Protoporphyrin IX (PpIX) is the most commonly used clinical photosensitizer in the photodynamic therapy (PDT) of cancer and skin diseases, and it is endogenously produced in cells from its precursor $\delta $-aminolevulinic acid (5-ALA) within the mitochondrial membranes [1–4]. PpIX emits red fluorescence light and has been reported to have delayed fluorescence (DF) [5,6]. This DF signal results from spontaneous emission from the first excited singlet state (${{\rm S}_1}$), which was previously repopulated by a reverse intersystem crossing from the excited triplet state, as illustrated in Fig. 1(a). The DF spectrum is very similar or identical to normal prompt fluorescence (PF), but the timescale of the DF emission is the same as that of phosphorescence, taking many microseconds. The intermediate triplet states are strongly quenched by oxygen, which makes the DF lifetime inversely proportional to oxygen partial pressure (${{\rm pO}_2}$) [5]. Cellular ${{\rm pO}_2}$ is a very important parameter in PDT, as well as radiotherapy of cancer, and a tool for monitoring of intracellular oxygen would be useful. In this Letter, the ability to image this DF signal was explored with an intensified time-gated camera and pulsed laser.

 

Fig. 1. (a) Jablonski diagram showing the origin of delayed fluorescence of PpIX from reverse intersystem crossing (RISC). (b) Schematic representation of PF vs. DF kinetics. (c) Experimental setup.

Download Full Size | PDF

Since PpIX is naturally produced within mitochondrial membranes, the µs-time-resolved DF enables the ${{\rm pO}_2}$ measurement directly at the location where the oxygen metabolism takes place, as shown by Mik et al. [5,7–11]. This earlier work developed a clinical device to measure ${{\rm pO}_2}$ in human tissues after topical application of 5-ALA, and it has been proposed as a tool for monitoring patients during anesthesia [11]. This device uses the point measurement of the DF decay kinetics with a photomultiplier and pulsed 515 nm laser excitation, and the DF lifetimes report on the effective ${{\rm pO}_2}$ [7,8]. The DF kinetics of PpIX were also measured in the chicken egg chorioallantoic membrane [12]. DF of different sensitizers was imaged microscopically in polymeric nanofibers [13] and cell culture in vitro [14,15], but to date, to our knowledge, no macroscopic imaging has been done in mammals or humans.

Here we report on the first macroscopic imaging of PpIX DF in vivo. Given the importance and wide-spread clinical use of PpIX in fluorescence-guided surgery and PDT of cancer and a variety of skin diseases, the technique shown here could be a unique diagnostic tool for oxygen monitoring in tumors and skin lesions, because the outcome of PDT and radiotherapy is tightly linked to oxygenation of the treated tissue.

DF is inherently several orders of magnitude weaker than the PF. Further, the lifetime of PpIX PF is 5–16 ns, depending on the micro environment, [16], whereas the DF lifetime is in the µs–ms range [5,6]. A pulsed excitation source, with no after-pulses or tail, and fast time-gated detection are needed for macroscopic imaging of DF. For excitation, a compact laser was used (World Star Tech, Compact multiwave laser) modulated by a pulse generator (Agilent, 33250A) providing 635 nm pulses with the 20 µs temporal width at 2 kHz (4% duty cycle) and 1 mW average power. The laser was brought to the sample by a collimator-terminated optical fiber and filtered by a 650 nm short-pass to remove any potential laser emission interfering with the detection spectral window. To measure DF, a circular illuminated spot with a 1 or 2 cm diameter was used for irradiance of 1.2 or ${0.3}\;{{\rm mW/cm}^2}$ (average power), respectively. The luminescence was collected by an objective lens (Nikon #f 1/1.2, $\phi{49}\;{\rm mm}$), equipped with a spacer ring to shorten the working distance to ${\sim}{15}\;{\rm cm}$, through a 695/50 bandpass filter.

