A Multiplexed Assay That Monitors Effects of Multiple Compound Treatment Times Reveals Candidate Immune-enhancing Compounds.

Abstract

We previously developed a flow cytometry assay that monitored lytic granule exocytosis in cytotoxic T lymphocytes stimulated by contacting beads coated with activating anti-CD3 antibodies. That assay was multiplexed in that responses of cells that did or did not receive the activating stimulus were distinguished via changes in light scatter accompanying binding of cells to beads, allowing us to discriminate compounds that activate responses on their own from compounds that enhance responses in cells that received the activating stimulus, all within a single sample. Here we add a second dimension of multiplexing by developing means to assess in a single sample the effects of treating cells with test compounds for different times. Bar-coding cells before adding them to test wells lets us determine compound treatment time while also monitoring activation status and response amplitude at the point of interrogation. This multiplexed assay is suitable for screening 96-well plates. We used it to screen compounds from the National Cancer Institute, identifying several compounds that enhance anti-LAMP1 responses. Multiple treatment-time screening enabled by bar-coding and read via high-throughput flow cytometry may be a generally useful method for facilitating the discovery of compounds of interest.

INTRODUCTION:

One of the major goals of basic immunology research is to identify targets that can be used as the basis of strategies to manipulate the immune system for therapeutic purposes. We recently reported a flow cytometric assay that can detect compounds that inhibit or enhance T cell receptor (TCR) dependent activation of cytotoxic T lymphocytes (CTLs) ¹, cells that kill virus-infected or transformed cells by releasing granules that contain cytolytic agents ². In the assay, cellular responses were measured by binding of fluorescently-labeled antibody against lysosome-associated membrane protein 1 (LAMP1, also known as CD107a ^3,4) and polystyrene beads coated with anti-CD3 antibodies were used to stimulate cells by binding to and cross-linking a component of their TCRs. The beads we use are inherently fluorescent in all channels, and their binding to cells also causes a characteristic increase in side scatter, allowing us to measure by gating on scatter and bead fluorescence responses of cells that received the activating stimulus and cells that did not in a single sample. The assay was thus multiplexed in a single dimension, that of cell stimulation. This enabled us to discriminate compounds that augment TCR-stimulated responses from compounds that stimulate cells in the absence of TCR engagement. The former affect only cells that received a relevant activating stimulus and are potentially useful as immune-enhancing agents that could augment anti-viral and anti-tumor immunity. The latter affect all cells regardless of whether they have been stimulated via their TCRs or not, and would likely nonspecifically stimulate cells and would likely produce serious deleterious effects that would limit their utility.

There are two general ways in which we envision compounds could augment TCR-stimulated responses. The first involves acute effects occurring after short compound treatments, perhaps by effects on the signals that control the release of granules. It is well-established that increases in intracellular calcium ⁵ and the activation of protein kinase C and ERK ⁶ are required for granule release, so compounds that enhance these signals might be expected to act acutely to increase exocytosis. Consistent with this, using the assay to screen a plate containing extracts from tunicate-associated bacteria, we identified the PKC activator teleocidin A as an immune enhancer ¹. Compounds that activate MAP kinases or inhibit phosphatases that dephosphorylate PKC or ERK targets would likely produce similar effects. Compounds that enhance the activity of calcium channels or inhibit calcium extruding transporters could have similar effects. However, there is reason to believe that at least in some cases effects such as these might not be detected after longer compound treatment times. For example, some PKC activating compounds can trigger time-dependent down regulation of PKC isoforms ⁷, and effects would therefore only be observed at short treatment times. The second broad category of mechanism that could enhance immune responses involves changes in the level of expression of genes that control lytic granule exocytosis and biogenesis. Such effects would likely be observed only after long-term incubation, but would be missed if only shorter treatment times were assessed.

Based on these considerations, we reasoned that an assay with the maximum power to detect compounds with immune-enhancing effects would assess the effects of multiple compound treatment times in a single sample. In flow cytometry, it is possible to bar code cells (or beads) at channels other than those used to measure parameters of interest, a powerful strategy that has been used in multiple ways to extend the degree of multiplexing ⁸. Here, we describe a further modification of our assay that uses bar-coding to make it possible to assess effects of treating cells with compounds for various lengths of time prior to measuring effects on cellular responses, a development we call multiple treatment time (MTT) screening.

