabi) as an optical scatterer with scattering properties similar to human tissue, and 1% v/v whole porcine blood (Lampire Biologicals Inc), to match the average blood level of whole tissue (Reference Pogue and Patterson, JBO, 2008). These were held in optically clear plastic containers approximately 120 mm x 60 mm x 60 mm in size. containers were positioned in the radiation beam, with the beam edge stopping immediately at the external boundary of the medium. Immediately outside the edge of the phantom, a linear fiber array was placed in light contact with the surface, such that light emission could be directly collected from the emission, as shown in Figure 2(a) and (b). Typical Cherenkov emission spectra from the two types of homogenous phantoms are shown in Figure 2(c).

Figure 2.

The tissue phantoms used with 1% intralipid (a) and 1% intralipid with 1% porcine blood (b) are shown setup on the treatment table in the radiation treatment room with the optical fiber set up in contact with the plastic container. Representative Cherenkov spectra measured from LINAC irradiation of these phantoms in shown in (c). The measurements were taken with a single fiber bundle coupled to the spectrometer.

Ten 2-mL centrifuge tubes filled with fluorophores — IRDye 800CW (LI-COR Biosciences, Lincoln, NE) and IRDye 680 RD (LI-COR Biosciences, Lincoln, NE)— at concentrations 25 μM to 0.1 μM in 500 μL phosphate-buffered saline (PBS) were used to simulate fluorescent inclusions (see Figure 2 (c). Both fluorophores were present in all the tubes, and at the same concentrations i.e. a 1:1 concentration ratio was maintained between dyes for simplicity. Typical spectra of the Chernekov emission with the absorption and fluorescence emission spectra of the fluorophores used are shown in Figure 3(a).Each tube was then placed in the ‘background’ liquid phantom which contained PBS + 1% intralipid or PBS + 1% intralipid + 1% whole blood. Blank control tubes containing only background intralipid and blood were also prepared. The phantom setup was placed on a stirring plate, to prevent red blood cell-stacking and inhomogeneity in scatterer concentrations. The tube was positioned inside the ‘background’ phantom container such that it was at a distance of 5 mm from the container wall at which the optical fiber tip was fixed. The radiation beam cross-section shape was set to 1 × 1 cm², and was setup to irradiate the inclusion tube from the top, with the LINAC beam isocenter set to be inside the inclusion. This provided sufficient buildup region for the radiation beam passing through the phantom. Cherenkov emissions and fluorescence emission triggered by radiation delivery were acquired. Five frames of continuous wavelength spectra were generated with 100–3000 accumulations on the ICCD chip (AOC), or pulses. Since the dose rate was 10 cGy/s, and there were 360 pulses/s, this range of AOC values corresponds to roughly 2.8 cGy to 83 cGy.

Figure 3.

Dyes in phantoms and Intralipid (IL) with blood (B). In (a) the absorption (left) and emission (right) spectra are shown for Cherenkov emission in intralipid (black), Cherenkov emission in intralipid & blood (red) and IRDye 680RD (blue) and IRDye 800CW (pink). In (b), an example raw emission spectrum is shown from a combined phantom with Cherenkov and the two emission peaks convolved, with both temporal filtering, and then moving window wavelength filtering to reduce the noise. In (c) the tubes with serial dilutions are shown for both dyes, systematically decreased together with concentrations listed in micromolar. In (d) an example spectrum is shown with the wavelength fitting regions shown, for all 3 components.

2.3. Spectral Fitting and Data Processing

The collected spectra were temporally median-filtered and wavelength-averaged with a moving window filter, as shown pictorially in Figure 3(b). To process the spectra for signals related to dye concentration, the acquired normalized background spectra, from Figure 3(a), were used for spectral fitting, because the signals included all distortions to the Cherenkov emission and fluorescence spectra which could be caused by fiber and system autofluorescence and attenuation. These were thought to be minimal but still direct measurement of these spectra on the same system was completed. Then the detected spectra were used as the input to an algorithm for spectral decoupling, based on a linear least squares solver, fitting in the regions where each component was dominant, as shown in Figure 3(d).

