2.1.

Murine Models of Human Cancer

This study was approved by the University of Minnesota Institutional Animal Care and Use Committee (IACUC). To generate high-grade PNSTs, we generated genetically engineered Dhh-$Cre;Pte{n}^{Fl/Fl}$; Cnp-EGFR mice^41,42 through a focused breeding scheme. This model is faithful to the fundamental features of human PNSTs⁴¹ and results in development of plexiform neurofibroma-like lesions and high-grade PNSTs with 100% penetrance by $\sim 4$ weeks of age. To generate models of pancreatic adenocarcinomas, primary carcinoma cells were grafted into nude (Nu/Nu) mice. Primary carcinoma cells were established from autochthonous pancreatic tumors from the $Kra{s}^{LSL\text{-}G12D/+}$; $Tp{53}^{LSL\text{-}R172H/+}$; Pdx1-Cre (KPC) genetically engineered murine model, which is a faithful model for human pancreas cancer^20,43 and harbors the relevant oncogene and tumor suppressor gene mutations associated with human pancreatic cancer, as we have previously described.⁴⁴ Primary carcinoma-associated fibroblasts (CAFs) were isolated by differential trypsinization and magnetic bead depletion of carcinoma cells and immune cell populations. Cells were grown in high glucose Dulbecco-modified eagle’s essential medium (Corning) supplemented with 10% FBS and confirmed to be free of Mycoplasma contamination. To generate tumors, 100,000 carcinoma cells and 100,000 CAFs were mixed into a $2\text{-}\mathrm{mg}/\mathrm{mL}$ collagen matrix that was injected subcutaneously in the mouse flank. Tumors developed within 10 days and were harvested 28 days after engraftment. For each model, doxorubicin ($20\mathrm{mg}/\mathrm{kg}$; Sigma) was administered via intraperitoneal (i.p.) injection for PNST models or intravenous (i.v.) injection for pancreatic cancer models 15 min prior to euthanasia. While under sedation with isoflurane, formalin was injected into the left ventricle 5 min prior to euthanasia. Liver, intestine, and tumor tissues where then harvested and submerged in formalin for at least 24 h before being processed to generate FFPE histology sections. For doxorubicin imaging, slides were deparaffinized and mounted with Prolong Gold (Life Technologies).

2.2.

Multiphoton Microscopy and Multiphoton Fluorescence Lifetime Imaging

Doxorubicin solutions or FFPE sections from $n=5$ $Dhh\text{-}Cre;Pte{n}^{Fl/Fl}$; Cnp-EGFR mice or $n=3$ grafted pancreatic tumor models were imaged on a custom-built multiphoton laser scanning microscope (Prairie Technologies/Bruker), where multiphoton imaging relies on a nonlinear dependence on the excitation power and is inherently optically sectioning. A Mai Tai HP Ti:Sapphire laser (Spectra-Physics) that produces 100-fs pulse widths at an 80-MHz repetition rate to simultaneously generate multiphoton excitation and second harmonic generation (SHG) (Fig. 1) was used. This system was controlled using PrairieView software. Here samples were excited at wavelengths ranging from 730 to 920 nm, and the light focused on the sample using a universal M plan fluorite infinity corrected near-infrared $40\times$ objective ($\mathrm{NA}=0.8$). Emission was detected in the blue, green, red, and far-red channels (though bandpass filters at either 440/20, 460/50, 525/50, 525/70, 595/50, 605/70, or 690/50) with either multialkali (Hamamatsu) or GaAsP (Hamamatsu model 7422PA-40) PMTs. Mulitphoton FLIM was performed on this system using the GaAsP PMT connected to a time-correlated single photon counting (TCSPC) system (Becker & Hickl). The instrument response function (IRF) was measured using urea crystal to generate the second harmonic signal. Each image of $256\times 256$ or $512\times 512$ was acquired using an integration time of 60 to 180 s, with stable photon count rates during image acquisition confirming that photobleaching did not occur. The photon counts from each pixel are used to determine fluorescence lifetime after application of a deconvolution algorithm of the data and a known or estimated instrument response function.^37,45 Analysis was performed with SPCImage (Becker & Hickl) with fluorescence decay modeled as a single- or multiexponential by