MATERIALS AND METHODS:

Chemicals, Cells, and Solutions

TALL-104 human leukemic CTLs were maintained as described previously ⁹. Early experiments were conducted in complete culture medium. Later experiments were in culture medium without IL-2. Complete culture medium contains 40mL heat-activated fetal bovine serum, 4mL Pen-Strep, 8mL L-Glutamine, 80 μL IL-2, and 348mL Iscove’s modified Dulbecco’s medium. Dynabeads polystyrene beads coated with anti-CD3 or anti-CD8 monoclonal antibody (Thermo Fisher, Waltham, MA) were washed in medium according to the manufacturer’s protocol, and were added to the cell suspension at a ratio of ~one bead per cell. Alexa647-conjugated anti-LAMP1 antibody (0.5 μg/mL final concentration) was purchased from BD Pharmingen (San Jose, CA). Phorbol 12-myristate 13-acetate (PMA) was from Alexis Biochemicals (San Diego, CA). Thapsigargin (TG) was from AdipoGen (San Diego, CA). Cells were incubated with imiquimod (Adipogen) San Diego, California) at 10 uM or PMA at 100 nM for indicated time. The National Cancer Institute’s Diversity Set V (https://wiki.nci.nih.gov/display/NCIDTPdata/Compound+Sets) was used for screening, and powder resupply was from NCI. Compounds were prepared in DMSO at a final concentration of 10μM.

Fluorescent bar-coding with calcein

TALL-104 cells were incubated with calcein AM (Thermo Fisher, Waltham, MA) at 0, 20, 50, 75 or 200 nM for 25 minutes, washed with dye-free medium twice and then mixed together at indicated time points.

Flow Cytometry, plate format assays and screening

Some preliminary experiments were conducted on a BD FACSCAlibur flow cytometer equipped with 488 nm and 640 nm lasers. Most experiments were conducted on a five-laser (350 nm, 405 nm, 488 nm, 540 nm and 640 nm) BD Fortessa X-20 with a plate reading unit. Both instruments are housed in the University of Connecticut’s Flow Cytometry Facility. When the FACSCalibur was used, pulse height was measured. Pulse area was used for all other experiments. For experiments conducted in plates, including plate screening, sample volume was 120 μl, and samples were acquired a rate of 3 μL per second. 60 μl of sample was collected, unless 20,000 events were acquired first, in which case sampling was terminated. Samples were mixed 3 times with 60 μl per well at a mixing speed of 100μL per second. It took approximately 40 minutes to read a whole plate, and, for screening, generally only one plate was read per day. Data were analyzed offline using FlowJo software (Tree Star, Inc., Ashland, OR). Statistical analysis of data was performed using Prism 7 (GraphPad Software, LaJolla CA) .

For screening, 1 μL of compound at a concentration of 1mM in DMSO or 1 μL of DMSO was added to wells of test plates (Greiner 675076, Kremsmünster, Austria). One hundred microliters of medium-washed TALL-104 cells (1.25 × 10⁶/mL) were added to each well, mixed with the compound or DMSO, and incubated at 37°C for 24 hours. This comprised the 24 hour treatment time sample. Cells for the 30 minute treatment time were maintained in medium at 37°C for 23 hours, before being bar-coded with 20nM calcein AM and added to the plate in 10 μL cell solution (1.25 × 10⁷/mL). Each well was thoroughly mixed and incubated at 37°C for 30 minutes. To stimulate cells, ten microliters of CD8 bead solution (medium supplemented with anti-CD8 beads, anti-LAMP1, 4% DMSO), CD3 bead solution (medium supplemented with anti-CD3 beads, anti-LAMP1, 4% DMSO), or stimulation solution (medium supplemented with anti-CD3 beads, anti-LAMP1, 1 μM PMA, 20 μM TG) was added to the appropriate wells and mixed by repeated pipetting. All manipulations were performed with digital multichannel pipetters. Plates were incubated at room temperature for 1 hour with constant rotation (resulting in plate inversion), before it was sampled.

