2.1.

Cell Culture

Six breast cell lines were used in this study; two were normal mammary epithelial cells [MCF10A and human mammary epithelial cells (HMEC)] and four were breast cancer (MDA-MB-231, MDA-MB-361, MDA-MB-435, and MCF7). All cell lines, except for HMEC, were obtained from the American Type Culture collection (Manassas, Virginia). HMEC primary cells were obtained from Lonza (Basel, Switzerland) and infected with a retrovirus encoding human telomerase reverse transcriptase for immortalization. All cells remained free of contaminants and were propagated by adherent culture according to established protocols.⁴⁴ All breast cancer cell lines were cultured in $\alpha$ -MEM (Minimum Essential Media, Gibco, Carlsbad, California) supplemented with 6% fetal bovine serum, 1% hepes, 1% nonessential amino acids, 1% of $100\mathrm{mM}$ sodium pyruvate, $10\mu \mathrm{g}∕\mathrm{mL}$ insulin, $10\mu \mathrm{g}∕\mathrm{mL}$ hydrocortisone, and $5\mu \mathrm{g}∕\mathrm{mL}$ epidermal grown factor (EGF). All normal mammary epithelial cell lines were cultured in mammary epithelial basal media (MEBM) (Lonza, Basel, Switzerland) with 0.4% bovine pituitary extract, 0.01% hydrocortisone, insulin, and human EGF. Plated cells were incubated at 5% $\mathrm{C}{\mathrm{O}}_2$ and cultured every $3–4\text{days}$ . Cells were double washed in $3\mathrm{mL}$ of phosphate buffer saline and detached from flasks with 0.25% trypsin before centrifugation.

After $3\min$ of centrifugation at 120 relative centrifugal force (rcf), a standard hemocytometer was used to count the number of cells per milliliter. Approximately 200,000 cells were plated on a $35\text{-}\mathrm{mm}$ coverslip dish (Mat-Tek, Ashland, Massachusetts) to be tested.

2.2.

Confocal Microscope and Imaging Parameters

All confocal images were collected at the Duke University Light Microscopy Core Facility on a Leica SP5 laser scanning confocal microscope (Wetzlar, Germany). All cells were imaged in a temperature-controlled $(37°\mathrm{C})$ live cell chamber. A $405\text{-}\mathrm{nm}$ diode laser source with a mean power of $2.4\mathrm{mW}$ at the sample plane and scanning rate of $400\mathrm{Hz}$ was coupled to an inverted Leica DM16000CS microscope with a 40X oil-immersion objective (Leica Plan NeoFluor, $\mathrm{NA}=1.25$ ). The imaging field of view was $242\times 242\mu \mathrm{m}$ with $512\times 512\text{pixels}$ per image. An acousto-optic beamsplitter (AOBS) served as a dichroic mirror to allow the $405\text{-}\mathrm{nm}$ excitation light to reach the sample and to allow selection of emission wavelengths. Cooled photomultiplier tubes (PMTs) were used to measure fluorescence through a $121\text{-}\mu \mathrm{m}$ pinhole in a confocal arrangement with a theoretical axial resolution of $2\mu \mathrm{m}$ .

The emission bandpass for integrated fluorescence intensity measurements was set to $610–700\mathrm{nm}$ , which was chosen from preliminary spectroscopy data on cells that showed a fluorescence peak at $630\mathrm{nm}$ , corresponding to PpIX, as published in the literature.⁴² Images were taken at two separate fields of view and line averaged (16X per line) for improved image quality. Each fluorescence intensity image was acquired in $\sim 20\mathrm{s}$ per image ( $1∕400$ per second 16 averages per $\text{line}\times 512$ lines per image). Spectrally resolved fluorescence images at $405\mathrm{nm}$ excitation were collected at each of 34 emission wavelengths between 420 and $750\mathrm{nm}$ with a line average of 1. The AOBS had a collection bandwidth of $20\mathrm{nm}$ . A prism with a sliding mirror placed in the optical path prior to the PMT was scanned every $10\mathrm{nm}$ to ensure Nyquist sampling was fulfilled across all wavelengths. Total acquisition time was $43.5\mathrm{s}$ for all wavelengths in each spectral image ( $1∕400$ per second 1 average per $\text{line}\times 512$ $\text{lines}\times 34$ wavelengths per spectral image). Leica LAS AF 1.8.2 software (Wetzlar, Germany) was used for data acquisition. All images (intensity and spectral) were acquired with a gain of $1200\mathrm{V}$ , offset 0.7% and zoom of 1.6.

