Falcon^® 96 Well Clear Round Bottom Non-Treated Assay plates (NTA plates, Corning #353910) and White 96 Well Transparent Bottom CELLSTAR^® plates (white plates, Greiner Bio-One #655088) were purchased from Dominique Dutscher (Grenoble, France). Universal polystyrene lids (Thermo Fisher Scientific #5500), Dulbecco’s modified Eagle’s medium (DMEM, Gibco #41966), Roswell Park Memorial Institute medium (RPMI1640, Gibco #61870), Keratinocyte Serum Free Medium (K-SFM) kit supplied with bovine pituitary extract (BPE), and human epidermal growth factor (EGF) (Gibco #17005-042), Dulbecco’s phosphate-buffered saline (PBS) without calcium or magnesium (Gibco #14190), 0.05% trypsin–EDTA (Gibco #25300054), 10 kU/mL penicillin–streptomycin (Gibco #15140122), CellEvent Caspase-3/7 Green Reagent (Invitrogen #C10423), and LysoTracker Deep Red (Invitrogen #L12492) were obtained from Thermo Fisher Scientific (Courtaboeuf, France). Fetal bovine serum (FBS, PAN Biotech #P30-3302) was purchased from Dominique Dutscher (Grenoble, France). ViaLight™ Plus Cell Proliferation and Cytotoxicity BioAssay Kit (Lonza #LT07-121) were obtained from Lonza (Basel, Switzerland). For chemotherapy drugs, MLN4924 (MLN, MedChemExpress #LSG790) was purchased from Interchim (Montluçon, France), cisplatin (Cayman Chemical #CAYM13119-250) from VWR International (Fontenay-sous-Bois, France), docetaxel (Acros Organics, #15316097) from Thermo Fisher Scientific, Etoposide (Aldrich, #E1383) from Sigma-Aldrich (Saint-Quentin-Fallavier, France), and ARN-509 (Adooq Bioscience, #A11923-10) from CliniSciences (Nanterre, France).

The VCaP (androgen receptor-positive, androgen-sensitive, containing TMPRSS-ERG translocation, p53-R248W, PTEN-wt; ATCC #CRL-2876), LNCaP (androgen receptor-positive, androgen dependent, p53-wt, PTEN-mut; ATCC #1740), PC3 (androgen receptor-negative, androgen independent, p53-null, PTEN-null; ATCC #CRL-1435), DU145 (androgen receptor-negative, androgen-independent, p53-P223L-V274F, PTEN-wt; ATCC #HTB-81), and RWPE-1 (HPV18-immortalized normal adult prostate epithelial cells; ATCC #CRL-11609) cell lines were purchased from American Type Culture Collection (ATCC). The DuCaP cell line (similar origin as VCaP) was kindly provided by Prof. Jack Schalken from the Radboud University Nijmegen Medical Center, who originally received them from Kenneth J. Pienta, MD, Director of Research at The Brady Urological Institute, Baltimore, where this cell line was created (90). The PNT2 cell line (SV40-immortalized normal adult prostate epithelial cells; ECACC #95012613-1VL) was purchased from the European Collection of Authenticated Cell Cultures (ECACC).

Cell lines were tested on a semester basis for AR, PSA, SLC45A3, and ERG using quantitative RT-PCR. VCaP and DU145 cells were cultured in DMEM containing 10% FBS and 1% penicillin–streptomycin (complete growth medium). LNCaP, PC3, DuCaP, and PNT2 cells were cultured in RPMI1640 with the same supplements. RWPE-1 cells were cultured in K-SFM containing 0.05 mg/mL BPE, 5 ng/mL EGF, and 1% penicillin–streptomycin. The cells were grown in an incubator at 37°C, 5% CO₂, and 95% humidity. For different passages, the cells were washed twice with PBS followed by the addition of trypsin–EDTA and incubation for 3–10 min depending on the cell line. Subculture was performed depending on the density of the cells. Typically, for VCaP cells, subculture was performed once per week, at 1/2 dilution; LNCaP cells were divided twice per week, at 1/5 dilution.

