The propensity of human embryonic stem cells to die upon enzymatic disaggregation or low-density plating is an obstacle to their isolation and routine use in drug discovery and basic research. Equally, the very low rate of establishment of implanted cells hinders cell therapy. In the present study we have developed a high-content assay for human embryonic stem cell survival and used this to screen a range of libraries of ‘lead-like’ small molecules and known bioactives. From this we identified 18 confirmed hits with four structural classes being represented by multiple compounds: a series of 5-(acyl/alkyl-amino)indazoles, compounds with a 4-(acylamino)pyridine core, simple N6,N6-dialkyladenines and compounds with a 5-(acylamino)indolinone core. In vitro kinase profiling indicated that the ROCK (Rho-associated kinase)/PRK2 (protein kinase C-related kinase 2) protein kinases are of pivotal importance for cell survival and identified previously unreported compound classes that inhibited this important biological activity. An evaluation using an extensive panel of protein kinases showed that six of our hit compounds exhibited better selectivity for ROCK inhibition than the routinely used commercially available ROCK inhibitor Y-27632. In this screen we also identified the K+-ATP channel opener pinacidil and show that it probably promotes cell survival, by ‘off-target’ inhibition of ROCK/PRK2. We have therefore identified novel pro-survival compounds of greater specificity, equivalent potency and reduced toxicity relative to the routinely employed ROCK inhibitor Y-27632.
- drug discovery
- high-content screening
- human embryonic stem cell
- protein kinase inhibitor
- Rhoassociated kinase (ROCK)
Poor recovery after single-cell enzymatic passage of hES cells (human embryonic stem cells) is a considerable obstacle to accurate cell counting and efficient culture expansion. This is of particular importance when using hES cells for drug discovery, where accurate control over cell number is critical, with evenly distributed cells more amenable to high-content screening. Use of the small-molecule inhibitor Y-27632 has revealed a role for the ROCK1/2 (Rho-associated kinase 1/2) in preventing the severe loss of viability upon cell–cell dissociation . Subsequent studies demonstrated the utility of Y-27632 in enhancing the survival of cryopreserved hES cells [2,3]. The mechanism by which ROCK inhibition promotes survival is still not fully understood, but may occur via inhibition of apoptosis . ROCKs have been targeted by the pharmaceutical industry for their important role in vascular smooth muscle contraction and inhibitors, such as HA-1077 (fasudil), are used to treat cerebral vasospasm and in ischaemic preconditioning (reviewed in ). The most commonly employed tool compounds for ROCK1/2, such as Y-27632 and HA-1077, are only moderately selective . The related Rho-associated kinase, PRK2 (protein kinase C-related kinase 2) appears to share some common functions and is also targeted by these inhibitors .
In order to identify compounds promoting hES cell survival with higher efficacy, improved specificity and lower toxicity, we developed a 96-well format high-content screening assay. We screened the Prestwick library of known pharmacologically active agents (1200 compounds), a focused library of lead-like small-molecule scaffolds targeted for binding in the ATP site of protein kinases (4100 compounds) and a lead-like diversity set (15000 compounds). In the present paper we describe the identification and initial characterization of pro-survival ‘hit’ compounds and their in vitro kinase inhibition profiles using a panel of >70 protein kinases. Despite screening a wide range of compounds we demonstrate that the identified pro-survival compounds all target ROCK2/PRK2 kinases in vitro. Furthermore, we show that pinacidil, a K+-ATP channel opener, often used for ischaemic preconditioning in organ transplantation , has hES cell pro-survival effects via a mechanism unrelated to its effects on ion channel pharmacology; again this compound typifies the actives identified through this screen by inhibiting the ROCK2/PRK2 kinases in vitro. Our lead molecule is effective in promoting survival in three different hES cell lines. Together our results emphasize the critical importance of this signalling pathway in hES cell survival.
hES cell culture
hES cell lines SA461, SA121, SA181, SA167 and SA002 (Cellartis AB, Dundee, Scotland, U.K.) were maintained in a FF (feeder-free) system in a fully pluripotent state for prolonged periods by passaging every 3–4 days at a density of (3–5)×104 cells/cm2 on fibronectin (Merck Chemicals) in chemically defined medium  mixed at a 1:1 ratio with MEF-CM (mouse embryonic fibroblast-derived conditioned medium) or, if stated, HuWIL-CM (human foreskin fibroblast-derived conditioned medium), supplemented with 10 ng/ml bFGF [basic FGF (fibroblast growth factor); Invitrogen]. Karyotypically normal SA461 cells were used in the screening process and were fully pluripotent, as judged by Oct3/4 (octamer 3/4) and Nanog immunofluorescence, throughout the duration of the screening campaign.
