Measuring Multiplexed Signal Transduction

Amnis’ Platform Was Developed to Assist in Assessment of NF-kappa-B Translocation

The common goal of many drug discovery efforts is the modulation of immune responses, for example by the inhibition of tumor necrosis factor-alpha (TNF-a). TNF-a is a potent activator of many cells of the immune system and plays a large role in the pathogenesis of asthma and of autoimmune diseases such as rheumatoid arthritis. TNF-a mediates its pro-inflammatory effect by inducing translocation of NF-kB from the cytoplasm to the nucleus.

In this article, we present a multiplexed method for the measurement of NF-kB translocation in THP1 cells, a monocytic cell line, in response to TNF-a stimulation at various doses and exposure times. The Amnis ImageStream platform employed in this study images cells directly in flow, making it well-suited to the analysis of both primary and cultured immune cells. Imaging flow cytometry generates over 100,000 cell images per minute to achieve high assay robustness even with rare sub-populations.

Nuclear Localization of NF-kappa-B

Assessment of NF-kB translocation is generally performed using nuclear and NF-kB fluorescence images of a cell. These images are used to define the nuclear and cytoplasmic compartments, followed by an assessment of the relative amount of NF-kB fluorescence in each compartment. While this approach works well for cells with low nuclear to cytoplasmic (N/C) areas, it is less reliable for immune cells, which have characteristically high N/C ratios and correspondingly small cytoplasmic areas. These limitations can be overcome by using an image-correlation algorithm that does not rely on an accurate definition of the cytoplasmic compartment. Assay robustness is further improved by imaging large numbers of cells for increased statistical measurement accuracy.

To measure NF-kB nuclear localization, THP1 cells are labeled with propidium iodide and Cy3 anti-NF-kB antibodies. Nuclear localization of NF-kB was measured on a per-cell basis using the Similarity score, a measure of the correlation between the NF-kB and nuclear image pairs.

The Cy3 and PI images for an untranslocated cell look different from one another and thus score low for similarity (Figure 1). As NF-kB translocates to the nucleus the Cy3 and PI images look more alike, increasing the Similarity score. The Similarity score is plotted for thousands of cells from two samples, one exposed to 10 ng/mL TNF-α (Sample 30) and the other to only 0.1 pg/mL.

Figure 1. Similarity scores for cells spanning the range from untranslocated to translocated are shown as well as histograms of the Similarity scores for two samples.

The percentage of cells in each sample that exhibit translocation is derived by setting a threshold and calculating the fraction of the distribution above the threshold. Alternatively, the position and shape of the distribution can be defined using a wide array of statistical metrics (mean, standard deviation). With the ability to image thousands of cells per sample, even subtle changes in sample response due to dosage or time can be detected by shifts in the percentage of translocated cells or the mean Similarity score of a given sample.

In order to assess the suitability of the ImageStream for high-content screening, we applied the method of Krutzik & Nolan to multiplex 64 THP1 cell samples, each having a different TNF-α dose and exposure time.

After treatment, the individual samples were stained with three different Alexa Fluor succinimidyl ester dyes (AF405, 488, 660), each at one of four intensities, to yield 64 unique dye/intensity combinations. The samples were then mixed, labeled with a rabbit anti-NF-kB polyclonal antibody, a Cy3-anti-rabbit secondary antibody, and propidium iodide, before being run on the ImageStream. Six doses at five time points were tested in duplicate, plus two negative control samples. Imagery of >100,000 cells (about 600,000 images) was acquired in approximately 10 minutes.

Figure 2. Histogram shows resolution of four AF660 intensity levels within one sample, and a scatter plot showing resolution of 16 samples using AF405 and AF488

Figure 2 demonstrates the ability to resolve the multiplexed cell samples. The intensity histogram resolves four comingled cell populations labeled with Alexa Fluor 660, and the AF405 versus AF488 intensity scatter plot demonstrates the ability to resolve 16 samples. Extending this to three dimensions with four intensities for three dyes results in resolution of 64 samples per well.

Figure 3 includes a heat map table of the mean Similarity scores calculated from each of the multiplexed samples and organized by dose (vertically) and time point (horizontally).

Figure 3. Heat map of mean Similarity score for each sample after deconvolution

Each cell of the table is also color coded by mean Similarity score, with high Similarity scores shown in red and low scores shown in green. As expected, the highest Similarity scores corresponded to the highest doses and latest time points. Figure 3 also shows example imagery from two of the multiplexed samples.

The mean Similarity values were also plotted as dose-response curves (Figure 4), with each curve generated using a different TNF-α exposure time. Doses above 100 pg/mL and exposure times longer than 15 minutes caused significant NF-kB translocation.

The ImageStream platform can perform 64X multiplexing of NF-kB translocation assays in nonadherent cells through the use of intensity coding. The recent availability of an autosampler for the ImageStream system now allows the independent measurement of translocation within 64 samples per well, for up to 6,144 samples per 96-well plate.

Figure 4. TNF-alpha dose response curves for various exposure times



Shannon Henery ([email protected]) is scientist and Thaddeus George is director of biology at Amnis.