In vivo quantification of photosensitizer fluorescence in the skin-fold observation chamber using dual-wavelength excitation and NIR imaging
- First Online:
- Cite this article as:
- Kaščáková, S., de Visscher, S., Kruijt, B. et al. Lasers Med Sci (2011) 26: 789. doi:10.1007/s10103-011-0888-z
A major challenge in biomedical optics is the accurate quantification of in vivo fluorescence images. Fluorescence imaging is often used to determine the pharmacokinetics of photosensitizers used for photodynamic therapy. Often, however, this type of imaging does not take into account differences in and changes to tissue volume and optical properties of the tissue under interrogation. To address this problem, a ratiometric quantification method was developed and applied to monitor photosensitizer meso-tetra(hydroxyphenyl) chlorin (mTHPC) pharmacokinetics in the rat skin-fold observation chamber. The method employs a combination of dual-wavelength excitation and dual-wavelength detection. Excitation and detection wavelengths were selected in the NIR region. One excitation wavelength was chosen to be at the Q band of mTHPC, whereas the second excitation wavelength was close to its absorption minimum. Two fluorescence emission bands were used; one at the secondary fluorescence maximum of mTHPC centered on 720 nm, and one in a region of tissue autofluorescence. The first excitation wavelength was used to excite the mTHPC and autofluorescence and the second to excite only autofluorescence, so that this could be subtracted. Subsequently, the autofluorescence-corrected mTHPC image was divided by the autofluorescence signal to correct for variations in tissue optical properties. This correction algorithm in principle results in a linear relation between the corrected fluorescence and photosensitizer concentration. The limitations of the presented method and comparison with previously published and validated techniques are discussed.
KeywordsMeso-tetra(hydroxyphenyl) chlorin (mTHPC Foscan®) Ratio fluorescence imaging technique mTHPC pharmacokinetics Skin fold observation chamber
The primary response to photodynamic therapy (PDT) is determined by the tissue oxygenation, light fluence (rate), and local concentration of the photosensitizer. The biological activity of the photosensitizer within the illuminated volume is related to the concentration of the fluorescent active form of the photosensitizer. Thus, optical imaging of photosensitizer fluorescence (using light doses much lower than are necessary for PDT damage) aids the understanding of PDT by monitoring the photosensitizer spatial distribution and its fluorescence activity.
For many years, observation chambers implanted in various animal species and in humans have been used for intravital microscopy of living tissue [1–12]. Through the possibility to transplant neoplastic tissue within the chamber, the skin-fold observation chamber was especially developed to monitor the early vascular events, anti-tumor effects, and pharmacokinetics of photosensitizers.
However, the ability to accurately quantify the in vivo measured fluorescence is critical [13, 14]. The fluorescence signal originates not only from the photosensitizer but also from various other fluorescent molecules naturally present in the tissue that can cause an unknown and variable amount of background autofluorescence . Moreover, the measured fluorescence signal is also influenced by geometric factors (the distance and the angle of the excitation and detection source relative to tissue surface) and the tissue optics (scattering and absorption of the excitation and fluorescence emission light in the tissue). For example, tissues with higher background absorption coefficients, e.g., due to high melanin or blood content, can decrease the propagation of both the excitation and fluorescence emission light. This is further complicated by the fact that the absorption spectrum of blood depends on its oxygenation. In addition, tumor tissue is less scattering than normal tissue [16–18] and the thickness for the different types of tissue within the chamber varies as well. In general, the tumor within the chamber is thicker than the normal tissue. Furthermore, normal tissue can also show variability in thickness at the different sites within the chamber. Since the collected fluorescence image is influenced by all of the factors discussed above (optical properties of the tissue, tissue autofluorescence, chamber thickness variations, and geometric illumination and collection factors), an imaging methodology that corrects for these factors is necessary to obtain quantitative photosensitizer fluorescence images.