Compound collection and storage
Two in-house collections of lead-like molecules assembled from commercial suppliers  were utilized: a kinase-inhibitor-focused set (4110 compounds) and a small diversity collection (15677 compounds). We also utilized a 1200-compound bioactives set of marketed drugs and pharmacologically active agents (Prestwick Chemicals). In-house compound collections were formatted into 96-well daughter sets at 3 mM in 100% DMSO and active working sets were stored for up to 4 months under nitrogen gas and with low humidity at room temperature (26 °C) in custom-made storage pods (FluidX).
Primary screening was performed in single-point with a final compound concentration of 30 μM. CellBind® plates (96-well; Corning) were pre-coated with 0.1 mg/ml fibronectin (Calbiochem) diluted in PBS (with Ca2+ and Mg2+), stored overnight at 4 °C and then aspirated immediately prior to the addition of medium. Cell plates were prepared by dispensing 50 μl of FF medium with bFGF into each well using a Wellmate dispensing robot (Thermo Fisher Scientific). A total of 1 μl of each library compound (30 μM final concentration) was dispensed to 88 individual wells (columns 1–11) per plate using a BioMek-FX robot (Beckman Coulter). Y-27632 was used as the positive control (four wells in column 12, final concentration 10 μM) and DMSO as the negative control (four wells in column 12, final concentration 1%). Cells were washed with PBS, then dissociated for 8–12 min using TrypLE™ Select (Invitrogen) followed by the addition of FF medium to arrest dissociation. After trituration in a 10-ml pipette, clumps were removed using a 0.45 μm tube insert filter. Cells were counted using a haemocytometer or by using a Cellometer automated cell counter (Peqlab). Appropriate volumes were diluted with FF medium to give 100000 cells/ml. A total of 5000 cells/well in a 50 μl volume were dispensed into each well of the pre-prepared 96-well plates, containing the library compounds, using a Wellmate robot (Thermo Fisher Scientific). Plates were incubated for 24 h at 37 °C under a 5% CO2 atmosphere before being processed for imaging.
Processing of cell plates was performed using a BioMek-FX robot (Beckman Coulter) incorporating an ELX405 plate washer (BioTek Instruments). Cell medium was removed using the plate washer and wells were washed twice with PBS followed by fixation at 37 °C with 100 μl of warm 4% (w/v) paraformaldehyde in PBS (prepared fresh from 40% stock solution; Sigma). After 10 min at 37 °C, cells were washed and permeabilized in TBS (Tris-buffered saline) with 0.1% Triton X-100. After a further 10 min at 37 °C, cells were washed twice in TBS and incubated with 1 ng/ml DAPI (4′,6-diamidino-2-phenylindol) for 5 min at room temperature before two washes in TBS and aspiration. Then 50 μl of anti-fade mounting medium [0.5% p-phenylenediamine in 20 mM Tris/HCl, pH 8.8, in 90% (v/v) glycerol] was added to each well. Imaging was performed with an IN Cell Analyzer 1000 (GE Healthcare) using a 10×lens, 50 ms exposure and 15 fields/well.
Image data analysis and processing
Image analysis was performed to count nuclei using a customized multi-target algorithm within the IN Cell Analyzer 1000 Workstation software (GE Healthcare), with a filter applied to remove small DAPI-bright objects (dead adherent cells). Total ‘live’ cell counts were summed for the well and the data exported in a csv (comma-separated values) format. ActivityBase version 5.4 (ID Business Solutions) was used for numerical data processing and analysis. Screening plates were approved if a quality-threshold criterion (Z′>0.4) was attained. The activity of a particular compound was expressed as percentage activation, [(experimental value−mean of low control)/(mean of high control−mean of low control)]×100, where the high control was 10 μM Y-27632 and the low control was 1% DMSO.
Concentration curves (ten-point, consisting of half-log serial dilutions of compound in DMSO ranging between 10 mM and 0.508 μM stocks, and 100 μM and 5.08 nM final assay concentrations) were prepared in 96-well plates using a JANUS workstation (PerkinElmer Life Sciences). All curve fitting was undertaken using a 4 Parameter Logistic dose–response curve (XLFit 4.2 Model 205).
Kinase profiling and ROCK2 IC50 determination
Specificity profiling utilized a panel of over 70 protein kinases in the MRC National Centre for Protein Kinase Profiling Service at the University of Dundee (http://www.kinase-screen.mrc.ac.uk). Each compound was tested in vitro, in duplicate, at 10 μM final concentration using recombinant kinases and model substrates. Selected compounds were submitted for IC50 determination using ROCK2. A ten-point half-log dilution series (a maximum concentration of 100 μM) was used. Assays were performed in duplicate.
