2.1.

Imaging Instrumentation

Two separate-custom designed multiphoton time-domain lifetime systems were employed at the University of Wisconsin for the cell culture and in vivo experiments. The system used for cell culture experiments was constructed¹⁹ around a Nikon Eclipse TE300. A titanium-sapphire laser (Spectra-Physics-Millennium/Tsunami, $82\mathrm{MHz}$ , $100\mathrm{fs}$ ) pumped by a 5-W Millenia was used as an excitation source at 780-nm excitation for FLIM, and at both 780- and 890-nm excitation for intensity redox imaging. Other components include a Nikon $40\times$ Plan Apo oil-immersion lens [numerical aperture $\left(\mathrm{NA}\right)=1.3$ ], and a dichromatic mirror (BG39, Schott, Elmsford, New York) that enabled the 780-nm excitation light to reflect onto the sample and wavelengths between 400 and $600\mathrm{nm}$ to transmit to the detector. Intensity and FLIM data were collected pixel by pixel by a laser scanning unit (based on a BioRad, MRC-600) and a $\mathrm{GaAsP}$ photon-counting photomultiplier tube (PMT) (H7422, Hamamatsu) connected to a time-correlated single-photon counting (TCSPC) system (SPC-730, Becker & Hickl).

All in vivo multiphoton FLIM images were collected with a second system constructed around a Nikon Diaphot 200 microscope.²⁰ The excitation source is a titanium-sapphire laser (Coherent, Mira; $76\mathrm{MHz}$ , $120\mathrm{fs}$ ) pumped by an 8-W solid state laser (Coherent, Verdi). Other components include a $60\times$ oil-immersion objective (Nikon, PlanApo $\mathrm{NA}=1.4$ ), and a dichromatic mirror (650DCSP, Chroma Technology, Rockingham, Vermont) to reflect the 780-nm excitation light onto the sample and transmit wavelengths between 400 and $650\mathrm{nm}$ to the detector. FLIM data were collected pixel by pixel with a laser scanning unit (BioRad, MRC-600), a fast photon-counting PMT (Becker & Hickl, PMH-100), and TCSPC electronics (Becker & Hickl, SPC-830). To control the power coupled to the system, a variable neutral density filter wheel was positioned before the entrance to the scanning unit. Acquisition for both systems was done with WiscScan,²¹ a lab-developed software package.

2.2.

Cell Culture Perturbation Experiments

MCF10A cells were obtained from the American Type Culture Collection.²² The cells remained free of mycoplasma and other contaminants and were propagated by adherent culture according to established protocols.²³ MCF10A cells were grown in DMEM-F12 supplemented with 5% horse serum, 20-ng/mL epidermal growth factor, 10- $\mathrm{\mu }\mathrm{g}∕\mathrm{mL}$ insulin, and 0.5-mg/mL hydrocortisone. Plated cells were stored in a 10% ${\mathrm{CO}}_2$ incubator at 37°C and cultured every 3 to 4 days. Cells were detached from the flasks by trypsinization and triply washed in $10\mathrm{mL}$ of phospate-buffered saline (PBS). A hemacytometer was used to determine the concentration of the cells by counting the number of cells per milliliter. On determination of the concentration of the cells, the volumes required to plate 25,000 and 1,000,000 cells per dish were calculated, respectively. MCF10A cells at the two different densities were plated on lysine-coated cover-glass-bottom Petri dishes (size, $35\mathrm{mm}$ ; P35G-0-41-C, MatTek, Ashland, Massachusetts) and immersed in their standard growth media. Two different cell concentrations were used for these experiments because previous work has shown that the lifetime of NADH in cell culture depends on the cell confluency (due to changes in metabolism with cell confluency).²⁴

