Establishing Pluripotency in hiPSC Cohorts in High-Throughput Format

To test whether we could detect and characterize variable drug responses across a cohort of hiPSC lines from different donors, the high throughput Cell Painting assay [16] was first adapted for use with hiPSC lines grown in the pluripotent state. We reasoned that an image-based Cell Painting assay could be used to detect variable responses and stratify cell lines accordingly, followed by using quantitative MS-based proteomics to identify potential mechanisms causing the variable responses (see Fig 1A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. (A) Study overview.

Quality controlled hiPSC lines from 43 donors in their pluripotent state were used for Cell Painting and proteomic analysis, by mass spectrometry, following addition of 52 FDA approved drugs. 38.2 million cells were imaged across all donors and 871 morphological traits were quantified per donor/drug combination. Variability of drug response of the hiPSC cohort was studied using both phenotypic screening and proteomics. (B) Heatmap of RobustZ scores from Cell Painting shows variations in the magnitude of drug responses across hiPSC lines. Differential responses to afatinib, atorvastatin, rapamycin and simvastatin are shown. The response to each drug creates a characteristic pattern of features, but with differences in the magnitude of the response. Color bar shows RobustZ scores normalized to DMSO expressed as standard deviations.

https://doi.org/10.1371/journal.pone.0323953.g001

First, we established whether conditions could be established for culturing multiple hiPSC lines in the pluripotent state in 384 multi-well plates, suitable for high content imaging and analysis. Twenty-eight independent hiPSC lines, all derived from different donors (S1 Table, “Run1”), were propagated in mTeSR medium, plated at 3x10⁴ cells/cm² in 384 multi-well plates for 72 hrs and then fixed and analysed by immunofluorescence to detect the established markers of pluripotency, i.e., Oct-4, Sox-2, Nanog and TRA-1–81 (S1A Fig). All four markers showed strong, homogeneous staining.

We conclude, first, that multiple hiPSC lines can be cultured successfully in multi-well format suitable for large-scale, compound profiling by image analysis and second, that the panels of hiPSCs can be maintained in a pluripotent state throughout the 48–72 hours required for conducting the screening protocol.

Using cell painting to profile variable drug responses

Next, the set of 28 independent hiPSC lines from different donors, (hereafter, “donors”), were grown in 384 multi-well plates for 24 hrs (see Methods for details), then treated either with compounds from a library of FDA-approved drugs, or with DMSO carrier, each at a single concentration (S2 Table). After incubation for a further 24 hrs in the presence of the drugs, the cells were fixed and stained with Cell Painting markers (see Materials and Methods). Images were recorded in an automated, plate-based imager, then imported into OMERO (9). Images were segmented with respect to the ‘nuclei’, ‘cytoplasm’, ‘Golgi’ and ‘protrusions’, using a custom CellProfiler pipeline [17]. For each drug-donor combination, image features were calculated, using 871 separate feature measurements, from a total of 6,000-8,000 cells per donor-drug pair. After Robust Z’ normalisation [18–20], all measured features were scaled as the number of standard deviations different from DMSO for each of the hiPSC lines [16].

Fig 1B shows a heatmap of the measured features of the 28 donors, comparing treatment with 4 of the drugs from the library, i.e., atorvastatin, simvastatin, rapamycin and afatinib. The heatmap shows that each drug has characteristic feature patterns, but that different donors show varied magnitudes of response, visible as either darker, or lighter, horizontal stripes, respectively (Fig 1B). The feature patterns for atorvastatin and simvastatin are similar, consistent with them being competitive inhibitors of the same protein target, i.e., HMG CoA reductase, the rate-limiting enzyme of the cholesterol biosynthetic pathway. Interestingly, atorvastatin, which has a higher target affinity, also shows a stronger response than simvastatin in this assay.

These data demonstrate that the Cell Painting assay, which was originally developed for analysis of immortalised cancer lines [16], also generates characteristic phenotypic fingerprints in hiPSC lines. Moreover, the data show that Cell Painting fingerprints from hiPSC cohorts reveal variable responses between the lines from different donors. The major variation corresponds to differences in the magnitude of features.

Next, the assay was expanded to profile the hiPSC cohort responses to 52 FDA-approved drugs, which are used to treat a range of different diseases (S2 Table). A subset of the drugs tested are shown for visibility with full data available on Image Data Resource IDR [21]. To remove highly correlated features, we calculated the Spearman correlation coefficient, s, and removed all features where s > 0.98, which left 442 filtered features for further analysis.

