2.1.

Correction Theory

To measure intrinsic fluorescence, it is necessary to correct the measured raw fluorescence signal for the influence of ${\mu }_s^{\prime }$ and ${\mu }_a$. The method of determining ${\mu }_s^{\prime }$ and ${\mu }_a$ from MDSFR spectra has been described in detail elsewhere.^21,22,43 Furthermore, a semi-empirical model is developed, based on Monte-Carlo simulations and phantom studies to account for the influence of tissue optical properties on a fluorescence signal.^23,27 To obtain the fluorescence in the absence of absorption $F_{\mathrm{SF}}^0$ from the measured fluorescence $F_{\mathrm{SF}}$, a modified form of Beer’s law is used.²⁷

Here, ${\mu }_{a,\mathrm{avg}}$ is the average ${\mu }_a$ at the excitation and emission wavelengths and the fluorescence photon path length $L_{\mathrm{SFF}}$ is defined as

In this equation, $A_1$ is a proportionality factor that depends on the geometrical shape of the effective volume, and $A_2$ and $A_3$ are determined empirically to be $0.71\pm 0.01$ and $0.36\pm 0.01$, respectively.^23,27

A model is previously developed to relate $F_{\mathrm{SF}}^0$ to the intrinsic fluorescence spectrum $F({\lambda }_m)$ (${\mathrm{nm}}^{-1}{\mathrm{mm}}^{-1}$), which is the product of fluorophore absorption coefficient ${\mu }_a^f({\lambda }_x)$ and quantum yield at the emission wavelength $Q^f({\lambda }_m)$,²⁷

Here, $\langle V\rangle$, $\langle {\varphi }_x^V\rangle$, and $\langle H_m^V\rangle$ are the effective sampling volume, excitation fluence, and escape probability, respectively, which can all be described as a function of ${\mu }_s^{\prime }$ and $d_f$. The factor ${\lambda }_x/{\lambda }_m$ accounts for difference in photon energy between emission and excitation wavelength. $F_{\text{Ratio}}^{\mathrm{SF}}$ is the ratio between the total excitation photons launched (related to $P_{\text{laser}}$) and the emitted photons detected (related to $F_{\mathrm{SF}}^0$). The product of Eq. (4) and the expressions for $\langle V\rangle$, $\langle {\varphi }_x^V\rangle$, and $\langle H_m^V\rangle$ yields

2.2.

Measurement Setup

The illustration in Fig. 1 represents a single-probe unit of the combined MDSFR-SFF spectroscopic setup. Two of those probe units were used, with the only difference in fiber diameter of the probe and quadfurcation, to be able to perform measurements with both 0.4 and 0.8 mm probes. Positioning of the probes was done manually.

Fig. 1

Schematic diagram of the measurement setup. Reflectance and fluorescence are measured through a single fiber of either 0.4 or 0.8 mm. Two identical setups are used to accommodate the two fiber diameters.

The measurements consisted of subsequent reflectance and fluorescence measurements without probe repositioning. During a reflectance measurement, white light emitted by a halogen light source (HL-2000-FHSA; Ocean Optics, Duiven, Netherlands) was directed through the first leg of a quadfurcation into a solid core fiber-optic probe and guided onto the tissue. A small fraction of the white light was reflected from the tissue into the acceptance cone of the probe fiber and directed via a second leg of the quadfurcation into a spectrometer (SD-2000; Ocean Optics] for detection and further analysis. After the reflectance measurement, the fluorescence measurement was performed, using low power 405-nm excitation light coupled into one leg of the quadfurcation to illuminate the tissue through the same solid core probe as the reflectance. The emitted fluorescence was then guided through the last leg of the quadfurcation, which leads through a 450-nm long-pass filter into a second spectrograph (QE-65000; Ocean Optics) by the use of a solid core 0.2-mm fiber. The complete MDSFR-SFF system consists of two identical probe-units containing probe core diameters of 0.4 and 0.8 mm, combined with a $4\times 0.1\mathrm{mm}$ and a $4\times 0.2\mathrm{mm}$ quadfurcation, respectively. The fiber probes were sequentially placed in contact with the tissue under investigation. Both probes were polished under an angle of 15 deg to minimize internal specular reflections from the probe tip. The calibration procedure using a diluted Intralipid 20% phantom included correction for fiber alignment and transmission efficiency and is described elsewhere.⁴⁵

2.3.

