High Throughput Method to Quantify Anterior-Posterior Polarity of T-Cells and Epithelial Cells

The virologic synapse (VS), which is formed between a virus-infected and uninfected cell, plays a central role in the transmission of certain viruses, such as HIV and HTLV-1. During VS formation, HTLV-1-infected T-cells polarize cellular and viral proteins toward the uninfected T-cell. This polarization resembles anterior-posterior cell polarity induced by immunological synapse (IS) formation, which is more extensively characterized than VS formation and occurs when a T-cell interacts with an antigen-presenting cell. One measure of cell polarity induced by both IS or VS formation is the repositioning of the microtubule organizing center (MTOC) relative to the contact point with the interacting cell. Here we describe an automated, high throughput system to score repositioning of the MTOC and thereby cell polarity establishment. The method rapidly and accurately calculates the angle between the MTOC and the IS for thousands of cells. We also show that the system can be adapted to score anterior-posterior polarity establishment of epithelial cells. This general approach represents a significant advancement over manual cell polarity scoring, which is subject to experimenter bias and requires more time and effort to evaluate large numbers of cells.


Introduction
Anterior-posterior (AP) cell polarity is defined as the asymmetric distribution of macromolecules and organelles in the x-y cell plane. This process, which becomes initiated by localized activation of receptors at the cell surface, is essential for diverse functions, including embryonic morphogenesis, tissue repair, and immune surveillance [1][2][3][4]. AP cell polarity is also exploited by HIV and HTLV-1 to form the virological synapse (VS), where virus cell-cell spread takes place [5,6]. Normal AP cell polarity is also disrupted by viruses [6][7][8][9]. For example, Barnard et al. showed that in T-cells, HTLV-1 blocks normal AP cell polarization while promoting AP cell polarization associated with VS formation [10].
The immunological synapse (IS) is a specialized cell-cell contact membrane junction formed by the interaction between a T-cell and antigen-presenting cell. Formation of the IS triggers activation of cell signaling cascades that promote AP polarization and contribute to activation, survival, and migration of T-cells [6,11,12]. During the process of AP polarization in T-cells, the microtubule-organizing center (MTOC) changes from a random orientation to an orientation near the IS of the cell and therefore represents a useful marker to score AP cell polarity [5,12,13].
Previous studies have scored AP cell polarity establishment in microscopic images using a time consuming and laborious manual process [5,10,[14][15][16][17][18]. To improve upon this method, we have developed a new automated, high throughput technique to score AP polarity establishment in both T-cells and epithelial cells. This automated system should facilitate all studies of AP polarity, as well as those directed at investigating how viruses perturb and exploit this important cellular process.