Redox ratio images,²⁵ which are defined as the fluorescence intensity of FAD divided by the sum of the fluorescence intensities of FAD and NADH, can be used to measure relative changes in the oxidation-reduction state of cells. Redox ratio imaging was used to verify that the concentrations and incubation times of 3-bromopyruvate- and ${\mathrm{CoCl}}_2$ -produced changes in cellular metabolism. Cells were either left untreated (control), treated with 300- $\mathrm{\mu }\mathrm{M}$ 3-bromopyruvate (Sigma-Aldrich, St. Louis, Missouri) for 60 to $90\min$ to inhibit glycolysis, or treated with 200- $\mathrm{\mu }\mathrm{M}$ ${\mathrm{CoCl}}_2$ (Sigma-Aldrich, St. Louis, Missouri) for 60 to $90\min$ to inhibit oxidative phosphorylation, as verified by changes in the redox ratio. NADH lifetime and intensity images were collected at a two-photon excitation wavelength of $780\mathrm{nm}$ and FAD intensity images were collected at 890-nm excitation.²⁶ At the completion of the final image acquisition, cell viability was confirmed under light microscopy (using the same microscope) by trypan blue exclusion. All cell dishes remained viable throughout any given imaging session.

The average power for the cell imaging experiments was approximately $15\mathrm{mW}$ , and the scan area for each image plane was $256\times 256$ pixels $(165\times 165\mathrm{\mu }\mathrm{m})$ . Single-photon counting was done at a rate of approximately $500\times {10}^5$ $\mathrm{photons}∕\mathrm{s}$ for a total of $60\mathrm{s}$ with a pixel dwell time of approximately $11.5\mathrm{\mu }\mathrm{s}$ (i.e., $11.5\mathrm{\mu }\mathrm{s}$ dwell time on each of $256\times 256$ pixels). Thus, over $\sim 80$ frames were averaged to generate each image. Photon count rates at the beginning and end of image acquisition were monitored to ensure that photobleaching did not occur. The instrument response function (IRF) was measured using a second-harmonic-generated signal from a $\beta \text{-}{\mathrm{BaB}}_2{\mathrm{O}}_4$ crystal.²⁰ The full width at half maximum (FWHM) of the IRF was $0.313\mathrm{ns}$ for the cell culture experiments. This represents the IRF of the overall system where the components that mainly contribute to temporal degradation are the transient time spread of the detector, optical dispersion at mirrors and through lenses, and light scattering at diaphragms.

2.3.

DMBA-Treated Hamster Cheek Pouch Model of Oral Cancer

A total of 13 DMBA-treated and 9 control male Golden Syrian hamsters were evaluated in this study (average weight: $156\pm 13\mathrm{g}$ ). Animal care and procedures were in accordance with the guidelines in the U.S. Department of Health and Human Services and National Institutes of Health “Guide for the Care and Use of Laboratory Animals” and approved by the Institutional Animal Care and Use Committee at the University of Wisconsin. For each hamster, the right cheek pouch was either treated three times per week with 0.5% DMBA (by weight, Sigma-Aldrich, St. Louis, Missouri) in mineral oil (DMBA-treated animals), or the right cheek pouch was treated at the same frequency with mineral oil only (control animals) for a total of 16 weeks. The treatment schedule and procedures were established from previous studies.^{27, 28} At 16 to 21 weeks after the commencement of DMBA or mineral oil treatment, the cheek pouch of each animal was imaged using multiphoton FLIM at an excitation wavelength of $780\mathrm{nm}$ , which falls within the absorption band²⁶ of NADH. This wavelength was also selected because statistically significant differences in tissue fluorescence intensity have been observed²⁸ between normal tissues, dysplasias, and carcinomas in the hamster cheek pouch model with multiphoton microscopy at a two-photon excitation wavelength of $780\mathrm{nm}$ .

Prior to imaging, each hamster was anesthetized with an intraperitoneal injection of a mixture of $2.5\mathrm{mg}∕\mathrm{kg}$ acepromazine, $200\mathrm{mg}∕\mathrm{kg}$ ketamine, and $5\mathrm{mg}∕\mathrm{kg}$ xylazine. Next, the cheek pouch was inverted and stretched over an 8-mm-diam cork, pinned to the sides of the cork with two 27-gauge needles, and then wiped with saline. A cover slip was secured on the objective side of a 10-mm-diam opening in the imaging stage using Vaseline. The animal was placed on its stomach, and the cork was secured to the imaging stage of the inverted microscope such that the cheek was flush against the cover slip. The microscope objective was centered within the 10-mm opening of the imaging stage prior to image collection.