Fig 2A shows a heatmap of the Cell Painting induction, calculated as the fraction of filtered features with RobustZ > 2σ from DMSO (i.e., significant with >95% certainty) [22]. This effectively reduces the filtered features for each drug-donor pair to a single statistic, which may underestimate the complexity of phenotypes. Despite this limitation, Fig 2A shows clearly that, first, hiPSC lines derived from different donors exhibit different levels of induction when treated with the same drugs and second, that examples of such differential response are observed between cell lines for all of the 52 different drugs tested.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. (A) Induction plot of drug responses for each hiPSC line shows variations in response magnitude.

Induction is the fraction of features for each combination of drug and hiPSC line that differ in value by > 2σ (standard deviation) away from the DMSO control. (B) UMAP embedding of Cell Painting features measured for 28 hiPSC lines treated with different drugs. Cell lines are represented on the graph with different shapes and drugs with different colours. Representative separated clusters are indicated by arrows. For visibility, a subset of drugs used in this screen are shown. (C) PCA analysis of Cell Painting features measured for human iPSC lines from phenotypic screens performed in 2021−02 (28 lines) and 2023−12 (18 lines). Runs are represented on the graph with different shapes (“Run 2021-02” solid squares and “Run 2023-12” crosses) and drugs with different colours. The reproducibility of the assay results on different dates is indicated by the overlap of drug treatments in the different assay runs.

https://doi.org/10.1371/journal.pone.0323953.g002

In all but one case, the individual cell lines show varied responses and appear to be either hyper- or hyposensitive to specific drugs. The exception was the cell line voce2, which showed a higher induction level to several, but not all drugs. Combining the results shown in Fig 1B (feature heatmap) and Fig 2A (induction heatmap), we conclude that a significant source of variable responses between donor cell lines seen in the Cell Painting assay results from differences in the magnitude of response shown by each line.

We next asked if phenotypes caused by different compounds could be distinguished across the cohort of hiPSC lines. The filtered features from all drug-donor pairs were subjected to embedding with UMAP, which fits a manifold to a high-dimensional dataset and then projects this to a 2D plane [23] (Fig 2B). This showed that compound responses clustered according to their known mechanism of action. For example, statins, everolimus and rapamycin, topoisomerase inhibitors, cytotoxics and microtubule depolymerizers all formed separate, individual clusters. Interestingly, fluphenazine, an antipsychotic that acts by blocking specific dopamine receptors in the brain and that is used to treat schizophrenia and other psychotic disorders, clustered near the statins. While clustering in UMAP is not always a reliable indicator of similarity, we also noted the pattern of induction scores for fluphenazine was similar to that observed for atorvastatin (Fig 2A). These results are consistent with reports that fluphenazine, like statins, affects lipid metabolism in human patients [24].

Besides showing that many drugs with similar modes of action cluster in this assay, the data also reveal cell lines that show variable responses to the same drugs. For example, when treated with rotenone, which disrupts microtubules [25], most hiPSC lines cluster together (“MT Modulators, Fig 2B). An exception is the nufh4 cell line (pink and blue dot), which clusters instead with hiPSC lines treated with fenbendazole. This is a drug which also causes microtubule depolymerisation, but by a different mechanism to rotenone, which inhibits microtubule assembly dynamics [26] (filled circles with trefoil, Fig 2B). The nufh4 line consistently shows a lower induction score when treated with microtubule deploymerizers (Fig 2A), which is consistent with nufh4 having a different response to rotenone than the other hiPSC lines.

To test whether these results were reproducible, a selection of the previously analysed hiPSC lines (S1 Table) were regrown from frozen stocks and analysed in Cell Painting assays after treatment with the same set of FDA-approved drugs, but this time using a different High Content Screening (HCS) imaging system (data collected on a Yokogawa CV7000 instead of InCell 2200). Using UMAP for visualisation, we observed similar clustering of Cell Painting features generated in this repeat set of assays to those collected from the original experiment that was performed 6 months previously, using independent cultures of each hiPSC line and with data acquired using a different type of HCS imaging system (S1B Fig).