Cells were grown in appropriate complete growth medium for at least 3 days in standard T75 flasks to reach approximately 70% confluence. For reproducible results, cells of a similar passage number (1 < N < 15) were always used. To prepare the cell suspension, a cell monolayer was washed with 10 mL of PBS and treated with 1.5 mL of a trypsin–EDTA solution for 5 min at 37°C. Trypsin was then neutralized with 8 mL of complete growth medium, and the cell suspension was centrifuged at 400 × g for 5 min. After supernatant removal, the cell pellet was resuspended in 1–3 mL of complete growth medium, and the cell density was determined using a Scepter™ Cell Counter (Merck Millipore). Note that it is critical to ensure that the cell suspension is homogeneous (single cells) and does not contain aggregates, because this determines the size and uniformity of spheroids. The optimal cell density must be determined in preliminary spheroid generation tests (see Figure 1C, for example).

The suspension was then diluted with complete growth medium to obtain the final density and dispensed (200 μl/well) into 96-well NTA plates. To reduce the effect of uneven evaporation on spheroid formation, the edge wells of the plates were not used and filled with PBS. This resulted in 6 × 10 = 60 wells per plate to analyze. The plates were covered with the universal polystyrene lids and kept in an incubator (37°C, 5% CO₂, 95% humidity) for at least 2 days before imaging. During this period, it is important not to disturb the plates as this can displace spheroids from the center of the well. In addition, although previous reports (70, 83, 91) have shown that mild centrifugation can promote single spheroid assembly, no significant effect of centrifugation was observed with VCaP and LNCaP cells.

Finally, the analysis of spheroid assembly was conducted immediately after the cell suspension addition when performed on a video microscope (Zeiss Axiovert 200 M) equipped with a cell culture chamber or on a lens-free microscope [VideoCell (92)] within the cell culture incubator.

After 2 days of spheroid formation, the plates were scanned once a day on a CellInsight NXT High Content Screening Platform (Thermo Scientific) equipped with a 10× objective. The acquisition parameters were set using the instrument software HCS Studio 6.5.0. The bright-field channel was used for autofocusing, and the exposure time was set within the 25–35% saturation range according to the HCS Studio recommendations. Acquisition of “one field of view per well” was sufficient for the majority of spheroid measurements and significantly reduced the time required for plate analysis. Using 4 × 4 binning, an 896.6 µm × 896.6 µm field of view was converted into a 552 px × 552 px image. With these settings, the scan of a 96-well plate took approximately 5 min. The images were saved automatically and could be analyzed later.

Image analysis was performed with the instrument software HCS Studio 6.5.0—Morphology.V4. Up to 20 different “Cell-selected features” were chosen for Channel 1. Then, the bright-field images were imported as “Channel 1-Object.” This generated the object masks for each image. The masks were further adjusted by imposing the size range for the counted objects (“Object selection” > 1,000 px² for VCaP, and >3,000 px² for LNCaP) and “Cell-selected features” were calculated for each of the adjusted masks. The object “Area” and “Shape P2A” were the most informative morphometric parameters (see Results). Shape P2A = (Perimeter)²/4π × (Area) represents the object roundness and is equal to 1 for a perfect circle; for less spherical objects, P2A becomes larger than 1. Note that, apart from these features, other calculated parameters, such as object “CircDiameter,” “Sphere Volume,” and “Aspect ratio LWR” were also useful for evaluating spheroid morphology. Next, a table of all calculated features was generated as a .csv file using the “Cellomics-View” application. The data were imported into “R” statistical software to create a fully annotated .csv data file by fusing with the “Experimental Design” table. The mean, median, SD, and quartile values were calculated with “R” software.

For manual acquisition, the images from each well were recorded using an inverted microscope (Zeiss Observer Z1). A 10× objective was primarily used, switching to a 5× objective when an object area was too large. The entire acquisition took approximately 30 min for a 96-well plate. To analyze the images taken on standard and video microscopes, an in-lab ImageJ macro was used. Alternatively, open-source tools, e.g., based on MATLAB (39), ImageJ (83), or Fiji (93), could also be utilized to produce object masks and calculate object area and perimeter.