Fig. 1

Multiphoton FLIM setup. Excitation light from the Ti:Sa laser (illustrated in red) is polarized and shaped before entering the galvanometric scanners. Emission signal (illustrated in orange) is captured in the epiconfiguration and sent to a detector array containing multialkali (Hamamatsu) or GaAsP (Hamamatsu model 7422PA-40) PMTs through bandpass filters that include 440/20, 460/50, 525/50, 525/70, 595/50, 605/70, and 690/50 filters. Mulitphoton FLIM was performed on this system using the GaAsP PMT connected to an FLIM system employing TCSPC (Becker & Hickl).

2.3.

Statistical Analysis

Two group data were analyzed with unpaired $t$-tests. For multigroup comparisons, one-way analysis of variance with a post hoc Tukey–Kramer test was used. For plots of mean lifetime data, boxes are the 25th to 75th percentiles and bars represent the maximum and minimum values.

3.1.

Two-Photon Excitation of Doxorubicin

Doxorubicin is commonly employed to combat many cancers and is frequently utilized to evaluate chemotherapy distribution in tumors and normal tissue due to its fluorescent properties. The one-photon absorption and emission spectra of doxorubicin has been characterized with peak excitation in the $\sim 480$ to 500 nm range and a broad emission in the range of $\sim 520$ to 720 nm (${\lambda }_{\mathrm{em}}$ peak in the 580 to 600 nm range).^39,47,48 The behavior of doxorubicin in response to multiphoton excitation is less well described. Examining multiphoton excitation across distinct wavelengths ranging from 730 to 920 nm (Fig. 1), we observed that at equal input power the highest emission intensities result following excitation at 760 to 800 nm [peak ${\lambda }_{\mathrm{ex}}=780\mathrm{nm}$; Fig. 2(a)]. As expected, emission was detected in the red and far-red channels though bandpass filters at 595/50, 605/70, and 690/50, and the two-photon nature of doxorubicin excitation was confirmed by examining the nonlinear dependence of fluorescent intensity scaling with the square of the excitation power [Fig. 2(b)]. Thus, doxorubicin has an effective two-photon absorption cross section and, as expected, emits fluorescence into the red/far-red spectrum following two-photon excitation.

Fig. 2

(a) Emission intensity from doxorubicin excited at wavelengths ranging from 730 to 920 nm. (b) Emission of doxorubicin following excitation at 800 nm showing the nonlinear dependence of emission on input power that is characteristic of two-photon excitation. Right inset: Log–Log plot of the two-photon-induced fluorescent intensity versus excitation power ($R^2=0.99$).

3.2.

Fluorescence Lifetime Behavior of Doxorubicin

The performance of our FLIM setup was initially determined by measuring the IRF and subsequently the fluorescence lifetime of known standards. The IRF was first measured using urea crystals that exhibit strong SHG after interaction with 800-nm pulsed light. SHG is a nonlinear wave mixing process that can occur when light interacts with a material possessing second-order susceptibility. The result is a second-order polarization to generate light at double the excitation frequency (i.e., emission at half of the excitation wavelength) that, in contrast to fluorescent signals, is instantaneous and thus has a zero lifetime. Utilizing the urea crystal-generated IRF, we were able to accurately measure the lifetime YG beads, a well-established standard, to be $2.3\pm 0.04\mathrm{ns}$ ($\text{mean}\pm \mathrm{s.d.}$; $n=10$), consistent with other reports.^49–52 Evaluation of the fluorescence lifetime of doxorubicin in PBS solution following multiphoton excitation at 800 nm revealed a lifetime of $1.01\pm 0.008\mathrm{ns}$ ($\text{mean}\pm \mathrm{s.d.}$; $n=7$), which is consistent with previous reports using linear excitation modalities,^9,11,38,47 and was well described with a single exponential model. Thus, multiphoton FLIM provides the capability to robustly generate fluorescence following two-photon excitation and accurately measure its fluorescence lifetime.

3.3.