2.3. In Vivo Proof-of-concept work: Subcutaneous Matrigel-based fluorescent inclusions in murine model

Before the above approach could be applied and tested to quantify fluorescent-tracer ratios in animal lymph nodes, a simulated in vivo experiment utilizing fluorophores injected subcutaneously in a mouse was performed. In this experiment, IRDye 680RD and IRDye 800CW were mixed with clear phenol red-free and growth factor-reduced matrigel medium (#356230, Corning) such that their concentration was 3 μM each. The matrigel that was previously stored in 500-μL aliquots at −20°C had to be thawed in an ice tub to raise its temperature to 4°C. Care had to be taken during the handling to prevent overheating the matrigel as this causes it to cross-link and solidify. Dyes dissolved in PBS were cooled and mixed with the matrigel by pipetting. Insulin syringes with 27-gauge needles were chilled on ice prior to loading. Animals were anesthetized using 1.5 – 2.5% isofluorane in 1.5 L/min oxygen, and placed on a heating pad. A 200 μL injection of matrigel mixed with fluorescent dyes was performed subcutaneously such that the matrigel solidified into a spheroid with minimal flattening. Matrigel is a highly viscous substance, care needed to be taken to ensure bubbles are not present in the gel.

For the spectra acquisition, the mouse was placed on a black sheet over layers of solid water phantoms (added to 1.5 cm thickness), to provide sufficient build-up region. The optical fiber position was fixed and the animal was adjusted to maintain contact between the fiber tip the matrigel bump (Figure 6(a)). A 6 MV radiation beam was directed from below the animal such that the beam cross-section size was 1×1 cm², with the isocenter near the optical fiber tip.. A background spectrum was obtained by measuring a matrigel-free skin region away from the liver and gut which are known sources of high autofluorescence. Spectra were acquired using the spectrometer and ICCD camera, and processed in a similar fashion as above to estimate the ratio of IRDye 800CW emission to IRDye 680RD emission. For the sake of simplicity of radiation treatment, and optical fiber-based measurement, the animal was euthanized prior to any measurement by cervical dislocation. All animal work was performed under protocols approved by the Institutional Animal Care and Use Committee (IACUC) at Dartmouth College.

Figure 6.

Preliminary in vivo imaging of a nude mouse injected subcutaneously with 200 μL phenol red-free matrigel basement membrane matrix with 3 μM IRDye 800CW and 3μM IRDye 680RD. The mouse was positioned on the radiation treatment table and the optical fiber was placed in contact with the matrigel injection (a). The measured spectrum was spectrally decoupled to estimate the contributions of each fluorophore.

3. Results

Figure 2(c) shows the detected Cherenkov emissions from the background phantom with 1% intralipid and no blood, and 1% intralipid along with 1% whole blood; note the change in spectral shape when blood was present. The decrease in signal at wavelengths < 620 nm is apparent and matches the known attenuation of blood well. Notably, the signal loss is not strong in wavelengths above 620 nm, because the Cherenkov emission is being produced throughout the entire volume of the sample, and so the transport distance of light getting into the detection fiber was moderate. This is the rationale behind using IRDye 800CW and IRDye 680RD in this study.

Figures 4 and 5 show the processed spectra for various fluorophore concentrations (IRDye680:IRDye800 concentration was 1:1) in 1% intralipid in the background (Figure 4) and the 1% intralipid + 1% blood background medium (Figure 4). The IRDye 680RD measurement is on the left in both figures, while that of IRDye 800CW is on the right in all panels in Figures 4 and 5. The resulting integrated intensity values over the fitted fluorophore peaks for every measurement are plotted as a single point on each graph for each concentration and each radiation dose. Because of high Cherenkov background signals, there is always a lower limit on the concentration of dyes which could be detected with any accuracy. Also, at the highest dye concentrations, there could be self-quenching of the dye, limiting the linearity of response. Concentrations above and below these extremes were eliminated from the data, to focus on reporting those data which are monotonically related to the reporter signal.

Figure 4.

The measured integrated fluorescence signal intensity from intralipid solutions are shown for IRDye 680RD (left) and IRDye 800CW (right) for varying amounts of radiation dose (a) at varied concentration of 0.1 to 25 μM. Then the variation in signal with varying concentration was examined (b) for varied radiation dose of 2.8 to 83 cGy (AOC 100 to 3000). Green dashed lines represent linear fits.

Figure 5.

The effects of integrated signal from intralipid + blood solutions are shown for IRDye 680RD (left) and IRDye 800CW (right) for varying amounts of radiation dose (a) at varied concentration of 0.1 to 25 μM. Then the variation in signal with varying concentration was examined (b) for a varied radiation dose of 13.9 to 83 cGy (AOC 100 to 3000)

Figures 4(a) and 5(a) show the dependence of each fluorophore signal with respect to radiation dose used, at varied concentration of 0.1 to 25 μM. The signals in the blood-bearing phantoms (Figure 5) were considerably noisier than in Figure 4, largely because the background Cherenkov of course varies with dose as well, and so the entire signal to noise of the fit degraded more significantly with less dose delivered. To better interpret the data, fits of signal versus dose were plotted in green dotted lines for these figures. In both figures, simultaneous fitting of IRDye 680RD and 800CW was possible over some range of concentrations (Figure 4 (b) and Figure 5 (b)). In comparing Figure 4 (a) to Figure 5 (a), it is apparent that the contribution of blood merely limits the output signal by about a factor of 2, while the full range of concentrations could be detected with the radiation dose used here.