Measuring Enhancement of ANTI-LAMP1 Fluorescence

After data compensation for fluorescence spillover, the geometric mean of anti-LAMP1 fluorescence for cells bound to beads was obtained. Values from sample compound-containing wells (S) were compared to DMSO-treated wells from CD3 bead (CD3) or CD8 bead (CD8) controls. Values from CD8 controls represent the inherent fluorescence of non-responding cells bound to beads, and we subtracted their average values before calculating the percent enhancement (% Enhancement) using the following equation (equation 1):

$$\text{\% }Enhancement=\frac{S-CD8}{CD3-CD8}\text{*}100\text{\%}$$

Plate data were also normalized using the Z-score ¹⁰, calculated as (equation 2):

$$Z=\frac{X_i-\mu }{\sigma }$$

where X_i is the value of a given sample, μ is the mean of all the samples in the plate, and σ is their standard deviation.

Dose response of active compounds

Potential enhancing compounds were obtained from NCI in powder form and dissolved in DMSO. The same procedure described in “plate format assay and screening” was used for dose response testing except for compound NSC 125095, which was tested at only 24 hour. Serial fold dilutions were created from 10 or 3 mM DMSO stock for each compound. 1 μl of compound or DMSO was added to plates. Cell-containing solutions were added, mixed and incubated at 37°C for 23 hours. Cells loaded with 20 nM calcein AM were added to the wells, mixed, and incubated at 37°C for 30 minutes. CD8, CD3 or stimulation solutions were prepared as described and added to appropriate wells. Percentage enhancement was calculated and the dose response was fitted to Hill equation for compounds that demonstrated dose dependence.

RESULTS:

Bar-coding cells with calcein

We reasoned that we could implement an assay in which the effects of treating cells with compound for various lengths of time before stimulating them by adding cells to compound-containing wells at different times prior to the start of the assay (Figure 1A). As long as the volume in which cells are added is small relative to the volume of compound-containing solution in the assay wells, compound concentration will only be changed by a few percent with each subsequent addition. Bar-coding would allow the time at which cells were added to compound containing wells to be determined at the same time as responses are measured. In our assay, lytic granule exocytosis is measured with an alexafluor 647-conjugated anti-LAMP1 antibody, and the anti-CD3 beads we use for stimulation are inherently fluorescent in multiple channels. Maximal fluorescence is observed with 488 nm excitation and 670 nm emission. We thought it would be possible to bar-code the time of addition using calcein AM, which is efficiently excited at 488 nm and emits at wavelengths >530 nm, provided three requirements are met. First, we must be able to compensate fluorescence signals for spillover. Calcein spills into the 488/ 670 channel used to measure bead fluorescence on the FACSCalibur, and, as stated above, there is inherent fluorescence of beads in the channel used for calcein on both of the cytometers to which we had access. Compensation for bead fluorescence in the channel used for alexafluor 647-conjugated anti-LAMP1 antibody is also necessary on all cytometers for accurate measurement of anti-LAMP1 fluorescence. Second, calcein fluorescence levels must be stable enough so that the fidelity of bar-coding is retained. Finally, calcein should not affect cellular responses.

Figure 1. Barcoding for multiple treatment time screening.

A) Schematic of a multiple treatment time screening experiment. B) Discriminating three calcein loaded populations on the basis of fluorescence and scatter on a FACSCalibur. C) Cells loaded with 0, 50, or 200 nM calcein were washed and incubated in dye-free medium. Samples were collected on the flow cytometer at the indicated time points, and the geometric mean of calcein fluorescence for live cells were calculated. Loading is stable for many hours. D) Fluorescence spillover can be compensated allowing bead-binding status and anti-LAMP1 fluorescence to be measured. In this experiment, cells from C) were combined and incubated with beads coated with anti-CD8 antibodies, which bind to cells but do not trigger lytic granule exocytosis. Cells were sampled on two-laser FACSCalibur flow cytometer, with uncompensated plots shown on the left and compensated plots shown on the right. E) Calcein loading does not affect responses triggered by either anti-CD3 beads or TG+PMA. Cells were loaded with the concentration of calcein indicated to the left of the histograms, and (left) treated with DMSO (blue) or stimulated with TG+ PMA (red), or (right) treated with anti-CD8 beads (blue) or anti-CD3 beads (red).