Doxorubicin Intensity in Archival Nerve Sheath Tumor Sections

PNSTs are poor-prognosis soft tissue sarcomas that can arise spontaneously or develop in patients with mutations in the NF1 gene that causes neurofibromatosis type 1.^53,54 Although they are notoriously difficult to treat with both chemical and/or radiation therapies, these tumors are frequently treated with doxorubicin.^1,2,54 To generate PNSTs, we employed a genetically engineered murine model of human cancer. PNSTs were generated in $Dhh\text{-}Cre;Pte{n}^{Fl/Fl}$; Cnp-EGFR mice that develop plexiform neurofibroma lesions and high-grade peripheral nerve tumors with 100% penetrance by $\sim 4$ weeks of age that show strong similarity to their human counterparts.⁴¹ As such this is an ideal model for studying therapy dynamics. Mice were administered doxorubicin by i.p. injection 15 min prior to euthanasia. Under anesthesia, mice were rapidly perfused with formalin via intracardiac (i.c.) injection. Tissues were harvested and further fixed in formalin for at least 24 h and then processed to generate FFPE sections.

Two-photon excitation of doxorubicin in FFPE samples at $\lambda =800\mathrm{nm}$ revealed robust doxorubicin fluorescence following fixation and processing. Consistent with analysis of free doxorubicin in solution, maximum emission intensities result following two-photon excitation at 760 to 800 nm with emission detected in the red and far-red channels though bandpass filters at 595/50, 605/70, and 690/50. In normal FFPE tissues, the doxorubicin signal is strongest from the nucleus, consistent with the intercalation/binding of doxorubicin to DNA,^6–8,38 with additional signal from the cytoplasm and the extracellular space that vastly exceeds background fluorescence levels observed in regions of poor doxorubicin perfusion or from untreated mice (Fig. 3). Likewise, in PNSTs, examination of the outer boundaries of tumor that are well perfused with doxorubicin revealed that doxorubicin fluorescence is commonly strongest from the nuclei with additional signal from the cytoplasm and the extracellular space [Figs. 3(b), 4(a), and 4(b)]. However, we note that the inner regions of the PNSTs are either not perfused or not well perfused with doxorubicin, consistent with our understanding of poor molecular transport phenomena in solid tumors.^20,23,25,27 Furthermore, most well perfused regions reflect a spectrum of doxorubicin compartmentalization behavior, including areas in which fluorescent signal from the nuclei is stronger, similar to, and less than that in the cytoplasm [Figs. 3(b1)–3(b2) and 4(c)]. Yet, because the antitumor mechanisms of action of doxorubicin have been described as occurring in both the nucleus and the cytoplasm^55–58 and extracellular or cytoplasmic doxorubicin may enter the nucleus at later time points, measurement of the total fluorescence intensity in a region is very meaningful, particularly when also considering analysis of molecular transport into a tumor region.^20,23 However, if detecting doxorubicin in a particular cell compartment or extracellular region is needed, dissecting out the signal localization from fluorescence intensity images can at times be very challenging for instances where the signal intensity is not strongly localized to a distinct cell or extracellular compartment.

Fig. 3

Multiphoton excitation of doxorubicin at ${\lambda }_{\mathrm{ex}}=800\mathrm{nm}$ in FFPE sections of (a) liver and (b) PNST from $Dhh\text{-}Cre;Pte{n}^{Fl/Fl}$; Cnp-EGFR mice systemically administered $20\text{-}\mathrm{mg}/\mathrm{kg}$ doxorubicin 15 min prior to euthanasia. Boxed regions in 1 and 2 in (b) are magnified in panels b1 and b2, respectively. Arrows in b2 highlight examples of cells with greater doxorubicin signal in the cytoplasm than the nucleus. (c) Quantification of average fluorescence at ${\lambda }_{\mathrm{ex}}=800\mathrm{nm}$ in sections without and with doxorubicin (background and dox, respectively), showing robust doxorubicin fluorescent above background levels (data are $\text{mean}\pm \mathrm{s.d.}$). $\text{Scale bars}=25\mu \mathrm{m}$.