An intensified time-gated camera PI-MAX4 (Princeton Instruments) was synchronized with the laser pulses and allowed setting an arbitrary gate-delay with ns precision and gate-widths spanning from 3 ns to seconds. Therefore, it is suitable both for the DF and PF detection. To obtain a DF lifetime image, a series of time-gated frames with an increasing delay was captured [Fig. 1(b)]. One frame typically consisted of 2000 on-sensor accumulations.

${{\rm pO}_2}$ is related to the DF lifetime $\tau $ in the following way, predicted by the Stern–Volmer equation [5]:

The ability of the system to acquire DF and PF lifetime images and visualize ${{\rm pO}_2}$ was tested on the PpIX tissue phantoms consisting of 50 µM PpIX (Frontier Scientific), 1% intralipid (Fresenius Kabi), and 1% bovine whole blood (Lampire) in phosphate buffer saline (PBS) in tubes. All in vivo procedures in this work followed the protocol approved by the Institutional Animal Care and Use Committee. Nude mice (Jackson Labs, Maine) were used and placed on low chlorophyll diet to reduce autofluorescence. The mice skin had the Ameluz cream containing 10% 5-ALA (Biofrontera AG, Germany) applied on several spots along the spine after wiping with ethanol. While the mice were still under isoflurane anesthesia, the cream was left to dry for 10 min, and then a bandage was put on to prevent the mice from removing the cream. The mice were kept in dark for 3 h and then imaged under isoflurane anesthesia. Before imaging, any leftover Ameluz was washed off. All procedures for human skin testing followed an approved protocol by the Institutional Review Board. The human index finger was washed, and subtle abrasion of stratum corneum [18] and alcohol wash were applied, prior to Ameluz application and occlusion. After 5 h, the finger was washed, and fluorescence imaging was performed.

Figure 2 shows the time-resolved DF images from PpIX tissue phantoms (1% Intralipid and 1% blood in phosphate buffered saline) at different oxygen concentrations. A sequence of six time frames with 5 µs gate-width and increasing delay were acquired (1st and 5th frames are shown). The first row is the air-saturated sample, and the second row is the sample after oxygen scavenging by glucose-oxidase/catalase system [18]. (${{\rm O}_2}$ removal by nitrogen bubbling provides similar results; data not shown). The samples with reduced oxygen have a much longer DF lifetime than the air-saturated one, which shows the ${{\rm pO}_2}$ sensing ability of PpIX DF.

 

Fig. 2. DF images of PpIX tissue phantoms. First row, air saturated; Second row, after oxygen scavenging by glucose-oxidase/catalase.

Download Full Size | PDF

 

Fig. 3. Mouse and human skin treated with the ALA-cream Ameluz. (a)–(d) Effect of the mouse skin pinching on tissue oxygenation and PpIX delayed fluorescence lifetime. The DF lifetime is prolonged, and oxygen depleted right after the pinch, but later the oxygen recovers and the DF lifetime shortens. (e) and (f) Effect of human finger strangling with a rubber band provides similar results.

Download Full Size | PDF

Figure 3(a) presents the white-light image of a mouse, the PF image, and their combination. Ameluz was applied along the spine. The background fluorescence of the skin untreated with Ameluz was very low. The oxygen concentration in the mouse skin can be modulated by local pressure on the skin, which leads to the closure of capillaries, oxygen supply shut-down, and hence a decrease in ${{\rm pO}_2}$ and increase in the PpIX DF lifetime [9]. Local pressure was applied by pinching the mouse skin in between fingers for 1 minute at the bottom part of the investigated spot. The pinch was then released, and a series of DF time-resolved sequences was collected. One sequence consisted of five 50-µs-wide gate-windows in the time interval 10–260 µs. Two of these gate windows are shown in the first two columns of Fig. 3(b). The third and fourth column display the resulting DF lifetime and ${{\rm pO}_2}^*$ maps. Images are shown before the pinch, right after release and 3 min later when ${{\rm pO}_2}$ is recovered again. The temporary local depletion of oxygen is clearly visualized by the prolonged DF lifetime and enhanced intensity. Figure 3(c) shows the DF kinetics from the selected numbered areas, and Fig. 3(d) displays the recovery of oxygen during the capillary refill in the spot number 2. The total acquisition time per one time-gate frame was typically 1 s, corresponding to 2000 on-sensor accumulations with the 2 kHz laser repetition. The acquisition time per a five-frame sequence was then about 5 s (with ${6}\;{{\rm mJ/cm}^2}$ radiant exposure).