Several techniques have been developed to correct the measured fluorescence for tissue autofluorescence, absorption, and scattering properties of the tissue and variations in irradiance, excitation geometry, and detection efficiency. Profio  calculated the ratio of fluorescence marker signal over the reference autofluorescence. Baumgartner et al.  and Witjes et al.  described a subtraction method where the autofluorescence background signal is subtracted using dual-wavelength excitation. In our group, Sinaasappel and Sterenborg developed the double ratio technique based on dual-wavelength excitation and dual-wavelength detection  and more recently Saarnak et al.  published a ratiometric method based on measuring the autofluorescence signal prior to marker administration.
Bogaards et al.  reviewed the performance of these correction methods for various input parameters over ranges that can be expected during in vivo imaging around a standard set of optical properties representing those for human skin. The study revealed that the subtraction method of Baumgartner et al.  corrected the detected fluorescence signal of the marker for variations in autofluorescence, but was unable to correct for variations in irradiance, excitation geometry, and tissue optical properties. The ratiometric methods of Profio , Sinaasappel and Sterenborg , and Saarnak et al.  completely corrected for variations in irradiance, excitation geometry, and detection efficiency. The correction for tissue optical properties was to a great extent also achieved using ratiometric methods. However, the method of Profio  remained dependent on changes in autofluorescence. The method of Saarnak et al.  demonstrated the best quantification performance, as it depends only on the concentration of the fluorophore. This method is based on image acquisition before and after fluorophore administration, and it requires that these subsequent images be taken from the same tissue site under an identical geometry. Here the assumption was made that the optical properties and autofluorescence may change spatially within the image, but remain constant over the time interval between image acquisitions. However, these conditions are typically not met during window chamber pharmacokinetics experiments. Especially the assumption that autofluorescence, tissue optical properties, and tissue thickness remain constant over the time period for which marker pharmacokinetics is studied (typically over periods of up to 7 days) may not be valid. In this case, the method of Sinaasappel and Sterenborg , which does not have such restrictive conditions with respect to window chamber changes over time, was revealed to be a better option for fluorophore quantification. However, the double ratio method of Sinaasappel and Sterenborg  suffers from a significant drawback: the relation between marker fluorescence and marker concentration is non-linear and saturates for high marker concentrations. This severely limits the applicability of the double ratio correction technique for determining absolute fluorophore concentrations for the study of photosensitizer pharmacokinetics where large variations (several orders of magnitude) in fluorophore concentrations need to be quantified [14, 22].
In summary, two important factors limit the quantification performance of state-of-the-art ratiometric methods : (1) the remaining dependence of the methods on time-dependent variations in autofluorescence, tissue optical properties, and tissue thickness , and (2) the relation between marker fluorescence and marker concentration is non-linear and saturates for high marker concentrations. In the present study, we address both issues using a novel ratiometric method that corrects for time-dependent variations in tissue thickness and tissue optical properties and features a linear response even for high photosensitizer concentrations. In addition, in our method, we minimize the dependence of the fluorescence signal on tissue optical properties by selecting the excitation and emission wavelengths in the red and in the near infrared (NIR), where the tissue absorption and scattering coefficients are relatively small and do not vary much with wavelength . We have applied our method to measure the pharmacokinetics of the second–generation photosensitizer 5,10,15,20-tetra (m-hydroxyphenyl) chlorin (m-THPC) in the rat skin-fold observation chamber.
mTHPC (Foscan®) (c = 4 mg mTHPC/ml dissolved in PEG, EtOH, water-free solution) was obtained from Biolitec pharma (Edinburgh, The United Kingdom). Alexa Fluor 720 was obtained from Invitrogen (Breda, The Netherlands). Polyethylene glycol 400 (PEG400), Evans Blue, and Titanium dioxide (TiO2) were obtained from Sigma-Aldrich (Zwijndrecht, The Netherlands) and 96% ethanol (EtOH) was purchased from Merck (Amsterdam, The Netherlands).