The spheroids were formed following the protocol described earlier by incubating 2,000 cells for 7 days (VCaP) or 500 cells for 4 days (LNCaP) in NTA plates. Three replicate plates were prepared per experiment. MLN solutions were prepared immediately before use by diluting 2 mM DMSO stock solution in complete growth medium to a 2× final concentration. The drug solution of the highest concentration was made first and used to prepare all drug dilutions of the series. To avoid the effect of DMSO on spheroid growth, it is critical to ensure that the highest DMSO concentration in spheroid medium does not exceed 0.5%, which corresponds to 1% in drug solutions. Using “6 × 10 wells per plate” and “six spheroids per condition” experimental formats, nine drug concentrations and the DMSO vehicle control could be tested with one NTA plate. This corresponded to 600 µL total volume per plate for each drug dilution.

For apoptosis analysis, CellEvent was added to each of the prepared drug solutions to obtain a 2 µM concentration. Then, 100 µL of spheroid medium was carefully removed from each well (trying not to displace the spheroids from the center of the well) and replaced with 100 µL of CellEvent-drug solution. The plates were kept in an incubator (37°C, 5% CO₂, 95% humidity), and image acquisition was performed once a day.

For viability measurements, a 0.4 µM solution of LysoTracker Deep Red dye was prepared just before use by diluting 1 mM DMSO stock solution in complete growth medium. Then, 10 µL of this solution was added to each well, and imaging was performed at time points defined in the “Experimental Design” table.

The effect of MLN on spheroids was followed using a CellInsight NXT High Content Screening Platform as described earlier. The acquisition parameters were as follows: bright field for Channel 1, CellEvent fluorescence (FITC filter set used) for Channel 2, and LysoTracker Deep Red fluorescence (Cy5 filter set used) for Channel 3. Autofocus was performed in bright field (Channel 1), and the plates were scanned using “one field of view per well” acquisition with 4 × 4 binning. With these settings, the scan of a 96-well plate usually took less than 7 min.

The image analysis was performed using HCS Studio 6.5.0—Morphology.V4 software as described earlier. Up to 30 different “Cell-selected features” were measured for Channels 1–3. The bright-field images were imported as “Channel 1—Object,” CellEvent images as “Channel 2—Member,” and LysoTracker Deep Red images as “Channel 3—Spot/Fiber.” The corresponding masks were generated for each image, and “Cell-selected features” were calculated for each mask. The preferred parameters used for fluorescence analysis of the drug effects were fluorescence “Area,” “Total Intensity,” and “Average Intensity.” The data were analyzed using “R” software.

For manual acquisition, the images were recorded on the inverted microscope with a 10× (5×) objective using bright field, FITC, and Cy5 channels. The analysis was performed using an in-lab ImageJ macro.

Spheroid viability was analyzed by measuring ATP with a ViaLight™ Plus Cell Proliferation and Cytotoxicity BioAssay Kit (Lonza). The manufacturer’s protocol was applied with some modifications. Thus, 100 µL of spheroid medium was carefully removed from each spheroid-containing well of the NTA plate, and 50 µL of cell lysis reagent was added. The plate was agitated on an orbital shaker for 20 min at room temperature. Note that this protocol has been optimized for VCaP and LNCaP spheroids. For other cell types, the optimal lysis time must be determined in preliminary tests by measuring the efficiency of ATP extraction (using chemiluminescence).

Then, 100 µL of room-temperature ATP monitoring reagent plus (AMR plus) was added to each well, and the plate was incubated for 2 min at room temperature. Blank solution was prepared by adding cell lysis reagent and AMR plus to complete growth medium. Luminescence was measured on a GloMax^®-Multi Detection System (Promega) with a 1 s integration time. Optionally and when higher sensitivity is required, the samples can be transferred into a white plate with a flat transparent bottom (Grenier Bio-One) and remeasured. This may result in an up to threefold increase in signal intensity.

Cell viability in monolayer cultures was measured according to the manufacturer’s protocol.

The effect of chemotherapy drugs (cisplatin, docetaxel, etoposide, and ARN-509) on VCaP and LNCaP spheroids was examined following the protocol described earlier for MLN. Drug stock solutions were prepared in DMSO, with the exception of cisplatin, for which DMF was used (94). Drug dilutions used to generate dose–response curves are given in Table S1 in Supplementary Material.

