Single-cell nonphotochemical hole burning of ovarian surface epithelial carcinoma and normal cells

Abstract

Persistent spectral nonphotochemical hole-burning (NPHB) spectroscopy has recently been applied to dye molecules in cells. The sensitivity of NPHB to the nanoenvironment of the probe is well established. It has been shown that NPHB applied to bulk suspensions of cultured human cells can distinguish between normal and cancer cells. Thus, NPHB has potential as a diagnostic cancer tool. For this reason, the methodology is referred to as hole-burning imaging, by analogy with MRI. The optical dephasing time (T₂) of the dye in hole-burning image replaces the proton T₁ relaxation time in MRI. In addition to the T₂ mode of operation, there are four other modes including measurement of the spectral hole growth kinetics (HGK). Reported here is that the selectivity and sensitivity of NPHB operating in the HGK mode allow for distinction between normal and carcinoma cells at the single-cell level. The ovarian cell lines are ovarian surface epithelial cells with temperature-sensitive large T antigens (analogously normal) and ovarian surface epithelial carcinoma (OV167) cells. The mitochondrial specific dye used was rhodamine 800 (Molecular Probes). This carbocationic dye is highly specific for the outer and inner membranes of mitochondria. In line with the results for bulk suspensions of the two cell lines, the hole-burning efficiency for OV167 cells was found to be significantly higher than that for normal cells. Theoretical analysis of the HGK data leads to the conclusion that the degree of structural heterogeneity for the probe–host configurations in OV167 cells is lower than in the normal cells. Possible reasons for this are given.

In seeking to expand on available diagnostic tools for the detection of cancer while improving the level of resolution, this group has previously introduced low-temperature intracellular persistent spectral nonphotochemical hole burning (NPHB) (1, 2) by using probe dye molecules (3–6). We refer to this technique as hole-burning imaging (HBI), by analogy with MRI (see below). Those works, which involved the study of bulk samples of different cell lines, established that NPHB of the inhomogeneously broadened S₀ → S ₁ absorption band of probe molecules in an intracellular environment is efficient and, moreover, that HBI can be used to distinguish between normal and carcinoma cells. In refs. 3 and 4, NPHB with the mitochondrial dye rhodamine 800 [MitoFluor Far Red 680 (MF680), Molecular Probes], which is selective for the outer and inner membranes of mitochondria, was used in two modes of operation to distinguish between normal and carcinoma human ovarian surface epithelial cells. We report here that the sensitivity and selectivity of NPHB allow for distinction between single normal and carcinoma ovarian cells.

