1. Introduction

Cancer mortality presents one of the leading medical challenges in the United States, accounting for almost 25% of all deaths [1]. While research has been extensive in this area, loss of life due to cancer was unchanged from 1950 to 2002 (193 deaths per 100,000), a trend explained to some extent by increased diagnosis and awareness in the general populace [1]. Cancer prevention through lifestyle changes and early detection is the most effective means of lowering these mortality rates.

Given the success of endoscopy as a minimally-invasive imaging modality for early cancer detection in the esophagus, the identification of an optically discernable mechanism that is consistently altered in a malignant process could prove very useful, not only for further developing therapeutic targets, but also for understanding disease pathogenesis and developing minimally invasive optical technologies for early detection. A recent clinical study presented evidence for the presence of detectable levels of NAD(P)H fluorescence in human epithelial tissues in vivo and demonstrated that NAD(P)H may be used as a quantitative fluorescence biomarker for in vivo detection of dysplasia in the esophagus [2]. Here, the underlying biological basis for endogenous fluorescence changes that occur during the course of esophageal cancer progression was investigated by employing a unique, time-resolved optical molecular imaging approach that provides a molecular snapshot of metabolic function in living human esophageal cancer cell models.

Knowing the complicated nature of cellular machinery, it is not surprising that pervasive changes in metabolic function almost always manifest as ill health at the anatomic level [3]. The oxidative phosphorylation cycle is an attractive point of analysis in this regard: several key biomolecules such as NAD(P)H, FADH₂, oxygen, and ATP are consumed/generated during this process. Of these, NAD(P)H provides a bright, endogenous fluorescence signature that is detectable optically without the use of exogenous dyes [4] (the term NAD(P)H is used to describe the similar absorption/emission characteristics of NADH and NADPH). Oxygen, while not fluorescent, can quench fluorescence in transition-metal complexes [5] and a class of ruthenium dyes has emerged as the biological probe of choice for oxygen sensing that has been studied for photochemistry, oxygen response, and toxicity [6, 7].

Developments in imaging technologies allow researchers to extract information from biological systems, including cell morphology, intracellular ionic concentrations, and membrane integrity [8]. In particular, fluorescence lifetime studies have gained prominence as molecular timers that can be used to study a plethora of cellular events. Lifetime is an inherent property of the excited electronic state of the fluorescent molecule (fluorophore) and is defined as the average time the fluorophore spends in the excited state before returning to the ground state. While lifetime is independent of fluorophore concentration, photobleaching, absorption, and scattering, it is influenced by the local microenvironment of the fluorophore (e.g., pH, ions such as Ca²⁺, molecular associations) and hence can be used to probe the intracellular milieu [9]. While fluorescence lifetime imaging has been put to a variety of uses, the most common has been for studying energy transfer during molecular associations in living cells as a function of intermolecular distance, thereby measuring molecular gaps below the resolution of current imaging capabilities [10, 11].

A wide-field, time-domain fluorescence lifetime imaging microcopy (FLIM) system was developed to probe cellular metabolic function and detect molecular activity in living cells (Fig. 1(a)) [12]. The large temporal dynamic range (600 ps – infinity) of this system is unique and essential for the purpose of this study: UV excitation and an ultrafast gated camera enable imaging of endogenous fluorophores such as NAD(P)H with sub-nanosecond lifetimes, while visible excitation from a low repetition rate laser source and large gates enable measuring lifetimes for ruthenium dyes, which are typically hundreds of nanoseconds (Fig. 1(b)). FLIM has been successfully applied to the study of NAD(P)H fluorescence, as well as lifetime modulation of oxygen sensitive ruthenium tris(2,2’-dipyridyl) dichloride hexahydrate (RTDP), in living human bronchial epithelial cells [12, 13]. Here, the oxygen sensitivity of RTDP was calibrated to provide quantitative oxygen measurements in living cells under strict, temperature-controlled conditions using a fiber-optic oxygen sensor. Approaches for oxygen measurements in biological tissue have included, but are not limited to, electrochemical Clark-type electrodes (consume oxygen), NMR (requires high signal levels) and SECM (invasive, requires electrolyte media) [14, 15]. Fluorescence, on the other hand, is an established technique that meets the demanding criteria of high sensitivity and spatial resolution required for intracellular oxygen sensing. Oxygen affects biological systems at all strata of organization, from the subcellular (ATP production) to the cellular (proliferation) and supracellular level (e.g., arterial oxygen), so an accurate and reliable method to quantitatively measure oxygen levels is critical for understanding as well as controlling systemic response.

The FLIM system was applied towards studying metabolic function in two related cell lines: normal human squamous esophageal epithelial cells (HET) and Barrett’s adenocarcinoma esophageal cells (SEG). SEG were used as a model cell line for assessing the effect and capability of RTDP in the intracellular environment. Both cell lines were imaged for NAD(P)H and assessed for oxygen levels. Measurement of multiple components of an ordered pathway such as oxidative phosphorylation not only provide a more complete perspective on cancer progression, but also offer endogenous targets for clinical optical diagnostic technologies for cancer prevention, without the need for exogenous contrast agents.