The average power incident on the sample ranged from approximately 3.3 to $7.2\mathrm{mW}$ . Single-photon counting was done at a rate of approximately $500\times {10}^5\mathrm{photons}∕\mathrm{s}$ for $45\mathrm{s}$ to $5\min$ with a pixel dwell time of approximately $6.25\mathrm{\mu }\mathrm{s}$ . The scan area at each image plane was $256\times 256$ pixels $(140\times 140\mathrm{\mu }\mathrm{m})$ . Photon count rates at the beginning and end of image acquisition were monitored to ensure that photobleaching did not occur. The image $z$ stacks were generated using a 10- $\mathrm{\mu }\mathrm{m}$ $z$ step size.

To ensure instrument consistency between experimental days, the lifetime of a known and stable instrument standard, 20- $\mathrm{\mu }\mathrm{m}$ -diam Fluoresbrite YG (yellow green) microspheres (Polysciences Inc.) was imaged at the beginning of each day of the in vivo experiments $(n=14)$ . The microspheres were measured under the same experimental conditions as the in vivo measurements, except the power at the sample and the integration time were adjusted to achieve a single photon count rate of approximately $500\times {10}^5$ from the significantly brighter microspheres. The lifetime decay curves measured from the microspheres were fit to a single exponential decay model using SPCImage software (see later). The average lifetime for the microspheres $(n=14)$ was $2.13\pm 0.11\mathrm{ns}$ . The measured value for the lifetime of these spheres is consistent with that reported in other studies^{24, 29} $(\sim 2.2\pm 0.1\mathrm{ns})$ . The IRF measured using a second-harmonic-generated signal from a $\beta \text{-}{\mathrm{BaB}}_2{\mathrm{O}}_4$ crystal had an FWHM of $0.190\pm 0.002\mathrm{ns}$ for the in vivo experiments. The IRF for the in vivo experiments differs from that of the cell culture experiments due to different detectors and optical paths in the two multiphoton FLIM systems.

After in vivo imaging, a biopsy was taken from the imaged site (center of the 10-mm stage opening) with a 3-mm-diam dermal biopsy punch. The biopsy was placed in 10% buffered formalin and submitted for histopathology. The tissue biopsies were cut and stained with hematoxylin and eosin (H&E), and read by a board certified veterinary pathologist (Annette Gendron-Fitzpatrick). Diagnosis was based on established criteria,³⁰ with tissues assigned to one of the following categories: normal, hyperplasia, mild dysplasia, moderate dysplasia, severe dysplasia, papillary hyperplasia/papilloma, CIS, or squamous cell carcinoma (SCC). The diagnosis for a given biopsy was determined from an evaluation of multiple sections cut from that biopsy. If any of these tissue biopsies were found to have more than one diagnosis (for example, severe dysplasia and SCC), the most severe diagnosis was assigned to that biopsy. FLIM images were taken from a microscopic field of view $(140\times 140\mathrm{\mu }\mathrm{m})$ , thus it was difficult to mark the precise image site on the biopsy. The approach taken in this study was to assign a diagnosis to each image stack based on the most severe pathologic diagnosis of the entire corresponding biopsy sample. This approach was consistent with that reported in our previous study, which employed multiphoton microscopy to characterize the NADH intensity in the hamster cheek pouch model of carcinogenesis.²⁸

2.4.