Figure 1 shows that the prerequisites for using calcein for bar-coding up to three samples are met, even when using the FACSCalibur, an instrument in which spillover is more of a problem. Cells could be loaded with two dye concentrations such that three populations- one unloaded- can be distinguished (Figure 1B). Loading was stable over the course of many hours (Figure 1C). Provided we leave cells treated for the longest time (24 hours) unloaded, which would likely be a good choice given potential effects of long-term dye loading for cell health, these results suggest that calcein loading is stable enough for bar coding. Figure 1D shows that spillover between fluorescence channels can be adequately compensated, even under the relatively adverse circumstances that pertain when using a two laser instrument. In this experiment, cells were incubated with beads coated with anti-CD8 antibodies, which bind to cells but do not trigger exocytosis ¹. Without compensation, fluorescence from calcein contributed so much signal in the channel used to measure bead binding that discriminating cells bound to beads was impossible. After compensation, however, the different populations were clearly discernible. Figure 1E shows that loading cells with calcein per se does not affect lytic granule exocytosis as measured with anti-LAMP1 antibodies. Cells were treated by adding 1 μM thapsigargin (TG) and 50 nM PMA, soluble compounds that stimulate maximal responses in TALL-104 cells ¹, comparing responses to cells treated with DMSO as a vehicle control. No differences were observed at different levels of calcein loading (including no loading). We also examined effects of incubating cells with beads coated with anti-CD8 (non-stimulating) or anti-CD3 (stimulating) antibodies. Again, no differences were apparent in cells loaded with different amounts of calcein.

Detection of compounds with actions occurring on different time scales

Our major rationale for seeking to develop an MTT assay is to be able to detect compounds that might only be active after either short or long compound treatments. We identified two compounds that we thought might serve as exemplars. Phorbol myristate acetate (PMA) activates PKC and enhances lytic granule exocytosis after short duration treatment ¹, but downregulation of PKC isoforms that occurs between 2 and 6 hours after stimulation ¹¹ could result in inhibition of lytic granule exocytosis. Imiquimod, in contrast, has been reported to increase perforin levels in CTLs after ~12 hours ¹², suggesting that it might enhance lytic granule exocytosis in a time-dependent manner. Figure 2 shows results of an experiment in which we examined the effects of treating cells with PMA or imiquimod using our BD Fortessa X20. We placed cells into wells containing DMSO, PMA or imiquimod 24, 6 or 0.5 hours prior to adding either CD3 beads or CD8 beads to DMSO-containing wells, or CD3 beads to PMA and imiquimod containing wells. Cells added at 24 hours were not loaded with calcein, cells added at 6 hours were loaded with 7.5 nM calcein and cells added at 30 minutes were loaded with 20 nM calcein. After adding beads, plates were rotated end-over-end for 60 minutes prior to analysis. We found that, as expected, PMA greatly enhanced responses when cells were treated for 30 minutes (~8-fold enhancement compared to CD3 controls after subtracting the fluorescence of cells bound to anti-CD8 beads, which is a measure of intrinsic cell/ bead fluorescence). The effect was substantially smaller after 6 hours (~2.4 fold), while after 24 hours responses were actually reduced compared to DMSO-treated CD3 controls. Imiquimod, by contrast, enhanced responses in cells treated for 24 hours by 1.5 fold, but was without effect at earlier times. The effect of imiquimod after 24 hours was statistically significant as assessed by one-way ANOVA, as were the effects of PMA at all time points.

Figure 2. Detecting compounds with different modes of action via multiple time point measurements.

Plots of anti-LAMP1 fluorescence for the conditions indicated from cells treated with imiquimod (IMQ) or PMA for 30 minutes, 6 hours or 24 hours. Cells treated for 24 hours were not loaded with calcein, and cells treated for 6 hours or 30 minutes were loaded with 7.5 nM and 20 nM calcein, respectively. For each time point, samples are arranged in the order (from left to right): anti-CD8 beads, anti-CD3 controls, anti-CD3 stimulated cells treated with imiquimod, and anti-CD3 stimulated cells treated with PMA. Note that the left axis is discontinuous to allow the large enhancing effect of PMA after 30 minutes of treatment (~8-fold) to be displayed while allowing the differences between other conditions to be visible.