2. Materials, instrumentation, and methods

2.1. Fluorescence lifetime imaging microscope (FLIM)

 

Fig. 1. (a) Fluorescence Lifetime Imaging Microscopy (FLIM) setup. Abbreviations: CCD–charge coupled device; HRI–high rate imager; INT–intensifier; TTL I/O–TTL input/output card; OD–optical discriminator. Abbreviations for optical components: BS–beam splitter; DC–dichroic mirror; FM–mirror on retractable ‘flip’ mount; L1, L2, L3, L4, L5–quartz lenses; M–mirror. Thick solid lines–light path; thin solid line–electronic path. The FLIM system can excite in the UV-NIR range, from 337–960 nm depending on the laser dye used. The nitrogen laser is a pulsed source with peak energy of approximately 1.3 mJ with reproducibility within ±2%. FLIM has a spatial resolution of 1.4 µm and (with structured illumination) can achieve an optical section down to 10 µm. A key advantage of this FLIM implementation is its ability to measure lifetimes of long-lived fluorophores, ranging from 750 ps to almost 1µs with a resolution of 50 ps. Two different culture dish systems provide temperature controlled units for cellular study under more physiologically-apt conditions. (b) Illustration of data collection for RTDP fluorescence at delays of 40, 140, 240 and 340 ns. The gate width was set to 50 ns in order to obtain sufficient fluorescence signal. Note that the excitation is very short compared (<1ns) to the long lifetimes observed for RTDP (hundreds of nanoseconds) and is hence depicted as a pulse. (c) RTDP Calibration Curve: Plot of relative lifetime (τ₀/τ) vs. oxygen levels (µM). RTDP calibration indicated a linear relationship between oxygen levels and relative lifetime, which was in good agreement with the Stern-Volmer equation. Over multiple runs K_q=4.5±0.4×10^-3 µM^-1. The intercept ≠1, indicating some degree of experimental variance. The calibration could differentiate between oxygen levels differing by as little as 8µM.

Download Full Size | PDF

A schematic of the FLIM system is shown in Fig. 1(a) [12]. FLIM is configurable to use either the nitrogen laser (GL-3300, PTI, Lawrenceville, NJ) output for UV excitation at 337.1 nm or either of two tunable dye lasers (GL-301, PTI) for excitation in the vis-NIR range (337–960 nm), depending on the setting(s) of the flippable mirror(s) (FM). For RTDP experiments, a dye laser was set to 460 nm excitation (Coumarin 460, Exciton, Dayton, OH). Pulse energy from the dye laser was 0.3 mJ at the output and 20 µJ at the sample. For NADH experiments, the nitrogen laser was used to directly provide UV excitation. Pulse energy from the nitrogen laser was 1.3 mJ (±2%) at the output and 60 µJ at the sample. Lifetime measurements were possible in the range 750 ps – 1 µs with a temporal resolution of 50 ps.

Excitation light was coupled into a 20 m long optical fiber (SFS600/660N, Fiberguide, Stirling, NJ). The light exiting the fiber was projected into the back port of a Zeiss Axiovert S100 inverted microscope via a structured illumination lens assembly. A dichroic mirror (Q495lp, Chroma Technology Corp., Brattleboro, VT) delivered the beam into a 40× Fluar (1.3NA, Zeiss, Jena, Germany) objective for sample excitation with an illumination diameter of 150 µm. A reference beam split from the main laser served as the timing reference via an electronic pulse generated by means of a constant fraction optical discriminator (OD) (OCF-400, Becker&Hickl GmbH, Berlin, Germany), the output from which was sent to a picosecond delay generator (DEL350, Becker & Hickl GmbH) that provided an adjustable time delay to trigger an ultrafast gated (min. 200 ps), intensified CCD camera (Picostar HR, La Vision, Goettingen, Germany) that was used for image detection with a spatial resolution of 1.4 µm. For RTDP experiments, gate widths were controlled by slaving the ICCD gate to an external logic signal. Intensifier (MCP) voltage was set at 700 V for the RTDP solution and 800 V for cell samples. For imaging cellular NADH, the ICCD was operated in Comb mode and the dichroic mirror (350lpDC Chroma Technology Corp.) was changed accordingly.

2.2. Image Collection and Analysis

FLIM imaging and analysis was done via an intensified CCD camera controlled by the DaVis 6 software (LaVision, Goettingen, Germany). A dark background was taken to account for thermal current in the CCD and was subtracted from each intensity image. The CCD camera was operated within its specified linear intensity-response range, which was controlled by two separate settings: 1) Adjusting the gain on the intensifier via the HRI and 2) modulating the gate width via the software interface. No adjustments were made to the intensity images other than those during image capture. Lifetime maps were calculated from multiple intensity maps by implementing the Rapid Lifetime Determination (RLD) algorithm on a per-pixel basis by fitting the lifetime of pixel p to the logarithm of the intensity (assuming single-exponential decay):

where Ii,_p is the intensity of pixel p in image i, t _i is the time delay of image i, and C is a constant. Least squares lifetime fits were:

where N was the number of images. All sums are over i. At each delay, the average of five intensity images was recorded to compensate for excitation laser intensity. All images were pre-processed via a uniform threshold filter where pixels below a minimum, non-zero intensity value were set to zero. The nitrogen laser was operated at about 3 Hz and 640×480 pixel lifetime images using 4 gates were generated using a PC with an Intel Pentium-IV processor at 3.2 GHz, for total acquisition and analysis times <15 s. Accurate lifetimes were obtained by creating a histogram based on the lifetime map, fitting it to log-normal functions and calculating the average time from the regression. Further threshold filters were applied which set pixels with improbable values of lifetime (low and high) to zero. This is reasonable, since the range of possible lifetime values for RTDP and NADH have been well documented in literature [13, 16]. The resulting image is the final lifetime map that is evaluated for oxygen levels. The linear scale of the lifetime map was converted to a binary, two-color scale for the sole purpose of easier visualization. The oxygen distribution image was generated by directly applying the calibration (described below) to the lifetime map without pre-processing. Finally, threshold filters were applied that set pixels with unlikely oxygen levels (both low and high) to zero. Again, this is reasonable given the range of physiological oxygen values [17].

2.3. Temperature Control

A schematic of the modified microscope stage configuration is shown as a blowup in Fig. 1(A). For temperature control, the Delta T controlled culture dish system (Bioptechs, Butler, PA) was implemented along with the Delta T perfused heated lid (Bioptechs) and was used for all experiments, unless noted otherwise. The heated lid had two ports which could be used for perfusion purposes. A remote controller was used to adjust temperature settings. For closed chamber experiments the FCS2 chamber system and controller (Bioptechs) was used, where cells were enclosed between a cell plate and a perfusable gasket with a media holding capacity of 1 ml. Due to the large thermal mass of the objective, a separate Objective Heater (Bioptechs) with its own controller was also installed. All temperature control systems had an operating range from ambient to 45°C with accuracy within ±0.2°C.

2.4. Confocal Microscopy

An independent Zeiss LSM 510 microscope (Zeiss, Jena, Germany) was used to acquire confocal NADH fluorescence images via 364 nm excitation (emission: 435nm–485nm). Images were acquired via LSM 5 Software Release 3.2 (Zeiss) in real time. Detector gain and offset were configurable via the software interface and were adjusted to avoid detector saturation. Offline image analysis was done using LSM Image Browser 3.2 (Zeiss). The only image parameters adjusted for enhancement were brightness and contrast and these changes were uniformly applied across the entire image. A range indicator function in the software provided information on color saturation (via brightness enhancement) or zero pixel values (via increased contrast). The function used a pseudo-color scheme where all saturated pixels were colored red, all zero pixels were colored blue and all other were plotted on a linear grayscale. Adjusting brightness and contrast while maintaining the entire image within the grayscale range allowed preservation of image information while avoiding loss or generation of data.

2.5. RTDP Characterization

0.025 g of RTDP (ruthenium tris(2,2’-dipyridyl) dichloride hexahydrate) powder obtained from Sigma-Aldrich (#224758, St. Louis, MO) was dissolved in 25 ml phosphate buffer saline (PBS, Invitrogen, Carlsbad, CA) to create a stock solution of 1 mg/ml, or 1.34 mM. Absorption spectra of RTDP were obtained from a Beckman-Coulter DU-800 Spectrophotometer (Ontario, Canada) over a range of 200–800nm. The fluorescence emission spectrum was extracted using a Jobin-Yvon Fluorolog-3 (Edison, NJ) in a range of 300–700 nm. Temperature response of RTDP lifetime was studied using 2 ml of stock solution and FLIM. Initially, ambient temperature was 22°C and the solution was exposed to normal atmospheric conditions. Temperature was varied in the range of 25–45°C in steps of 2.5°C. At each temperature point, lifetime images were acquired three times and averaged. Multiple runs were used to validate the slope of the plot obtained. FLIM was used to study RTDP lifetime independence of its concentration. Six different RTDP concentrations in the range 0.5–5 mg/ml were used for intensity and lifetime studies. The temperature was set to 37°C (physiologic value), gate width at 50 ns and the HRI gain at 700 V. The settings were selected to maximize signal at the CCD without saturation when the highest concentration of 5 mg/ml was measured for lifetime. Below 0.5 mg/ml, the intensity was too weak to reliably evaluate lifetimes without changing gain settings to manipulate intensity. Three readings were obtained per concentration and were averaged for intensity and lifetime.