Analysis of the Fluorescence Lifetime Decay Curves

SPCImage software (Becker & Hickl) was used to analyze the fluorescence lifetime decay curves.²⁴ First, the measured lifetime decay curve was binned over the pixel of interest and the eight-nearest neighbor pixels. Each set of binned pixels was fit independently. Next, the measured instrument response was convolved with the model, and this reconvolved model was fit to the measured lifetime decay curve for a given set of binned pixels. This produced a lifetime decay curve with a peak value of approximately 100 counts. The lifetime decay curve of the pixel of interest was then fit to a double-exponential decay model:

2.5.

Quantification of Variables from the FLIM Images for Statistical Analyses

3.1.

Cell Culture Perturbation Studies

Figure 1 shows multiphoton FLIM images of control cells, cells treated with ${\mathrm{CoCl}}_2$ , and cells treated with 3-bromopyruvate for cells plated at high density. This figure shows qualitative differences between the control and the two treated groups. ${\mathrm{CoCl}}_2$ decreases the protein-bound NADH lifetime and 3-bromopyruvate increases the protein-bound NADH lifetime relative to that in the control group.

Fig. 1

Multiphoton FLIM of high-density MCF10A cells imaged in cell culture. Cells were untreated (control), treated with ${\mathrm{CoCl}}_2$ , or treated with 3-bromopyruvate. The color-coded value is ${\tau }_2$ , the protein-bound NADH lifetime (780-nm excitation).

Perturbation with 3-bromopyruvate produced a statistically significant increase in the redox ratio and ${\mathrm{CoCl}}_2$ produced a statistically significant decrease in the redox ratio, reflecting an alteration in the metabolism with these perturbations [Fig. 2c , $p<0.05$ ]. The results for high- and low-density cells are separated because previous work has shown that the cellular lifetime of NADH depends on the cell confluency²⁴ (due to changes in metabolism with cell confluency). The fluorescence lifetimes at 780-nm excitation ( ${\tau }_1=0.30\pm 0.03\mathrm{ns}$ and ${\tau }_2=2.40\pm 0.07\mathrm{ns}$ , averaged for control cells at low and high cell densities) are consistent with free and protein-bound NADH, respectively.^{8, 31} The fluorescence lifetimes at 890-nm excitation ( ${\tau }_1=0.15\pm 0.01\mathrm{ns}$ , ${\tau }_2=2.44\pm 0.06\mathrm{ns}$ , ${\alpha }_2=17\pm 2\%$ , averaged for control cells at low and high cell densities) are consistent with FAD fluorescence.³³ The focus of these experiments is changes in protein-bound NADH with metabolic perturbations, so the statistical analyses of the lifetime of free NADH $\left({\tau }_1\right)$ and the lifetimes of FAD are not shown. Perturbation with 3-bromopyruvate caused an increase in the fluorescence lifetime of protein-bound NADH $\left({\tau }_2\right)$ and a decrease in the relative contribution of protein-bound NADH to the overall lifetime decay $\left({\alpha }_2\right)$ in cells plated at both low and high densities [Figs. 2a and 2b, $p<0.05$ ]. Perturbation with ${\mathrm{CoCl}}_2$ caused a decrease in the fluorescence lifetime of protein-bound NADH $\left({\tau }_2\right)$ and a decrease in ${\alpha }_2$ in cells plated at low and high cell densities [Figs. 2a and 2b, $p<0.05$ ].

Fig. 2

NADH fluorescence lifetimes and redox ratios of MCF10A cell monolayers. Bar plots include the mean and standard error of (a) ${\tau }_2$ (protein-bound NADH lifetime), (b) ${\alpha }_2$ (percent contribution of protein-bound NADH), and (c) the redox ratio [fluorescence intensity of $\mathrm{FAD}∕(\mathrm{NADH}+\mathrm{FAD})$ ] for cells plated at either high or low density. Values are plotted for control cells (high density, $n=8$ ; low density, $n=5$ ), cells treated with 3-bromopyruvate (high density, $n=6$ ; low density, $n=6$ ), which inhibits glycolysis, and cells treated with ${\mathrm{CoCl}}_2$ (high density, $n=6$ ; low density, $n=6$ ), which inhibits oxidative phosphorylation. Statistically significant differences ( $p<0.05$ , unpaired Wilcoxon rank-sum tests) exist between control versus 3-bromopyruvate-treated and control versus ${\mathrm{CoCl}}_2$ -treated cells for the variables marked with an asterisk ${(}^{*})$ . The multiphoton FLIM excitation wavelength is $780\mathrm{nm}$ .