Compound mass and purity analysis
The compounds were made up at 3–10 mM in DMSO (Chromasolv®; Sigma) and diluted with Optima grade acetonitrile/water (50:50, v/v) to give a final concentration of between 15 and 50 μM. Separations were performed by gradient elution on a 2.1×50 mm, 1.7 μm BEH C18 column on an ACQUITY UPLC (ultra performance liquid chromatography) system (Waters). Samples were eluted with a flow rate of 0.6 ml/min. Mobile phase A was acetonitrile with 0.1% formic acid, mobile phase B was MilliQ water with 0.1% formic acid. The flow was directed through a photodiode array (scanning UV between 210 and 400 nm) and then sequentially on to the mass spectrometer (Quattro Premier XE™ triple quadrupole system; Waters), using positive electrospray ionization in full-scan mode. Compound dilutions were analysed alongside a true DMSO blank to identify the compound-related peak in both the total ion count trace and the UV output. Extraction of the associated spectrum for these peaks confirmed identification by mass of the compound to one decimal place. UV output confirmed purity and the presence/absence of geometric isomers (where appropriate).
SA461 cells were plated at 5000 cells/well in fibronectin-coated 96-well plates in fully supportive medium with 10 μM Y-27632 (Calbiochem). The following day the medium was changed to fully supportive medium containing compound or controls, with eight wells used per condition. Proliferation was assayed at 48, 72, 96 and 120 h, using a Rapid Cell Proliferation Kit (Calbiochem). Kinase inhibitor/Prestwick library compounds were used at the concentrations shown in the Figures, the diversity library hit compounds were used at 30 μM and Y-27632 was used at 10 μM. Hits from the initial screen of the kinase inhibitor/Prestwick sets were tested using MEF-CM. However, as this is not an ideal culture system for the development of ‘Good Manufacturing Practice’ compliant or clinically acceptable compounds, the second set of hit compounds (the diversity set) was tested in medium containing only human components (HuWIL-CM).
Long-term compound culture
SA461 cells were cluster-passaged using TrypLE™ Select on to fibronectin-coated 12-well plates in FF medium. The following day, medium was changed to FF medium plus compounds. The medium (FF medium plus compounds) was changed every 2 days and cells were cluster-passaged when they reached confluency. Control cells were maintained up to passage 5. Kinase inhibitor/Prestwick library compounds were all used at 5 μM with the exception of DDD00066570 (2.5 μM) and tested in FF medium with MEF-CM. Diversity set hit compounds were of lower potency and therefore were used at 30 μM and tested in FF medium with HuWIL-CM.
RNA isolation and qRT-PCR (quantitative real-time PCR)
Total RNA was extracted using RNeasy kits (Qiagen) and DNaseI treatment used Turbo DNA-Free (Ambion). The absence of genomic DNA was confirmed by qRT-PCR targeting GAPDH (glyceraldehyde-3-phosphate dehydrogenase), with Taqman® primers and probe (MWG Biotech). cDNA was synthesized with up to 2 μg of total RNA in a 20 μl volume using Superscript II (Invitrogen). The micro-fluidic 384-well TLDA (Taqman® low-density array) system was used to analyse samples, and either the Taqman® human stem cell pluripotency array (Applied Biosystems) with primer/probe sets for 96 genes or custom-designed plates for 24 genes were used. Samples were run and analysed on the 7900HT real-time PCR system (Applied Biosytems).
Cellartis cell lines SA461, SA121, SA181, SA167 and SA002 were grown in FF conditions and single-cell suspensions were prepared as described above, adding either Y-27632, DDD00033325 or DMSO to the cell suspensions (10 μM final concentration or 1% DMSO) and then seeding them at 0.5 ml/well in fibronectin-coated 24-well plates, in triplicate. After 24 h, each well was washed and a single-cell suspension generated by using extensive trypsinization using TrypLE™ Select. Cell numbers were counted using a Cellometer (Peqlab) and cells/ml calculated. Fold survival values for DDD00033325 and Y-27632 were calculated relative to the DMSO control.
Development of a high-content single-cell hES cell survival assay
In order to identify novel small molecules that act as pro-survival agents we developed a high-content 96-well format screening assay. The methodology was developed to extensively disaggregate and filter hES cells to give single cells; this enabled accurate counting and plating using a bulk dispensing robot to give low variance in cell numbers between wells. Y-27632, a commonly employed ROCK inhibitor that promotes stem cell survival  was used as a positive control compound and DMSO used as a negative control. In the presence of the reference compound (Y-27632) over 90% of cells survive and are retained as viable cells on the substratum, in contrast with the control conditions where very few cells were adherent and viable (Figure 1A). Nuclei from non-specifically adhered dead cells were observed at a low level in the absence of the reference compound which, if counted, would reduce the screening window. However, these were excluded from the measurements on the basis of nuclear size and signal intensity using the analysis software filtering tools, underlining the flexibility of using a high-content screening strategy over a plate reader-based approach. Seeding density was found to be critical since at very high densities (>20000 cells/well) a self-supporting effect was apparent, possibly due to rapid reaggregation or growth factor secretion (results not shown). Compound addition into the medium in fibronectin-coated wells of the screening plates was performed immediately prior to the addition of the cells and cell-processing time was kept to an absolute minimum.