Preparation of mTHPC and administration
Stock solution of mTHPC was dissolved in a solution of PEG400:EtOH:water = 3:2:5 (v/v) to a concentration c = 0.126 mg/ml. The solution was stored in the dark for at least 30 min prior to injection, after which the sample was determined to be stable, i.e., no changes in fluorescence peak intensity were detected. To investigate the state of photosensitizer after 30 min from mixing with PEG:EtOH:H2O mixture, the absorption spectra of mTHPC were recorded in the concentration range c = 0.11 – 200 μM. Over the concentration range, the spectrum of mTHPC did not change and the solution accurately followed Beer’s law with no evidence of spectral peak broadening or shifts in absorption maxima. The same species of mTHPC, i.e., the monomer, is present at the concentration administrated to the animal. The prepared mTHPC solution was administered intravenously at a dose of 0.15 mg kg−1 body weight under isoflurane/O2/N2O anesthesia.
The animal experiment committee of the Erasmus Medical Center approved the experimental design for this study. Skin-fold observation chambers were prepared using a slightly modified technique to the one that has been described previously [10, 23]. Briefly, the chamber was prepared on the back of female Fischer-344 rats (weight m = 144 ± 3 g) in four operations (carried out under general isoflurane/O2/N2O anesthesia) during the period of 2 weeks. As a result of these operations, a thin layer of subcutaneous tissue was clamped between mice and a cover slide. In the first operation, sterile air (12 ml) was subcutaneously injected on the back of the rat to gently separate the skin from the underlying tissue. In the second operation, a plastic ring and a cover slide was positioned under the skin above the subcutaneous tissue matching the vessels. The third operation was proceeding on the second week. During this operation, the skin was folded, prepared and fixed in the splint. Finally, during the last operation, the mammary adenocarcinoma (R3230AC) tumor was transplanted in the layer of normal tissue, reached by unscrewing the cover slide on top of the chamber. Within 5–7 days, the chamber was ready for treatment, i.e., the tumor was supported by blood vessels and had visibly grown. Ideally the chamber contained fat cells and capillaries over an area of approximately 1 cm in diameter with some supporting arterioles and venules.
Experimental set-up for fluorescence and transmission imaging
Principle of correction method
The method presented here is based on dual-wavelength excitation and dual-wavelength detection: One excitation wavelength is chosen to be at an absorption maximum of mTHPC and the other at its absorption minimum. The two emission wavelengths are chosen to be at the secondary fluorescence maximum of mTHPC (at λ = 720 nm) and in the region of no photosensitizer fluorescence. The reason for such a wavelength selection was to detect the changes in the amount of mTHPC fluorescence and monitor the autofluorescence changes in real time. The excitation of mTHPC with 652-nm wavelength (Q-band absorption maximum of mTHPC (Fig. 2)) will produce the highest fluorescence compared to excitation by any other wavelength in the red region of the mTHPC spectra. The second excitation wavelength was 629 nm. This excitation wavelength corresponds to an absorption minimum of mTHPC (Fig. 2). Therefore, excitation of mTHPC with 629-nm wavelength leads to only small amounts of mTHPC fluorescence emission at 720 nm. Furthermore, the fluorescence of mTHPC decreases rapidly beyond 730 nm (with a minimal fluorescence for wavelengths above 750). Therefore, excitation at 629 nm and detection of fluorescence for wavelengths longer than 763 nm will give us information on how background autofluorescence changes with time. It is important to note that fluorescence emission detected in the band-pass filter 720 ± 10 nm (λexc = 652 nm) is a spectral convolution of mTHPC fluorescence and background autofluorescence. Therefore, correction for background autofluorescence in the band-pass filter 720 ± 10 nm is necessary. For this purpose, the subtraction of fluorescence signal detected in the band pass filter 720 ± 10 nm by excitation at 629 nm was included. Subsequently, to ensure correction for tissue optical properties, this difference is divided by the autofluorescence signal excited by 629-nm wavelength and detected in the region of wavelengths longer than 763 nm.
Two assumptions are made: (1) the excitation at 629-nm and 652-nm wavelengths lead to the same autofluorescence, i.e., difference in yield of fluorescence for the tissue fluorescence by excitation at 652 nm and 629 nm is minimal; (2) the excitation by 629-nm wavelength does not lead to mTHPC fluorescence detection at wavelengths longer than 763 nm.