3.2.

Hamster Cheek Pouch Pathology

Histopathology was used as the gold standard for tissue diagnosis. Of the 22 cheek pouches imaged (one per animal), 9 were diagnosed as normal (the 9 control animals), and of the 13 DMBA-treated animals 2 were diagnosed with mild dysplasia, 4 with moderate dysplasia, 4 with severe dysplasia, 2 with CIS, and 1 with SCC. For the purpose of statistical analyses, the tissue samples were divided into three categories: normal $(n=9)$ , low-grade precancer (mild dysplasia and moderate dysplasia, $n=6$ ), and high-grade precancer (severe dysplasia and CIS, $n=6$ ). High-grade precancers are more likely to progress to cancer than low-grade precancers.³⁴ The single SCC sample was not included in the statistical analysis.

3.3.

Multiphoton FLIM Images of Normal, PreCancerous, and Cancerous Tissue in Vivo

The hamster cheek pouch has three distinct layers: an acellular superficial layer, an epithelium, and an underlying stroma. Three-dimensional image stacks of fluorescence lifetimes and the fractional contribution of the long-lifetime component of these different layers (Fig. 3 ) were used to determine whether multiphoton FLIM can differentiate histologically confirmed normal and dysplastic tissues.

Fig. 3

Three-dimensional multiphoton FLIM images of the normal and precancerous hamster cheek pouch in vivo. Multiphoton FLIM of the normal hamster cheek pouch (a), (b), and (c), low-grade precancer (d, mild dysplasia), high-grade precancer (e, severe dysplasia), and SCC (f) measured in vivo at 780-nm excitation. Images are color coded to values of ${\alpha }_2$ [contribution of long-lifetime component; (a)], ${\tau }_1$ [short-lifetime component; (b)], and ${\tau }_2$ [long-lifetime component; (c), (d), (e), and (f)]. The color bar range is at the bottom of each montage. Note that ${\tau }_1$ , ${\tau }_2$ , and ${\alpha }_2$ each have a different color scale. The numbers in the lower left corner of each image indicate the depth below the tissue surface in micrometers, and each image is $100\times 100\mathrm{\mu }\mathrm{m}$ .

In the normal cheek pouch [Figs. 3a, 3b, 3c], the acellular superficial layer fluorescence $\left(0\mathrm{\mu }\mathrm{m}\right)$ has a fairly uniform distribution of long-lifetime contribution $\left({\alpha }_2\right)$ , short lifetime $\left({\tau }_1\right)$ , and long lifetime $\left({\tau }_2\right)$ . However, in low-grade [Fig. 3d] and high-grade precancers [Fig. 3e], the ${\tau }_2$ of the superficial layer is nonuniform. There was no acellular superficial layer in the one SCC sample [Fig. 3f]. The cytoplasmic fluorescence of the cells in the epithelium of the normal cheek pouch [Figs. 3a, 3b, 3c between 10- and 30- $\mathrm{\mu }\mathrm{m}$ depths] has a short $\left({\tau }_1\right)$ and long $\left({\tau }_2\right)$ lifetime component consistent with free and protein-bound NADH, respectively^{8, 31} (percent contribution of protein-bound NADH, ${\alpha }_2)$ . The nuclear to cytoplasmic ratio of the cells increases with depth in the epithelium. However, this trend is not observed in the high-grade precancer [Fig. 3e] or the SCC [Fig. 3f]. The cellular fluorescence distribution in the high-grade precancer and SCC is also more perinuclear compared to normal tissue. The predominant lifetime component in the normal and precancerous epithelial cells is the free NADH lifetime $\left({\tau }_1\right)$ . The protein-bound NADH lifetime $\left({\tau }_2\right)$ of precancerous and cancerous cells appears to be shorter than normal cells. The lifetime of structural proteins in the stromal layer is much different than the lifetime of NADH in the epithelial cells. In the stromal layer of the normal cheek pouch [Figs. 3a, 3b, 3c, 40- $\mathrm{\mu }\mathrm{m}$ depth] ${\tau }_1$ is the predominant lifetime component (the contribution of ${\tau }_2$ is negligible) and its value is similar to the temporal response of the FLIM system and thus is attributed to the second-harmonic generation (SHG) of fibrillar collagen (which has a zero lifetime³⁵). The stromal layer in the precancers and SCC is at a much greater depth than $40\mathrm{\mu }\mathrm{m}$ .