Results
To induce IS formation and AP polarity establishment, we chose the established method of incubating human Jurkat T-cells with anti-CD3/CD28 coated beads [5]. Microscopic imaging of cell:bead conjugates requires attachment to a slide or plate. While centrifugation of non-adherent T cells is often used for this purpose [5], this method causes Jurkat cells to have an abnormal morphology and irregular size (data not shown), which interferes with automated analyses detailed below. This problem is circumvented by allowing cell:bead conjugates to settle by gravity onto plates, thereby retaining normal Jurkat cell morphology. Binding of the anti-CD3/CD28 bead to the Jurkat cell subsequently induces AP polarity establishment characterized by redistribution of the MTOC near the IS formed at the bead-cell contact point. Thus, to assess cell polarity establishment, we must visualize the MOTC, which is stained with anti-pericentrin antibody [19]. DAPI staining additionally permits visualization of nuclei, whereas beads are detected directly due to their red color. The automated high throughput system then captures microscopic images of cell fields and uses the relative locations of the MTOC, nucleus, and bead of each cell:bead conjugate to calculate a MTOC:bead angle. Finally, MTOC:bead angles output from the automated system are analyzed to score T-cell polarity establishment for a cell population. The overall procedure is outlined in Figure 1A.
In the first step of the automated system, confocal microscopy images are collected and compiled by a Pipeline Pilot algorithm. The algorithm determines the eligibility of each cell for MTOC:bead angle determination based on two strict criteria. (1) The cell must contact only one bead and must not contact another cell to ensure that the cell receives only one polarization signal. (2) The cell must contain only one MTOC. Cells meeting both criteria are subjected to the second step of the automated system, in which a reference line, designated the 0 line, is drawn between the center of the nucleus and the center of the bead attached to this cell ( Figure 1B). The system also draws a second line, designated the MTOC line, between the center of the nucleus and center of the MTOC. In each cell, the 0 and MTOC lines intersect at the center of the nucleus to define the MTOC:bead angle, which is the smallest angle formed between the two lines and which is accurate to four significant digits, with the dimension of a pixel representing the limiting factor. The Pipeline Pilot algorithm additionally allows one to view compiled images showing the MTOC:bead angle for each cell ( Figures 1C,D). Images include the white cell mask, green-stained MTOC, blue-stained nucleus, and red-colored bead. The algorithm creates the cell mask to simulate the cell perimeter based on information obtained about the unique size and shape of Jurkat cells.
To assess cell polarity establishment, MTOC:bead angles calculated for cell sample populations are compiled and sorted into four different 90 regions: region 1 includes two adjacent 45 sectors bisected by the 0 line; region 2 includes two separate 45 sectors adjacent to and on either side of region 1; region 3 includes two separate 45 sectors adjacent to and on either side of region 2; region 4 includes two adjacent 45 sectors next to region 3 ( Figure 1E). Thus, by definition, region 1 is closest to the bead whereas regions with higher numbers are progressively further from the bead. Sorted data are then subjected to an ANOVA analysis to determine whether the number of cells found in any region is statistically different from that of other regions. Theoretically, unpolarized cells should show equal (25%) distribution within regions 1-4, whereas polarized cells should show higher distribution within region 1 and concomitantly lower equivalent distributions in regions 2-4.
We first used the automated system to establish the kinetics of AP polarity establishment following incubation of Jurkat cells with anti-CD3/CD28 beads for 5, 10, 15, 20, 30, 60 or 120 min ( Figure 2). The results showed that the percentage of cells containing an MTOC in region 1 progressively increased from 0 min, peaked at 20 min, and remained approximately steady thereafter through the 120 min time point. As expected, we also observed that a smaller percentage of cells contained the MTOC in regions 2-4 than in region 1 at all time points (Figure 2). Differences between the numbers of cells containing an MTOC in region 1 and those with an MTOC in regions 2-4 were statistically significant for the 20, 30, 60 and 120 min time points. These findings show that T-cell polarization occurs rapidly, as it is detected as early as 5 min after addition of anti-CD3/CD28 beads. Also notable was that the automated system permitted imaging of large numbers of cells fitting our strict scoring criteria at each time point. For example, the lowest number of scored cells was 474 at the 5 min time point, presumably due to the short time available for cells to adhere to the plate surface and to attach to beads. In contrast, the number of scored cells increased to a maximum of 2,417 at 30 min of conjugation time. We considered scored cell numbers greater than 1,000 to be optimal for statistical         Results in nternal mon of 33.3% ex egion 3, sim egions 1 an significantly correspondin As was Figure 5D in cells had th Figure 5E). cell-bisectin differs from offset from validated th polarization