Blinded mock screening was performed at five different concentrations of Y-27632 (0.1 μM, 1 μM, 2 μM, 5 μM and 10 μM) to test for assay performance and this showed that at concentrations above 2 μM, 100% of the spiked wells were identifiable as ‘hits’ (Figure 1B). The screens involving the Dundee compound collections (i.e. the kinase and diversity sets) were performed at a final concentration of 30 μM, whereas the average screening concentration of the bioactives library was 62.5 μM (from a stock concentration range of 1.2–19.59 mM).
Hit identification and screen performance
On the basis of a promising assay window and high-quality mock screening, we undertook a proof-of-concept screen using the compound library of 4110 kinase-targeted lead-like ATP-competitive inhibitor scaffolds . The performance of the screen, measured by the Z′ value  for each plate, is shown in Figure 2(A). The assay was found to be robust and achieved a signal-to-background ratio of >10:1 with an average Z′ value for the screen of 0.76±0.09, and a mean CV (coefficient of variation) percentage for the reference compound of 5.9±2.3% (Figure 2B). Plotting a frequency distribution of percentage activation (the increase in viable cell count over the DMSO control, normalized to the difference between high control and low controls) showed that the data lay within a normal distribution (Figure 2C). The cut-off for hit definition was set at compounds returning a percentage activation value in excess of 3 S.D. away from the mean of the negative control (i.e. background). On this basis, and manual image inspection, all 51 putative hits were ‘cherry-picked’ and tested in dual-point retests and subjected to mass and purity confirmation by UPLC and LC (liquid chromatography)–MS/MS. Of these, six were deemed confirmed using ten-point dose–response curve retests with potencies that ranged from 3 μM to >30 μM, giving a hit rate for the screen of 0.15%. A combination of repurchasing and resynthesis enabled further retesting of fresh material using a higher concentration of 100 μM and allowed confirmation of the activity and efficacy of the confirmed hits (see Figure 2D for example curves, and Figure 3 and Supplementary Table S1 at http://www.BiochemJ.org/bj/432/bj4320021add.htm).
We also screened the commercially available 1200-compound bioactives library of drugs and known pharmacologically active agents (Prestwick). A single hit was identified giving 60% activation relative to the reference compound Y-27632 (this particular compound is supplied at 8 mM in the library). The compound was found to be pinacidil, a K+-ATP channel opener and vasodilatory drug used in the reduction of blood pressure . Pinacidil was repurchased and verified in a ten-point dose–response curves, giving an EC50 of 46 μM (Figure 3 and Supplementary Table S1).
In order to extend our coverage of chemical space and perhaps identify hits with a non-kinase mode of action, a small diversity collection (15667 compounds), a representative subset of a larger diversity set (65000 compounds) , was screened. At total of 18 compounds with a percentage activation of >3 S.D. over the low control were identified as putative hits and taken into dual-point retesting alongside mass/purity determinations. Of these initial putative hits, 12 compounds were verified as positive and retested for potency. Resupply material was obtained for all of the confirmed 12 hits, their mass/purity was confirmed, and retesting performed up to a higher maximum concentration of 100 μM, giving EC50 values ranging from 16 μM to 49 μM (Supplementary Table S1).
In vitro kinase profiling of hit compounds
In order to provide insight into the kinase inhibition profiles of the hits identified, compounds were subjected to panel screening by the MRC National Centre for Protein Kinase Profiling at the University of Dundee (Dundee, Scotland, U.K). Screening was performed at 10 μM, in duplicate, using recombinant kinases and model substrates . The full results of the analyses are listed in Supplementary Table S2 (at http://www.BiochemJ.org/bj/432/bj4320021add.htm), with examples shown in Figure 4(A) for the reference compound, the apparently most selective hit (DDD00033325) and pinacidil. The striking observation derived from these results was the fact that ROCK2 and/or the closely related PRK2 were inhibited to a significant degree by all survival compounds tested (Figures 4B and 4C). DDD00082896 and DDD00074187 were exceptional in not significantly inhibiting PRK2. In fact, in general, the hit compounds exhibited a higher degree of inhibition against ROCK2 than PRK2 in the in vitro assay (Figure 4C). Beyond this, there was little other commonality in kinases inhibited (Figure 4B). The inhibitory activity of the compounds against ROCK1 was not defined in the present study, as this kinase, which is very closely related in its kinase domain to ROCK2, was not present in the profiling panel. An IC50 analysis of several hits with hES cell survival EC50 values of <20 μM was subsequently performed to compare their ROCK2-inhibitory activities with Y-27632. As expected, the in vitro-determined IC50 values for the compounds tested are all 5–20-fold lower than the equivalent EC50 values determined phenotypically (Figure 3 and Supplementary Table S1) and, although the most potent ROCK inhibitor in vitro is also the most potent compound in the survival assay, the data set available is not large enough to delineate a clear correlation.