By way of background, NPHB is a laser-based site excitation energy-selective spectroscopy that eliminates the large contribution of inhomogeneous broadening (a few hundred cm⁻¹) to the ground to excited-state origin and vibronic absorption bandwidths of a chromophore in complex (“glassy”) host matrices (7). For dye molecules with nanosecond excited-state lifetimes, the improvement in resolution at ≈4 K can be as high as 10⁴, as determined by the ratio of the inhomogeneous broadening of the origin absorption band to the width of a laser-limited hole burned into that band. Such a hole is referred to as a zero-phonon hole (ZPH) and corresponds to the zero-phonon line in absorption (7). The zero-phonon line in a solid is analogous to a pure electronic transition of a gas-phase molecule to which rotations and molecular vibrations do not contribute. For the above rhodamine dye (hereafter MF680) in ovarian cells, the ZPH widths at 2 K are ≈0.1 cm⁻¹, whereas the inhomogeneous broadening is ≈800 cm⁻¹ (3, 4). (It follows from theory that the width of the ZPH is twice the width of the zero-phonon line.) Glass-like structural disorder of the host, such as exists in glasses, polymers, and proteins, is an essential ingredient for NPHB. Production of a spectral hole is a consequence of tunneling between nearly isoenergetic bistable host–chromophore configurations that is triggered by electronic excitation of the chromophore. The bistable configurations are often referred to as two-level systems (TLS). On completion of the ground-state to excited-state to ground-state cycle, the chromophore is trapped in a nanoenvironment that is different from that before excitation. As a result, the absorption frequency of the chromophore is shifted to another position within the inhomogeneously broadened absorption band, thus the production of a ZPH at the laser burn wavelength λ_B. The ZPH can persist for months, provided the sample is kept at or below the burn temperature and in the dark to avoid light-induced hole filling. The resulting positive absorption at other wavelengths, referred to as the antihole, usually has a width on the order of the inhomogeneous width. The reader is referred to Reinot and Small (8) for a detailed discussion of the NPHB mechanism. Briefly, the TLS can be subdivided into those intrinsic to the glassy host (TLS_int) and those associated with the dye molecule and its inner shell of solvent molecules, extrinsic TLS (TLS_ext). Probe excitation results first in tunneling of TLS_int in the outer shell, which are intimately connected with the excess free volume of glasses (9). This tunneling leads to diffusion of excess free volume to the inner shell, opening the way for the rate-determining step of NPHB, which is phonon-assisted tunneling of the TLS_ext. The tunneling rate is written as R = Ω exp(−2λ), where λ is the tunnel parameter, and Ω is proportional to the square of the tunneling frequency of the TLS_ext (10). NPHB has been widely used for investigating the spectral dynamics of chromophores in glasses and polymers (11−18), due to tunneling of TLS_int and for excited-state electronic structure and excitation energy transfer/electron transfer processes of photosynthetic protein–chlorophyll complexes (19−24). NPHB is very sensitive to the nanoenvironment of the probe chromophore, and thus its use in differentiation between normal and cancer cells is a logical application.

Four properties of the probe molecules can be determined by using the ZPH. (i) Its optical dephasing time (T₂) that is temperature dependent and determined by the homogeneous width of the ZPH (25): T₂ replaces the T₁ proton relaxation time in MRI. (ii) The dispersive (nonsingle exponential) hole-growth kinetics (HGK) that reflect structural heterogeneity (10). (iii) The permanent dipole moment change (Δμ) of the S₀ → S₁ transition that is obtained by measuring the response of the ZPH (broadening and/or splitting) to an external electric (Stark) field (26). Here it is the matrix-induced contribution to Δμ that is of interest. (iv) The local compressibility (κ) that is determined from the linear pressure shift rate of the ZPH frequency (27). In addition, one can monitor the phonon (low-frequency intermolecular mode) sideband hole structure associated with the ZPH. It is the strength of the transition electron–phonon coupling that determines the relative intensities of the ZPH and phonon sideband hole structure.

The studies of Walsh and colleagues (3, 4) with MF680 used two human epithelial ovarian cell lines: one, a carcinoma cell line [ovarian surface epithelial carcinoma cells (OV167)] derived from a high-stage high-grade serous tumor, the most commonly diagnosed form of ovarian cancer; and the second, a line derived from normal ovarian surface epithelium (28) (see below). Specificity of MF680 for mitochondria was demonstrated by mean of confocal microscopy and spectroscopy. (MF680 is a carbocationic lipophilic dye that locates preferentially in mitochondrial membranes due to their lipid-enriched composition and lowered membrane potential relative to extramitochondrial areas.) Stringent cryopreservation protocol was used. Of the five aforementioned NPHB properties, it was found that NPHB operating in the HGK mode and Stark (Δμ) mode provided the highest degree of differentiation between the two cell lines. The Δμ value for the OV167 cells was found to be a factor of 1.5 larger than that of the normal cells. It was suggested that the difference may result from the membrane potential (Δψ_m) of the carcinoma cells being larger than that of the normal cells (3). The HGK data revealed that the OV167 cells burn significantly more efficiently than the normal cells. Theoretical analysis of the data revealed that the degree of structural disorder for the OV167 cells is the lower of the two, as is the average value of the tunnel parameter (λ_o). It was emphasized that HBI operating in the HGK mode has the advantage that only inexpensive diode lasers are required rather than tunable lasers required for the other modes of operation (4).