A two-time point assay may be suitably robust for screening

In high throughput flow cytometry, a fixed sample volume is typically acquired from each well ¹³. This means that each treatment time added to the assay necessarily decreases the number of events acquired for any given treatment time, unless the cell density is increased to compensate. We decided to focus on developing a two time-point assay in which cells are treated for either 30 minutes or 24 hours, as this seemed likely to offer the optimal ability to identify acute-acting and long term-acting compounds. It also made the work flow much easier, as a 6 hour time point is difficult to incorporate into most work days. To determine whether a two-time point version of the assay might be robust enough for screening (Figure 3), we laid out 48 wells of a half-area 96 well plate with two repeats of two columns containing DMSO and one containing TG + PMA. Unloaded cells were added 24 hours prior to the start of the stimulation period, and calcein-loaded cells were added 30 minutes prior. CD8 beads were added to one set of columns of DMSO-containing wells, and CD3 beads were added to the other column of DMSO containing wells and also to the TG + PMA containing wells. The plate was read on the Fortessa X-20. Samples were mixed 3 times before 60 μL was sampled to collect ~20,000 events. Responses of cells bound to beads were measured and used to calculate the assay performance parameter Z’ ¹⁴ for both treatment times. We found Z’> 0.5 computed from 16 samples each comparing DMSO-treated CD3-stimulated cells to TG+PMA-treated CD3-stimulated cells for both 30 minute and 24 hour-treated cells, indicating that the assay is likely suitable for screening in our hands using relatively low-throughput manipulations. At each time point, responses of CD8 bead-treated cells, CD3 bead-treated cells and TG + PMA-treated CD3-stimulated cells were significantly different from one another as assessed using one-way ANOVA.

Figure 3. Assay performance is suitable for screening in low density format.

A) Plot of calcein fluorescence for a single well used to determine treatment time. B) Histograms of anti-LAMP1 fluorescence for cells bound to CD8 beads (dashed black line), cells bound to CD3 beads (blue) and cells bound to CD3 beads in wells that contained TG+PMA (red) from cells treated for 24 hours (1) or for 30 minutes (2). C) Plots of anti-LAMP1 fluorescence intensity for cells treated for 24 hours (1) or 30 minutes (2) for the conditions indicated. CD3+ indicates CD3 + TG + PMA. Data (16 points per condition per time point) were used to calculate Z’. All conditions from a given treatment time were significantly different as assessed by one-way ANOVA.

Screening an NCI compound library probes assay performance and reveals candidate immunomodulators

We screened 12 plates from the NCI’s Diversity Set V containing a total of 960 compounds using the two-treatment time assay format (Figures 4 and 5). In a first limited set of experiments, we screened two plates from Diversity set V at a final compound concentration of 10 μM. For these plates, we included as controls 4 wells each of: 1) anti-CD8 bead stimulation, 2) anti-CD3 bead stimulation, 3) PMA + anti-CD3 bead stimulation and 4) TG + PMA + anti-CD3 bead stimulation. Two compounds seemed to enhance anti-LAMP1 staining in cells treated for 24 hours but not for 30 minutes. Data from a plate that contained the one compound (NSC 125095) that was subsequently confirmed active is shown in Figure 4A. Powder resupply of NSC 125095 was obtained and retested. Two time-point dose-response measurements confirm that NSC 125095 was without effect at any concentration tested after 30 minutes (data not shown). Pooled dose-response measurements conducted after 24 hour compound treatment indicate that the compound enhances anti-LAMP1 responses after 24 hours with an EC₅₀ of 6.9 μM (Figure 4B).

Figure 4. Identifying a potential immune-enhancing compound in a preliminary screen of two plates from NCI’s Diversity set V.

A) Plot of anti-LAMP1 fluorescence for cells not bound to a bead (open circles) or bound to a single bead (closed circles) from compound-containing wells of plate 4822 from NCI Diversity set V. The top plot shows data from cells treated with compound for 24 hours, which were not loaded with calcein) while the bottom plot shows data from cells that were loaded with calcein and added to wells 30 minutes prior to the start of the assay. Compound effects were tested at a final concentration of 10 μM. Values from controls are also displayed. Squares represent wells to which anti- CD8 beads were added instead of anti-CD3 beads, triangles are from wells that contained PMA to which anti-CD3 beads were added, inverted triangles are anti-CD3 bead controls and diamonds are wells that contained TG + PMA to which anti-CD3 beads were added to generate maximal signals. Closed symbols indicate responses from the bead-bound population, and open symbols are responses from the no-bead population, as determined by changes in forward and side scatter and bead fluorescence (not shown). B) Dose-response curve for the effects of cells treated for 24 hours with powder resupply of compound NSC 125095. Data are mean ± S.E.M. from 4 experiments. The solid line is a fit to the Hill equation, with a Hill number of 1.3 and a K_1/2 of 6.9 μM. The calculated maximal effect was a 2.5-fold enhancement.