2.6. Calibration of Oxygen Sensitivity of RTDP

Independent oxygen measurements for RTDP calibration were performed using a commercial fiber-optic oxygen sensing system (FOXY, Ocean Optics, Dunedin, FL). The FOXY system is a fluorescence-based, spectrometer-coupled chemical sensor used for spectral analysis of both dissolved and gaseous oxygen with a manufacturer specified range of 0–40 ppm (>1 mM) for oxygen dissolved in water, a resolution of 0.02 ppm, and response time < 1 sec. The FOXY was calibrated using a two-point linear fit. The first point was obtained by immersing the FOXY probe in a solution of sodium hydrosulfite (dithionite) dissolved in DI water through which nitrogen had been bubbled, yielding a solution with near-zero oxygen concentration (sodium dithionite is a potent oxygen scavenger and bubbling nitrogen removed virtually any traces left). The second point was obtained by immersing the FOXY probe in a solution of DI water, which has a known oxygen concentration of 7.1 ppm or 222 µM at 37°C under normal atmospheric O₂. Data from these two points was used to calibrate the FOXY via a software interface.

Next, a lifetime-oxygen calibration of RTDP was performed on the stock solution using FLIM. Gas tubing was attached to one port on the heated, perfused lid while the FOXY was inserted into the solution via the other port. Under equilibrium conditions, the oxygen level was similar to the calibration value, i.e., 7.1 ppm. Nitrogen was flowed over the sample and a drop in oxygen was recorded via FOXY. Once the FOXY output stabilized at <0.1 ppm, the lifetime was recorded, the flow of nitrogen was stopped and the solution was allowed to recover by equilibrating with atmospheric oxygen. During this process, several concurrent lifetime and oxygen measurements were taken.

Once the solution reached equilibrium, the solution was perfused with oxygen and an increase in levels was correspondingly observed by the FOXY. The oxygen flow was then cut-off and while the solution stabilized toward equilibrium, readings were taken in a manner similar to that described above. The data was verified by repeating the calibration multiple times. Average values were used for cell studies.

Data analysis - Stern-Volmer Equation: Oxygen is one of the best known collisional quenchers and its effect on RTDP can be effectively described by using the Stern-Volmer Equation (Eq. 4) [5]. Since increased temperatures lead to faster diffusion for both fluorophore and quencher and hence higher collision quenching rates, it is characteristic for K_q to increase with temperature [5]. To evaluate K_q, the highest lifetime value attained (when [O₂] <0.1 ppm) was set equal to τ₀. This was reasonable, since the FOXY had a minimum detection limit of 0.02 ppm and it is near impossible to completely deplete oxygen. For all other values, τ₀/τ was evaluated and plotted as a function of [O₂]. Line fitting of all data points and regression analysis was carried out using statistical software. The slope of the line yielded a single value of K_q. Over multiple runs, K_q was averaged to provide a mean value that was used for analyzing cell lifetimes (see below). A key advantage of the relative lifetime, Stern-Volmer approach was that any variability in RTDP lifetime associated with aging or other ambient conditions was systematically eliminated by determining τ0 from the known (stock) solution for each experiment.

2.7. Cell preparation

The HET-1 is a human squamous epithelial cell line immortalized by transfection of the SV40 T antigen early region gene [18]. The SEG-1 (Barrett’s) adenocarcinoma esophageal cell line was derived from Barrett’s-associated adenocarcinoma of the distal esophagus [19]. HET and SEG were propagated in 100 cm² culture flasks in DMEM media (#11995-065, Invitrogen) containing 10% fetal bovine serum (FBS) and 1% antibiotic-antimycotic (#15240-062, Invitrogen) and cultured at 37°C under standard conditions. Upon reaching 70% confluence, the cell cultures were split using trypsin and divided at a ratio of 1:4. For open-air studies, the cells were plated a day before experimentation. About 6 hours before analysis, 1 ml of RTDP stock solution was added to the cells. The cells were washed with PBS or media (without phenol red) before being transferred to the microscope stage. No significant fluorescence was observed in the extracellular space, indicating low culture media fluorescence. This process was mirrored for both cell lines whenever comparative studies were done. Cell viability tests (Trypan Blue) yielded >85% viability after 6 hrs of incubation with RTDP, indicating that significant cytotoxicity was not occurring within the incubation period. No staining was necessary for NADH studies.

For closed chamber RTDP studies, the cells were plated a day before on cell plates (see Temperature Control) and the RTDP stock solution was added about 6 hours before experimentation. After washing with PBS, the plates were assembled into the closed chamber system, sealed shut with media and mounted onto the microscope stage.

2.8. RTDP lifetime measurement in cells

All controllers were set at 37°C after the cells were loaded onto the stage. The objective was used to focus on an appropriate region with the cells clearly visible and lifetime measurements were taken. Each measurement was repeated multiple times. For time-lapse experiments, measurements were taken periodically. Lastly, for each experiment, the lifetime of the stock solution was also obtained. The calibration (K_q) was applied to determine intracellular oxygen concentration. Values of lifetime (τ) was evaluated for the stock solution of known oxygen level and fed into the Stern-Volmer equation to obtain τ₀. A software macro was implemented to generate oxygen distribution images based on the calibration data and using the lifetime images as input. The oxygen distribution was evaluated on a pixel by pixel basis for each lifetime image by feeding back the value of τ into the Stern-Volmer equation. For each cell visible in the lifetime image, τ_cell was evaluated as the average of all pixels comprising the image of the cell. Hence, oxygen levels were assumed to be uniform within a cell. Average intracellular oxygen levels were evaluated from the oxygen distribution in a similar manner. For each experiment, oxygen levels from at least 5 cells were averaged to obtain mean values.