Image analysis was performed using the Labview 7.1 (National Instruments Corporation, Austin, USA). Images were first corrected for background and minor variations in fluence rate of the corresponding excitation light sources. In the second step, the sequence of fluorescence images from each animal was registered by translation and rotation using anatomical landmarks identified in the corresponding transmission images. In the next step, binning of the pixels was performed (4x4) to increase the signal-to-noise ratio. After the pixel binning, the images were resized on the same size of 16-bit image at a resolution of 1344 × 1024 pixels. The registration of images from the second step of analysis enabled us to determine the fluorescence intensity of each tissue type from the same area as follows: in the corresponding transmission image the regions of interest were chosen for each tissue type. Tumor and normal tissue regions of interest were chosen so that no large vessels were in, or close to, the region. Thus in every animal, three regions of interest were chosen for normal tissue, three regions of interest selected inside the tumor, and, depending on the vessel’s content, three up to five regions of interest were selected within the vessels. The same selected regions of interest were also applied for images corrected by ratio imaging technique, where the subtraction and division of the images was done according to Eq. (1).
To investigate the validity of our ratiometric imaging correction technique, we analyzed vessels of various diameters. Blood vessel diameter was estimated from the transmission images. According to determined square field of view of macro lens, one picture element (pixel) corresponds to approximately 3 μm.
In the present study, three female Fischer-344 rats were used to determine autofluorescence and transmission-intensity profiles. However, the mTHPC pharmacokinetics profile was evaluated just in one animal. The aim of our current study is to present the correction method and not to validate the mTHPC pharmacokinetics profile. Presentation of the results from one animal avoids the increase in standard deviations due to intra-animal (biological) variations.
Student’s t test was used to determine significance for the difference in autofluorescence signal measured by two different excitation light sources. Results with a p value below 0.05 were considered significant. Statistical analysis was done using Microcal Origin, version 6.0 (Microcal Software, Inc., Northampton, MA).
Window chamber tissue optical properties and autofluorescence
Fluorescence pharmacokinetics of mTHPC
Figure 7c shows the mTHPC uncorrected fluorescence pharmacokinetic profile within the different tissue types. Five minutes after mTHPC administration, the fluorescence intensity is highest in the vessels and decreases for longer time points. Four hours after mTHPC administration the difference of fluorescence observed in vessels and other tissue types reduces and from time points beyond 8 h after mTHPC administration the fluorescence distribution in tumor and normal tissue is similar. Normal and tumor tissue showed no significant difference in fluorescence intensity over the investigated period and follow the same type of kinetic profile (Fig. 7c): between 5 min and 8 h, the fluorescence intensity does not change significantly, while above the 8-h time point, an increase in fluorescence is observed. The intensity increased up to 48 h after mTHPC administration and decreased for the 96-h time point. For all incubation time points, the standard deviations are higher for normal tissue than for tumor tissue or vessel area.
The time-dependent evolution of the corrected fluorescence images of mTHPC (images corrected by ratio imaging technique, formula ) within the chamber is shown in Fig. 7d. According to transmission images (Fig. 7a), at time point 5 min from mTHPC administration, only vessels show fluorescence. The intensity of fluorescence within the vessels decreased for longer incubation time points. At t = 4 h the contours of the vessels are not as clear anymore due to mTHPC penetration through vessel walls into the tissue. For t > 8 h after mTHPC administration the fluorescence is now within all structures of the chamber. Note that the fluorescence of the plastic ring (Fig. 7b) is not visible in any of corrected images (Fig. 7e).
The time dependence of the corrected fluorescence intensity for the different tissue types is plotted in Fig. 7e. The signal detected from the vessels shows high variations (large error bars) for times below 24 h. The variation was highest for 5 min followed by decreasing variations for longer incubation time points. Beyond 24 h, these variations minimized. In contrast, variations of the corrected fluorescence signal for normal tissue and tumor tissue are small throughout the time-course of the experiment.