Processed and fitted spectra for the murine matrigel inclusion loaded with 3 μM IRDye680RD and IRDye800 CW are shown in Figure 6 (b). The results indicated that it was possible to resolve signals for the ratio of 800:680 down to 1, as expected.

4. Discussion

This study focused on the particular question of utilizing fluorescent molecular probes in Cherenkov excited molecular sensing. The results have large implications as to the utility of spectrally resolved CELSI sensing in vivo, and the uniqueness of the geometry shown in Figure 1 (e), where selective areas of tissue can be probed with careful geometric and spectral choices. The concept of injecting x-rays as a way to excite molecular probes in vivo with Cherenkov light has not been widely exploited, but with the right molecular sensor, the signals can be detected from deeper than most areas which could otherwise be sampled with light. However, the range of concentration and radiation dose required for sensing will directly determine what this imaging tool can be used for, and so these two factors in particular were studied here.

Several previous studies had focused on phosphorescent agents, which have emission lifetimes in the 10’s of microseconds range, allowing for temporal gating of the signal, effectively eliminating Cherenkov completely from the signal. In such a case, luminescence can be sampled well down into the sub-nanomolar concentrations (3), because the background signal has been eliminated. In the case of time-independent spectra detection, the fluorescent signal is embedded on top of the Cherenkov background, and so the deconvolution of the two signals is dependent upon the system noise and variability, and inherently is bound to be orders of magnitude less sensitive. The data shown in figures 3 and 4 support this, indicating that the sensitivity range of IRDyes detected in this manner is about 1-10μM, in tissue equivalent media. The upper end of the detection is likely higher than this, but just dependent on the microchemical quenching effect of the dye and the microenvironment that it is in. However, the lower end of this detection range is realistic for in vivo imaging, as it has more to do with the detection methodology than the tissue and dye.

Some interesting observations can be taken from the results beyond these simple limits though. For example, the effect of blood versus no blood is quite subtle in the signal. For example, note that the signals in Figure 5 (a) are not that much lower than Figure 4 (a). This is perhaps obviously given by the different amounts of Cherenkov light seen in emission from the two solutions, as graphed in Figure 2 (c). The reason that these are not all that different is likely because the fiber collecting the light picks up through the whole volume, but the Cherenkov is launched through the whole volume as well, and because of the scattering the fiber will preferentially pick up light from near the surface. So, while the Cherenkov light transmission through longer distances of tissue will undoubtedly be more attenuated in the NIR by blood, in the short pathlength, it does not appear to be so. Given that fluorophores are excited locally by the Cherenkov light originating right within the phantom (tissue), the distance effect of attenuation is less of an issue than just the amount of Cherenkov generated. In addition to the tissue-simulating phantom studies, the preliminary in vivo testing showed that ~1:1 ratio of fluorophore concentrations was measurable when 3 μM of each dye was injected simultaneously and subcutaneous in the mouse body suggesting that the presented approach may be promising in ratiometric fluorescent tracer quantification (8, 17).

The fact that the detectable concentration range is limited to the micromolar regime, will mean that this style of sensing will not be sufficient for immune-tagged fluorophores, where typical cell surface receptors or cell receptor concentrations are routinely in the range of 1-100 nM (18). However most metabolic events such as enzymatic proteins, minerals, respiratory cell components are present in tissue at the micromolar range, and these are primary targets to delineate cancer tumors. So fluorescent sensing of these reporters should be inherently possible for CELSI sensing. Additionally, at least two fluorophores should be detectable. The two dyes chosen here were based upon the results of our previous study, and so are thought to be optimally separated in wavelength to allow maximal accuracy in spectral fitting (3).

5. Conclusions

The observations from this tissue phantom and preliminary in vivo study indicate that Cherenkov is attenuated by blood, but predominantly in wavelengths < 620 nm, and so choosing fluorophores with emissions above these wavelengths is optimal. Demonstration of simultaneous detection of two NIR emitting fluorophores in the Cherenkov spectrum illustrates what is achievable in future tissue studies. The concentration range feasible to detect signals is the 1-10 μM, indicating that metabolic tracer imaging with multiple reporters would be feasible with IRDyes and Cherenkov excitation. Further work to optimize the signal to background might lower this detectable range, either by better time-gating or other means, but in this current state of spectrometer-based detection, it would appear that this range is approximately the working region of use in vivo. The doses required are in the range of 3-80 cGy, which is at the upper end of a diagnostically useful dose.