2.1.

General Design and Optics

A schematic representation of the apparatus is shown in Fig. 2. The excitation light pulses at 405 nm are provided by a dye laser (1) (LTB model UDL-200), pumped by a pulsed N₂ laser (LTB model MSG 803-TD). The laser beam enters the main, light-tight, blackened experimental enclosure (2) through a small hole and then passes through an iris diaphragm (3^′) which blocks most of the spontaneous emission of the dye cell. The beam passes through a small channel drilled in the parabolic mirror (4) (Edmund Optics NT47-103, 90° off-axis, aluminium coated, effective focal length (EFL) = 50.8 mm), while being focused with a low luminescence plano-convex lens (3) (Edmund Optics NT47-276, [TeX:] \documentclass[12pt]{minimal}\begin{document}$\varnothing$\end{document} $⌀$ 20 mm, EFL = 150 mm) onto the tip (5) of a multimode optical fiber (6) (Laser Components HCG-M0550T, core diameter = 550 μm, NA = 0.22). For in vitro experiments, a quartz cuvette (Hellma, 174-QS) is held inside the apparatus at the irradiation location (5), using a thermostable cuvette holder (6^′), see Fig. 3. The plano-convex lens (3) is mounted on a 5-degrees-of-freedom holder, allowing for easy focusing of the laser beam on a pre-defined point, (5). The luminescence emitted from the irradiated sample is collected by the far-end of the optical fiber (6) and returned to the enclosure, where a large fraction (probably more than 80%) of this light falls on the parabolic metal mirror (4) and is reflected and focused by it on a gatable photomultiplier (PMT) (7) (Hamamatsu R955). The reflected laser light and unwanted optical background noise are filtered out by a bandpass filter (Chroma HQ645/75) in the filter wheel (8). The PpIX DF signal is recorded with a digital storage oscilloscope (DSO) (Lecroy LT342), connected to a PC (9). Note that a signal delay generator (10) (Stanford Research Systems, DG535) is used to synchronize the gating of the PMT with the laser pulses and to trigger the DSO with the DF signals.

Fig. 2

Optical design and overall geometry of the spectrofluorometer. (1) dye laser; (2) metal-and-wood experimental enclosure, x = 430 mm, y = 230 mm, height = 250 mm; (3^′) diaphragm; (3) lens, F = 150 mm; (4) parabolic mirror with cylindrical hole; (5) irradiation location; (6) optical fiber; (6^′) thermostable cell holder; (7) gatable photomultiplier; (8) rotatable optical filters holder; (9) digital storage oscilloscope and PC; (10) signal delay generator.

Fig. 3

Thermostatable quartz cell holder with dual, water-heated side-blocks. Cast aluminum, x = 55 mm, y = 32 mm, z = 45 mm.

The present optical design, based on the use of a parabolic mirror, presents some advantages as compared to standard configurations for optical-fiber-based spectrometers. These include: i. the absence of spectral distortions due to the fact that no refraction — but only reflections — take place in this part of the optical setup; ii. no autoluminescence is induced in this optical element which, due to its metallic nature, does not interact with the excitation pulse; iii. the solid angle of acceptance of the parabolic mirror (≈0.6 sr) is larger than the entrance solid angle of the optical fiber (≈0.15 sr), which leads to a high light recovery fraction; iv. the mirror's parabolic shape eliminates the need of an additional lens to focus the luminescence emitted by the fiber tip (or the quartz cuvette) on the detector. This is because the detector's window is conveniently located at the imaging point of this mirror. Note: the setup is designed for a 1:1 magnification.

The fiber-based measurement mode is well suited for in vivo studies, where the light must be delivered to tissue, outside of the setup's light-tight enclosure.

When measuring a sample in the thermostable quartz cell (6^′), placed at the focal point of the laser beam (5), again a sizable fraction of the luminescence from the liquid sample falls on the parabolic mirror. The measurements in that case are thus completed inside the light-tight enclosure (2). This leads to higher sensitivity and less background noise. In addition, it is possible, if desired, to control the temperature, the humidity, and gas composition in the enclosure.

2.2.

Delayed and Prompt Fluorescence Discrimination: Detectors and Electronics

As the whole setup is constructed in a light-tight enclosure, undesired light is eliminated. The time-resolved detection of the weak DF signal nevertheless requires gating to avoid saturation of the detector by the much stronger PF signal. To achieve this, the photomultiplier tube can be electrically switched on, about 3 μs after the laser pulse. During this time, the intensity of any PF decreases to the noise level. Because the intensity of the PF is at least 2 orders of magnitude stronger than that of the DF, and because the photons of the PF hit the inactive photocathode during the ≈3 μs offset-time, fast recovery of the high gain as well as proper correction for PMT artifacts are required. For these reasons, we used an R955 Hamamatsu PMT with a broad wavelength response (185 to 900 nm), low noise, and high gain.