A simple study comparing the PpIX production and DF in the human skin relative to that of the mouse skin was completed. Figures 3(e) and 3(f) show the PF and DF signal from a human finger, and the effect of temporary strangling of the finger with a rubber band. A reversible lengthening of the DF lifetime and depletion of ${{\rm pO}_2}$ is illustrated, similarly to the mouse skin in Fig. 3(b). Three sequences of four frames were summed together to obtain each ${\rm lifetime}/{{\rm pO}_2}$ image, with a total capture time of 12 s and ${14}\;{{\rm mJ/cm}^2}$ radiant exposure.

In Fig. 4, the 5-ALA-treated skin region is shown with the mouse alive, and then again at 5 min after sacrifice. The amplitude of the DF increased several-fold, and the DF lifetime got much longer, as expected, because oxygen was quickly metabolized and not resupplied in the dead mouse, leading to oxygen depletion and a prolonged DF lifetime. This again illustrates that DF is capable of sensing cellular ${{\rm pO}_2}$. Four sequences of nine frames (total radiant exposure ${10}\;{{\rm mJ/cm}^2}$) were summed, with delay increasing from 10 to 410 µs and 50 µs gate.

 

Fig. 4. Time-resolved DF of a live mouse (second row) and the same spot 5 min after sacrifice (third row). The signal amplitude and lifetime clearly increased after death.

Download Full Size | PDF

It is well documented that the PpIX DF lifetime is sensitive to the local oxygenation of the tissue [5,7–9,11], and here the ability to spatially image this lifetime has been explored in the skin with a pulsed laser and a time-gated intensified camera. The combination of a fast-pulsed laser, ns-gated intensified camera, and production of PpIX DF in the skin has never been demonstrated before, to the best of our knowledge, but this is in fact a very efficient way to image tissue ${{\rm O}_2}$, where the signal originates from the cellular content in which PpIX is generated.

One of the major challenges in this technique is that PpIX is a therapeutic photosensitizer, as well as a diagnostic fluorescence probe, and so the process of illumination and imaging can itself alter the ${{\rm pO}_2}$ and microenvironment. Thus, the required intensity and imaging time were carefully analyzed here to allow diagnostic use without appreciable PDT effect. Based on the measurements, time-resolved macroscopic imaging of 5-ALA-induced PpIX delayed fluorescence was feasible in vivo with low excitation power (${0.3}\;{{\rm mW/cm}^2}$) at tissue-penetrating red wavelengths of 635 nm. The typical acquisition time for a time-resolved sequence was ${\sim}{5 - 20}\;{\rm s}$ with radiant exposure of only ${\sim}{10}\;{{\rm mJ/cm}^2}$, which is four orders of magnitude below therapeutic doses used in the PDT treatment (${\sim}{100}\;{{\rm J/cm}^2}$). No appreciable oxygen depletion due to irradiation was observed during the repeated ${{\rm pO}_2}$-image acquisitions for radiant exposures $ {\lt} {100}\;{{\rm mJ/cm}^2}$. The irradiation-induced changes of ${{\rm pO}_2}$ during PDT are currently under investigation.

To shorten the acquisition time, it is possible to decrease the number of the sampled time-windows. It is quite common to derive lifetime images based on the ratio of just a few time-windows [19] with wider windows to provide more signal. Further, stronger excitation sources are available, thus enabling a larger illuminated area, decreased acquisition time, and improved signal-to-noise ratio, although there is a tradeoff with affecting the tissue as previously mentioned. Shorter laser pulses would allow for better time resolution.