Because cell attachment is inhibited in NTA plates, the cells clump together and form 3D spheroids. The U-profile of the wells ensures that only one spheroid per well is formed (Figure 1A; Movie S1 in Supplementary Material). As with ULA plates (21), not all cells form spheroids under these conditions. For instance, VCaP and LNCaP prostate cancer cell lines form spheroids while PC3 cells produce only loose aggregates [Figure 1A; see also Ref. (21, 33, 95)]. In some cases, modification of the cell culture medium composition could promote spheroid formation in NTA plates. Thus, among the seven prostate cancer cell lines examined in the present work, the majority produced individual spheroids after brief optimization of the culture conditions (Figure S1 in Supplementary Material). Two cell lines, DU145 and PNT2, formed multiple non-uniform spheroid-like aggregates (Figure S1 in Supplementary Material). Because VCaP and LNCaP cells reproducibly form spheroids and have different sensitivities to MLN in monolayer culture (77), we focused our study on these cell lines.

To avoid drug interference with spheroid assembly, spheroids must be well formed before drug treatment. The criteria used in the literature are rather visual and arbitrary (21, 31). We therefore decided to define a measurable parameter of spheroid assembly. Spheroid formation was followed over time with a CellInsight NXT HCS platform and analyzed with the instrument software (Figures 1B,C). We found that “Area” and “Shape P2A” are the best direct morphometric parameters to characterize spheroid assembly and growth. For slow-proliferating (in 2D) cells, such as VCaP cells, the minimum of the “Area” curve may represent a characteristic time of spheroid assembly (Figure 1B). However, for the faster growing LNCaP cells, the “Area” occupied by the aggregate steadily increased (Figure 1C). The best parameter for both cell lines appeared to be “Shape P2A”: when it attains the minimum, the aggregates are considered round and spheroid (Figures 1B,C). Using this parameter, a period of 4–7 days was found to be optimal for spheroid assembly. It should be noted that when measuring spheroid assembly, particular attention must be paid to the composition of the culture medium. Thus, we found that small changes in medium FBS, e.g., different concentrations, suppliers, and even lots, may significantly influence spheroid assembly. In addition, charcoal stripping of the serum reduced the rate of spheroid formation by both VCaP and LNCaP cells (Figure S1 in Supplementary Material). Therefore, it seems that spheroid assembly in NTA plates is, at least partially, an active cellular process that depends on specific cues within the culture medium.

The NTA plates ensured the formation of highly uniform spheroids (Figure 1D). Increasing the density of cell seeding resulted in a linear increase in spheroid volume. For VCaP cells, the slope of the trendline corresponded to approximately 12 pL of occupied volume per seeded cell (Figure 1E), which fell within the range of 5–20 pL/cell estimated for other spheroids described in the literature (21, 47, 83). This significantly exceeds the size of a VCaP cell (1.3 pL) measured with a Scepter™ Cell Counter, suggesting that cells undergo multiplication upon spheroid assembly. This could also indicate the presence of some extracellular space within the spheroids.

The optimal conditions required to form spheroids of approximately 400 µm in diameter (21, 31, 83) were determined to be 2,000 cells/well for 7 days with VCaP cells. Because the fast-growing LNCaP spheroids rapidly developed a hypoxic core that interfered with phenotypic measurements (see below), 500 cells/well and a duration of 4 days were used to form LNCaP spheroids.

Having established the optimal conditions for spheroid assembly, we examined the effect of MLN on preformed VCaP and LNCaP spheroids. Of note, in monolayer culture, these cell lines display a marked difference in MLN sensitivity: IC50 of 500 and 50 nM for VCaP and LNCaP cells, respectively.

First, the time course of changes in spheroid size and shape induced by 50 and 500 nM of MLN was analyzed (Figures 2A,B). For VCaP cells, at both concentrations, the drug arrested the growth of spheroids without affecting their roundness (Figure 2A, “Area” vs “Shape P2A”). By contrast, automated analysis of LNCaP spheroid growth was not conclusive (Figure 2B, “Area”). Visual examination revealed that at 500 nM the drug caused physical disruption of LNCaP spheroids, resulting in an unexpected increase in apparent spheroid size (Figure 2D, “Bright field”). The disruption starts on the second day of MLN treatment and is easily detected automatically by measuring spheroid roundness (Figure 2B, “Shape P2A”).