A gated D-type socket assembly (Hamamatsu C1392 series) was used together with this PMT. The PMT is “gated” by holding the photocathode at a potential of about +10 V with respect to the first dynode. The gate function is driven by an external user-triggered TTL pulse. In this way, the PMT is blinded during the laser excitation and prompt fluorescence emission. The signal delay generator (10, Fig. 2) thus turns the PMT off, about 5 μs before triggering the laser pulse, for about 8 μs. The rise time for the gating is typically 400 ns (see Fig. 4).

Fig. 4

Chronogram of the PMT gating and signal acquisition process.

2.3.

Data Acquisition and Analysis

The DF signal is transformed by the PMT into an analog electrical signal which is supplied to the DSO for averaging. The analog signal from about 30 transients is digitized in 24-bits floating point values and transferred to the PC for further analysis. The parameters describing the luminescence decay are obtained by fitting the time-dependent digitized signal f(t) with a sum of exponential functions:

The set of exponential lifetimes (τ_i) characterizes the PS molecule and its physical environment. Each pre-exponential factor A _i is proportional to the intensity of the i'th component to the total luminescence intensity.

The Levenberg–Marquardt (L–M) method is used for non-linear least-squares fit of the measured data.^{45, 46, 47, 48, 49} We wrote and used an Octave [GNU GPL] L–M algorithm implementation. The calculated mathematical fits were graphically checked by plotting the residuals of each fit and the autocorrelation of the residuals,^{50, 51} as shown, e.g., in Fig. 5.

Fig. 5

Analysis of the fitting results: the semi-logarithmic scale used to plot the experimental data and the fitting results help to visualize a bad fit. Random dispersions, around zero, of the residuals and their autocorrelation, confirm the quality of the fit.

2.4.

Setup Calibration and System Sensitivity

We have validated and calibrated the setup through luminescence measurements on well-characterized molecules. The luminescence lifetime of reference samples depends on the fluorochrome's environment, on small changes in the molecular structure, on the presence of quenchers, and on temperature. For these reasons, we have chosen two different crystalline lanthanide(III) complexes whose properties and behaviors were known to be stable in the chosen environments.

The first reference compound was a powdered samarium complex: [Sm^III -(hfa)₃(4-cypNO)]₂ (hfa⁻ = hexafluoroacetylacetonate; 4-cpyNO = 4-cyanopyridine N-oxide).⁵² This sample was used to verify the validity of our measurements close to the temporal resolution limit of our system (≈0.7μs, see Sec. 2.5). When the luminescence of this powder was measured with our setup, we observed a signal with a mono-exponential lifetime of 1.27 ± 0.09 μs. This compares well with the literature value of 1.13 ± 0.05 μs.⁵³ The second reference compound was a 2:3 stoichiometric solution of an europium complex [Eu^III-:H₂L^C2] in water/glycerol (9/1, v/v) (H₂L^C2: homoditopic ligand 6.6^′ — for very long luminescence lifetimes) whose lifetime is expected to be in the milliseconds range. With a fiber-to-quartz-cell distance of 100  μm, we measured a mono exponential decay with a lifetime of 2275 ± 127 μs, as compared to the published value of 2200 ± 100  μs.⁵³ The small differences of <10% resp. <5% observed between our measurements and those made by Chauvin et al are compatible with the stated error limits. They might be explained by small differences in i. the sample excitation wavelengths: 355 nm versus 405 nm, and ii. the actual physical environments of the lanthanide complexes.

The sensitivity of the system was determined by measuring the time-resolved luminescence decay of solutions of PpIX in dimethyl sulfoxide (DMSO) and of Pd-meso-tetra(4-carboxyphenyl)-porphyrin (PdTCPP) in phosphate buffered saline (PBS). PpIX is a frequently used photosensitizer and presents a delayed fluorescence,^{54, 55, 56} while PdTCPP is a commonly used oxygen sensor for in vivo measurements.^{29, 31, 32, 49, 57} Luminescence decays were thus measured at decreasing concentrations ranging between 10⁻⁴ and 10⁻⁷ M, within 1 h of sample preparation. The sensitivity limit was then determined for the lowest photosensitizer concentration, still yielding measurements with a signal-to-noise ratio (SNR) allowing for the retrieval of accurate lifetimes. The latter is defined here as the standard deviation on the measured lifetime being <10%. The SNR was expressed as the ratio of the signal to the rms noise level. Table 1 shows the calculated SNR at the different concentrations. The sensitivity of the system is therefore sufficient to measure the weak delayed fluorescence of PpIX in typical in vivo conditions where its concentration might be on the order of 1 × 10⁻⁵ M.