We have investigated the use of a ratiometric imaging technique for monitoring the kinetics of mTHPC fluorescence in the rat skin-fold observation chamber. The chamber model was specifically designed for monitoring the pharmacokinetics of photosensitizers used in PDT. However, a problem associated with fluorescence measurements is the difficulty of obtaining quantitative fluorophore fluorescence, due to varying optical properties of tissues and differences in tissue thickness. This has important consequences for the general interpretation of fluorescence measurements in the window chamber. For example, in a previous study, our group determined the spatial distribution of the kinetics of protoporphyrin IX (PpIX) fluorescence during ALA-PDT . PpIX fluorescence kinetics were measured in different tissue types and conclusions were based on data that was not corrected for differences in tissue optical properties and differences in the thickness within and between window chambers. Although the conclusions were based on determining the rate of fluorescence increase in each tissue type separately, temporal variations in tissue optical properties and differences in the thickness of different tissues were not considered. This may be of particular importance, especially when the relationship between the increase of PpIX fluorescence with distance from an arteriole and venule was investigated. Thus in the present study, our intention was to investigate the difference in optical properties of different types of tissue and its time changes in the chamber model. The methodology proposed here accounts for the wavelength dependence of tissue optical properties and overcomes the non-linearity of previously published ratiometric methods.
Tissue optical properties of chamber
Figures 4, 5, 6 clearly show the differences between the fluorescence and transmission intensity acquired from different types of tissue during the course of the experiment. The fluorescence intensity detected from vessels was lower than that from normal or tumor tissue. Considering the blood content within these vessels and the absorption spectra of oxy- and deoxy hemoglobin, the result observed in Fig. 5a or b is not surprising. The same effect is also evident in the transmission-intensity profile from different tissue types (Fig. 4), and in the transmission images of the chamber (Fig. 7a). The chamber vasculature represents the tissue with the highest absorption coefficient, which significantly attenuates the propagation of light. In contrast, tumor tissue exhibits higher autofluorescence intensity than the normal tissue area (Figs. 5, 6). This might be caused by (1) a lower absorption coefficient of tumor tissue; (2) a different scattering coefficient of tumour tissue (it is not evident whether a higher or lower scattering coefficient would result in a higher fluorescence yield); (3) a larger thickness of tumour tissue; (4) a higher native fluorophore concentration in the tumor tissue. Most likely, a combination of these factors results in the observed difference. Since the raw fluorescence signal is confounded by each of these factors, the differences in (and changes to) optical properties of the tissue under interrogation will lead to significant quantification errors of photosensitizer fluorescence within the chamber. Clearly a method that corrects for the influence of these effects would be a step forward.
Validation of assumptions
In the study of Bogaards et al. , it was shown that the performance of imaging techniques can be improved by selecting the excitation and emission wavelengths towards the NIR. Thus, in this study, the excitation wavelengths were carefully selected between 620 and 840 nm, where the tissue absorption and scattering are relatively small. In addition, the combination of exciting tissue fluorescence at a wavelength where the photosensitizer absorbs minimally and detection at the wavelength of no photosensitizer fluorescence, allowed us to monitor kinetics of tissue autofluorescence during the course of the experiment. According to the principle of the method we have presented (Eq. 1), the assumption was made that both excitation wavelengths used in this study lead to the same autofluorescence. This assumption is confirmed in Fig. 5, where the fluorescence intensity detected from different tissue types of control animals is plotted. Within the same tissue type, the fluorescence intensity was not significantly different using 629- and 652-nm excitation.