Although a number of the compounds possessed a degree of selectivity for inhibition of ROCK2 and PRK2, most notably DDD00033325 (see Figure 4B), many of the hits showed low ROCK2/PRK2 specificity. This is unsurprising for compounds issuing from the focused kinase set, since all of the compounds in this library are designed to be ATP-competitive kinase inhibitors. The reference compound Y-27632, as expected, also inhibited ROCK2/PRK2 but, additionally, inhibited several other kinases in the panel (Figure 4 and Supplementary Table S2), which is consistent with previous studies showing that this purported ROCK-specific inhibitor does have significant off-target actions on other protein kinases [6,12]. Most commercially available, so-called ROCK-specific inhibitors also show poor selectivity towards ROCK . In marked contrast with this, however, six of our hit compounds (DDD00033325, DDD00033207, DDD00082896, DDD00074187, DDD00074102 and DDD00076032) showed reasonably clean profiles against non-ROCK/PRK kinases in the panel employed.
Of particular interest is the K+-ATP channel opener pinacidil, which displayed moderately selective inhibition of ROCK2/PRK2 (Figure 4 and Supplementary Table S2). This selective, albeit nominally off-target, effect is potentially of importance as ROCK inhibitors are, like pinacidil, used in ischaemic preconditioning. Indeed, pinacidil shares a common functionalized 4-aminopyridine core structure with both Y-27632 (Figure 3) and other established ROCK inhibitors , so this previously unreported ROCK-inhibitory activity of pinacidil has a reasonable structural basis. To confirm that the effect of pinacidil on hES cell survival is probably due to kinase inhibition, rather than ion channel activity, the K+-ATP channel blocker glibenclamide, known to reverse the effects of pinacidil on K+ flux, was used. Glibenclamide had no adverse effect on hES cell survival alone, nor was able to block the effect of either pinacidil or Y-27632 on survival, as shown in Supplementary Figure S1 (at http://www.BiochemJ.org/bj/432/bj4320021add.htm). One other compound in our screen output, DDD00079566, shared the ROCK-inhibitory 4-(acylamino)pyridine structure (Figure 3). This compound and pinacidil are therefore expected to share a common binding mode with Y-27632, for which co-crystal structures with ROCK1 (PDB code 2H9V) and ROCK2 (PDB code 2ETR) established that the pyridyl nitrogen acts as a key hydrogen bond acceptor to engage the backbone of the enzymes' MEYMPG hinge sequence [14,15].
Chemistry-based analysis of the full hit set revealed numerous core structures that are commensurate with ROCK-inhibitory activity. Thus DDD00074187 is structurally related to the established ROCK inhibitor HA-1077 (fasudil; Figure 3), for which ROCK1 (PDB code 2ESM) and ROCK2 (PDB code 2F2U) co-crystal structures exist [15,16]. In these structures the pyrido ring of the isoquinoline superimposes very closely on to the position occupied by the pyridine ring of Y-27632 in the co-crystal structures of the latter. Many established ROCK inhibitors, such as fasudil and Y-27632, feature a basic nitrogen positioned 9–12 Å (1 Å=0.1 nm) from the hydrogen bond acceptor of the core subunit that anchors the inhibitor to the kinase hinge. This feature allows engagement of one or other of a number of aspartic acid residues located at the rim of the ATP-binding site, including the aspartic acid of the conserved kinase DFG motif. The structures of both DDD00033325 and DDD00033207 embody a basic nitrogen at a suitable distance from the N-1 centre in their respective pyrazolo[1,5-a]pyrimidine and imidazo[1,2-b]pyridazine core structures to allow adoption of such a binding mode. Interestingly, DDD00033325 is isosteric with an indazole (Iwakubo 5b; Figure 3), which has previously been found to be a potent ROCK inhibitor [13,17]. The proposed binding mode for indazole ROCK inhibitors of this type  exploits the respective hydrogen bond donor and acceptor capacities of both N-1 and N-2 centres for engagement of the kinase hinge sequence to provide strong inhibitory potency. The strong intrinsic ROCK-inhibitory activity of the 5-substituted indazole core is reflected in the identification of a series of six 5-(acylamino)- and 5-(alkylamino) indazoles in our hES cell-survival screen output (DDD00046328, DDD00046780, DDD00047425, DDD00074083, DDD00074102 and DDD00074270; Figure 3).
Two other less well-populated series of hits were also identified in our screens. First, a group of N6,N6-dialkyladenines, a class of compounds not reported as ROCK inhibitors previously, was discovered in DDD00066444, DDD00066508 and DDD00081921 (Figure 3). Secondly, a pair of related 5-(acylamino)-1-methylindolinone (or thiaindolinone) compounds was found in DDD00076032 and DDD00082896 (Figure 3). A third indolinone (DDD00066570) may be related in that the 2-oxo group of its core heterocycle may serve as a hydrogen bond acceptor feature for interaction with the kinase hinge sequence. In this latter compound the unsubstituted indolinone nitrogen may additionally act as a hydrogen bond donor to the kinase hinge in a manner analogous to that proposed for the indazoles [13,17]. Indeed, such interactions are supported by the co-crystal structures of related indolinones, SU6656 and SU5402 (Figure 3), with the kinase domains of CaMKII (Ca2+/calmodulin-dependent protein kinase II) (PDB code 2WEL) and FGFR1 (FGF receptor 1) (PDB code 1FGI) [18,26]. DDD00066570 proved the least selective compound in our hES cell screen output, showing strong inhibition of FGFR1 and numerous other kinases in the test panel (Figure 4B and Supplementary Table S2).