Figure 5. Screening the remainder of Diversity set V probes assay performance and generates additional candidates.

A) Plots of normalized responses (top) and Z-score (bottom) for the entire compound collection at 24 hour and 30 minute treatment time. For the top pair of traces, data were normalized according to equation 1 in the Materials and Methods using both CD8 and CD3 controls. For both sets of traces, compounds that were selected for retesting are indicated by open circles. The rest of the compounds are represented by closed circles. B) Analysis of the effects of cherry-picked compounds on normalized anti-LAMP1 responses. Open bars represent data from 30 minute compound treatment, while closed bars are data from samples treated for 24 hours. Data are pooled from 5 initial experiments at 10 μM and single concentration data-points taken from four dose-response measurements. Data were analyzed using repeated measures ANOVA performed on raw data, which are not shown; * denotes p < 0.05, ** p < 0.01, *** p < 0.001. C) Two time-point dose-response measurements were performed on cherry-picked compounds. One, NSC 11643, shows dose-dependent effects that can be described by the Hill Equation. Best fits are shown as black lines. See text for details.

We later screened the other 10 plates in the Diversity set, also at a compound concentration of 10 μM. For these experiments, we included all of the controls described above except the PMA + anti-CD3 bead condition. We instead added four additional anti-CD3 bead control wells. We were able to assess assay performance during our limited plate screening (Figure 5 and Table 1). 7 out of 12 plates gave Z’ > 0.5 (measured by comparing 8 anti-CD3 controls to 4 controls stimulated with anti-CD3+ TG + PMA) for 30 minute treated cells, while only 2 out of 12 did for 24 hour treated cells. Note that for this analysis we did not exclude outliers, If we instead compute robust Z’ (Z’_R) ¹⁵, a variant of Z’ that uses median and robust standard deviation (1.48 x median absolute deviation) instead of mean and standard deviation and is thus more resistant to outliers, 10/12 plates at 30 minute treatment time and 4/12 after 24 hour treatment had values > 0.5. There was a correlation between Z’ for the two treatment time points (r² = 0.49). The enhancement of anti-LAMP1 fluorescence signals caused by adding soluble stimuli in addition to anti-CD3 beads varied between plates, and there was a strong correlation between the magnitude of enhancement for 30 minute treatment and 24 hour treatment times (r² = 0.65). However, Z’ and Z’_R values were only weakly correlated with enhancement produced by soluble stimuli (r² = 0.002 and r² = 0.1).), respectively), suggesting that sub-optimal response amplitude (low signal too high or high signal too low) was not responsible.

Table 1: Assay Performance during screening of NCI Diversity set V.

	30 Minutes					24 hours
Plate	Z’	Z’r	CD3c	CD3s	E	Z’	Z’r	CD3c	CD3s	E
4812	0.5	0.6	1591	2282	12.7	0.6	0.7	1522	3395	10.9
4813	0.5	0.6	876	1507	17.1	0.1	−0.1	1023	2137	11.8
4814	0.0	0.9	274	1241	39.6	0.0	0.3	623	1931	12.4
4815	0.6	0.7	1047	1492	15.5	0.2	0.4	1018	2485	11.2
4816	0.1	0.1	815	616	15.2	−0.1	0.1	750	2252	8.6
4817	0.8	0.8	976	1690	18.9	0.8	0.8	1109	2591	13.5
4818	0.6	0.6	3908	5107	7.2	0.3	0.2	1846	3315	6.7
4819	0.3	0.6	2143	3491	5.9	−0.4	−0.4	1988	4782	3.9
4820	0.3	0.2	1594	1831	8.7	0.1	0.0	1465	3288	6.1
4821	0.5	0.4	2872	1969	5.6	0.0	0.0	3431	2789	3.3
4822	0.4	0.6	6945	4109	2.8	0.3	0.6	8726	5294	2.2
4823	0.5	0.9	3820	4464	3.9	0.3	0.7	6103	3137	2.6

Z’_r is the robust Z’, calculated using media and robust SD.