These results demonstrate that within the given time-frame simple bright-field analysis of spheroid morphology is not sufficient to characterize the effect of MLN. Indeed, it does not distinguish between two very different drug concentrations in the case of VCaP spheroids, whereas in LNCaP spheroids, it does not reliably detect the toxic effect of 50 nM MLN that corresponds to the IC50 value in monolayer culture. More prolonged drug treatment can partially help but requires reevaluation of culture conditions to prevent spheroid overgrowth.

Because MLN induces apoptosis in prostate cancer cells (77), we measured the kinetics of apoptosis induction using CellEvent, a fluorogenic caspase-3/7 substrate (23, 77) (Figures 2C,D, “CellEvent”). The reagent was added to spheroids at the same time as MLN. Drug-induced apoptosis was readily detected in both VCaP and LNCaP spheroids starting from the second day of spheroid treatment with 500 nM MLN (Figures 2A,B, “CellEvent”). In addition, independently of the treatment, all spheroids larger than approximately 450 µm in diameter showed a CellEvent-positive central region consistent with apoptosis induction within the nutrient deprived spheroid core (23, 62, 77). When this signal is so strong that it interferes with the drug cytotoxicity assay, as in LNCaP spheroids (Figure 2D), the measurement of the “fluorescence area” and normalization of the CellEvent signal to the spheroid area may help to distinguish the drug-specific effect. Thus, although the measurement of CellEvent “total intensity” did not distinguish between control and 50 nM MLN conditions (Figure 2B, “CellEvent”), some increase in “fluorescence area” was observed upon drug treatment (Figure 2D). Finally, and consistent with our previous observation (77), VCaP spheroids growth arrested with 50 nM MLN showed negligible apoptosis, even compared with control conditions (Figure 2C, “CellEvent”).

Compared with the morphometric parameters, the use of CellEvent provides complementary information regarding how the drug induces apoptosis within spheroids. Notably, for all tested concentrations (<2 µM), CellEvent was non-toxic to prostate cancer spheroids (did not change morphometric parameters or ATP content) and ensured a measurable and informative signal for more than 2 weeks of spheroid culture.

Because CellEvent analysis of apoptosis did not detect any significant effect of 50 nM MLN on LNCaP spheroids, we looked for other vital probes that were more sensitive. We found that spheroid staining with LysoTracker Deep Red was inversely proportional to MLN concentration. When added to the spheroids, the fluorescence of LysoTracker steadily increased and then started to plateau after 8 h of incubation (Figures 3A,B). The plateau level in MLN-pretreated spheroids was maximal in control conditions and decreased with increasing MLN concentration. The staining pattern was complementary to that of CellEvent, with the LysoTracker accumulating in the outer and presumably metabolically active layers of spheroids (Figures 3C,D). This result was unexpected because LysoTrackers are considered to be markers of cytotoxicity, cell senescence and apoptosis (78–80) (Table 2). One possible explanation for this discrepancy is that LysoTracker Deep Red belongs to a different chemical class than LysoTracker Red DND-99, which was used in previous studies (78–80).

Nevertheless, the use of LysoTracker Deep Red allowed for fair quantification of the effect of different doses of MLN with both VCaP and LNCaP spheroids. Similar to CellEvent, LysoTracker Deep Red dye is non-toxic (at tested concentrations <50 nM) and can be used for long-term monitoring of the drug effect in a 3D spheroid model.