Fig. 4

Multiphoton FLIM imaging of doxorubicin at ${\lambda }_{\mathrm{ex}}=800\mathrm{nm}$ in FFPE sections of (a, d, g) intestine, (b, e, h) liver, and (c, f, i) PNST from $Dhh\text{-}Cre;Pte{n}^{Fl/Fl}$; Cnp-EGFR mice administered $20\text{-}\mathrm{mg}/\mathrm{kg}$ doxorubicin 15 min prior to euthanasia. Using multiphoton FLIM with TCSPC, we simultaneously collected (a–c) intensity and (d–f) fluorescence lifetime data. In all cases, (g, h, i) the lifetime of doxorubicin in nuclei is significantly longer than lifetimes observed in the cytoplasm. For the box and whiskers plots, the box extends from the 25th to 75th percentiles, while the error bars span the min and max values. *$p<0.05$; $\text{scale bars}=25\mu \mathrm{m}$.

3.4.

Doxorubicin Fluorescence Lifetime Imaging in Solid Tumors

To test the capability of multiphoton FLIM to capture doxorubicin intensity and lifetime and quantitatively evaluate drug localization, we imaged FFPE archival samples from two distinct tumor models (autochthonous PNSTs and grafted pancreatic adenocarcinomas). Similar to PNST sarcomas, carcinomas of the pancreas have a poor prognosis and respond poorly to chemotherapy.^19,20 To generate pancreatic tumors, we grafted purified primary carcinoma cells derived from KPC mice into host mice. After 28 days, all tumors reached predefined critical dimensions and doxorubicin was administered systemically 15 min prior to euthanasia, similar to the endpoint protocol used in Dhh-$Cre;Pte{n}^{Fl/Fl}$; Cnp-EGFR mice. Tissues were again fixed in formalin and processed to generate FFPE sections.

Using multiphoton FLIM with TCSPC, we simultaneously collected both fluorescence intensity and fluorescence lifetime (Figs. 4 and 5). Consistent with standard two-photon excitation imaging of doxorubicin intensity in FFPE tissues from young mice harboring PNSTs (Fig. 3), FLIM data from intestine and liver (tissues that are commonly perfused and impacted by chemotherapy and that are frequently employed to confirm chemotherapy delivery in experimental models), and tumor (PNST) sections imaged at 800 nm demonstrate that doxorubicin fluorescence is commonly strongest from the nuclei, with additional signal from the cytoplasm and the extracellular space [Figs. 4(a)–4(c)]. Further, heterogeneity in doxorubicin fluorescence within the tumor section remains readily apparent with some cells possessing strong signal from the nuclei while others display more robust intensity from the cytoplasm [Fig. 4(c)]. Extending intensity imaging to the detection of fluorescence lifetimes using multiphoton FLIM reveals strong separation of doxorubicin fluorescence lifetimes in different regions in normal tissue and tumor sections [Figs. 4(d)–4(f)].

Fig. 5

Quantifying the number of photons for a particular range of lifetimes. Shown here is an example of the photon counts for each fluorescence lifetime in a PNST tumor that is determined from the frequency of a particular lifetime multiplied by the corresponding number of photons in each pixel. Gray boxes represent the region between the 25th and 75th percentiles for the cytoplasmic and nuclei compartments.

The lifetimes for doxorubicin were well described by a single exponential fit, similar to free doxorubicin in solution, but are longer than the lifetime in solution ($\sim 1.2$ to 1.5 ns versus $\sim 1\mathrm{ns}$). In particular, a significant separation of doxorubicin lifetimes is observed for doxorubicin in the nucleus versus the cytoplasm in all cases and the distribution of lifetimes in each compartment is quite homogeneous relative to intensity data [consistent with the concentration-independent nature of lifetime imaging; Figs. 4(g)–4(i)]. Interestingly, these characteristics allow for quantification of the relative amount of doxorubicin in each compartment by isolating the number of photons within the range of lifetimes for each compartment (Fig. 5), similar to information obtained by isolating the areas as regions of interest but in a much more efficient way on a larger scale. Furthermore, multicomparison statistical analysis also reveals a significant difference between doxorubicin fluorescence lifetimes in the nuclei of cancer cells versus the nuclei in normal tissues ($p<0.05$ for both tumor versus intestine and tumor versus liver), while lifetimes in the nuclei of the normal tissues are not significantly different ($p\ge 0.05$), raising interesting questions about a potential diagnostic role for doxorubicin lifetime imaging to distinguish normal versus transformed cells.