Discuss
We desc polarity esta criteria, calc addition, fle cell regions key features providing hi The vers Jurkat T-cel approximately equally (25% of cells/region) within each region, whereas polarized cells should instead distribute predominantly in the region encompassing the polarization stimulus (anti-CD3/CD28 coated bead contact site). Indeed, in Jurkat cells, we observed a significantly increased accumulation of MTOC in regions adjacent to the site of bead contact, an effect that was maximal after a 20 min incubation period. This finding led us to hypothesize that the MTOC migrates relatively rapidly to a position immediately adjacent to the cell polarization stimulus. We investigated this idea by re-sorting cells into smaller 20 or 30 regions. To our surprise, we found that the MTOC does not concentrate closest to the bead contact site, but rather to a 20-30 region on either side of this site (region 3 in Figure 3A or region 2 in Figure 3B). This finding led us to observe that many cells become dimpled at the point of bead contact, and to suspect that this altered cell architecture may restrict MTOC movement into this region. Also possible is that cellular factors or pathways restrict MTOC migration into this region or anchor the MTOC at a position offset from this region. Determining the molecular basis for this effect requires additional study.
To determine whether the automated system could be adapted to other cell types, we modified the algorithm to score polarity in adherent MCF-10A epithelial cells. At present, a reference point is manually added to the images of epithelial cells. To fully automate polarity scoring of migrating epithelial cells, we plan in the future to use an antibody to mark the leading edge membrane of wound edge cells and to develop an algorithm that places a dot at the center of the leading edge membrane. We succeeded in scoring polarity from large numbers of epithelial cells. Furthermore, contrary to polarized Jurkat T cells, the MTOC of polarized MCF-10A cells is known to concentrate in the region directly opposite from the polarizing stimulus. Indeed, we observed maximal MTOC accumulation in the region directly opposite from the polarizing stimulus, rather than offset from the predicted pattern seen in Jurkat cells. The difference between MTOC accumulation patterns in Jurkat and MCF-10A cells may depend on the physical nature of the stimulus itself as no equivalent to the bead-induced dimple in Jurkat T cells is seen in scratch-wounded MCF-10A epithelial cells. It is also possible that MTOC migration away from the polarizing stimulus in MCF-10A cells does not enforce the same physical restriction for MTOC movement postulated in Jurkat cells.
The new automated system also has important virological applications. The loss of cell polarity is a feature of epithelial-derived cancer cells, and there is increasing evidence that this defect in polarity may play a pivotal role in the pathogenicity of cancer [22]. In support of this idea, several oncogenic human viruses are known to target and affect cell polarity proteins [22][23][24][25]. For example the E4-ORF1 oncoprotein from human adenovirus (Ad) type 9 binds cell polarity proteins Dlg1, PATJ, and ZO-2 [22]; the E6 oncoprotein of high-risk human papillomavirus (HPV) binds cell polarity proteins Dlg1, PATJ, and Scribble [22,24]; and the Tax oncoprotein of human T-cell leukemia virus type 1 (HTLV-1) binds cell polarity proteins Dlg1 and Scribble [26]. Thus, our original goal was to develop an assay to facilitate scoring of polarity in T-cells and epithelial cells, the natural host cells of HTLV-1, Ad, and HPV. We propose that this automated system can be used to provide key mechanistic insights about how cells polarize and how viruses disturb this process to alter cell growth and spread cell to cell. In summary, the high resolution use of this flexible, high throughput system will likely help to reveal new biologic mechanisms.

Jurkat T-Cell IS Formation and Immunofluorescence Assays
Exponentially growing Jurkat cells and Dynabeads conjugated to anti-CD3/CD28 (Invitrogen Dynal AS, Oslo, Norway, catalog # 111.31D) or anti-CD19 (catalog # 111.43D) antibodies were separately washed in PBS (Invitrogen catalog # 14190-144) containing 0.5% FBS (PBS FBS buffer). Washed Jurkat cells (8 × 10 4 ) mixed with beads (4 × 10 4 ) in 0.1 ml total volume of PBS FBS buffer were added to each poly-L-lysine-coated well of a 96-well optical grade plate (Greiner Bio-One, Monroe, North Carolina, catalog # 655892P) and allowed to settle for 20 min at 37 C in a tissue culture incubator. After aspiration of the PBS-FBS buffer, cells were fixed in 3.5% paraformaldehyde (PFA) in PBS for 10 min at room temperature (RT), rinsed, and permeabilized with 0.1% Triton-X 100 in PEM buffer for 30 min at RT. Permeabilized cells were washed three times for 5 min, blocked in filtered IF buffer containing 10% goat serum for 1 h at RT, and incubated with rabbit anti-pericentrin (1:1000) (Abcam, Cambridge, MA, catalog # ab4448) in KB buffer for 1 h at 37 C in a tissue culture incubator. After incubation with the primary antibody, cells were washed three times for 5 min, incubated with goat anti-rabbit AlexaFluor-488 (1:5000) (Invitrogen, Carlsbad, California catalog # A-11008) in KB buffer for 45 min at 37 C in a tissue culture incubator, washed three times for 5 min, counterstained with 1 µg/µL DAPI, and rinsed with PBS. Unless otherwise indicated, rinses and washes mentioned above were done with PEM buffer (80 mM PIPES, pH 6.8, 5mM EGTA, pH 7.0, and 2 mM MgCl 2 ). Images were collected on a Beckman IC100 high throughput microscope at the Baylor College of Medicine Integrated Microscopy Core.