Prior to imaging, a calibration intensity image of standard fluorescent beads was obtained (FocalCheck Fluorescence Microscope Test Slide No. 2, Invitrogen, Carlsbad, California) to correct for daily variations in microscope throughput. The mean fluorescence intensity over all experiments was $208\pm 5\text{units}$ . Integrated fluorescence intensity images were normalized by dividing the measured intensity by the intensity of the calibration standard measured on the same day.

2.3.

Cell Imaging

Twenty-four hours after plating, the original cell media was removed from the culture dish, and cell lines were treated either with $500\mu \mathrm{g}∕\mathrm{mL}$ of ALA in standard cell media or standard media alone and incubated for $2\mathrm{h}$ prior to imaging. Because ALA is not fluorescent, it was not removed before imaging. The concentration of ALA (data not shown) was chosen experimentally. At $500\mu \mathrm{g}∕\mathrm{mL}$ of ALA, no phototoxic effects were seen in any cell line tested for this paper. Cell viability was verified by a trypan blue exclusion assay as determined by a previously tested protocol.⁴⁵ Viable cell concentration was then determined by counting the ratio of cells that were still viable (unstained) to all cells, and $89\mathrm{\%}\pm 6\mathrm{\%}$ of ALA-treated and $85\mathrm{\%}\pm 8\mathrm{\%}$ of untreated cells were found to be viable. Finally, $500\mu \mathrm{g}∕\mathrm{mL}$ of ALA was found to be within the range of concentrations used in previous cell culture studies. ^{18, 40}

To determine the optimal time to measure the fluorescence of PpIX (data not shown), a subset of cells [normal mammary epithelial MCF10A, MDA-MB-435 $(\mathrm{ER}-)$ , and MCF7 $(\mathrm{ER}+)$ ] was tested. The fluorescence intensity increased linearly from $0\text{to}6\mathrm{h}$ post-ALA treatment, as seen in previous studies,⁴⁰ and therefore, the peak intensity window was assumed to be $>6\mathrm{h}$ post-ALA treatment for all cell lines. The $2\text{-}\mathrm{h}$ time point was chosen because it is clinically practical and showed distinct differences between normal mammary epithelial and breast cancer cell lines (see Sec. 3).

The sample size for each cell line included 12 treated and seven untreated plates. Fluorescence intensity images were obtained from six of the treated and six of the untreated plates. Confocal spectral imaging was completed on six additional treated plates and a single untreated plate for each cell line. Sample size calculations for imaging of integrated fluorescence and confocal spectral imaging are summarized in Table 1 .

Table 1

Experimental sample size for confocal microscopy.

2.4.

Analysis of Integrated Fluorescence Intensity Images

Integrated fluorescence intensity images were analyzed using NIH ImageJ software. Integrated fluorescence intensity for each breast cell line was determined by manually segmenting each cell in the image and then computing the average fluorescence intensity. Background intensity was subtracted, and the resulting data normalized by the fluorescence of the calibration was standard. The background was defined as an area without cells that was approximately equivalent to the size of an entire cell ( $2000–5000\text{pixels}$ , depending on the cell line). Fluorescence intensity values for all cells within each image were then averaged to represent one sample (each image had approximately 10–50 cells within a field of view, depending on the cell line). A total of $n=12$ fields of view (two images per sample) were imaged, from which the average integrated fluorescence intensity per cell line was derived for both treated and untreated cells.

2.5.