To further asses the reproducibility of the assay, we compared the measured phenotypes of hiPSC lines in an additional run of the assay, executed after an interval of 2 yrs 8 months (both datasets were collected on the same InCell 2200 imager) but with a different cohort of hiPSC lines (S1 Table). Fig 2C shows the comparison of the two assays, using Principal Component Analysis (PCA). This further validates the comparison between the two different runs and also avoids the effects of clustering in high-dimensional space that can affect UMAP [27]. PCA analysis shows that similar compounds are indistinguishable in the two assay runs. S1 (C and E) Fig shows a comparison of the two assays and again reveals that phenotypic responses are similar. As expected, PCA and UMAP show different clustering properties and different variations across cell lines, but the specific clustering responses of cell lines treated with different classes of drugs, as well as the variable responses observed to the same drug, between cell lines derived from different donors, are reproducible in this assay format. We conclude this represents intrinsic properties of the respective cell lines, independent of variables specific to individual assay runs, such as specific cell line stocks, prepared media, or the instrumentation used for image acquisition.

While many of the drug treatments resulted in tight clustering of the responses of different cell lines on the UMAP and PCA plots (Fig 2), a subset of the drugs tested, i.e., statins, 5-FU and erlotinib, showed more dispersed clustering of cell line responses. We hypothesised that this behaviour reflected an increased degree of variable response to these drugs between the respective cell lines. To test this hypothesis, we calculated the Euclidean distance of the feature vectors measured for each compound and DMSO, for each donor. S3B Fig shows the result of this calculation, along with the mean and coefficient of variation (CoV) of the Euclidean distance for each drug across donors. This metric is elevated for simvastatin, atorvastatin, erlotinib and fenbendazole, consistent with the evidence from UMAP and the induction plot that response to these drugs varies across the donors [27].

The variable responses shown in Fig 2 and S3B Fig might be due to intrinsic variations within an assay. To directly assess the variations in Cell Painting features in different assays, we calculated the Earth Mover Distance (EMD, see Methods) which has been shown previously to provide a useful measure of differences between phenotypic features in HCS assays [28]. S2A Fig shows the distribution of EMDs for all features from all compound treatments common between these two assay runs. Most show EMDs similar to DMSO, suggesting the differences are in the range of the assay controls. Cytotoxic compounds (e.g., bortezimib, dasatinib, colchicine) show a broader distribution of EMDs, which is likely due to cell death caused by these compounds during incubation. Inspired by the overall similarities in features between the two assays, we calculated induction scores for the drugs and donors that are common across these assays. Comparison of heatmaps of these induction scores shows the variations between drugs, but also between donors within individual drugs, are similar (S2B, C Fig). These data suggest that the variations we observe in drug responses (e.g., Figs 1 and 2) across hiPSC lines generated from different donors are robust and reproducible.

Linking variable responses to protein expression & stoichiometry

Our previous studies on this HipSci cohort have demonstrated that variations in protein expression levels between lines derived from different donors can be controlled by specific genomic loci, i.e., protein Quantitative Trait Loci (pQTLs) [13]. This suggested the hypothesis that a potential source of differential drug responses between hiPSC lines might be due to epistatic interactions, i.e., resulting from genomic variation between donors causing differences in protein expression that can modulate drug-induced phenotypes, for example by affecting the stoichiometry of components of drug response pathways [29]. Importantly, this hypothesis takes into account drug-cell interactions outside of the direct interaction of a drug with its protein target. Furthermore, it can potentially explain why patterns of drug response features measured by Cell Painting appear similar across cell lines, while the magnitude of responses between lines differ, as seen in the feature and induction heatmaps (Figs 1B and 2A).

To test this hypothesis, we first used the induction plot (Fig 2A) and a graph of induction values (S3A Fig), to classify cell lines as, respectively, “low” and “high” responders, for two well-characterised drugs, i.e., atorvastatin and simvastatin. These drugs share the same target, but are well-known to have different responses and efficacies in different individuals [30]. Representative cell lines denw6 and zaie1 (“low responders”) and hayt1 and tuju1 (“high responders”), were each treated separately with either DMSO (control), or with either atorvastatin, or simvastatin, each at 5 µM, for 24 hrs, as in the Cell Painting assay. Following treatment, cell extracts were prepared and proteomes analysed by LC-MS/MS (see Methods; quantitative proteomics were performed on a small subselection of donors and compounds because the culture of cells required a larger culture format). From the resulting data, differences in protein expression levels between DMSO- and drug-treated lines were visualised in ‘volcano plots’ (Fig 3). Combining data for treatment of all four cell lines (tuju1, hayt1, denw6, zaie1), with either atorvastatin (Fig 3A), or simvastatin (Fig 3B), showed that many of the proteins whose expression increased after drug treatment are involved in lipid metabolism (red dots in Fig 3A, B). Further, GO term enrichment analysis also showed a significant enrichment for expression of protein factors involved in lipid metabolism (S3 and S4 Tables).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Volcano plots showing differential protein expression relative to DMSO in all four (tuju1, hayt1, denw6 and zaie1), hiPSC lines, treated with either atorvastatin (A) or simvastatin (B).