Effects on long-term growth and pluripotency
In order to determine whether our pro-survival compounds had an effect on cell physiology if used beyond a short-term culture, cell proliferation was monitored for up to 5 days at a compound concentration around the EC50 value in the survival assay. Two compounds (DDD00066444 and DDD00066570; Figure 5A) strongly suppressed proliferation; the remaining hits showed only modest or no effect on cell proliferation (Figures 5A and 6A). Interestingly, we consistently observed a positive effect of pinacidil on cell proliferation. hES cells were next treated persistently with the compounds under normal growth and passaging conditions to determine how sustained exposure affected cell proliferation (Figures 5B and 6B). As seen in the short-term proliferation assay described above, compounds DDD00066444 and DDD00066570 adversely affected cell growth, whereas the cells behaved as the control cultures in the presence of all other compounds tested. Cells could be maintained in DDD00066444 by passaging cells at a higher density up to passage 5; however, cells grown in DDD00066570 were arrested in their growth at passage 1 and high concentrations were toxic (Figure 5B and J. Gilmour, unpublished work). In addition cells grew particularly slowly in compound DDD00074187 and only reached passage 2, cells grown in DDD00074102, DDD00074083 and DDD00036524 reached passage 3 and cells grown in DDD00046780 reached passage 4.
To assess the maintenance of pluripotency in the presence of the compounds, cells were stained for Oct3/4 and also for cytokeratin-7, as an early differentiation marker for trophectoderm. Oct3/4 staining was maintained throughout the experimental time-course in all samples (Supplementary Figure S2 at http://www.BiochemJ.org/bj/432/bj4320021add.htm). Occasional small patches of differentiated cells staining positive for cytokeratin-7 (<5% of the total cells) were observed in all samples including untreated and Y-27632-treated cultures. This level of spontaneous differentiation is not unusual during hES cell maintenance and was not altered by the compounds.
Gene expression analysis of mRNA extracted from cells exposed to compounds was performed using TLDAs. Results of the analysis for the cells treated with kinase inhibitor/Prestwick compound sets are shown in Figure 7(A). Key pluripotency genes were mostly unaffected by the compound treatment; compound or Y-27632 resulted in less than 2.5-fold difference in gene expression when compared with the untreated control. Cells treated with DDD00066570 and DDD00066444 showed increased expression of some endoderm and mesoderm-specific genes, such as eomesodermin, Sox17 [SRY (sex-determining region Y)-box 17], brachyury, GATA4 and GATA6, indicating that these compounds may be inducing differentiation in addition to reducing the proliferation of hES cells. Gene expression analysis for the diversity library hits was performed over the course of the culture at passages 1, 3 and 5. As shown in Figure 7(B), pluripotency was maintained and most compounds induced very little change in expression of the pluripotency markers Oct3/4, Sox2, Nanog and DNMT3B [DNA (cytosine-5)-methyltransferase 3β]. There were some small changes in gene expression for several of the differentiation markers; however, these probably reflect the small patches of differentiated cells observed in the stained cells.
Since it is widely known that distinct hES cell lines can behave differently in culture we tested our lead compound DDD00033325 and Y-27632 against five separate cell lines (SA461, SA181, SA167, SA121 and SA002). We observed that DDD00033325 promoted survival and was as effective as, or more effective than, Y-27632 in all the cell lines tested (Figure 8).
Unlike mouse embryonic stem cells, hES cells die if dissociated into single cells by enzymatic treatment, a property that hampers their isolation and manipulation for both drug discovery and basic research. In the present study we performed a high-content screening campaign using fully pluripotent hES cells, growing in FF monolayer cultures in a 96-well format, in order to identify small molecules that promote single-cell survival. We screened a bioactives library in addition to two libraries of lead-like small compounds, one designed for chemical diversity and the other designed to target the active site of protein kinases. We identified a number of hit series, some with structural similarity to known ROCK inhibitors and others with structural novelty. Some compounds were found to be less toxic than existing ROCK inhibitors and some induced differentiation in hES cells on long-term exposure. Remarkably, despite using three distinct screening libraries, the small molecules identified all inhibited ROCK2 and/or PRK2.