Animal Model

For this study, 19 mice (BALB/c nu/nu female mice, aged 4 to 6 weeks; Charles River Laboratories, Chatillon-sur-Chalaronne, France) were injected metastatic oral squamous cell carcinoma cell line OSC19. This cell line has been lentivirally transfected with luc2 (Biocat, Heidelberg, Germany) and Cop-GFP (Evrogen, Moscow, Russia). $6\times {10}^4\text{cells}$, diluted in 20-μl phosphate buffered saline (PBS), were injected to induce an oral squamous cell carcinoma at the tip of the tongue after an incubation period of 10 to 13 days.^43,46–48 GFP was included in the cell line to easily assess the inclusion of tumor tissue in the measurement volume. After tumor incubation, animals were restrained and injected with chlorin e6 ($16\mathrm{mg}/\mathrm{kg}$ in 100-μl PBS, Frontier Scientific, Inc., Logan, Utah) or Bremachlorin ($20\mathrm{mg}/\mathrm{kg}$, form stock solution of $3.5\mathrm{g}{\mathrm{l}}^{-1}$, Radapharma International B.V., The Netherlands). At $t=3$, 4.5, 6, 24 or 48 h after photosensitizer injection, animals were anesthetized with an injection of ketamine/xylazine and spectroscopic reflectance measurements with both fiber diameters and fluorescence measurements with the 0.4-mm fiber were performed on normal tongue tissue, OSC tumor (both five measurements per fiber diameter), skin and liver (both three measurements per fiber diameter), after which the animals were sacrificed. The consecutive measurements were taken by removing and carefully repositioning the fiber between measurements, without imparting undue pressure on each tissue. For tumor, the fiber tip was positioned directly on top of the visible tumor. The study protocol was approved by the Animal Welfare Committee of the Leiden University Medical Center. Housing of BALB/c nu/nu female mice, the experiments and euthanazation were performed in accordance with the guidelines of this committee. Chlorophyll-free food and sterilized water were provided without restriction.

2.4.

Data Analysis

All data analyses were performed using custom applications running in MATLAB (release 2009b, The Mathworks, Natick, Massachusetts). Reflectance spectra were fitted with a combination of basis spectra by using a Levenberg–Marquart nonlinear fit procedure to estimate ${\mu }_a$. ${\mu }_s^{\prime }$ was determined by averaging the absorption-corrected reflectance spectra per fiber diameter and fitting a model through those averaged reflectance spectra of both fiber diameters, as previously described.⁴³ Individual measurements that were excluded from the MDSFR fit were excluded from SFF analysis as well. Spectra were excluded from the analysis when (1) fluorescence measurements did not indicate the presence of tumor tissue in the tumor measurement, (2) we were unable to fit ${\mu }_a$ to the individual reflectance spectrum, and (3) we could not successfully fit ${\mu }_s^{\prime }$ or $\gamma$ to the averaged absorption-corrected reflectance spectra. The fluorescence spectra from all individual measurements were corrected for absorption and scattering as described above. These corrected fluorescence spectra were then divided into autofluorescence, GFP fluorescence, and photosensitizer fluorescence by spectral deconvolution. The autofluorescence and GFP were then subtracted to obtain the intrinsic fluorescence spectrum. Basis spectra for GFP and autofluorescence were constructed from fluorescence measurements on more than five animals before fluorophore injection, where normal tongue and tumor were fitted with an identical autofluorescence spectrum. Skin and liver were both analyzed with their tissue-dependent autofluorescence basis spectrum. Spectral deconvolution was again based on a Levenberg–Marquart nonlinear fit procedure. The resulting intrinsic photosensitizer fluorescence spectrum was related to the product of photosensitizer absorption coefficient and quantum yield ${\mu }_a^f{Q}^f$ by integration over wavelength. To quantify variations in the shape of the spectral emission, two Gaussians were fitted to the main ce6/Bremachlorin peak present in the intrinsic photosensitizer fluorescence spectrum; one covering the high-intensity emission peak, fitted between 650 and 680 nm and a skewed Gaussian to determine the width and intensity of the emission shoulder at the right side of the maximum between 680 and 800 nm. This skewed function is defined as

The parameters $b_1$ to $b_4$ were fitted using a least-mean-square method to describe the shape of the right hand sight of the spectra. From these fitted Gaussian curves, we were able to determine the intrinsic fluorescence intensity, peak wavelength, FWHM of both fitted Gaussians, and the ratio between different fit parameters. The underlying reasons for implications of these spectral shifts are discussed below.