3.4.

Statistical Analyses of Fluorescence Lifetime Parameters

Features that were qualitatively observed to change with precancer development in the multiphoton FLIM images were next quantified to determine if these observed differences were statistically significant. First, morphological variables were quantified, and statistical tests revealed that the epithelial thickness of high-grade precancers $(52\pm 19\mathrm{\mu }\mathrm{m})$ are greater than normal $(18\pm 5\mathrm{\mu }\mathrm{m})$ . All hamster cheek pouches were imaged to the depth of the bottom plane in the epithelium, except for one low-grade precancerous sample, in which the signal was completely attenuated before the last epithelial plane was imaged. Thus, the epithelial thickness of low-grade precancers is slightly underestimated $(34\pm 19\mathrm{\mu }\mathrm{m})$ due to this one sample. Statistical tests also revealed that the nuclear diameter of low-grade $(7.5\pm 0.5\mathrm{\mu }\mathrm{m})$ and high-grade $(7.5\pm 0.4\mathrm{\mu }\mathrm{m})$ precancers are greater than normal $(6.3\pm 0.3\mathrm{\mu }\mathrm{m})$ $(p<0.05)$ . More extensive analysis of morphological changes quantified from multiphoton intensity images of the DMBA-treated hamster cheek pouch at 780-nm excitation is included in a previous paper.²⁸

The range of ${\tau }_1$ in normal epithelial cells was $0.29\pm 0.03\mathrm{ns}$ (consistent with free NADH). The focus of this study is changes in cellular protein-bound NADH with precancer development ( ${\tau }_2$ in the cells), so the statistical analysis of the lifetime of free NADH is not shown.

Changes in cellular lifetime with depth in the epithelium are shown in Fig. 4 . Changes in ${\tau }_2$ and ${\alpha }_2$ with depth were evaluated only within the epithelium of the hamster cheek pouch, where precancer arises. The change in lifetime with depth within the epithelium was determined from a paired comparison of the first (superficial layer) versus the last (basal layer) imaged plane of the epithelium within an animal where the greatest differences in morphology are expected. The bottom plane imaged in the epithelium of normal tissue $(18\pm 2\mathrm{\mu }\mathrm{m})$ was less than that of low-grade precancers $(34\pm 19\mathrm{\mu }\mathrm{m})$ and that of low-grade precancers was less than that of high-grade precancers $(52\pm 19\mathrm{\mu }\mathrm{m})$ . This is expected, since one of the hallmarks of epithelial precancer development is thickening of the epithelial layer with neoplasia. Paired statistical tests revealed no change in cellular ${\tau }_2$ or ${\alpha }_2$ with depth within the epithelium of normal tissues and high-grade precancers $(p>0.05)$ . However, cellular ${\tau }_2$ and ${\alpha }_2$ were found to decrease with depth within the epithelium of low-grade precancers $(p<0.05)$ . Note that the depth of the last imaged plane in the epithelium of normal tissues is much less than precancerous tissues, and thus it is not possible to compare the changes in ${\tau }_2$ and ${\alpha }_2$ over the same depth between these two tissue types.