The protein expression fold change (FC) and p value were calculated using Perseus. All sample runs (n = 12), from each drug, were considered as one group, and the DMSO control set (n = 12) was considered as a separate group. Representative proteins that play a role in the HMG-CoA reductase pathway are highlighted in red and show increased expression following drug treatment.

https://doi.org/10.1371/journal.pone.0323953.g003

These data suggest that homeostatic mechanisms in the hiPSC lines respond to drug mediated inhibition of cholesterol metabolism by increasing the expression of enzymes required for cholesterol synthesis. While a common type of response to inhibition of cholesterol synthesis is seen across the hiPSC lines, differences were evident in the magnitude of responses between individual cell lines. This is illustrated by the response to either atorvastatin, or simvastatin, as detected by Cell Painting induction (Fig 2A). We therefore tested whether variation in response magnitude might be linked to differences in protein expression between the respective high and low response lines. Volcano plots comparing protein expression in the high and low response lines, after treatment with either atorvastatin (Fig 4A), or simvastatin (S4A Fig), show that multiple proteins involved in the cholesterol synthesis pathway (highlighted in purple), together with some other classes of proteins, have increased expression specifically in the high response lines. These data are consistent with the hypothesis that differential protein expression may explain, at least in part, the differences in the magnitude of drug responses between individual hiPSC lines.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. (A) Volcano plot of differential protein expression in four hiPSC lines, stratified by their respective high, or low, response to atorvastatin (see

S3A Fig). High response lines: tuju1 and hayt1. Low response lines: denw6 and zaie1. Protein FC and p value for each cell line were calculated as in Fig 3. Sample runs from the drug treatment set (n = 6), were considered as one group and the DMSO control set (n = 6), were considered as a separate group. Members of the cholesterol biosynthesis pathway are represented as purple dots and listed on the right side of the graph. (B) Heat map analysis of proteins showing differential expression (log2FC_high-low > 0.263) between the respective high and low response hiPSC lines, following atorvastatin treatment. Heatmap generation steps: the protein fold change (FC_high) and p value (P_high) for high response cell lines were calculated using sample runs from both high response cell lines to Atorvastatin set (n = 6), considered as one group and the DMSO control set (n = 6) considered as a separate group. The protein fold change (FC_low) and p value (P_low) for low response cell lines were also calculated. Only proteins with both P_high ≤0.05 and P_low ≤ 0.05 were considered further to compare the difference between high and low response. log2FC of these filtered proteins was calculated as (log2FC_high-low = log2FC_high- log2FC_low). Heatmap shows FC between drug and DMSO control of the filtered list z-score normalised. StringDB analysis of filtered proteins shown in (B), in high response lines (C) and low response lines (D).

https://doi.org/10.1371/journal.pone.0323953.g004

Next, we compared the expression of individual proteins in each of the four drug-treated lines (Fig 4B). Proteins were identified whose expression differed significantly in high vs low responders (|log₂FC| > 0.263 and p < 0.05, see Materials and Methods). This reveals clear differences in the responses to the two statin drugs, suggesting that there may be variation in the mechanisms of response to statin treatment between each hiPSC line (see also S4B Fig; S9 Table).