In this paper, HGK differences are demonstrated for single cells of the model cell lines, with trends similar to our previous observations for bulk suspensions containing 0.5−2.0 × 10⁶ cells, thus demonstrating the potential of HGK diagnostics with resolution at the single-cell level. As in refs. 3 and 4, the HGK curves were obtained by monitoring the decrease in fluorescence that accompanies burning of spectral holes.

Materials and Methods

Cell Culture.

Pertinent information regarding the choice of model cell lines was covered in a recent article (4). Only relevant growth culture information is given here.

Ovarian surface epithelial carcinoma (OV167) cells were grown in Alpha MEM Earle's salts with nucleosides (Irvine Scientific), supplemented with 20% FBS, penicillin/streptomycin (100 units/ml and 100 μg/ml, respectively) and l-glutamine (2 mM final concentration, resupplemented every 2 wk). For each hole-burning experiment, cells were cultured for 6–7 days (medium resupplementation on day 3) in a 25-cm² culture flask and then transferred to an 18-mm square glass poly-l-lysine treated coverslip for 24 h before staining and cryofixation. Cells were incubated at 37°C and 5.5% CO₂ atmosphere.

For analogously normal cells, short-term normal OSE cell cultures were infected with pZipSVtsA58, a retrovirus encoding a temperature-sensitive mutant of the simian virus 40 large T antigen; for simplicity, we will refer to them as OSE(tsT)-14. OSE(tsT)-14 cells were cultured in a 1:1 mixture of Medium 199 and MCDB105 medium mix, supplemented with 15% FBS, penicillin/streptomycin (100 units/ml and 100 μg/ml, respectively), and l-glutamine (2 mM final concentration, resupplemented every 2 wk). Cells were expanded at the permissive temperature of 34°C, 5.5% CO₂ for 6–7 days (medium resupplementation on day 3), transferred to a glass poly(l-lysine)-treated coverslip 24 h before use, and cultured at 39°C, 5.5% CO₂ for 12–18 h (allowing degradation of the heat-sensitive large T antigen) before staining and cryofixation.

Cells were stained by using MF680 (Molecular Probes) for 15 min at a concentration of 100 nM. MF680 was initially dissolved in DMSO. At concentrations used for staining (after diluting with PBS and subsequently growth medium), the DMSO component of the staining medium was negligible (0.0002%). OV167 and OSE(tsT)-14 cells were removed from the culture flask by using 2× and 1× trypsin/EDTA (2.5 g/liter/1 mM stock) solutions, respectively. For multiple runs, cells were used at the same passage number. Unless otherwise noted, all chemicals were purchased from Sigma.

Laser System.

A schematic of the fluorescence data collection system is shown in Fig. 1. In brief, a ring dye laser using LD 688 dye (Exciton, Dayton, OH) was pumped by a Coherent Innova 90-6 argon ion laser (Coherent, Santa Clara, CA). The laser system provided 100–500 mW of power over a wavelength region of 660–720 nm. Laser intensity was stabilized at 50 mW by an LS100 laser power stabilizer (Cambridge Research and Instrumentation, Cambridge, MA), and the laser beam was focused into a multimode fiber optic with a ×5 microscope objective. At the output end of the fiber optic, polyethylene sheets were used to reduce the spot power to ≈1 mW and to diffuse the laser light for uniform illumination of the entire area imaged. Laser intensity during burns was ≈250 μW/cm². Two low-pass filters, one a 720-nm cutoff filter (model no. 720ALP, Omega Optical, Brattleboro, VT) and the other a 750-nm cutoff filter placed between the cryostat window and charge-coupled device (CCD) served to remove excitation wavelengths. For burns, the laser was operated with intracavity etalons [linewidths <0.0003 cm⁻¹ (<10 MHz)] installed. Laser fluence for hole burning was varied with the use of a series of neutral density (gray) filters. Excitation wavelengths are noted in figures.

Figure 1.

