Localization and Relative Quantification of Carbon Nanotubes in Cells with Multispectral Imaging Flow Cytometry

An overview of the principle of the ImageStream device is given in Figure 1. It produces multiple high resolution images of each cell in flow, including bright-field (BF), dark-field (DF or side-scattered light), and fluorescence channel(s), as exemplified in Figure 2 with three different cells labeled with CNTs. The overlays of these channels are also shown, at first sight illustrating the correlation. Given that all the post-processing analysis is based on cell imaging, the selection of focused cells is crucial for the study (Figure 3). Actually, the presence of CNTs induces changes on both the contrast and the gradient compared to nonlabeled cells: focused cells, having internalized an important quantity of CNTs, tend to have a low rms gradient (which is usually the unique parameter used for the discrimination of focused versus unfocused cells) but also an important contrast, explaining why a biparametric plot is essential. On Figure 3 is shown the discrimination between unfocused (yellow area) versus focused cells (blue area). Cell-internalized CNTs are clearly distinguishable – particularly on BF appearing as strongly absorbing dark spots (see the effect of diverse labeling concentration, Figure 4). Moreover, the mean pixel signal on BF was analyzed, giving a relative quantification of the uptake: graphs on Figure 5 show its significant increase with the CNT-labeling concentration (results for about 7,000 cells per experimental condition). Second, quantification of CNTs is also possible through the use of masks, applied either on a restricted range of dark pixels of the BF or on the pixel of high intensity of the DF (see examples on Figure 6), corresponding to the spots resulting from the presence of CNTs. Thus, plotting the area of these masks allows highlighting as well as the increase of the uptake (exemplified on Figure 7 with the area of the mask on BF, selecting pixels ranging from 0-533). In addition, the correlation between spots on BF and DF can be evaluated by plotting the DF intensity versus the BF mean pixel signal (as in Figure 8) (mask-independent analysis) or the area of the mask on BF (Mask 1) versus the area on DF (Mask 2). Regarding the procedure created for the localization of CNTs inside the cells, Figure 9A displays the different masks that were created. Note that the robustness of the localization masks (based on erosion of default cell mask) must be checked visually on a collection of labeled and unlabeled cells. The same masks are used for all samples. Plotting the total area of black spots on the cell membrane versus that on the entire cell provides a score of internalization for cells which have been incubated for 20 hr at 37 °C in comparison to 2 hr incubation at 4 °C (see Figure 9B); it indicates that CNTs are mostly located on the membrane at 4 °C, whereas at 37 °C CNTs are internalized in more than 90% of cells. Finally concerning the fluorescence, this parameter was not used as indicator for the quantification of CNTs due to a too low correlation between fluorescent spots and spots on BF (and DF). Indeed, the correlation between the FITC intensity and the bright field mean pixel indicates a Pearson correlation coefficient of -0.2. On Figure 10, two cases are illustrated, in which the fluorescence signal matches properly or not with the bright field black spot. Figure 1. Overview of the ImageStream- general operating. The multispectral imaging system acquires up to 12 images/cell in three different modes: brightfield, darkfield, and fluorescence. The setup consists in 5 excitation lasers, including 405, 488, 560, and 658 nm. Light is collected from the cells flowing in the cuvette with different objective lens (in this study, the 40X objective lens was used) and relayed to the spectral decomposition element, which consists in a set of longpass filters in an angular array. The CCD camera collecting cell images is operated using a technique called time-delay-integration in order to avoid image streaking. Extended depth of field (EDF) is an option. Figure 2. Image gallery of cells labeled with carbon nanotubes (50 μg/mL) captured by ImageStream through diverse channels (Bright Field (BF), Dark field (DF), fluorescence (FITC), and their overlays). CNTs confined within intracellular lysosomes appear as large black spots on bright field images. Conversely they create bright spots on the dark field and on the fluorescence images. Figure 3. Selection of cells in focus. Bivariate dot plot representing root mean square (rms) gradient versus contrast on BF images allows the exclusion of unfocused cells. Here the yellow area represents unfocused cells, whereas focused cells are represented in blue. The unfocused cells are characterized by both low contrast and low gradient rms on the bright field image. In addition to rms gradient, the contrast feature is require to discriminate focused cells after their labeling with CNTs. Figure 4. Dose-dependent uptake of CNTs visualized by ImageStream. The number and area of black spots on the bright field images are increasing with the dose of CNTs during incubation with cells (ranging from 0-50 µg/ml). Note that the control cells are devoid of black spots. Figure 5. Relative quantification of CNT-uptake based on the bright field images. Histograms of the mean pixel feature applied to the mask "object" featuring the whole cell clearly show a dose-dependent behavior, which can be used for quantification. R9 range encloses the CNT-positive cells. Figure 6. Creation of masks selecting black spots on BF and bright spots on DF images. As CNTs appear as strongly absorbing dark spots on BF image, a threshold mask is created to select a restricted range of pixels with a low intensity comprised between 0-533 in this study (Mask 1, notation: Intensity(M01, BF, 0-533)). Conversely, a mask fitting with the pixels of high intensity (150-4095 in this case) on DF image (Mask 2, notation: Intensity (M06, DF, 150-4095)) is used to select bright spot on DF, corresponding to high scattering areas. Figure 7. Dose-dependent uptake of CNTs is demonstrated by quantification of the black spot area on BF images. Using threshold mask (mask 1) applied on BF, the area of black spots can be quantified for each cell, showing a dose-dependent distribution. Figure 8. CNT quantification methods based on BF (absorption) and DF (scattering) images are correlated. On the left, mean pixel on the BF image versus intensity on the DF image, show a good correlation for cells labeled with CNTs (50 µg/ml) (Pearson correlation coefficient r = -0.68). On the right, area of the mask fitting dark spots on BF versus area of the mask fitting bright spots on DF for cells labeled with CNTs (50 µg/ml). Pearson correlation coefficient is r = -0.83.Values from 7,000 cells are plotted. Figure 9. CNT localization in cells using specific masks. (A) Masks are created to fit a particular location of the cell, namely from left to right – the entire cell, the membrane only, and the inside of the cell. Another mask fits black spots thanks to a threshold intensity (Mask 1, notation: Intensity(M01, BF, 0-533)). (B) Assessment of the localization of CNTs can be obtained by measuring the area of black spots on different localizations of the cells using the localization masks described in A. Incubation with CNTs (50 µg/ml) was performed both at 37 °C and at 4 °C to inhibit internalization pathway. The graph representing black area on the membrane versus black area on the entire cell shows a dose-dependent absorption of CNTs on the membrane at 4 °C, while the full internalization of CNTs is effective at 37 °C. Figure 10. Colocalization of CNT-induced black spots on BF with fluorescence bright spots on FITC image. The fluorescence bright spots are not always colocalized with the black spots on BF. They match properly for the first two cells displayed, but not for the last two cells.