3.1.

Magnitude of the Correction for Tissue Optical Properties

The correction factor was calculated over the whole spectral range for both absorption and scattering for the four different tissues, averaged over all mice. These are shown in Figs. 3(a) and 3(b). The error bars indicate the SD of all measurements per tissue. The correction factors were dependent on the tissue type and illustrated the importance of our technique for the recovery of intrinsic fluorescence in different tissue types: absorption was significantly stronger in liver where there was a high concentration of blood, whereas skin had the lowest absorption. Scattering correction did not vary significantly with tissue type [Fig. 3(b)]. The spectral variation of oxygenated and deoxygenated blood was observed in tumor and normal tongue absorption coefficients. The variation in the total correction factor over the visible wavelength range for both the absorption and scattering is shown in Fig. 4, where the maximum value of the absorption- or scattering-based correction factor is divided by the minimum value. This figure shows clearly that correction for tissue absorption was most important with regards to analysis of spectral shape. Scattering varied less over wavelength, although the contribution of the scattering correction to the total correction procedure was larger than that of absorption. Therefore, incorrect scattering correction will result in an incorrect quantification of the fluorophore but much less in incorrect conclusions on spectral shape. Applying both absorption and scattering correction results in the intrinsic fluorescence spectrum.

Fig. 3

(a) Average absorption correction factor spectrum for four different tissues. Since the absorption coefficient was wavelength dependent, the absorption correction is, too. Calculation is based on Eq. (1). (b) Average scattering correction factor spectrum for four different tissues, resulting from Eq. (5).

Fig. 4

Variation over the visible wavelength range (400 to 800 nm) of the corrections for absorption and scattering. Maximum value in that range divided by minimum value.

When comparing intrinsic fluorescence spectra of Bremachlorin and ce6 of normal and tongue tumor tissue in Fig. 5, Bremachlorin showed higher fluorescence intensity than ce6 did for both the tissues. Furthermore, the tumor tissue intensity was much higher than that of normal tissue. To illustrate the spectral variations of ce6 and Bremachlorin in different tissues at different time points, Figs. 6(a) and 6(b) show normalized spectra acquired from different tissues and time points. Between 3 and 24 h after injection, the ce6 peak shifted toward longer wavelengths, whereas for Bremachlorin, a shift toward lower wavelengths was measured. Furthermore, clear differences in peak width and the contribution of the longer wavelength shoulder were observed over time after photosensitizer administration.

Fig. 5

Typical ce6 and Bremachlorin spectrum in normal tongue and tumor, 3 h after photosensitizer injection.

Fig. 6

(a) Characteristic intrinsic photosensitizer fluorescence spectrum for varying tissue, time, and photosensitizer, normalized to $\text{peak}=1$. (b) Close up of peak position.

To quantify these spectral variations in the intrinsic photosensitizer fluorescence spectrum, the fluorescence was characterized by two fitted Gaussian distributions as shown in Fig. 7; one covered the high-intensity emission peak, fitted between 650 and 680 nm and a skewed Gaussian determined the width and intensity of the emission shoulder at the right side of the maximum between 680 and 800 nm.

Fig. 7

Fluorescence intensity is fitted with a Gaussian distribution between 650 and 680 nm, and a skewed Gaussian distribution from 680 to 800 nm.

3.2.