CD3_c is the average of CD3 responses calculated from plate controls.

CD3_s is the average of the sample compound-treated wells.

E is the fold enhancement obtained by treating cells with TG + PMA in addition to CD3 bead stimulation.

There was for most plates a quantitative discrepancy between the amplitude of responses in anti-CD3 bead-stimulated controls and the average values obtained from compound-treated cells stimulated with anti-CD3 beads that made normalizing data problematic (see Figure 5A, top, for data from all plates in the collection normalized to CD3 control values after subtracting CD8 control values). As a result, we found it most useful to normalize data from different plates using the Z score as defined in the Methods. Plots of Z score for all of the compounds we screened are shown in the bottom panel of Figure 5A. Compounds that were selected for retesting are shown as open circles.

23 compounds that produced Z scores > 1.5 at either 30 minutes, 24 hours or both in the initial screen were selected for retesting. Compounds that enhanced anti-LAMP1 fluorescence in cells that did not contact beads, or were fluorescent in any of the channels collected were excluded. We did not select all of the compounds that produced the highest Z-scores, as we felt exploring the effects of compounds that produced a range of responses might be informative. Powder resupply of 10 selected compounds that produced the highest levels of anti-LAMP1 responses upon retesting was obtained, and these compounds were then retested at 10 μM. In 5 replicate experiments, seven compounds that enhanced anti-LAMP1 responses after 24 hour treatment were identified. Five of these also enhanced responses at 30 minutes, but the effect appeared smaller than at 24 hours (data not shown). We investigated the dose-dependence of inhibition of the seven compounds, using a two-timepoint protocol that assessed effects of compounds after 30 minutes and 24 hours of treatment using concentrations ranging down from a maximum concentration of 30 μM. Four such experiments were performed. As these experiments included 10 μM doses, we added these additional data points to the five single-point measurements described and used the pooled sample to assess statistical significance of effects. Note that one compound appeared to result in significant cell death at both 10 μM and 30 μM as assessed by measuring the number of events in the live-cell forward and side-scatter gate for the calcein-negative (24 hour treatment) population (Supplemental Figure 1), so we decided to exclude this compound from further analysis. We found that all six of the remaining compounds significantly enhanced anti-LAMP1 fluorescence compared to controls at both treatment times as assessed by repeated measures one-way ANOVA conducted on non-normalized data. (Figure 5B shows normalized results from these measurements). Inspection of the dose-response relations (Figure 5C) indicated that only one compound, NSC 11643, produced saturable dose-dependent augmentation that could reasonably be fit to the Hill equation. Concentrations of NSC 11643 producing half maximal enhancement were 33 nM for cells treated for 30 minutes and 150 nM for cells treated for 24 hours. Responses were significantly smaller for 30 minute treatment as compared to 24 hour treatment (P < 0.01 using a paired t test). Other compounds produced dose-response relations that were either flat or increased monotonically over the concentration range examined. These seem less likely to us to reflect genuine or useful biological effects than the effects of NSC 11643. We believe based on the results we present that NSC 120595 and NSC 11643 represent promising candidates for immune augmenting compounds that can be pursued in future studies.

DISCUSSION:

Multiplexing a cell-based assay so that it examines the effects of multiple compound treatment times has a number of possible uses. First, it increases the assay’s ability to detect compounds with different kinds of activity. This was the use to which we put the method. We envision that a second use might be to increase an assay’s reliability. In such an implementation, assessing effects of multiple treatment times could be thought of as analogous to quantitative HTS (qHTS) ¹⁶, where compounds are screened at different doses as a means of reducing the number of false positives and identifying the most promising compounds in the initial screening step. The assumption behind such an implementation of MTT screening would be that compounds with genuine activities should exhibit graded time-dependent changes in responses. One advantage of assessing multiple treatment times in this manner as compared to qHTS would be that higher throughput should be achievable, as the time series would all be contained in a single well.