Table 1

Setup sensitivity assessment with λex = 405 nm, and λem = 645 ± 75 nm FWHM.

After these preliminary validating results, we compared the radiant DF energy of PpIX, which is proportional to the area under the recorded signal — and therefore to the totality of detected DF photons — to its total luminescence. The ratio of these energies yields an estimate of the DF's detectability. The DF signal was recorded and measured as already described. The signal of the total luminescence (proportional to the sum of all molecule's emissive de-excitation) was measured by i. adding neutral density (OD 3.9) filters between the parabolic mirror and the PMT, and ii. recording the luminescence emitted by the PpIX sample from the beginning of the laser excitation pulse. The ratio of the total PpIX's luminescence to its DF was found to be 145:1 for in vitro measurements performed on a DMSO solution at pH = 7.2, and 300:1 for in vivo measurements performed on a chorio-allantoic membrance (CAM) membrane topically treated with 5-aminolevulinic acid (ALA), a precursor of PpIX. The values obtained are in reasonable agreement with the estimates given in the literature. Indeed, Mauzerall and Feitelson determined that this ratio is equal to about 40 for a similar porphyrin measured in an oxygen-free environment.⁵⁸

2.5.

Setup Temporal Resolution

The two devices that limit the temporal resolution of the setup are the PMT and the current-to-voltage converter circuit connected in series between the PMT and the DSO. The loss of temporal resolution due to the PMT can be approximated by the sum of the anode pulse rise time and the electron transit time spread ≈2.2 ns + 1.2 ns, for a total of ≈3.5 ns, which is negligible with respect to the contribution made by the current-to-voltage circuit. The latter is characterized by an RC response time. A load resistance value of R _load = 500 Ω was chosen in order to have a comfortable signal intensity during in vivo measurements. With the help of a very short light pulse (the nearly delta excitation laser pulse: ≈300 to 400 ps) and with two different load resistances, the parasite capacitance of the circuit was measured to be 145 pF. The characteristic temporal response of the setup is therefore set to be R _load × C = 72 ns, which is the setup's temporal resolution. This response time of ≈72 ns is negligible if the measured signals have typical lifetimes larger than 10 μs, as is the case for the delayed-fluorescence lifetime of PpIX.

2.6.

Background Optical Noise

To generate a well-defined maximum optical background noise, we placed a nonfluorescing white reflecting coated plate (SphereOptics, White Reflecting Coating, reflection >98% of incident light) perpendicular to and in contact with the end of the optical fiber, and recorded the signal generated by this diffusely reflected light. The aim was to evaluate if this maximum optical background noise could still be considered as negligible when compared to the weak DF signal expected from an in vivo measurement of the PpIX's DF. The backscattered signal was recorded after a delay of 3 μs from the excitation and averaged over 30 sweeps.

An appreciable optical background signal, probably due to both the parasitic luminescence of the optical fiber itself and the SphereOptics plate, plus the luminescence of the bandpass filter located in front of the PMT, was observed. The intensity of this background noise was in the worst case 1 order of magnitude smaller than the expected experimental signal. The background noise created by the reflection of the laser excitation on the sample can therefore be neglected.

3.1.

Examples of in vitro Lifetime Measurements

The luminescence lifetime of two porphyrins, PdTCPP and PpIX, in aqueous (PBS) and DMSO solutions, respectively, were carefully characterized in vitro to confirm that the sensitivity of the setup would allow to observe luminescence lifetime changes due to variations in O₂ quencher concentration. By bubbling during 30 min different, well-defined, oxygen–nitrogen gas mixtures, the O₂ dissolved in the porphyrin solutions was changed. All measurements were carried out at room temperature (22°) and the solutions were freshly prepared (within 1 h from the measurement) and stored in quartz cuvettes.