Fig. 4

Change in cellular lifetime with depth in the normal and precancerous hamster cheek pouch in vivo. Percent change in cellular ${\tau }_2$ [protein-bound NADH lifetime; (a)] and ${\alpha }_2$ [percent contribution of protein-bound NADH; (b)] with depth within the epithelium of the hamster cheek pouch in vivo measured with multiphoton FLIM at 780-nm excitation. Within each animal, ${\tau }_2$ and ${\alpha }_2$ of three cells per image plane were averaged, and then the percent changes in ${\tau }_2$ and ${\alpha }_2$ of each plane were calculated relative to those of the most superficial plane in the epithelium (defined as the first layer of cells below the acellular superficial layer). Next, the percent change in ${\tau }_2$ and ${\alpha }_2$ at each depth were averaged for all animals. Solid circles represent the mean percent change for the last plane of the epithelium for all animals, plotted at the mean depth of the last plane in the epithelium of all animals relative to the most superficial plane in the epithelium. Paired Wilcoxon signed rank tests of the first versus the last plane in the epithelium within an animal indicated a statistically significant decrease in ${\tau }_2$ and ${\alpha }_2$ with depth $(p<0.05)$ in low-grade precancers.

Statistical tests also revealed that normal and dysplastic tissues are differentiated with the mean cellular protein-bound NADH lifetime ( ${\tau }_2)$ (Table 1 ). The protein-bound NADH lifetime ( ${\tau }_2)$ of both low- and high-grade precancerous cells are statistically lower than normal $(p<0.05)$ . The relative contribution of protein-bound NADH $\left({\alpha }_2\right)$ for low-grade precancerous cells is also statistically lower than normal.

Intracellular variability was assessed with the cellular coefficient of variation³² (Fig. 5 ). Intracellular ${\tau }_2$ and ${\alpha }_2$ variability for high- and low-grade precancers is statistically greater than normal $(p<0.05)$ , and intracellular ${\alpha }_2$ variability for low-grade precancers is statistically greater than high-grade precancers $(p<0.05)$ .

Fig. 5

Intracellular fluorescence lifetime variability of normal and precancerous cells in the hamster cheek pouch in vivo. Mean and standard error of the cellular coefficient of variation (cell standard deviation/cell mean) for ${\tau }_2$ (protein-bound NADH lifetime) and ${\alpha }_2$ (percent contribution of protein-bound NADH) for all normal cells ( $n=81$ cells), low-grade precancerous cells ( $n=72$ cells), and high-grade precancerous cells ( $n=111$ cells) measured with multiphoton FLIM in the hamster cheek pouch in vivo at 780-nm excitation. The coefficient of variation assesses the extent of variability within a cell relative to its mean³² (intracellular variability). Unpaired Wilcoxon rank sum tests (and $t$ tests) revealed statistically significant differences $(p<0.05)$ in the coefficient of variation of ${\tau }_2$ and ${\alpha }_2$ between normal cells and all precancerous cells ${(}^{*})$ , and in the coefficient of variation of ${\alpha }_2$ between low-grade and high-grade precancerous cells $(†)$ $(p<0.05)$ .

Table 1

NADH fluorescence lifetimes of normal and precancerous cells in the hamster cheek pouch in vivo. Mean and standard deviation of cellular τ2 (protein-bound NADH lifetime) and α2 (percent contribution of protein-bound NADH) averaged for all cells within an animal for all normal animals (n=9) , low-grade precancerous animals (n=6) and high-grade precancerous animals (n=6) from multiphoton FLIM at 780-nm excitation. Statistically significant differences ( p<0.05 , unpaired Wilcoxon rank-sum tests) exist between normal and low-grade precancer (*) for τ2 and α2 , and between normal and high-grade precancer (*) for τ2 .

The mean ${\alpha }_2$ $(1.5\pm 0.6\%)$ and ${\tau }_1$ $(0.10\pm 0.003\mathrm{ns})$ of normal and dysplastic collagen confirms that the collagen signal is 99% SHG (which has zero lifetime,³⁵ the minimum calculated lifetime constrained by the fit model is $0.1\mathrm{ns}$ ).