The high-content screening assay described in the present paper was robust and gave good signal-to-background values using the known ROCK inhibitor Y-27632 as the reference compound . High-content analysis allowed the exclusion of dead cells from the analysis and demonstrated the advantages of using this platform over a plate-reader-based analysis. We initially screened our 4000-compound library of kinase inhibitor scaffolds. This screen identified five small molecules with pro-survival activity. Each hit was verified with resupplied/resynthesized material and dose–response curves revealed potencies ranging from 2 to 50 μM. We noted that at the highest concentration tested (100 μM) cell toxicity was observed, including for the reference compound Y-27632 (Figure 4). Indeed Y-27632 had the optimal effect on survival at 3 μM, with higher concentrations leading to >60% reduction in cell numbers. Two compounds (DDD00033325 and DDD00066508) were efficacious up to 30 μM and only toxic at 100 μM. Several of the hit compounds showed toxicity when present above 30 μM that may not necessarily be related to ROCK2/PRK2 inhibition. Kinase profiling demonstrated that Y-27632 inhibited ROCK2 and PRK2 to 80% (at the 10 μM final concentration used in the assay), but also inhibited RSK1 (ribosomal S6 kinase 1), MSK1 (mitogen- and stress-activated kinase 1), MNK2 [MAPK (mitogen-activated protein kinase)-interacting kinase 2], PHK (phosphorylase kinase) and BRSK2 (BR serine/threonine-protein kinase 2) to a similar degree, as reported previously . This lack of specificity supports previous experiments, which showed that several kinases (including ROCK1, ROCK2 and PRK2) contribute to the global effect of Y-27632 on cellular responses . The most ROCK2/PRK2-selective hit was DDD00033325, which showed strong inhibition of ROCK2/PRK2, as well as the cAMP-dependent protein kinase PKA (protein kinase A), and weak inhibition of MSK1.
We also screened the Prestwick bioactives library, which includes known pharmacological agents and marketed drugs. The single low-potency hit identified was the ATP-sensitive K+-channel opener pinacidil, a drug used as a vasodilator and in ischaemic preconditioning of organs in transplantation . Geron used pinacidil in its pro-survival cocktail, employed to sustain transplanted cardiomyocytes derived from hES cells in a rat infarct model, although its inclusion was to mimic ischaemic preconditioning rather than promote cell survival it itself . Our results indicate that the pro-survival effects of pinacidil are not due to the compound's effect on K+ channel opening. Given the structural similarity of pinacidil to established ROCK inhibitors including Y-27632 (see above), we considered that the compound's hES cell survival effect was likely to be due to off-target kinase-inhibitory action. This was subsequently confirmed by panel profiling that revealed modest activity against ROCK2 and the related kinase PRK2 (60% and 70% inhibition respectively at 10 μM), but little or no activity against the other 70+ kinases tested in the panel.
To expand the coverage of chemical space and possibly uncover novel effectors of cell survival, a small diversity library was screened identifying several hits. A total of 12 compounds with low-micromolar potencies were verified. Examination of the hit chemical structures revealed similarities to known ROCK inhibitors in several instances (see above), and kinase profiling confirmed that the most potently inhibited kinases were ROCK2/PRK2. This result was striking since the diversity library, unlike the kinase inhibitor library, was not assembled to target kinases specifically. This result shows that a phenotypic screening approach identified specific kinase inhibitors and underlines the pivotal importance of the ROCK pathway in the regulating the cellular response to lack of homotypic and heterotypic interactions.
Several compounds (DDD00082896, DDD00076032, DDD00074187, DDD00033325 and DDD00033207) were more selective for ROCK2/PRK2 than the reference compound Y-27632, and these might be of interest to dissect the role of ROCK/PRK in cytoskeletal rearrangement. An observation worthy of further study was that three compounds (DDD00082896, DDD00076032 and DDD00074187) were relatively selective for ROCK2 over PRK2, but exhibited low potency in the phenotypic assay compared with compounds such as DDD000033325, which inhibited the two enzymes to a similar extent. Further work is required to determine their in vitro selectivity profiles. Since both ROCK1/2 and PRK2 are thought to act in concert, at least in cytoskeletal signalling , it is possible both these related kinases have a role to play in hES cell survival. Of further interest is the observed inhibition of EPH-B3 (ephrin type-B receptor 3) receptor tyrosine kinase by both DDD00074102 and DDD00074187, since this receptor family is known to be involved in signalling in cell–cell adhesion . Evidence also suggests that signalling via the FGF receptor plays a role in sustaining self-renewal, partly via the cytoskeleton and in the prevention of anoikis . Results from a recent screen  for pro-survival small molecules have shed further light on this effect with the identification of a ROCK inhibitor (thiazovivin) structurally unrelated to those described in the present paper. That study provided evidence that contraction of the cytoskeleton mediated by the Rho/ROCK pathway leads to cell death .
One of the concerns in the stem cell field is the potential for adverse effects of small molecules on cell growth and proliferation, and, if present in the culture medium for extended periods, their effects on stem cell pluripotency. In the context of transplanted progenitor cells this could compromise their cellular identity and differentiation capacity. We noted that in our survival assay, which was of a 24-h duration, certain compounds, including Y-27632, showed toxicity at higher concentrations (between 10 and 30 μM). In order to address these potential issues we cultured hES cells with compounds (at 30 μM) from 48 up to 120 h and monitored cell proliferation. A subset of the compounds suppressed growth on long-term exposure (DDD00066570, DDD00066444, DDD74102, DDD00036524 and DDD74083); however, the majority did not have a marked effect.