Interestingly, changes in protein expression in the high response lines include increased expression of HMGCR and HMGCS1, which are directly involved in the cholesterol mevalonate synthesis pathway. In contrast, both “low response lines”, show increased expression of proteins involved in regulating the stability and recycling of lipoprotein receptor complexes. This includes PCSK9, a protein that provides post translational regulation of LDL receptor and directs it for lysosomal degradation, rather than regular recycling potentially limiting the efficacy of statins. In addition, SCARB1, a cell surface HDL receptor that mediates HDL-cholesteryl ester uptake, as well as ApoE, a plasma protein that participates in the transport of cholesterol and other lipids between cells and has been linked to the effect of statins [31–39]. Lipoprotein receptor complexes play an important role in the regulation of plasma and intracellular cholesterol homeostasis [40]. These differences in protein expression between the respective high and low responder hiPSC lines to statin treatment suggest that variable activity of biosynthetic and/or influx mechanisms may be involved in determining how each hiPSC line responds to statins. In addition to identifying well-established targets of statins, characterising the proteins whose expression increases in the high response lines, using GO enrichment and STRING analysis (https://string-db.org/), identified factors involved in Rab protein signal transduction and in the metabolism of RNA (Fig 4C and S4C Fig; S5 and S7 Tables). These are all known to change expression upon statin treatment [41–45]. In contrast, the low response lines showed increased expression of proteins linked to the cytoskeleton, along with several ubiquitin E2 and E3 enzymes, suggesting that ubiquitin-conjugation levels may affect the degree of statin response (Fig 4D and S4D Fig; S6 and S8 Tables).

These data support the view that a diversity of factors and pathways can affect the cellular response to well-defined drug perturbations and these differences are due, at least in part, to population-level variations in gene and protein expression between the respective lines.

To explore further the observed variable drug responses, we calculated the pairwise correlation matrix for the induction metric (see Fig 2A) for 32 drugs across donors. This provides a measure of the pattern of variable response across a given set of donors. S5 Fig shows a visualisation of these data in a heatmap and identifies sets of compounds that share common patterns of variable response. For example, erlotinib and afatinib, both ErbB family receptor inhibitors and the statins and fluphenazine, behave similarly. Everolimus, etoposide and irinotecan share a similar variable response pattern, which may be linked to a role for the mTor pathway in controlling the DNA damage response [46,47]. In addition, warfarin, omeprazole, niclosamide and metoclopramide, which have very different targets, nonetheless share a similar variable response pattern across hiPSC lines. We hypothesise that this may be linked to a shared off-target effect, e.g., mitotoxicity [48]. This is an interesting point to explore in further studies in the future.

Cell culture

Human iPSC lines used in this study were from the HipSci cohort (4). Feeder-free human iPSC lines were cultured in Essential 8 (E8) medium (E8 complete medium supplemented with (50x) E8 supplement ThermoFisher-A1517001) on tissue-culture dishes coated with 10 µg/cm² of reduced Growth Factor Basement Membrane matrix (Geltrex, ThermoFisher A1413202 resuspended in basal medium DMEM/F12 Thermo Fisher 21331020). Medium was changed daily.

To passage feeder-free hiPSC lines, cells were washed with PBS and incubated briefly (3–5 min), with 0.5 mM PBS-EDTA solution. After PBS-EDTA solution was removed, cell clusters were resuspended in E8 medium and seeded at ratios of 1:4–1:6, depending on their confluency, on Geltrex-coated tissue culture dishes. Once established, undifferentiated, feeder-free human iPSC lines were expanded and aliquots frozen for future use, prior to transition to TeSR medium for further study. To transition to TeSR medium supplemented with bFGF (Peprotech 100–1813, at a final concentration of 30 ng/ml), Noggin (Peprotech 120-10C, at a final concentration of 10ng/ml), Activin A (Peprotech 120-14P, at a final concentration of 10ng/ml), E8 medium of hiPSC lines was exchanged daily with medium containing increasing amounts of TeSR medium, until human iPSC lines were completely maintained in TeSR medium alone, typically over 1–2 weeks.

To passage human iPSC lines after transition to TeSR medium, cells were washed with PBS and incubated briefly with TrypLE Select (ThermoFisher 12563029), to create a single cell suspension prior to resuspension in TeSR medium supplemented with ROCK kinase inhibitor (Tocris 1254, at a final concentration of 10 µM), to enhance single cell survival, bFGF (Peprotech 100–1813, at a final concentration of 30 ng/ml), Noggin (Peprotech 120-10C, at a final concentration of 10ng/ml), Activin A (Peprotech 120-14P, at a final concentration of 10ng/ml) and seeded at a concentration of 5x10⁴ cells/cm² on Geltrex-coated tissue culture dishes. 24 hrs after replating Rock kinase inhibitor was removed from the culture medium. Pluripotency of TeSR-transitioned hiPSC lines was verified by immunofluorescence (see below) prior to either cell banking, or experimental use. Cells were grown at 37^oC and 5% CO₂. Medium was changed daily.

hiPSC line ID number and details of the cell lines are listed in S1 Table.