Multiphoton FLIM of FFPE sections from grafted pancreatic ductal adenocarcinomas in adult mice revealed similar behavior to normal tissue and PNSTs in young mice (Figs. 4 and 6). Examination of fluorescence lifetimes from FFPE sections derived from adult intestine, liver, and tumor imaged at 800 nm shows strong doxorubicin fluorescence in nuclei, with additional signal from the cytoplasm and the extracellular space, with greater heterogeneity in the localization of fluorescence intensity again found in the tumors [Figs. 6(a)–6(c)]. Indeed heterogeneity in doxorubicin fluorescence within the tumor section remains readily apparent, with some cells possessing strong signal from the nuclei while others display more uniform fluorescence or more robust intensity from the cytoplasm [Fig. 6(c)]. Analysis of fluorescence lifetime with multiphoton FLIM reveals strong separation of doxorubicin fluorescence lifetimes in different regions in both normal tissue sections and tumor sections, particularly between nuclei and the cell cytoplasm [Figs. 6(d)–6(i)]. Again, multicomparison statistical analysis reveals a significant difference between doxorubicin fluorescence lifetimes in the nuclei of cancer cells versus the nuclei in normal tissues ($p<0.05$ for both tumor versus intestine and tumor versus liver while liver and intestine are not significantly different, $p\ge 0.05$), further supporting the conclusion that there is a distinct behavior for doxorubicin in the nuclei of cancerous cells versus their nontransformed normal counterparts.

Fig. 6

Multiphoton FLIM imaging of doxorubicin at ${\lambda }_{\mathrm{ex}}=800\mathrm{nm}$ in FFPE sections of (a, d, g) intestine, (b, e, h) liver, and (c, f, i) grafted pancreatic ductal adenocarcinoma from mice administered $20\text{-}\mathrm{mg}/\mathrm{kg}$ doxorubicin 15 min prior to euthanasia. Using multiphoton FLIM with TCSPC, we simultaneously collected (a–c) intensity and (d–f) fluorescence lifetime data. In all cases, (g, h, i) the lifetime of doxorubicin in nuclei is significantly longer than lifetimes observed in the cytoplasm. For the box and whiskers plots, the box extends from the 25th to 75th percentiles, while the error bars span the min and max values. *$p<0.05$; $\text{scale bars}=25\mu \mathrm{m}$.

The observations here that fluorescence lifetimes in FFPE sections are consistently significantly longer in the nuclei regions than for doxorubicin in the cytoplasm are distinct from findings in some studies using live cancerous cells,^9,38,39 but they are consistent with other findings in live cancerous cells.⁴⁰ Interestingly, in the cases showing an opposite trend,^9,38,39 the lifetimes in nuclei are not substantially different between our data in fixed tissues and the published live cells data ($\sim 1.3$ to 1.7 ns in FFPE sections versus $\sim 1.5$ to 1.8 ns in live cells). The difference in lifetime in the cytoplasm however is quite different ($\sim 1.0$ to 1.4 ns in FFPE sections versus $\sim 1.8$ to 3 ns in live cells). This may result from a number of factors that include the time that the cells are exposed to doxorubicin, which has been shown to influence lifetimes,^38,39 and the environmental conditions, such as changes associated with the in vivo versus in vitro condition and fixing and processing the tissues, which are known to impact fluorescence lifetime behavior.³⁷ However, here we set out to specifically evaluate doxorubicin in archival sample sections and make comparisons across formalin-fixed archival samples. Notably, the lifetime values here for doxorubicin in the cytoplasm and nuclei are reasonably consistent across each normal tissue from both young and mature mice. The same is true for the values and shifts in distinct tumor types. As such, it is clear that multiphoton FLIM can be effectively implemented on FFPE sections to obtain reliable fluorescence lifetime data for doxorubicin and quantify DNA-bound doxorubicin localized in the cell nucleus.