Analysis of Spectrally Resolved Fluorescence Images

For spectrally resolved fluorescence images, a spectrum was calculated by the Leica software from an area of 3–5 adjacent cells ( $6000–\mathrm{15,000}\text{pixels}$ , depending on the cell line). Three groups of adjacent cells were manually segmented, and all spectra were averaged to obtain a single spectrum (three per image). A background spectrum was obtained from a cell-free region in each image ( $2000–\mathrm{90,000}\text{pixels}$ , depending on the cell-line growth pattern) and subtracted from each of the three spectra. These three spectra were imported into MATLAB (Mathworks, Inc., Natick, Massachusetts) to calculate Fractional PpIX contribution (FPC). (See Section 3 for further discussion of this calculation). The three FPC values from each image were averaged to obtain a single FPC per image. A total of $n=12$ FPCs (two images per sample) was calculated for all six treated cell plates, from which the average FPC for the cell line was calculated.

2.6.

Statistics

All statistics were computed with JMP (SAS, Cary, North Carolina) software. ANOVAs and t-tests were completed with a Tukey–Kramer correction for multiple comparisons to determine statistical significance. Exact two-sided $p$ -values were computed, and all $p<0.05$ were considered significant. All calculated correlations are reported with a Pearson coefficient.

3.1.

Integrated Fluorescence Intensity Images

Figure 1 shows representative integrated fluorescence intensity images of untreated and treated cell lines with different ER phenotypes. The untreated normal mammary epithelial cell line image has much higher pretreatment endogenous fluorescence than either breast cancer cell line, which is likely due to the endogenous fluorophore, FAD. Bright specks outside of the cells in the normal column are cellular debris caused by plating on untreated glass and were not included in subsequent analyses. Untreated $\mathrm{ER}-$ and $\mathrm{ER}+$ breast cancer cell lines qualitatively show similar distributions of endogenous fluorescence, which is diffusely distributed throughout the breast cancer cells prior to treatment and a few areas of high intensity around the nucleus. All cells showed an increase in PpIX fluorescence after $2\mathrm{h}$ of ALA treatment, as seen in the “Treated” row in Fig. 1. The normal mammary epithelial cell line shows bright points of PpIX fluorescence within an even distribution of cellular fluorescence. Breast cancer cell lines show brighter fluorescence in the perinuclear cytoplasm, and although the edges of the cells are clearly visible in the post-treatment images, the edges are not as intense. PpIX fluorescence is only in the cellular cytoplasm (as opposed to the nucleus) in all cell lines.

Fig. 1

Representative confocal integrated fluorescence intensity images (uncalibrated) of normal mammary epithelial (Normal) and breast cancer (Malignant) cells treated with ALA for $2\mathrm{h}$ . Excitation was at $405\mathrm{nm}$ , and emission was collected between 610 and $700\mathrm{nm}$ .

3.2.

Cellular Fluorescence Spectra

Figure 2 shows normalized representative fluorescence spectra from one untreated [Fig. 2] and one ALA-treated plate [Fig. 2] of normal mammary epithelial, $\mathrm{ER}-$ , and $\mathrm{ER}+$ cell lines at $405\mathrm{nm}$ excitation. Excitation at $405\mathrm{nm}$ is expected to elicit fluorescence from FAD and PpIX.⁴⁶ Normalized representative spectra in Fig. 2 show a single broad emission, which is due to FAD, and spectra from treated cells show an additional distinct ALA-induced PpIX emission peak at $630\mathrm{nm}$ with a shoulder at $700\mathrm{nm}$ [Fig. 2], similar to that observed by others.⁴⁶ It should be noted that the tail of the FAD fluorescence spectrum overlaps with the PpIX fluorescence between 610 and $700\mathrm{nm}$ [Fig. 2], and therefore, it is assumed to contribute to integrated fluorescence intensity images collected over a bandpass of $610–700\mathrm{nm}$ .

Fig. 2

Representative normalized fluorescence spectra at $405\text{-}\mathrm{nm}$ excitation of (a) untreated cells and (b) ALA-treated cells. Spectra were normalized to the peak intensity.

3.3.