The data from the experimental runs are included in this report. Run 2021−02, 52 drugs, 28 donors 1,456 drug-donors pairs). Run 2021−10, 8 donors, 52 drugs (416 drug-donor pairs). Run 2023−12, 18 donors, 52drug (936 drug-donor pairs).

Cell staining

384-well tissue culture plates (CellCarrier-384 Ultra Microplates, Perkin Elmer 6057300) were coated with human plasma fibronectin (Merck Millipore FC010) at a concentration of 5 µg/cm². Cells were passaged manually as described above with TrypLE Select, counted and seeded on the fibronectin-coated wells, at a concentration of 3x10⁴/cm². Cell line plating was in rows, with six wells per condition for each cell line (six technical replicates). Each replicate was sampled at 9 locations and repeated in a separate biological replicate separated by a week (i.e., 2 bioreplicates per assay run).

Cells were incubated for 24 hrs prior to drug treatment, followed by a further 24 hrs incubation before final fixation, staining and high content imaging. Antibody staining of a separate 384-well plate plated for each biological replicate, at 48 hrs, confirmed that pluripotency was maintained at 48 hrs.

Mitotracker (Thermo Fisher M22426), staining was performed in live cells with a 30 min incubation at 37^oC prior to fixation. Fixation, permeabilisation and immunostaining were performed at room temperature. 8% Paraformaldehyde in PBS (PFA, Sigma F8775), was added to an equal volume of cell culture medium in wells of a 384-well plate for a final concentration of 4% Paraformaldehyde and incubated for 15 min. Fixed cells were washed in PBS and permeabilised with 0.1% v/v Triton-X100 (Sigma) for 15 min. Following permeabilisation, cells were washed in TBS and stained with DAPI (Thermo Fisher D1306 at 5 µg/ml), SYTO 14 (Thermo Fisher S7576 at 2.5 µM), Concanavalin A (Thermo Fisher C11252 at 20 µg/ml), Phalloidin Alexa (Thermo Fisher A12381 at 66 nM), and WGA (Thermo Fisher W11262 at 11.8 µg/ml) for 1hr. Staining volume was 30 µl/well. Cell fixation, permeabilisation, staining and washing steps (2x, wash volume 50 µl), were performed using automated liquid handling systems (405 plate washer, BioTek– Tempest, Formulatrix and HighRes Biosolutions Automation systems).

Drug treatment

Drug dosing on 384-well plates was performed using an Echo acoustic liquid handler and corresponding software (Labcyte). Screened compounds were selected from the appropriate chemical library plates containing either Cloud library (Enamine), or an in-house library of publicly available compounds. Drug dosing was performed at final concentrations of; etoposide 1.7 µM, Berberine Chloride 2.68 µM and Fenbendazole 3.34 µM. For all other compounds drug dosing was performed at a final concentration of 5 µM. S2 Table lists the compounds and their final concentrations used in this study. For proteomics analysis drug treatments with either simvastatin (Tocris 1965), or atorvastatin; (Selleckchem S5715) were performed in triplicate in 6-well plates at a final concentration of 5 µM for 24hrs.

Cell staining and analysis of pluripotency markers

All cell lines used in this study were quality controlled for pluripotency prior and during the HCS assays. For each experiment 8 wells of each cell line were individually stained with pluripotency marker antibodies.

All fixation, permeabilisation and immunostaining analyses were performed at room temperature, apart from primary antibody incubation, which was carried out at 4^oC overnight. 8% Paraformaldehyde in PBS (PFA, Sigma F8775), was added to an equal volume of cell culture medium in wells of a 384-well plate for a final concentration of 4% Paraformaldehyde and incubated for 15 min. Fixed cells were washed twice in PBS and permeabilised with 0.5% v/v Triton-X100 (Sigma) for 10 min (apart from TRA1–81 staining). Following permeabilization cells were washed twice in PBS, then blocked in 10% normal donkey serum for 1 hr prior to antibody incubations. Pluripotency marker antibodies used in this study are: Nanog, Oct4, Sox2, Tra-1–81 all at 1:500, (Oct4 (C30A3), Sox2 (D6D9), Nanog (D73G4) and TRA181 (TRA-1–81) Cell Signalling StemLight Pluripotency Antibody Kit 9656). All affinity purified donkey secondary antibodies (either Alexa 488 or Alexa 594) were purchased from Jackson Immunoresearch Europe. Hoechst 33342 (Thermo Fisher) was used to stain DNA. Images of cells (for each hiPSC line, 9 fields/well, 16 wells per pluripotency marker), stained with pluripotency marker antibodies were segmented based on DAPI signal and the integrated nuclear intensity of each marker was quantified using CellProfiler. Distributions of integrated intensities of all nuclear pluripotency markers were monotonic and correlated with integrated DAPI staining.