The four parameters described earlier, “Area,” “Shape P2A,” “CellEvent,” and “LysoTracker Deep Red,” allowed for high-content monitoring of the drug effect in real-time. This is an important advantage compared with standard endpoint assays, because it reveals dynamic aspects of drug action and provides a flexible choice to measure dose–response curves. Using this approach, the complexity of the MLN effect, which is not obvious in monolayer culture (77), becomes readily dissected (Figures 4A,B). Thus, in VCaP spheroids, the two major outcomes of MLN treatment were (i) spheroid growth arrest with EC50 = 18 nM (Figure 4A, “Area”) and (ii) a cytotoxic effect with EC50 > 500 nM seen by an increase in apoptosis (Figure 4A, “CellEvent”) and a decrease in LysoTracker Deep Red uptake (Figure 4A, “LysoTracker”). A slight decline in total LysoTracker fluorescence at MLN concentration <100 nM most likely reflected the decrease in spheroid size. The quiescent state, characterized by cell cycle arrest and apoptosis inhibition, is easily detected between 25 and 100 nM MLN (Figure 4A). As we previously showed (77), this state is reversible, and removal of the drug results in spheroid regrowth at a normal pace. In the more MLN-sensitive LNCaP spheroids, these two effects could not be separated because of the earlier onset of cytotoxicity manifested by an increase in “CellEvent” (EC50 = 180 nM, Figure 4B) and a decrease in LysoTracker fluorescence (EC50 = 70 nM, Figure 4B). The major outcome of MLN treatment of LNCaP spheroids was spheroid disruption (Figure 4B, “Area”: EC50₂ = 190 nM, and “Shape P2A”: EC50 = 250 nM). The growth arrest induced by low <100 nM concentrations of MLN was also detected (Figure 4B, “Area”: EC50₁ = 18 nM), although it was much less defined than in VCaP spheroids.

The phenotypic analysis was complemented by evaluation of spheroid metabolic status using an ATP-based endpoint assay. Usually, this requires spheroid lysis and transfer of the lysate into an opaque-walled flat-bottom plate to measure chemiluminescence. We found that analysis can be performed in the same U-bottom NTA plate. Although this causes an approximate threefold decrease in signal intensity, the high sensitivity and dynamic range of the method ensure very similar robust results independently of the plate (Figures 5A,B). The ATP response curves most closely matched those measured with LysoTracker Deep Red, with the biphasicity of the VCaP response becoming even more obvious (Figure 5A). Biphasic drug dose–response curves have been documented previously and are attributed to the heterogeneity of cell populations within spheroids (83). However, phenotypic analysis of the MLN effect in VCaP spheroids and the results of our previous work (77) suggest the complexity of the drug action mechanism is the prime cause. Notably, ATP measurements in VCaP monolayer culture did not show a biphasic response to MLN (Figure 5A), and additional assays were required to detect the difference in drug effect at low and high concentrations (77). Alternatively, for LNCaP cells, similar MLN dose–response curves were obtained with either spheroids or monolayer culture (Figure 5B).

Collectively, these results show that high-content spheroid monitoring provides information that goes beyond a simple cytotoxicity assay. Thus, despite a unique target [NAE, with an IC50 of approximately 5 nM (87)], inhibition of neddylation by MLN affects multiple cellular processes with different effects on cell viability (77). This explains why MLN toxicity in various cell lines varies significantly. LNCaP cells are 10 times more sensitive than VCaP cells, despite a similar efficiency of neddylation inhibition (EC50 = 8–10 nM, Table 3). Some of these effects can be addressed by high-content monitoring. Our analysis shows that spheroid growth is the most sensitive process affected by MLN (EC50 = 18 nM, Table 3). Because the proposed mechanism of MLN toxicity is proliferation dependent (96), it seems plausible that in VCaP spheroids inhibition of cell proliferation results in protective quiescence [Figure 4A, and Ref (77)], while in the faster-proliferating LNCaP cells, this inhibition is not efficient and does not prevent the onset of apoptosis and cytotoxicity (Figure 4B).

To further evaluate the method, the effect of four anticancer drugs used for prostate cancer chemotherapy was examined. The following compounds with distinct MOAs were chosen: docetaxel, a microtubule-binding anti-mitotic agent, which is the most commonly used “standard of care” drug for castration-resistant prostate cancer (CRPC) (97, 98); cisplatin, a DNA-crosslinking chemotherapeutic agent, and etoposide, a topoisomerase II inhibitor, which are evaluated in combination therapies for some resistant CRPC phenotypes (98, 99); and ARN-509, a novel antiandrogen for prostate cancer treatment (100), which was recently submitted to the FDA for approval.

The experimental conditions established in the MLN study were used. Four parameters, “Area,” “Shape P2A,” “CellEvent,” and “LysoTracker Deep Red,” were measured to build dose–response curves with VCaP and LNCaP spheroids (Figures S2–S5 in Supplementary Material). At the end of the treatments (2 days for LNCaP and 4 days for VCaP), cell viability within spheroids was measured with a Vialight ATP-based assay, and the resulting dose–response curves were compared with those obtained with monolayer cultures (Figure S6 in Supplementary Material).