3. Results

3.1. NADH Measurements in HET and SEG

NADH reduction by complex I to NAD+ along with electron transfer to the carrier coenzyme-Q (CoQ) is one of the first steps in oxidative phosphorylation (Fig. 2(a)). This step contributes not only to oxygen consumption downstream but also to the proton gradient which ultimately results in ATP production. Since NADH is generated during glycolysis and consumed during oxidative phosphorylation, it is one measure of completion for these processes.

The term NAD(P)H is used to point out the spectral similarity of NADH and NADPH, both of which have a wide excitation range centered at approximately 350 nm and an emission peak at 460 nm. Since other endogenous fluorophores have overlapping excitation in the UV range, preliminary studies were conducted to observe the sub-cellular origin of fluorescence. Figure 2 presents confocal fluorescence intensity images of SEG incubated with a commercial mitochondrial stain (Mitotracker Red). Figure 2(c) shows that NAD(P)H fluorescence clearly has the same origin as the Mitotracker fluorescence (Fig. 2(b)), indicating that fluorescence observed arose mostly from mitochondrial NADH and not NADPH.

Figures 2(d), 2(e) provide confocal NADH fluorescence intensity images from HET and SEG, respectively, under identical measurement settings. It is visually evident that the SEG have a starkly brighter fluorescence signature than the HET. Similar results were also obtained with FLIM and overall the SEG exhibited 2–5 fold higher fluorescence signals than the HET. This difference in signal from NADH was observed to be statistically significant (p<0.05, Fig. 2(f)).

 

Fig. 2. (a) Oxidative phosphorylation in mitochondria. Complex I (NADH dehydrogenase) converts NADH to NAD+ and passes on electrons to carrier CoenzymeQ (CoQ) while pumping hydrogen ions into the intermembrane space. Complex II (succinate dehydrogenase) can also generate and pass on electrons to CoQ via a complex internal mechanism which is initiated by conversion of succinate to fumarate, a step in the Krebs cycle. CoQ transfers the electrons to an intermediary complex III (CoQ – cytochrome C oxidoreductase), which in turns enhances the proton gradient by pumping hydrogen ions against the gradient. The electrons are transferred to complex IV (cytochrome oxidase) via cytochrome C (CytC) which in turn hydrolyses oxygen to water and also pumps protons against the gradient. The ATP synthase complex (complex V) moves protons down the gradient, converting osmotic energy to chemical energy via ATP synthesis from ADP in a process known as chemiosmotic coupling. A more detailed explanation of the process and the unique structure of ATP synthase can be found in any standard biochemistry text. (b) NADH fluorescence and (c) Mitotracker-stained SEG images. For the Mitotracker, excitation=543 nm and emission=636 nm. Mitotracker Red is a commercially available stain used for tracking mitochondria within living cells. Fluorescent signals from both markers were found to co-localize. Confocal intensity images of NADH fluorescence in HET (d) and SEG (e): Illustration of differences observed with the Zeiss 5 LSM. The SEG consistently presented a brighter signature than the HET by approximately 2.5-fold. (f) Plot of differences in NADH fluorescence intensity and lifetime observed between the HET and SEG over multiple measurements with the FLIM system. No significant differences in NADH lifetime were observed.

Download Full Size | PDF

FLIM measurements were made for NADH lifetime in both cell lines. Values obtained were τ_HET=2.51±0.16 ns and τ_SEG=2.21±0.16 ns and were within the range of values reported in literature [16]. The differences in lifetimes were not statistically significant, therefore it can be established that the fluorescence intensity differences between the HET and the SEG were mainly attributable to differences in intracellular NADH levels. The observation reported here that esophageal cancer cells have higher absolute NADH levels than normal esophageal cells could explain the higher levels of NAD(P)H reported in dysplastic vs. nondysplastic esophageal tissues via clinical studies described earlier (see introduction) [2].

3.2. RTDP calibration

RTDP has been studied for its photophysical and photochemical properties and recently found application in biomedical research as a molecular oxygen sensor [6]. With a long fluorescence lifetime (hundreds of nanoseconds), easy uptake by cells, and minimal cytotoxic and phototoxic effects, RTDP provides a nearly ideal means of assessing intracellular oxygen.

Measurements made with the FLIM system indicated that RTDP lifetime decreased with increasing temperature and that this decrease was almost linear:

where τ=lifetime and T=temperature in degree Celsius and K was a constant, emphasizing the need for temperature control in lifetime studies. The slope was in good agreement with previous work, which measured RTDP lifetimes within a similar temperature range [7].