Intensity and Ratio

Both ce6 and Bremachlorin showed the highest intensity of the intrinsic fluorescence spectrum, hereafter referred to as fluorescence intensity, in liver and lowest in normal tissue. Furthermore, fluorescence intensity of Bremachlorin was significantly higher in all tissues at all time points (except 3 and 6 h measurements in liver, where the difference is not significant), as can be seen in Fig. 8(a). Every bar in Fig. 8 indicates the average measurement value of all measurements taken on all mice for that particular tissue and time point, which corresponds to an average of 6 to 10 measurements. The fluorescence intensity was measured as the maximum value of the fluorophore intrinsic fluorescence spectrum. It is evident from the data [Fig. 8(a)] that there was relatively less ce6 present in time as it was cleared more rapidly than Bremachlorin. This may have significant implications for the use of Bremachlorin as a photosensitizer and will be discussed in more detail elsewhere. Figure 8(b) shows the intensity ratio of the main peak and the shoulder on the right side. For this, the maximum values of both fitted Gaussian distributions were divided by each other. Ce6 showed ratios between 8.4 and 11.0, significantly larger than Bremachlorin (ratios between 5.9 and 7.5) for 3, 4.5, and 6 h after injection in all tissues, indicating a lower shoulder for ce6 compared with Bremachlorin ($P<0.02$ for all time points and tissues). The 24- and 48-h time points were not considered in this analysis because the fluorescence intensity of the shoulder was, in many cases, too low to be fitted accurately. For each photosensitizer, differences over time and tissue were small and rarely significant, as indicated in Fig. 8(b).

Fig. 8

(a) Maximum value of intrinsic fluorescence spectrum. This parameter correlates with the product of photosensitizer absorption coefficient and quantum yield at the peak wavelength. (b) Ratio between maximum value of intrinsic fluorescence spectrum and the fitted intensity parameter of the skewed Gaussian. This represents the ratio between peak and shoulder of the ce6/Bremachlorin fluorescence spectrum. (c) Wavelength that represents maximum intrinsic fluorescence peak position. (d) FWHM of the Gaussian representing the main intrinsic fluorescence peak.

3.3.

Width

When comparing the width of the main fluorescence emission peak of both photosensitizers in Fig. 8(d), it can be seen that in general the width of the spectrum of ce6 peak was significantly smaller than that of Bremachlorin. It should be noted that the 24 and 48 h measurements of ce6 were associated with large uncertainty caused by the very low fluorescence intensity, compared to Bremachlorin, as shown in Fig. 8(a). This was reflected in an increased fit uncertainty. Over the first 6 h after injection, Bremachlorin showed a stable peak width for all tissues under investigation, whereas ce6 showed a steady but not a significant decrease over time for tumor and normal tissue.

4.1.

Spectral Variability in ce6

It has been shown in vitro that both the fluorescence peak intensity and the peak position of ce6 are strongly influenced by pH of the solution.³² For ce6 in PBS, a decrease in pH in the range of 7.5 to 3.5 results in a peak shift from 661 to 647 nm, combined with a decrease in intensity of the main fluorescence peak when the spectrum is normalized at 720 nm.³² The range of peak positions for ce6 that we have observed was in the range of 665 to 671 nm, which is higher than reported by Cunderliková et al. It is shown for the spectral properties of ce6 in more physiologically relevant environments containing serum (by adding FCS, BSA, LDL/HDL) or measured in vivo that the binding of ce6 significantly shifts the peak position to higher wavelengths.⁴⁰ Studies of ce6 or ce6 conjugates in solution show a peak wavelength range comparable with what was found in this study.^35,40,49 Changes in pH influence binding affinity and thus the spectral shape represent the relative contributions of the differently bound ce6. The data in Figs. 8(b) and 8(c) showed a trend of increasing peak/shoulder ratio corresponding with increasing peak position wavelength in time after photosensitizer injection for tumor, normal tissue, and skin. According to Cunderliková et al.,³² the increase of peak/shoulder ratio and peak position suggests a transition into a higher pH environment in an in-vitro situation. In the range of biological relevant pH values, the peak shift associated with pH variations is small. How this transfers to the in-vivo situation remains to be determined.

4.2.