As a validation measurement, the lifetimes of a PdTCPP solution in PBS (83 μM, pH = 7.4) were measured, by the analysis of the averaged luminescence decay signal over 30 sweeps at five different pO₂ (See Fig. 6). An excellent linear Stern–Volmer relationship was verified and good agreement was found between our measurement at pO₂ = 0 and the values reported in the literature for oxygen-free PdTCPP solutions. No attempt was made to compare published lifetimes at other O₂ concentrations as the strong dependency of Henry's coefficients on temperature and exact solvent nature cannot be easily corrected for (see Table 2).^{59, 60, 61}

Fig. 6

Verification of the Stern–Volmer equation (τ₀ = 0.358 ms, [TeX:] \documentclass[12pt]{minimal}\begin{document}$k_{q} = 2.04 \, (\rm mm\,Hg \; ms)^{-1}$\end{document} $k_q=2.04{\left(\mathrm{mm}\mathrm{Hg}\mathrm{ms}\right)}^{-1}$ ) for a solution of PdtCPP in PBS (83 μM, pH = 7.4)

Table 2

PdTCPP phosphorescence quenching by oxygen: Lifetime measurements of an 83 μM solution of PdTCCP in PBS at room temperature, pH 7.4, at different pO2 values. Standard atmospheric pressure is 760 mm Hg but in our lab, the standard is 740 mm Hg due to elevation.

There are few in vitro PpIX's delayed fluorescence lifetime measurements in the literature and they are not directly comparable, due to the sensitivity of PpIX's DF lifetime to pH, temperature, concentration of the solution, and quencher concentration.^{54, 62, 63, 64, 65} Therefore, a direct comparative study could not be performed for this molecule. A PpIX solution in DMSO (1.6 μM, pH=7.2) was prepared and stored in a quartz cuvette as described before. The measurements reported in Table 3 allow us to confirm the possibility (in vitro) of measuring changesin PpIX's lifetime ascribable to changes in the O₂ concentration. Our measurement at pO₂ = 0 is compared here with the one published by Mik ⁵⁶ and found to be in reasonable agreement with it.

Table 3

Influence of variations in pO2 on PpIX's DF lifetime: Luminescence lifetime measurements of a 1.6 μM solution of PpIX in DMSO at room temperature, pH = 7.2, at different pO2 values.

The same PpIX solution was measured (after 3 h of pure N₂ bubbling) with a commercial (Perkin Elmer LS 50B) time-resolved spectrofluorometer to compare the spectrum of the measured luminescence with that of the prompt fluorescence. This allowed us to confirm that the luminescence measured was really the delayed fluorescence, as it exhibited the same spectroscopy as that of the prompt fluorescence, but displayed lifetimes characteristic of the triplet state (see Fig. 7).

Fig. 7

PpIX's luminescence spectrum. Measurements performed with a Perkin-Elmer LS 50B spectrofluorometer. (a) prompt fluorescence; (b) delayed fluorescence measured 1.3 ms after the excitation. The gain difference is approximately a factor of 100.

3.2.

Examples of in vivo Measurements

When using PpIX as a PS, the usual approach consists in administering to the patient a precursor of the drug. The most commonly used precursor molecules are ALA or certain of its esterified derivatives.^{66, 67, 68} The biosynthetic pathway involves the combination of two ALA molecules to form a pyrrole unit, and four of these are finally combined to form one PpIX moiety, which itself is a part of haemoglobin.^{69, 70} Because of the enzymatic activity in diseased or cancerous tissues, the production of PpIX is locally and selectively strongly enhanced, as compared to the surrounding normal tissue.^{71, 72, 73}

Our setup must allow, in this case, to measure through the PpIX's delayed fluorescence, the pO₂ at any location where the PS is being formed. We report three examples of in vivo delayed fluorescence measurements, performed on the chick CAM model.

Fifteen fertilized eggs underwent routine handling in order to obtain healthy developing CAMs until day 10.^{74, 75, 76} Four hours before the measurements and after removal of the upper part of the shell, as described by Lange,⁷⁶ the chick embryos where topically administered a droplet of ALA solution (20 μl, 152 mM ALA–H₂O solution, adjusted to pH 6.0, corresponding to 40 mg/kg). As described in Sec. 1, the DF lifetime is related to the concentration of molecular O₂ which acts as a quencher in this case.

It should be noted that the oxygen concentration in the CAM is also influenced by molecular O₂ diffusing from the atmosphere. In order to significantly reduce this O₂ in-flow and make the CAM's tissular [O₂] decrease measurable, just before start of irradiation, we placed an impermeable barrier on the CAM: the region to be measured was covered by a round microscope cover glass ( [TeX:] \documentclass[12pt]{minimal}\begin{document}$\varnothing$\end{document} $⌀$ 20 mm, [TeX:] \documentclass[12pt]{minimal}\begin{document}$0.15\,\,\textrm{\rm to}\,\,0.19\,{\rm mm}$\end{document} $0.15\mathrm{to}0.19\mathrm{mm}$ thick, Schott D263M). The optical fiber probe was made to point toward the middle of the cover glass (≈100 μm from the surface, probed surface diameter [TeX:] \documentclass[12pt]{minimal}\begin{document}$\approx\! \varnothing$\end{document} $\approx ⌀$ 500 μm).