RTDP calibration results at 37°C indicated a linear relationship between oxygen levels and relative lifetime, which was in good agreement with the Stern-Volmer equation (Fig. 1(c)):

where K_q is the Stern-Volmer quenching constant. Relative lifetime at a given oxygen level [O₂]_x was evaluated as the ratio of uninhibited RTDP lifetime (i.e. 0% oxygen, or τ₀) to τ_x. Over multiple runs, the value of K_q was evaluated to be K_q=4.5±0.4×10^-3 µM^-1. This is higher than other reported values which were measured at room temperature, confirming that K_q increases with temperature [17]. System resolution was determined by a) the resolution of the oxygen sensor used for calibration (±0.6 µM) (see Methods section) and b) the lifetime variance of the FLIM system (±2% of lifetime, or ±6 ns for RTDP). Using these values in Eq. 4, it was estimated that the calibration could reliably differentiate between oxygen levels differing by as little as 8 µM.

3.3. Oxygen Measurements in HET and SEG

 

Fig. 3. Clockwise from top left - (a) confocal fluorescence and (b) Differential Interference Contrast, or DIC images of SEG. SEG were incubated with the RTDP dye prior to imaging (see Methods section). FLIM images of SEG incubated with RTDP: (c) DIC, (d) fluorescence intensity in counts, (e) lifetime in ns and (f) oxygen in µM. Note that one cell in the bottom of (c) shifted position in (d) in the time lapse between these two images. (g) The results of depletion experiments on SEG (see Methods section). Cellular viability is compromised with the passage of time and this results in the lifetime/oxygen content leveling off towards the end.

Download Full Size | PDF

Confocal measurements indicated that RTDP distributed uniformly in the cytoplasm without any evident aggregation or adherence to organelles (Fig. 3(a,b)). It yielded a strong intracellular luminescent signal without visible cytotoxic effects or membrane damage. Similar results were obtained with the FLIM system for both HET and SEG with excellent cell viability (>85%).

Figures 3(c–f) presents representative images of SEG cell lines after analysis for lifetime and oxygen levels. The cell positions (c) overlaps with RTDP fluorescence (d). Note that the fluorescence intensity image (D) is non-uniform, indicating that the cells differentially absorb RTDP. The lifetime map (e) generated from the intensity map (d) indicated a clear uniformity across different cells, signifying similar oxygen levels and reflecting intensity independence of lifetime. RTDP calibration was used to generate an oxygen distribution map (f) that implied uniform oxygen levels of approximately 285 µM. The oxygen levels were found to be [O₂]_SEG=260.16±17 µM. While higher than reported in vivo values, these levels were comparable to values generally observed in living biological samples during ex vivo measurements [13, 17]. This is a leap forward from measurements of extracellular oxygen alone, which assume linear correlation with intracellular oxygen. Gradients between intracellular compartments and extracellular space are known to exist but cannot be explained by simple diffusion rates alone [20].

Time-lapse FLIM measurements were made on SEG under airtight conditions to verify that decrease in oxygen levels due to cellular consumption would be reflected in RTDP lifetime increase. The results are plotted in Fig. 3(g). As expected, RTDP lifetime increased by 35 ns over the course of an hour, reflecting a decrease in oxygen levels (Δ≈-50 µM) due to gradual consumption by cells. These experiments demonstrate the sensitivity of FLIM for oxygen sensing in living cells.

RTDP calibration was applied to measure quantitative differences in intracellular oxygen between HET and SEG. Figures 4(a–f) presents comparative images of HET (a,b,c) and SEG (d,e,f) incubated with RTDP. Intensity images (a,d) do not yield any useful information. Lifetime maps (b,e) are plotted on the same bicolor lifetime scale to better illustrate disparities: the SEG clearly exhibit lower lifetimes than the HET, which in turn translates to higher oxygen levels (c,f). This trend was consistent, with [O₂]_HET=182.08±9.38 µM and [O₂]_SEG=260.16±17 µM (Fig. 4(g)). The difference between [O₂]_HET and the extracellular media (222 µM) was within the range of gradients reported in literature for various other cell models [20, 21]. Interestingly, the opposite trend is observed in the SEG. The difference between HET and SEG for the images provided was approximately 60 µM.

 

Fig. 4. RTDP fluorescence intensity (a,d), lifetime in ns (b,e) and oxygen in µM (c,f) maps of HET (a,b,c) and SEG (d,e,f). The intensity images (a,d) could not be reliably used to discriminate between the two different cell lines. The binary lifetime maps (b,e), on the other hand, plainly indicate different lifetimes for these two cellular species, with the SEG recording lower lifetimes than the HET. For the given case, τ_HET=225 ns and τ_SEG=170 ns. The mean lifetime difference was found to be Δ_τ=44±7.48 ns. Logically, this translated into higher oxygen levels in the SEG vs. the HET using the calibration derived earlier, as can be seen in the oxygen distribution maps (c,f). (g) illustrates the differences in oxygen levels within the HET and SEG as measured over multiple runs and assessed using the RTDP calibration. The mean difference between the two cell lines was hence evaluated as Δ_[O]2=78±13 µM and this value was statistically significant (p<0.001).