The measured EC50 values are summarized in Table 4. The endpoint ATP assay showed that all drugs suppressed cancer cell growth. As often observed in 3D models (101), spheroids showed similar or increased resistance to drug treatments compared with the corresponding 2D cultures (Figure S6 in Supplementary Material).

High-content analysis (HCA) confirmed complementarity of the morphometric (“Area” vs “Shape P2A”) as well as the fluorescence (“CellEvent” vs “LysoTracker”) phenotypic parameters for characterization of the drug effects (Figures S2–S5 in Supplementary Material). Compared with the ATP assay, an analysis of the phenotypic changes revealed a more complex effect of the drug action. Thus, similar to MLN, the dose–response curves were not always sigmoidal or monotonic (see, for example, CellEvent apoptosis measurements, Figures S2 and S3 in Supplementary Material), which resulted in more than one EC50 value (Table 4). The sign of the effect could also vary depending on the drug. For instance, docetaxel induced an increase in the “Area” parameter in both VCaP and LNCaP spheroids, while all other drug treatments reduced the spheroid size (Figures S2–S5 in Supplementary Material). Furthermore, the obtained EC50 values differed, sometimes significantly, from those measured with the ATP-endpoint assay and often were smaller.

To summarize the HCA results, we plotted all EC50 values measured for a particular drug on the same diagram (Figure 6). Because up to two EC50s could be observed for each phenotypic parameter, the increments “−1” and “−2” were used to indicate lower and higher value, respectively. For each cell line, we traced a “static” threshold line that corresponded to the drug concentration at which no change in spheroid ATP content was observed within a given treatment time [2 days for LNCaP (red line) and 4 days for VCaP (green line)]. Using these thresholds, the observed EC50 values were tentatively assigned to cytostatic (value below threshold) or cytotoxic (above threshold) drug action. As a result, some treatments appeared mostly cytostatic (etoposide and ARN-509 in both cell lines, docetaxel in LNCaP), whereas others showed a concentration-dependent cytostatic-to-cytotoxic transition (cisplatin and MLN in both cell lines) or purely cytotoxic effect (docetaxel in VCaP) (Figure 6).

It should be noted that conditions underlying spheroid growth arrest were different depending on the drug. Thus, the cytostatic effect of low doses of MLN in VCaP and etoposide in LNCaP spheroids was characterized by stabilization of spheroid size, a slight increase in mean LysoTracker fluorescence and the absence of drug-induced apoptosis (Figures 7A,B). The formation of an apoptotic spheroid core was also completely prevented. As we previously showed for MLN, this rather stable quiescent state resulted from G0/G1 cell cycle arrest and could be reversible (77). By contrast, etoposide treatment of VCaP or docetaxel treatment of LNCaP spheroids was characterized by a significant apoptosis rate (Figures 7A,B). Many apoptotic cells were localized on the periphery and in protruding blebs. This suggests that the apparent cytostatic state was a result of a stationary equilibrium between cell proliferation and expulsion of dying cells, as it was previously described for tissue-derived colospheres and organotypic multicellular spheroids (5). This condition seems transient because extension of docetaxel treatment to 4 days resulted in a significant decline in viability of LNCaP spheroids, similar to that observed in VCaP cells.

Finally, despite the observation that the cytotoxic effects of different drugs shared many common characteristics, i.e., ATP decline, apoptosis, decrease in LysoTracker uptake, and compromised integrity (increase in “Shape P2A”), the corresponding spheroid phenotypes were very distinct (Figures 7C,D). This difference apparently reflects a particular MOA of the drugs. As a result, spheroids formed by different cell lines but treated with the same drug had greater resemblance than spheroids formed by the same cell line but treated with different drugs (see, for example, the characteristic “halo” of apoptotic cells in docetaxel-treated spheroids, Figures 7C,D). In our study, the most striking example of the unique drug effects was the asymmetric unilateral spheroid disruption seen in almost all LNCaP spheroids treated with 500 nM MLN (Figures 2D, 3D and 7D).