Schematic of laboratory setup for imaging cells by laser-induced fluorescence and subsequent calculation of HGK. “F” refers to a low-pass filter to remove excitation frequencies. Inset shows detail of the sample holder/manipulator and reflecting objective/prism configuration.

Cryostat/Fluorescence Signal Collection.

Coverslips with adherent cells of both types in separate wells were mounted on a sample manipulator (Fig. 1 Inset) and plunged into a Janis cryostat (model STVP-100, Janis Research, Wilmington, MA) precooled to near liquid helium temperature (≈5 K) (29). The sample manipulator allowed for movement in x, y, and z directions. A ×25-reflecting objective was mounted in the cryostat, with the image redirected out of the cryostat via a prism. The fluorescence image was collected by using a CCD camera (model LN/CCD-1340/400-EB/1, Roper Scientific, Trenton, NJ) controlled by WINVIEW32 software (version 2.4.1.15, Roper Scientific). Images were collected at equivalent intervals, with identically sized regions of interest. Typically, 900 images were collected during a burn.

Results and Discussion

Fig. 2 shows a series of ZPH burned at 1-nm intervals with a constant burn fluence of 10 mJ/cm² for a bulk suspension of MF680 in OV167 cells. This ZPH “action” spectrum was obtained in the fluorescence excitation mode and is the difference between the post- and preburn absorption spectra. For burn wavelengths (λ_B) ≳ 704 nm, the fractional ZPH depths were found to be constant within experimental uncertainty. (The intensity fluctuations in Fig. 2 for λ_B ≳ 704 nm are due to the step size of the laser scan being about five times larger than the width of the ZPH.) The action spectrum of OV167 in the OSE(tsT)-14 cells is similar (29). Thus, for both cell lines, HGK curves can be measured at a large number of wavelengths by monitoring the decrease in fluorescence intensity as hole burning proceeds. The decrease in hole intensities at wavelengths shorter than ≈704 nm is due to the absorption being increasingly dominated by vibronic/phononic transitions rather than the pure electronic transition. The burn fluence of 10 mJ/cm² is too low to reveal the phonon sideband hole structure that accompanies the ZPH. Such structure evolves at higher burn fluences (4).

Figure 2.

A set of zero-phonon holes (the action spectrum) of MF680 in OSE(tsT)-14 normal cells. Each hole was burned at T = 2.0 K with a fluence of 10 mJ/cm² at 1-nm intervals; burning started at 675 nm proceeding toward 715 nm. The structure of the dye molecule, MF680, is shown (Inset).

A representative image of OV167 cells stained with MF680 is presented in Fig. 3. HGK curves were generated by collecting sequential fluorescence images, followed by analysis of specific regions of interest for signal intensity, as depicted by the white box superimposed on one of the cells in Fig. 3A. For each frame of a series of images, each pixel contained within the box was summed and plotted against frame number, resulting in a HGK “image” as shown in Fig. 3C. By this method, numerous cells were analyzed individually for comparison. For simplicity, only frame numbers are quoted, regardless of frame exposure time used (which corresponds to 1 s in the present case). However, exposure time does not include the time taken to read and report the CCD chip, which is a significant time contribution. For a frame with 1-s exposure, 0.1–0.4 s of additional time is added by the read and report process, depending on the size of the image frame (regions of the CCD chip can be ignored during data collection, thus the region of interest must be set the same for accurate comparisons). The data presented in this case have not been corrected for this additional time. Because all data presented here were collected under identical conditions by using identical frame sizes, this is not seen as a significant problem.

Figure 3.

Initial image (A) of OV167 cells stained with MF680 and burned at λ_B = 710 nm. The white bar represents 20 μm. In B, individual images of the same cell at increasing frame number are shown, exhibiting the burn over time. The cell shown corresponds to the white box shown in A. In C, a hole growth kinetics curve has been created from the individual cell shown in B by summing the pixels contained within the box superimposed over the cell image at frame 1 and repeated for each of the 900 image frames. Other conditions: T = 6.8 K; time for each exposure: 1 s; laser intensity (P) = 250 μW/cm².