Imaging

Imaging data were collected on an InCell 2200 high content imager, with a 20x/NA 0.45 Plan Fluor lens and a Quad2 multichroic mirror (GE, Issaquah, WA) or a CV7000 high content imager, with a 20x/0.45 lens at epifluorescence mode (Yokogawa). For each assay, we performed 6 technical replicates (each in separate wells), with each replicate sampled at 9 locations, and repeated in a separate biological replicate separated by a week (i.e., 2 bio replicates per assay run). All of these data are then combined into a single well measurement, producing an n-dimensional data vector per well.

Data processing, analysis and visualization of multiple cell lines

Raw images were imported into OMERO Plus (Glencoe Software, Inc., [54]) and then processed using a custom pipeline in CellProfiler [16], which segmented nuclei, cytoplasm, Golgi and cortical protrusions and calculated a range of defined features. All further steps were executed using the Pandas Python library and Numpy libraries [55,56]. Features for each plate were Robust-Z normalised by the median of features for the DMSO control for that plate [16,57].The Robust Z-score converts the population of features derived from individual cells to a distribution of zero mean and unit variance. Robust Z is calculated according to the formula x’ _i =(x_i-µ)/Μ where x _i is the measured value for each feature, x’ _i is the normalised output value and µ and Μ are the median and median absolute deviation [18–20]. Features with a coefficient of variation >0.50 were removed from further analysis. We further removed all features with |Spearman coefficient| > 0.98. Data were then visualised using Uniform Manifold Approximation and Projection (UMAP [23]). UMAP parameters were n_neighbours = 15, mindist = 0.01 and a cosine metric was used for scaling the distance between points. GO enrichment analysis was carried out using WebGestalt (WEB-based Gene SeT AnaLysis Toolkit) [58,59]. Functional protein-protein interaction analysis was carried out using STRING-DB [60].

Nearest neighbour distances were calculated using Euclidean distance in scikit-learn [61]. The Earth Mover Distance (EMD) was calculated using sci-py [62] for filtered features from individual wells for each drug-donor pair across a pair of assays. The frequency distribution of EMDs for filtered features was calculated and plotted using Seaborn [63].

Proteomics sample preparation

To support parallel MS-based proteomics analysis in this screening platform in a 6-well plate format, we also tested whether the expression of pluripotency markers was maintained when hiPSC lines were plated at 5x10⁴ cells/cm², prior to any drug treatment and cultured as previously described. This showed that strong, homogeneous expression of the four pluripotency markers was maintained at this higher cell density (not shown).

All hiPSCs samples were washed twice with 1x PBS, on ice, then centrifuged at 300g to collect the cell pellets. 4% SDS lysis buffer (10 mM TCEP, 100 mM Tris-HCl, pH 7.4) was used to lyse the cells and extract proteins, with 1x Protease and Phosphatase Inhibitor (A32961, Thermo Fisher) added. 1 mL lysis buffer was added to the cell pellets. After 10 cycles of sonication, (30s on/off), using the Bioruptor® Pico, the protein samples were centrifuged at 20,000xg for 15 min. The supernatant proteins were denatured at 95^oC for 10 min, then alkylated with 40 mM IAA (Iodoacetamide), (final concentration) and incubated in the dark at room temperature for 30 min. The protein concentration was measured using the EZQ™ Protein Quantitation Kit following the manufacturer’s instructions.

500 ug proteins from each sample were further processed using the SP3 protocol, as described [64]. In brief, protein samples were mixed with SP3 beads (1:10, protein:beads) and digested with LysC/trypsin mixture (1:50, enzyme:protein). Peptides were eluted from the SP3 beads with 2% DMSO. The peptide concentration was measured using the Pierce™ Quantitative Fluorometric Peptide Assay following the manufacturer’s instructions. Peptide samples were stored at −20°C before LC-MS/MS analysis.

LC-MS/MS analysis

For each cell line, three biological replicates per condition were generated. They were combined in one sample, which was analysed in three separate, replicate MS runs. Each hiPSC line was treated three times with either vehicle control, or specified drug.