Download Full Size | PDF

4. Discussion

Patients diagnosed with esophageal cancer have one of the lowest 5-year survival rates (14%) [1]. Several reports describe the use of endogenous tissue fluorescence (e.g., autofluorescence from NAD(P)H, collagen, FAD) as a biomarker for dysplasia in the esophagus, and recent studies have corrected for signal distortions from other sources (e.g., hemoglobin absorption) to yield quantitative trends (differences) between fluorescence signatures of NAD(P)H and collagen in normal and dysplastic tissue in vivo [2, 22, 23]. Thus, it was possible to observe consistent changes in absolute NAD(P)H and collagen fluorescence levels of multiple patients and accurately correlate them with the degree of neoplastic change [2]. A key strength of tissue spectroscopy is the acquisition of data in the appropriate local environment. Although cellular NAD(P)H is a marker of esophageal cancer progression, the exact mechanisms are unknown in part due to the poor spatial resolution of fiber-optic based spectroscopic measurements. Thus, spectroscopy is unable to distinguish between cellular NADH in mitochondria and NADPH in the cytoplasm, due to their spectral similarity (hence the term NAD(P)H). Both these components, however, have contrary roles: NADH takes part in the electron transport chain in mitochondria to generate ATP, while NADPH is the reducing agent for anabolic reactions such as fatty acid synthesis that consume ATP. This article reports results from a dual confocal-FLIM approach towards measuring NAD(P)H fluorescence in cell models characteristic of normal and adenocarcinoma esophageal state (HET and SEG, respectively). High resolution confocal images of NAD(P)H and Mitotracker fluorescence (Fig. 2(b,c)) provided sub-cellular imaging of co-localization of these two signals, indicating that endogenous fluorescence arose primarily from mitochondria and was NADH. Using FLIM, differences in NADH fluorescence intensity were observed between HET and SEG, while there was no significant variation in NADH lifetime between cell lines (Fig. 2(f)), indicating that differences in fluorescence intensity were mainly attributed to differences in NADH concentration, and did not arise from fluorescence quenching [5]. Intensity results were corroborated with the confocal system (Fig. 2(d,e)) and confirmed that SEG contained significantly higher NADH levels than the HET, a trend similar to and supporting clinical observations [2].

Elevated NADH levels in SEG raises the question whether it is being excessively produced during glycolysis or consumed less due to possible mitochondrial dysfunction. Mitochondria provide a more logical starting point for further analysis: they are well characterized, complex organelles which play key roles in several cellular functions such as fatty acid metabolism, calcium homeostasis and most importantly, ATP production via the oxidative phosphorylation (OXPHOS) pathway (Fig. 2(a)) [24]. It is clear that defects/alterations in OXPHOS enzymes have links to some forms of cancer, neurogenerative diseases, diabetes and aging [24, 25]. The first step in OXPHOS involves the reduction of naturally fluorescent NADH to its non-fluorescent analog NAD⁺ via the membrane-bound complex I enzyme. Given the sequential nature of OXPHOS, it is logical that variations in NADH will affect the rate of downstream processes; i.e., consumption of oxygen and generation of ATP. ATP is the energy currency of the cell but is not an intrinsic fluorophore and commercial fluorescence indicators for ATP do no differentiate between ADP and ATP, making analysis difficult [26]. Oxygen, on the other hand, is likely the most pertinent indicator of completion of cellular respiration, in general, and OXPHOS, ATP production, in particular. Experiments conducted on HET and SEG indicated the potential of RTDP as an oxygen sensitive ruthenium dye that was well tolerated by living cells and yields a bright fluorescence signal to provide a molecular capability for analyzing intracellular oxygen levels in these cell models (Fig. 3(a–f)). After calibrating the oxygen sensitivity of RTDP lifetime via FLIM, modulation of intracellular RTDP lifetime was demonstrated via the gradual depletion of oxygen supply to SEG (Fig. 3(g)). Using FLIM, it was estimated that [O₂]_SEG=260.16±17 µM under cell culture conditions. Oxygen levels in HET were lower than SEG, observed as [O₂]_HET=182.08±9.38 µM, bearing an average difference Δ[O₂]_SEG-HET≈80 µM (Fig. 4(a–g)). This difference was statistically significant (p<0.05). It is important to note that cells in culture are exposed to higher oxygen levels than in vivo (especially for cancer cells which normally inhabit a hypoxic environment in vivo) and their behavior under these conditions can be meaningfully interpreted to understand causal behavior. This comparative tactic of tracking differences between normal and altered cell lines is easily extensible to other disease models and/or analytes other than NADH and oxygen.

In this study, one concern might be the validity of applying an oxygen calibration derived in PBS solution to measurements made in living cells, including how other intracellular factors might affect RTDP lifetime. Oxygen sensitivity of RTDP is widely accepted and exploited [7]. It has been demonstrated that RTDP lifetime is unaffected by solvent pH and ionic concentration (including Na⁺, Ca²⁺, Cl^-) within physiologic ranges, but is dependent on temperature, which is controlled by the FLIM approach [17]. Solvent viscosity is also known to affect fluorophore lifetime, not only by changing the microenvironment but also by altering diffusion rates of both fluorophore and quencher [7]. In one study, RTDP lifetime changes were reported over a range of solvent viscosities that is extreme in comparison to that encountered in biological systems (liquid media vs. rigid glass). Combined with previous studies, results presented here indicate that it is unlikely for viscosity to vary between HET and SEG or between cell cytoplasm and culture media/PBS. Localized binding events which may influence RTDP lifetime were also not observed (Fig. 3(a,b)).