Structural heterogeneity of amorphous solids leads to a distribution of values for the tunnel parameter λ of the extrinsic TLS. A Gaussian for the distribution function f(λ) has been shown to be physically reasonable (ref. 10 and refs. therein). The distribution center and standard deviation are denoted by λ_o and σ_λ. For each cell line, the HGK curves were fit (8) by using

1

where 1−D(t) is the fractional hole depth after a burn for time t at a burn frequency ω_B. P is the photon flux (number of photons cm⁻²·s⁻¹), α is the angle between the laser polarization and transition dipole of the chromophore, and σ is the peak absorption cross section [2.8 × 10⁻¹² cm² for MF680 (4)]. The NPHB quantum yield is given by

2

where Ω exp(−2λ) represents the phonon-assisted tunneling rate; Ω is set to 7.6 × 10¹² s⁻¹ (10); and τ, the fluorescence lifetime, was determined to be 1.8 ns at 77 K. The parameters λ and α each cause a distribution of hole-burning rates that leads to dispersive kinetics, and it has been shown that the λ distribution is by far the dominant factor for the first 80% of the total burn (8). σ_λ values for the λ_o parameter allow conclusions to be drawn regarding the relative order of the environment experienced by the chromophore, which in the present case consists of MF680 in the lipid-based mitochondrial membranes.

Representative curves fit by Eq. 1 are presented in Fig. 4, where the fits are superimposed over data curves. The value of P is given in the caption. λ_o and σ_λ were adjustable parameters. When fitting the kinetics data, it is necessary to know the Huang–Rhys factor (S), which defines the strength of the electron–phonon coupling, because the maximum depth of the ZPH is exp(−S) in the low-temperature limit. To obtain good fits that were also physically reasonable, an S value of ≈1.4 was required that is larger than the value of 1.1 determined in ref. 4 for both cell lines. In Fig. 4A, three HGK curves are presented for cells of each type of cell line, with the observation that the OV167 carcinoma cell line burns to a deeper fractional hole depth in the same number of frames (equivalent to the same amount of burn time) relative to the OSE(tsT)-14 normal analogs. This trend is in agreement with our previous observations on cells grown and analyzed in bulk. Noteworthy from Fig. 4A is that not all of the cells in the same cell line burn at identical rates. Rather, cells tend to show distinct rates by comparison, indicative of heterogeneities between cells in the same line but in keeping with the overall trends observed between the two cell lines.

Figure 4.

(A) Representative HGK data and fit curves generated for three of each cell line from subsequent images of OV167 carcinoma cells (C) and OSE(tsT)-14 normal analogs (N). In B, HGK curves from 15 different cells were summed and fit. P = 250 μW/cm²; T = 6.8 K.

Complete initial images of each cell line are presented in Fig. 5. Superimposed on individual cells are numbers corresponding to the fit parameters listed to the side of the image. From the fit parameters, heterogeneities in burn rates are evident, and further analysis of the fit parameters gives values for the OV167 carcinoma line of λ̄_o and σ̄_λ and values for the OSE(tsT)-14 normal line of λ̄_o and σ̄_λ. (The error limits are standard deviations of all of the measurements.) The values of λ̄₀ are statistically indistinguishable, indicating that the cellular environments of the dye in the two cell types are similar. On the other hand, the mean values of σ_λ for the two cell types are significantly different. At the 99% confidence level, the value for the OV167 cells would lie between 1.1 and 1.4, whereas the range for the OSE(tsT)-14 cells is from 1.7 to 2.0. However, for a single cell, there is about a 5% probability that the cell would be misidentified only on the basis of a measurement of σ_λ. One further point about the above averages is that the values of λ_o and σ_λ are uncorrelated [correlation coefficients = 0.2 for OV167 and 0.5 for OSE(tsT)-14, respectively]. In comparison to the single-cell averages, previously reported HGK fit parameters for bulk cell suspensions gave λ_o and σ_λ for OV167 cells and λ_o and σ_λ for the OSE(tsT)-14 cell line (4). The trend for lower σ_λ values in the carcinoma line is identical, but the values differ. This difference is likely contributed to by differences in experimental setups. The present method does not afford the signal sampling sensitivity that is capable with detection by a photomultiplier tube, as in ref. 4. In that study, HGK data points were collected every 0.1 s for the first 30 s, providing high resolution of the first 30–50% of the entire burn depth. After 30 s, data points were collected less frequently (1 s⁻¹). In the present case, however, images were collected every ≈1 s for the entire burn, limiting the resolution of the initial part of the burn, which is of primary importance in arriving at reliable values for σ_λ. As a result, a straight comparison of HGK fit parameters between bulk suspensions and single cells is not warranted. Another difference between the bulk and single-cell studies is that in the former, the samples were frozen under conditions that optimize cryopreservation, whereas in the latter, coverslips with adherent cells were frozen to optimize cryofixation (30). It is possible that the faster cooling rate with cryofixation traps a wider distribution of chromophore–host configurations, which translates into larger σ_λ values. Larger σ_λ values might also arise from straining of cells due to their adherence to the coverslips.