The probing laser (405 nm) excites the PpIX molecules produced during the 4 h of incubation and the resulting luminescence was collected by the fiber tip and analyzed as described earlier. It was demonstrated that the probing laser pulses did not photodynamically or otherwise damage the CAM and did not significantly affect the tissular oxygen concentration if the total number of consecutive pulses was limited to 30. Therefore, for the following in vivo measurements, only 20 shots were averaged.

The first in vivo experiment presented demonstrates the phenomenon of tissular oxygen consumption during laser irradiation, as monitored by measuring the PpIX's DF lifetime. Measurement of oxygen consumption was performed using the following experimental protocol: four series of 400 probing laser shots (at a pulse frequency of 10 Hz) were fired at the CAM sample, with a break of 1 min of darkness between each series. The DF measurements of the first 20 and the last 20 probing shots of each series were averaged, recorded, and compared. Using this experimental protocol, we observed an appreciable shift (toward longer values) in the PS's DF lifetime, between the first and the last group of 20 laser shots of any given series (400 shots in 40 s). The 60 s break between two consecutive series allows for (predominantly vascular) oxygen to rediffuse to the region of the probed spot, thereby restoring — at least partially — the tissular oxygen concentration, as confirmed by a shortening of the observed DF lifetime (see Fig. 8).

Fig. 8

PpIX's delayed fluorescence reciprocal lifetime change, as measured in CAM, showing PDT-induced tissular oxygen consumption and its regeneration during the treatment breaks. Results from 4 sequential series of 400 laser shots each (3.4 mJ/cm² per pulse, spot diameter [TeX:] \documentclass[12pt]{minimal}\begin{document}$\varnothing 500\,{\rm \mu m}$\end{document} $⌀500\mu \mathrm{m}$ ), separated by 1 min breaks. To demonstrate oxygen consumption, the initial and final oxygen levels are deduced from the DF recovered from the first 20 and the last 20 laser shots of each series. Their respective, averaged, reciprocal DF lifetimes are shown. The four different symbols in the graph corresponds to the four consecutive laser shots series.

The second in vivo example illustrates the relation between the PS's DF lifetime and the concentration of oxygen in the atmosphere surrounding the CAM sample: A small gas chamber (see Fig. 9) was built for easily regulating the oxygen content in the atmosphere surrounding the sample. A small hole (A) in the top part of the chamber allowed for positioning the fiber tip at the CAM's surface, while two lateral inlets–outlets (B and C) allowed to flush it continuously with known N₂/O₂ gas mixtures. We used it to measure the DF lifetime of PpIX (endogenously produced by the CAM, following the ALA administration protocol described above) under different O₂ atmosphere conditions. The CAM was introduced in the chamber and left for 3 min, in view of an approximate equilibration of O₂ in the CAM itself, immersed in a homogeneous atmosphere of known, stable composition. The linear relation obtained with 12 different samples (see Fig. 9) satisfies the Stern–Volmer equation [see Eq. 1], confirming and validating these in vivo measurements. These results show noticeable dispersion, increasing with the applied oxygen concentration, leading to the increasing error bars at 5, 10, 15, and 21% O₂ content, linked to intrasample and/or intersample variations, such as the actual vascular and micro-vascular density in the probed area.

Fig. 9

(a) Gas chamber to control the O₂ concentration on the atmosphere surrounding the egg. (b) Relation between PpIX's DF reciprocal lifetime, measured on the CAM's surface, as a function of oxygen concentration in the controlled atmosphere gas chamber.

The third in vivo experiment was designed to evaluate the scale/importance of any intra-sample dispersion, when locally measuring the pO₂. This experiment was conducted in the same way as the previous one, but the oxygen concentration was measured in eight different points of the same egg's CAM [see Fig. 10b]. The results showed sizable differences, increasing with the applied oxygen concentration, to be ascribed directly to local, intrasample, fluctuations (i.e., location-specific attributes which influence the measured pO₂ value, such as the actual vascular density in the probed area), as shown by the large error bars at 15% and 21% O₂ content [Fig. 10b]. As in the previous experiment, this dispersion increases in parallel to that of the pO₂. It may originate in i. the increasing measurement errors when the lifetimes become shorter (at high pO₂) and ii. in the fact that the amplitude of any pO₂ variation, linked to the density of the vasculature, is directly proportional to the pO₂ itself.

Fig. 10

Intra-sample fluctuations, measured by exciting the CAM of the same egg in eight different locations with different environmental oxygen concentration. (a) Measurement spots on the CAM. (b) The DF lifetime in the function of the amount of oxygen in the chamber's atmosphere in eight different locations of the same egg.