The assay was not as robust during plate screening (Table 1) as we would have anticipated based on its performance in control testing (Figure 3). We failed to reliably obtain Z’ values for control wells that were > 0.5, although robust Z’ values were somewhat higher. We note that Z’> 0.5 appears not to have been intended by its creators to be an absolute criterion for assay feasibility. Zhang et al. certainly seem to have noted that assays with Z’ > 0 could be fruitful. Based on the difficulties we encountered with the assay at low-throughput, we anticipate that it might be challenging to translate it to higher-density format. The main technical limitation is likely to be in mixing the activating beads and cells well enough, as beads tend to settle much more rapidly than cells. The inherent difficulties involved in trying to measure complex physiological responses from living cells will also present problems. However, we would argue that the unmet need for immunomodulatory small molecules makes the challenges worth confronting. Also, as we noted, there was a discrepancy between amplitude of anti-CD3 controls and the average response of compound-treated anti CD3-stimualted cells. As we would expect most compounds to have no effect on lytic granule exocytosis, we anticipated that the two measures would give similar results. The discrepancy was not related to the way we sampled plates. For the first two plates we tested (4822 and 4823), controls were sampled first, yet compound-treated samples had lower average fluorescence levels. For the remainder of the plates, controls were acquired interleaved amidst compound-treated samples. Controls were treated with DMSO at the same concentration as compound-treated cells. At present we have no explanation for the discrepancy.

The real test of whether our assay is useful is whether it identifies compounds as anticipated. We were able to identify 7 compounds in NCI compound collections that reproducibly enhance responses to CD3 beads as assessed by anti-LAMP1 staining. However, five of those compounds exhibit dose-response behavior that seems inconsistent with a plausible mechanism of action. Of the two others, NSC 125095 affects responses only after longer-term incubation, with an EC₅₀ of ~7 μM and a calculated maximum enhancement of 150% after 24 hour treatment, as derived from fits of the Hill equation to dose-response data, while NSC 11643 is effective after both short (EC₅₀ 33 nM, calculated maximum enhancement 59%) and long-term (EC₅₀ 150 nM, calculated maximum enhancement 115%) treatment. Thus, the assay reveals compounds with different temporal patterns of activity as we hoped. Assuming that compound penetration into cells is not the limiting factor, we suspect that the effects of NSC 125095 reflect changes in gene expression. Since NSC 11643, has much faster effects, we would imagine that it likely affects some aspect of the signaling cascade that controls lytic granule exocytosis. There is little information available upon which to formulate hypotheses regarding possible targets of these compounds. NSC 125095 has been tested in 771 assays reported in PubChem, and found to be active in only 6. One effect reported was promotion of myeloid differentiation in myeloid leukemia cells (PubChem AID 624256), which seems as though it might be relevant. NSC 11643 has been tested in 322 assays and found to be active in only two. One (AID 743287) was for compounds that enhance glucose-dependent insulin secretion, which seems promising. However, both of those assays yielded several thousand hits, and it is not clear that compound activities were confirmed.

It remains to be determined whether these compounds have useful effects on immune functions. Although LAMP1 externalization is believed to be a faithful measure of lytic granule exocytosis ^1,3,4, we have yet to confirm compound effects with an orthogonal measure, and are in the process of devising the needed assays to do so reliably. Furthermore, even if compounds are confirmed to increase lytic granule exocytosis, it is not clear what effect this would have on CTL-dependent target cell killing, or immune function more generally. It appears that CTLs release only a fraction of their lytic granules when they kill a target ⁹. Releasing more granules may not enhance overall target cell killing, in fact releasing more granules than needed to kill a single target could reduce overall killing by depleting granules that could have been used in later “lethal hits.” Of course, compounds that affect lytic granule exocytosis might also affect the activity of other immune cell types. There are sufficient similarities between the signals that activate CTLs and other immune cells that compounds that act on CTLs might be expected to have broader effects. For example, imiquimod, which we found enhanced anti-LAMP1 responses by about 50% (i.e responses were 1.5-fold higher than in control cells), somewhat less than the increases caused by NSC 125095 (~2.5-fold) and NSC 11643 (~2.2-fold), has effects on multiple immune cell types, including dendritic cells and B cells^17,18. It is possible that the compounds we found may do the same.