Figure 5.

Images of OV167 carcinoma and OSE(tsT)-14 normal cells and concurrent HGK fit parameters.

A second HGK comparison was also conducted by summing all of the kinetics curves for individual cells (Fig. 4B) and fitting the resulting curve, with the result of σ_λ = 2.0 and λ_o = 8.5 for the OSE(tsT)-14 cells and σ_λ = 1.2 and λ_o = 8.5 for the OV167 cells. These values are within the experimental uncertainty of the individual cell evaluations, exhibiting good agreement between the two methods used.

One familiar with Eq. 1 would immediately conclude from the experimental HGK curves in Fig. 4 (as well as those for bulk cell suspensions given in refs. 3 and 4) that the standard deviation σ_λ is smaller for OV167 cells, which indicates that heterogeneity of MF680 sites in the membranes of those cells is less severe than in the analogously normal OSE(tsT)-14 cells. At this time, we can only speculate on the reason(s) for this finding. One possibility emerges from the confocal fluorescence images presented in ref. 4. They showed that the OV167 cell line exhibits relatively small, more cobblestone-like shapes for the majority of the cells, and that the cells form highly aggregated structures relative to the OSE(tsT)-14 cells. This is presumably due, in part, to the very rapid proliferation rate of the carcinoma line. Long networks of mitochondria were not formed within the time frame used in those experiments. Such networks were observed in the normal cells, because with passage, these cells changed from the cobblestone morphology to a fibroblast-like morphology as characterized (31). The regular and aggregate cobblestone structure of the carcinoma line would seem to correlate with a decrease in structural heterogeneity. Another possibility is that the higher membrane potential of cancer cells relative to normal cells [by a factor of ≈1.5 (3, 4)] might lead to greater structural order due to enhancement of electrostatic interactions.

In conclusion, we have shown, to our knowledge for the first time, that persistent NPHB of a chromophore in single cells is feasible. In the HGK mode, HBI with the membrane-selective mitochondrial dye MF680 proved capable of distinguishing between single cancer and normal ovarian epithelial cells. However, a shortcoming of MF680 is that its hole-burning efficiency is too high, preventing us from using the other operational modes of HBI that require scanning of the burn laser to obtain hole-burned spectra. Efforts are underway to find less efficient dyes selective for specific organelles. We have recently shown that HBI can be applied to cancerous tissue (unpublished results). It will be interesting to see whether HBI can distinguish between normal and cancerous tissue.

Abbreviations

NPHB

nonphotochemical hole burning

HBI

hole-burning imaging

HGK

hole-growth kinetics

OSE(tsT)-14

ovarian surface epithelial cells with temperature-sensitive large T antigen

OV167

ovarian surface epithelial carcinoma cells

MF680

MitoFluor Far Red 680 (rhodamine 800)

ZPH

zero-phonon hole

TLS

two-level system

CCD

charge-coupled device