One subtlety of interpreting these data is that several different mechanisms of the DF formation have been reported for PpIX, some of which are second order in the excitation intensity [6]. Here it was found that a two-fold decrease in the excitation intensity led to a two-fold decrease in the DF intensity in mouse skin samples (data not shown). This confirms that due to low pulse energies, DF was produced mainly by thermally activated reverse intersystem crossing, which is first order in excitation intensity. This observation significantly simplifies the interpretation of the data.

Another important caveat is that the DF kinetics are not mono-exponential [7]. The detection of components with short lifetimes is limited by the width of the gate and excitation pulse. The method is unable to capture components with a lifetime less than approximately a half of the gate-width. As discussed above, the single exponential fitting leads to the underestimation of ${{\rm pO}_2}$ by a factor of 1.5–2. However, relative changes in the DF lifetime can be conveniently imaged this way.

The ${{\rm pO}_2}$ values previously measured by the DF lifetime technique were ${\sim}{50}\;{\rm mmHg}$ in the rat skin and ${\sim}{50}$ or 22 mmHg in the human skin [8,11]. The values of ${{\rm pO}_2}^*$ reported here were lower, probably due to a different analysis method. ${{\rm pO}_2}^*$ values measured in different mice were quiet variable and strongly influenced by respiration rate. The ${{\rm pO}_2}^*$ values recorded in the live mouse in Fig. 4 are relatively small, which could be explained by a very deep anesthesia and oxygen depletion due to scattered light from prolonged irradiation of an adjacent spot.

The dynamic range of ${{\rm pO}_2}^*$ measurements with the 50 µs gate, 20 µs laser pulse, and 2 kHz repetition was estimated to be 0.5–40 mmHg, but it can be adjusted by changing those parameters. The standard deviation of repeated ${{\rm pO}_2}^*$ readings in a single pixel was 2.9 mmHg, with the ${0.1}\;{{\rm mm}^2}$ pixel area, ${6}\;{{\rm mJ/cm}^2}$ radiant exposure (5 s capture time), and 16 mmHg average ${{\rm pO}_2}^*$. This gives an estimate of the resolution of ${{\rm pO}_2}^*$ measurements.

In general, the DF emission is much weaker than the PF emission. Here, the DF lifetimes are considerably long, which enhances the overall quantum yield of DF. It was found that the overall DF emission from the mouse skin in the time interval 20–470 µs was ${1000 - 1500\times}$ weaker than the PF emission (measured in the interval 0–48 ns after the excitation pulse). These measurements were performed with a sub-nanosecond pulsed laser, which allows arbitrary pulse sequencing (PicoQuant, Sepia II with LDH-D-C-635M). The interval 20–470 µs did not cover the whole DF emission time, so the overall DF emission is stronger than this. Given that the triplet quantum yield of PpIX is ${\sim}{0.75}$ [20], we can conclude that within the investigated time frame, about one in thousand triplets undergo RISC from ${{\rm T}_1}$ to ${{\rm S}_1}$ that can be followed by photon emission. Assuming that the quantum yield of PpIX PF is ${\sim}{8}\% $ [20], the overall quantum yield of PpIX DF is in the order of ${{10}^{ - 4}}$ in our skin studies. Given the emission lifetimes and intensities, the DF emission rate in the laser pulsed experiment appears about seven orders of magnitude smaller than the PF emission rate. Despite the low DF intensity, the DF lifetime images can be obtained with low excitation intensity within seconds.

The imaging method developed here is especially relevant to monitoring and dosimetry during the PDT of cancer, where ${{\rm pO}_2}$ could be mapped out across the tumor. DF imaging combined with the PF imaging provides a complete direct dosimetry dataset to map out skin lesions intended for PDT. We also verified that DF can be imaged after intraperitoneal injection of 5-ALA, which significantly broadens the scope of applicability of this method.