All peptide samples were analysed by using an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific), equipped with a Dionex UltiMate™ 3000 RSLCnano System, Thermo Fisher Scientific). Peptides from each sample were loaded onto a 75 μm × 2 cm PepMap-C18 pre-column and resolved on a 75 μm × 50 cm PepMap-C18 EASY-Spray temperature-controlled integrated column-emitter (Thermo Fisher Scientific), using either a 160 min or a 280 min multistep gradient from 5% B to 95% B with a constant flow rate of 300 nl/min. The mobile phases are: H₂O incorporating 0.1% FA (solvent A) and 80% ACN in H₂O incorporating 0.1% FA (solvent B). The details for 160 min gradient are: 0–6 min, 1%−5% B; 6–145 min, 5%−38% B; 145–155 min, 38%−95% B; 155–160 min, 95%B. The details for 280 min gradient are: 0–6 min, 1%−5% B; 6–255 min, 5%−38% B; 255–270 min, 38%−95% B; 270–280 min, 95% B.

For total proteome analysis, the data were acquired under the control of Xcalibur software in data-independent acquisition mode (DIA). For each sample, one run of 280 min method with 97 DIA windows was performed for boosting the identification of proteins [65] and another three replicate runs of 160 min method were performed for the quantification of proteins.

For the 160 min gradient method, the survey scan was acquired in the Orbitrap covering the m/z range from 375 to 1,175 Da with a mass resolution of 60,000 and an automatic gain control (AGC) target of 4.0 × 10⁵ ions. MS/MS was acquired in the Orbitrap covering the m/z range from 200 to 2,000 Da with a mass resolution of 30,000 and an AGC target of 1.0 × 10⁵ and 60 ms maximum injection time. The sequential DIA isolation windows are 10 m/z from 375 to 775 and 40 m/z from 775–1175. Fragmentation was performed using a HCD collision energy of 30%. For the 280 min gradient method, the survey scan was acquired in the Orbitrap covering the m/z range from 355 to 1,150 Da with a mass resolution of 120,000 and an automatic gain control (AGC) target of 4.0 × 10⁵ ions. MS/MS was acquired in the Orbitrap covering the m/z range from 200 to 2,000 Da with a mass resolution of 30,000 and an AGC target of 1.0 × 10⁵ and 60 ms maximum injection time. The sequential DIA isolation windows are 15 m/z from 350 to 410, 10 m/z from 410 to 440, 15 m/z from 440 to 645, 5 m/z from 645 to 775, 10 m/z from 775 to 875, 15 m/z from 875 to 965 and 27 m/z from 965–1150. The fragmentation was performed using a HCD collision energy of 30%.

DIA MS data analysis

For DIA analysis, the raw MS data of different sample runs from both drug and DMSO set were analysed together using DIA-NN v1.8.1 [66]. The in silico spectral library was generated by DIA-NN, using the Homo sapiens database from UniProt (SwissProt October 2021). The data were searched with the following parameters: stable modification of carbamidomethyl (C), N-term M excision, with a maximum of 1 missed tryptic cleavage threshold, match between runs (MBR) option checked. DIA-NN output reports were further processed with an R package, as described previously [67] to perform the MaxLFQ-based protein quantification, with precursors x samples matrix filtered at 1% precursor and protein group FDR. Statistical analyses were performed using Perseus v1.6.15.0 [68,69] In brief, the protein quantification result was imported to Perseus and the copy number of each protein was calculated based on its intensity using a “Proteomic Ruler” [70]. Only proteins with valid quantification values in at least 1/3 of total sample runs were considered for further comparison. Missing values were replaced from a normal distribution. After median normalisation, two sample student’s T tests were performed for p-value generation, using permutation-based FDR for truncation. Heatmap and K-means clustering analysis was performed based on the Euclidean distance, using the default settings in Perseus. In total, 5,895 proteins were quantified using DIA analysis and used for further analysis.

For heat map generation of Fig 4B and S4B Fig: At the first step of differential expression analysis, a two sample-test using Student’s T-test with the default setting in Perseus was performed to determine if the means of two groups (drug and DMSO, for either high or low response cell lines) are significantly different from each other.

The protein fold change (FC_{high or low}) and p value (P_{high or low}) were generated, respectively.

Proteins with p < 0.05 were considered further and used to create the results show in Fig 4B and S4B Fig. Z-score normalisation was used for Heat map analysis, according to the protocol in [69]. Data available at PRIDE [71].