Quantitative Integrated Fluorescence Intensity Measurements

In Fig. 3 , the six cell lines were grouped in pairs by ER expression (normal mammary epithelial, malignant $\mathrm{ER}-$ , and malignant $\mathrm{ER}+$ ). All cell lines exhibited significantly greater fluorescence intensity following treatment $(p<0.05)$ . The normal mammary epithelial cell lines (MCF10A and HMEC) untreated fluorescence intensity was significantly greater than all untreated breast cancer (MDA-MB-231, MDA-MB-435, MCF7, and MDA-MB-361) cell lines [Fig. 3, $p<0.01$ ]. Calibrated fluorescence intensity of the ALA-treated HMEC cell line was significantly greater than the treated MDA-MB-231 breast cancer cell line, but not significantly different compared to the other ALA-treated breast cancer cell lines. Figure 3 shows that the percentage change in PpIX fluorescence intensity following treatment was significantly greater in all breast cancer cell lines as compared to both normal mammary epithelial cell lines by at least 150% [Fig. 3, $p<0.05$ ]. No significant differences in the percent change of fluorescence intensity were observed between or within the $\mathrm{ER}-$ and $\mathrm{ER}+$ breast cancer cell lines.

Fig. 3

Average and standard error of the (a) calibrated integrated fluorescence intensity (c.u.) per cell line $(n=12)$ and (b) percentage change in integrated fluorescence intensity of treated compared to the untreated controls $(n=12)$ . Asterisks denote that each ALA-treated breast cancer cell line has a statistically higher percentage change in integrated fluorescence compared to ALA-treated normal mammary epithelial cell lines (Normal) $(p<0.05)$ .

3.4.

Quantitative analysis of the fluorescence spectral data

Although the breast cancer cells demonstrated a significant percentage increase in fluorescence following ALA treatment, it may be clinically impractical to obtain a pretreatment measurement. Post-treatment intensity measurements alone were not useful in discriminating normal mammary epithelial cell lines from breast cancer due to cell-to-cell variability in endogenous FAD fluorescence. A method for differentiating between normal mammary epithelial and breast cancer cells using only post-treatment measurements would be desirable. Fluorescence spectral images that capture both FAD and PpIX fluorescence have the potential to directly address this problem as presented below.

The contribution of FAD fluorescence to the PpIX fluorescence was quantified and removed as shown in Fig. 4 . First, the spectrum between $520–600\mathrm{nm}$ and $700–750\mathrm{nm}$ (○) was fit with a fourth-order polynomial to approximate an untreated spectrum (×). All polynomials were shown to have a goodness-of-fit coefficient $\left(r\right)$ of 0.93. The fitted curve (×) was then subtracted from the measured data (—), which resulted in an FAD-subtracted PpIX spectrum (--). The area under the curve of the resulting PpIX spectrum was calculated over $610–700\mathrm{nm}$ and represents integrated PpIX fluorescence intensity. The area under the curve of the fitted FAD spectrum was calculated to represent the integrated baseline FAD fluorescence intensity. Then, FPC was calculated as the ratio of integrated PpIX and FAD fluorescence intensities.

Fig. 4

Representative spectral data (—), spectral data used in fit (○), polynomial fit (×), and the resulting PpIX spectrum (--).

Figure 5 shows the FPC for each cell line. All breast cancer cell lines have a significantly higher FPC than normal mammary epithelial cell lines. It is not surprising to note that the FPC and percentage change in fluorescence intensity, (Figs. 3 and 5), are significantly and positively correlated (Pearson $\text{coefficient}=0.92,$ $p<0.05$ ). No differences in FPC were observed between $\mathrm{ER}+$ and $\mathrm{ER}-$ cell lines. Therefore, it can be said that PpIX fluorescence is preferentially accumulated in all breast cancer cells studied here, regardless of ER expression.

Fig. 5

Average and standard error of the FPC. Asterisks denote that the FPC of each breast cancer cell line is significantly higher than those of both normal mammary epithelial cell lines $(p<0.05)$ .