Comparison of ce6 and Bremachlorin

It is clear from Figs. 8(b) and 8(c) that although ratio and peak position increased for ce6, Bremachlorin showed the opposite; a blue shift in peak position accompanied by a decrease in shoulder ratio. This might suggest relocalization of the drug to an environment with a lower pH. Furthermore, binding of ce6 to low-density lipoproteins is also linked with a blue-shift of the peak position, compared to albumin bound ce6. This suggests that ce6 in Bremachlorin may be preferably bound to these LDL proteins, providing for uptake in (malignant) cells. Unbound ce6 is mainly accumulated in the plasma membrane of cells. The width of the peak, which could also be associated with pH, did not change significantly.

Bremachlorin consists of a mixture of chlorin e6, purpurin 5, and chlorin p6, in which both ce6 and purpurin 5 are known to have the photodynamic effect. In the administered mixture, ce6 also acts as a solvent, since purpurin 5 is too hydrophobic to be administered in aqueous solution. Purpurin has a high-binding affinity to low-density lipoproteins, allowing the drug to enter malignant cells via endocytosis.⁵⁰ Since ce6 accounts for the largest fraction in Bremachlorin, we expected the spectral shape to be dominated by ce6. The influence of the other two components, purpurin 5 and chlorin p6, on the spectral shape and behavior is unknown. Both purpurin 5 and chlorin p6 are known to have an emission spectrum similar to ce6 when analyzed separately.^51,52 Our results showed that a mixture of ce6, cp6, and purpurin 5 clearly behaved differently than pure ce6. Bremachlorin is more lipophilic than pure ce6,⁴² and it is therefore expected that the localization of Bremachlorin differs from pure ce6. It could not be determined from this experiment if the spectral variations between Bremachlorin and ce6 were a result of spectral influences of the two extra components, or if they were variations in ce6 spectra caused by changes in environment. However, our experiments did show that Bremachlorin has different spectral properties than pure ce6 and should therefore be investigated as a new photosensitizer, including optimization of the treatment protocol.

4.3.

Fluorescence Correction Method

The method to use MDSFR for measuring the absorption and reduced scattering in vivo to recover the intrinsic photosensitizer fluorescence is developed based on the use of Monte-Carlo simulations and phantom studies. In these optimized circumstances, theory is validated and empirical parameters are included in the model.²³ However, applying this method in vivo is much less straightforward considering spatial heterogeneity and the potential for the confounding effects of tissue layers. In this study, MDSFR was performed using consecutive placement of two fiber diameters on the same spot. Therefore, slight differences in measurement volume were not only due to different fiber diameters but could also occur from measuring a slightly different sampling volume in each measurement. Therefore, the ${\mu }_s^{\prime }$ used to correct the fluorescence was based on an average value and did not necessarily exactly represent the ${\mu }_s^{\prime }$ in the fluorescence sample volume. New technological developments should overcome this problem in the near future and provide true colocalized measurement positions for various fiber diameters. Although we repositioned the probe on each tissue up to five times per mouse, the variation of variables compared within measurements of one mouse was at least four times smaller than the maximum variation between animals. Therefore, the size of the error bars was predominately due to interanimal variations.

4.4.

Implications and Future Work

In this work, we have investigated spectral variations of chlorin e6 and Bremachlorin. Clear differences were observed in fluorescence intensity, peak/shoulder ratio and peak position for both sensitizers over time after injection. Furthermore, although Bremachlorin consists of approximately 80% of ce6, spectra showed distinct differences, suggesting variations in localization and/or clearance pathways. The more blue-shifted peak of Bremachlorin, together with the wider peak, suggests Bremachlorin was bound to LDL proteins more than ce6. This might result in a higher uptake of the photosensitizer in the cell. It is shown that internalization of the photosensitizer results in more PDT-induced damage than photosensitizers located on the cell membrane.⁵³ Furthermore, this may explain the prolonged detection of significant amounts of photosensitizer fluorescence in tissue compared to ce6.

Although results in this study look promising for using the ce6 in Bremachlorin as photosensitizer, purpurin 5 is known to have strong photodynamic activity as well. Spectral analysis showed us mainly variations in ce6. How this ce6 behavior relates to behavior and localization of purpurin 5 and influences PDT outcome remains to be investigated. In future studies, we aim to analyze the microscopic distribution of ce6 and Bremachlorin in tongue/tumor tissue.