To address further issues relating to continual exposure to compounds, hES cells were grown with compounds for up to five passages and a gene expression analysis was performed with TLDAs. This showed that most compounds had little or no effect on the expression of a variety of pluripotency genes and differentiation markers. The two compounds from the kinase inhibitor library that reduced proliferation also increased expression of mesoderm or endoderm markers; DDD00066570 predominantly up-regulated endoderm markers, including GATA4, GATA6 and Sox17, but also eomesodermin. Similarly DDD00066444 increased expression of eomesodermin, brachyury, Sox17 and GATA6. Two receptor tyrosine kinase inhibitors structurally related to DDD00066570 (SU5402 and SU6656; Figure 3) have been shown to have effects on hES cell differentiation [24–26], indicating a potential mechanism for the observed effects of DDD00066570. These results suggest that ROCK-inhibitory compounds that lack selectivity in targeting ROCK/PRK2 may be incompatible with maintaining pluripotency in culture over extended periods, but could have utility in steering differentiation to particular germ layers in some instances. However, minimal exposure of hES cells to small-molecule survival agents may, in general, be good practice to minimize risk of unwanted effects on the cells, and might be achieved if the medium change can be performed once cells are attached. The gene expression results from the present study were obtained using mRNA from cells growing in FF-conditioned medium derived from fibroblast cultures; it may well be that differences in medium recipes and sources of conditioning medium could influence the outcome of continued exposure to compounds. More work using different culture conditions would be needed to test the potential effect of compounds, in particular using chemically defined proprietary medium. In order to understand the general applicability of the ROCK-inhibitor effect we demonstrated that, at least for our lead compound DDD00033325 and Y-27632, five different cell lines were able to survive single-cell disaggregation when ROCK/PRK2 was inhibited, indicating a general efficacy towards hES cells.
In summary, we have identified novel pro-survival compounds of greater specificity, equivalent potency and reduced toxicity relative to Y-27632. Notably, our most ROCK/PRK2-selective compound (DDD00033325) promoted single-cell survival up to 20-fold over background, had little or no effect on hES cells in long-term culture and was effective on more than one cell line, and therefore has the potential for routine use requiring work with disaggregated hES cells. Clues as to the mode of action of these compounds in hES cells revealed a striking and consistent link to inhibition of ROCK kinases, despite identifying a range of hits with structural diversity. These results underline further the pivotal nature of the kinase ROCK in regulation of hES cell survival as single cells. ROCK's reported role in cytoskeletal rearrangement and adherence, and recent results from small molecule inhibitor studies , relate well to our observations using these phenotypically identified small molecules.
While this paper was in preparation we note that pinacidil was independently identified as a pro-survival molecule .
Paul Andrews and Melissa Becroft performed the assay development, high-content screening, data interpretation, hit validation and mode-of-action work. Anders Aspegren performed the cell production, delivery and cell line comparisons. Jane Gilmour performed the long-term proliferation assays and gene expression analysis. Martyn James was involved in hit validation and kinase profiling. Robert Kime performed the compound purity and identity analysis. Scott McRae was involved in early assay development. Robert Allcock, Achamma Abraham and Zhong Jiang performed medicinal chemistry, including compound synthesis and structural analysis. Raimund Strehl supervised the cell production programme. Graeme Milligan and Miles Houslay directed the initial development of the culture technology. Joanne Mountford supervised and interpreted the proliferation and gene expression analysis. Paul Andrews, Jane Gilmour and David Adams prepared the Figures. Paul Andrews primarily wrote the manuscript with contributions from David Adams, Julie Frearson and Jane Gilmour. Paul Andrews, Julie Frearson, David Adams, Graeme Milligan and Miles Houslay edited the manuscript before submission.
This work was supported by ITI Life Sciences (now Scottish Enterprise).
We thank Fergus McKenzie for project management.
Abbreviations: bFGF, basic fibroblast growth factor; DAPI, 4′,6-diamidino-2-phenylindol; DNMT3B, DNA (cytosine-5)-methyltransferase 3β; FF, feeder-free; FGF, fibroblast growth factor; FGFR1, FGR receptor 1; hES, cell, human embryonic stem cell; HuWIL-CM, human foreskin fibroblast-derived conditioned medium; MEF-CM, mouse embryonic fibroblast-derived conditioned medium; MSK1, mitogen- and stress-activated kinase 1; Oct3/4, octamer 3/4; PRK2, protein kinase C-related kinase 2; qRT-PCR, quantitative real-time PCR; ROCK, Rho-associated kinase; Sox, SRY (sex-determining region Y) box; TBS, Tris-buffered saline; TLDA, TaqMan® low-density array; UPLC, ultra performance liquid chromatography
- © The Authors Journal compilation © 2010 Biochemical Society