The second assumption underlying our correction algorithm is that only signals from tissue autofluorescence can be detected in the wavelength region > 763 nm, using 629-nm wavelength excitation. This choice of wavelengths was based on the minimal absorption of mTHPC at the excitation wavelength and on the lack of mTHPC emission in the wavelength region > 763 nm (Fig. 2). In Fig. 6, we did not detect any changes of autofluorescence within the first measured time points in any of the tissue types. In contrast, mTHPC fluorescence pharmacokinetics profile (excitation 652 nm, fluorescence detection at 720 ± 10 nm; Fig. 7c), revealed large differences in fluorescence intensity within the vessels 5 min after the mTHPC administration. It is very unlikely that there are changes of tissue autofluorescence over the timescale of these first few measurements compared to the first measurement of autofluorescence (at t = 0). Therefore, we conclude that the increase of fluorescence intensity observed in chamber vessels in the band-pass filter 720 ± 10 nm (Fig. 7b, c) is due to fluorescence emission of mTHPC. The fact that this effect was not seen in Fig. 6 confirmed that the signal detected in the wavelength region > 763 nm (using 629-nm wavelength excitation) is solely due to tissue autofluorescence.
Temporal changes of tissue optical properties within the chamber
The detection of autofluorescence within the timeframe of the experiment (Fig. 6) revealed that the autofluorescence kinetics for all tissues changed in the same manner, i.e., for measurement time points above 24 h there is a (steady) increase in autofluorescence intensity. The cause of the observed effect might be chamber thickness changes and/or the changes in tissue optical properties at later time points. Unfortunately, our data does not present conclusive evidence as to which of these effects is dominant. The fact that these changes occur within the time-course of the experiment demonstrates the necessity for a correction method, since the marker fluorescence will be affected by these chamber changes similar to the autofluorescence.
Fluorescence pharmacokinetics of mTHPC
Figures 4, 5, and 6 clearly illustrate the necessity of an appropriate correction technique for quantitative measurement of photosensitizer fluorescence in the window chamber. If we were to base conclusions on the pharmacokinetics profile of mTHPC from Fig. 7c (uncorrected fluorescence profile), then the tumor/normal tissue ratio is significantly higher at early time points after mTHPC administration. Taking into account Figs. 4, 5, and 6, this is actually due to a larger thickness and/or smaller absorption/scattering content of tumor tissue, rather than due to a higher mTHPC concentration within the tumor. The larger standard deviation associated with data detected from normal tissue can be explained by thickness inhomogeneities and due to the presence or absence of micro-capillaries. These variations are only visible at high magnification and are difficult to avoid in a model where normal tissue has a significant component of microvasculature.
The correction of raw fluorescence data using the ratiometric correction method that we presented results in a significant decrease in the spatial variation associated with a single measurement time point (evidenced by smaller error bars) and does not report a significant difference between the profile of mTHPC pharmacokinetics between tumor and normal tissue up to 24 h after the administration of mTHPC (Fig. 7e). In addition, we observe that in all uncorrected images, a fluorescent border to the chamber is clearly visible (Fig. 7b). This is due to the presence of a fluorescent plastic ring on which the chamber is mounted. Corrected fluorescence images do not show this artifact (Fig. 7d).
Limitations of the correction method
Optical phantoms and double ratio imaging
It is clear that the method we present as well as the DR correction technique have important limitations  and it is important to stress that the choice of the fluorescence detection technique and correction method should be based on a specific application. This is particularly true for the most appropriate selection of excitation wavelengths and detection wavelength bands. It is important to note that we have applied our correction method in a study investigating mTHPC detection. The application of this technique to different photosensitizers (or other fluorescent species) should involve the careful consideration of the spectroscopic properties of the photosensitizer, so that autofluorescence can be measured during the course of the study in real time and the method remains linearly dependent on photosensitizer concentration.
In summary, we have shown that the raw fluorescence intensity collected from tissues in the skin-fold observation chamber varied significantly between different tissue types and during the period of the investigation. Therefore, any method that does not correct for tissue optical properties and changes in autofluorescence leads to fluorescence quantification errors. We have shown that a ratiometric imaging method utilizing NIR autofluorescence detection can be applied and corrects for differences and changes in tissue optical properties and thickness. A limitation of our method is that for very early time points after the photosensitizer administration, a dependence on vessel size was found due to a mismatch in localization of marker fluorophore and background autofluorescence. Other than this limitation, the method we present shows a high sensitivity even for high photosensitizer concentrations.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.