Epithelial Cell Scratch Wound Assays and Immunofluorescence Assays
MCF-10A cells were seeded at 3 × 10 5 cells/well on 2-well glass slides in complete medium and, upon reaching confluency, were incubated overnight in complete medium lacking EGF. The cell monolayer was then scratch-wounded five times per well using a P200 pipette tip, washed with complete medium lacking EGF to remove detached cells, and incubated in complete media for 6 h. Immunofluorescence assays were conducted as described for Jurkat cells, except PFA was quenched with 7.5% glycine solution prior to blocking with goat serum, goat anti-rabbit AlexaFluor-488 (1:500) (Invitrogen, Carlsbad, California catalog # A-11008) secondary antibody was used, and images were captured on a Zeiss Axioplan upright microscope or a Nikon A1 confocal microscope at the Baylor College of Medicine Integrated Microscopy Core.

High Throughput Image Collection, Preprocessing, Segmentation, Filtering, and Angle Measurement Epithelial Cell Scratch Wound Assays and Immunofluorescence Assays
Background subtracted from images was approximated as m + m 1/2 , where m is the minimum pixel intensity of the image. Objects (sets of contiguous foreground pixels) in different channels were detected using a local threshold approach in which an image pixel is placed in the foreground if its intensity is greater than the mean intensity in a surrounding window plus some constant, C, which was empirically assigned based on its ability to exclude non-objects and retain real objects. Holes in objects were filled to make them solid. Any objects touching the border of an image field were excluded. Nuclei were detected in the DAPI channel using a window of 11 × 11 pixels and C = 3; MTOC were detected in the GFP channel using a window of 7 × 7 pixels and C = 10; and beads were detected in the red channel using a window of 7 × 7 pixels and C = 20.
Cell boundaries were approximated from the nuclear objects using a distance map, whose pixels have values based on their distance in pixels to the nearest nuclear foreground pixel. Contiguous pixels in the distance map with a value of five or less were used to define cell objects. Beads that did not overlap with cells were excluded. Next, the centers of all objects in an image field were calculated, and these measurements were used to associate a cell with the closest bead and MTOC. Cells associated with more than one bead were also excluded. For the remaining cells, the MTOC:bead angle, defined as the smallest angle between a line from the center of the nucleus to the center of the bead (0 reference line) and a line from the center of the nucleus to the center of the MTOC (MTOC line), was calculated. Automated image preprocessing, segmentation, filtering, and measurement were performed in Pipeline Pilot 8.0 using a custom-built protocol. The custom code for the Pipeline Pilot software (Accelrys) used to score T-cell and epithelial cell polarity will be provided upon request through a Materials Transfer Agreement (MTA) between the end-user institution and the Integrated Microscopy Core of Baylor College of Medicine.
The above protocol was modified slightly to process adherent epithelial cells, which were present at a higher density than Jurkat cells. Nuclear objects from the DAPI channel were detected using a local threshold with a window size of 60 × 60 pixels and C = 5. To identify overlapping nuclei, region splitting was applied to the objects by calculating an inverse distance transform on the thresholded image (that is, measuring a pixel's distance to the nearest non-foreground pixel) and then finding peaks in the resulting distance map, where each peak corresponds to a nuclear seed. Starting from these seeds, a watershed transformation was run on the distance map to identify overlapping nuclei, which were discarded from the analysis using a shape filter. MTOC were detected as described for Jurkat cells. Leading edge cells were located visually from the DIC channel and then labeled manually with a red dot to mark the center of the leading edge. The centers of the nucleus, MTOC, and leading edge were calculated and used to determine the MTOC:leading edge angle.

Sorting MTOC:Bead Angles and MTOC:Leading Edge Angles into Designated Regions
Angles generated by the algorithm in Pipeline Pilot were sorted and analyzed in Microsoft Excel.

Statistics
ANOVA with Tukey post-hoc analysis using R statistical software was used.

Conclusions
This paper describes a new automated, high throughput system to score cell polarity establishment by capturing cell images, applying strict criteria to exclude unwanted cell images, compiling data, and calculating MTOC:bead angles. This flexible system permits sorting and evaluation of MTOC positions within cells using a variety of cell sectoring strategies. The system rapidly analyzes thousands of cells, facilitating acquisition of statistically significant results from very small regions of the cell and high resolution insight into the establishment of cell polarity. The approach has a variety of applications, including visualization and analysis of the virologic synapse and/or the role and effect of viral oncoproteins on polarity, and can be adapted for use in other systems and cell types.