The phototoxicity of RTDP in living cells was extensively documented previously [6]. The protocol used in that study included supplementation of the culture media by a solution of RTDP immediately prior to 457 nm excitation in a Bio-Rad MRC1024 laser scanning confocal system. Based on the data provided, we estimate that each illumination delivered approximately 0.8 J/cm² of energy on the sample. Depending on the concentration of RTDP used, photodamage to living cells occurred when the cumulative light dose exceeded 20–80 J/cm² (2-0.2 mM). In contrast, the protocol employed here involved incubating the cells with 1.3 mM RTDP stock solution for 6 hours and washing away the dye with PBS prior to experimentation. The intracellular concentration of RTDP, while not measurable, was clearly lower than the stock solution as inferred from intensity based fluorescence measurements. For the study described here, the laser pulse energy at the sample was 15 µJ and the illumination area was approximately 0.5 mm in diameter. Hence, the energy per unit area was 1.9 mJ/cm² for a single measurement. Further, the lifetime analysis protocol required intensity images captured at six different delays, at five measurements per delay. Therefore, it required thirty measurements to deduce lifetime, resulting in a total energy exposure of 0.057 J/cm² at the sample. Lastly, six lifetime measurements were taken for oxygen depletion experiments and the cumulative energy deposited was 0.34 J/cm². This is significantly lower than the threshold observed previously [6]. Coupled with the extremely low levels of RTDP being excited, we reasonably infer that the FLIM-based experiments presented here do not cause photodamage.

The studies presented here confirm that optical measurements of fluorescence from endogenous mitochondrial NADH might be useful for non-invasively detecting altered metabolic function in esophageal cancers in vivo. A noteworthy result is that no significant NADH lifetime differences were observed between HET and SEG. One approximation made in this work, given that different forms of NADH exist (free and protein-bound, conformational variations), is the use of single-exponential analysis for NADH lifetime. This is, however, reasonable, since a) a clinical instrument measuring NADH fluorescence lifetime could use similar analysis, and b) fluorescence observed in this work arises mainly from within mitochondria and is likely the protein-bound form of NADH, which is also the NADH subset that enters the OXPHOS pathway and has lifetimes similar to those reported here.

Use of exogenous agents such as RTDP, provide a complementary, highly specific approach for quantitative, intracellular oxygen sensing in living cells. Results presented here indicate that a possible alteration in cellular metabolism in SEG early in the OXPHOS pathway could cause decreased NADH consumption (higher intracellular levels), which would then lead to lower oxygen consumption downstream and hence elevated intracellular oxygen levels relative to HET, as observed. While reports on basal cell carcinoma indicate that OXPHOS complex I gene and/or protein modification might result in impaired activity, no such report was found for esophageal cancers [27]. These experiments also provide avenues for further exploration: it might be useful to ascertain whether the parallel CoQ generating pathway via complex-II is also affected (impaired, or upregulated to compensate). Specifically, measurement of both NADH and FAD fluorescence would allow for redox ratio calculations which could provide further support for impairment of OXPHOS.

The flexibility of FLIM to detect endogenous cellular fluorescence from NADH, as well as exogenous fluorescent tags and labels, makes the method compatible with endoscopic clinical imaging studies in living human tissues. A large temporal dynamic range and a microscopy imaging modality provide the ability of measuring both fluorescence and phosphorescence with subcellular spatial resolution. High temporal resolution and single-shot imaging aid in resolving molecular events such as conformational and viscosity changes. Temperature and perfusion control of the cellular environment are beneficial for viable cell imaging under physiologic conditions and to observe cellular response to conditions such as hypoxia and heat shock. Further, thick biological specimens can be imaged in layers by rejection of out of focus light (optical sectioning) via structured illumination to yield 3-D distribution of analytes (e.g., oxygen gradients in thick tissues). The oxygen sensing capability of FLIM was recently exploited to provide one of the first few reports on oxygen gradients in microfluidic bioreactors containing living cells (submitted).

The approach described here employed time-resolved optical imaging to probe metabolic pathways in living cells using endogenous and/or exogenous fluorescence agents. A calibrated approach for quantitatively estimating oxygen levels in HET and SEG using a tunable, temperature-controlled FLIM was presented, with the potential to perform absolute oxygen measurements in living biological systems with a resolution of 8 µM over the entire physiological range (0–300 µM). Previous studies provided a calibration curve for RTDP dissolved in growth buffer, but measurements were performed at room temperature and were made using electrodes, which consumed oxygen and required constant stirring of the media [17, 28]. Given the temperature-dependence of lifetime, oxygen solubility, and cellular biochemical response, a controlled study like this provides more accurate results. Precise measurements of intracellular oxygen provide important information in areas such as pharmaceutical research or systems biology and may aid in a better understanding of systemic disorders and their control.