Salicylic diamines selectively eliminate residual undifferentiated cells from pluripotent stem cell-derived cardiomyocyte preparations

Clinical translation of pluripotent stem cell (PSC) derivatives is hindered by the tumorigenic risk from residual undifferentiated cells. Here, we identified salicylic diamines as potent agents exhibiting toxicity to murine and human PSCs but not to cardiomyocytes (CMs) derived from them. Half maximal inhibitory concentrations (IC50) of small molecules SM2 and SM6 were, respectively, 9- and 18-fold higher for human than murine PSCs, while the IC50 of SM8 was comparable for both PSC groups. Treatment of murine embryoid bodies in suspension differentiation cultures with the most effective small molecule SM6 significantly reduced PSC and non-PSC contamination and enriched CM populations that would otherwise be eliminated in genetic selection approaches. All tested salicylic diamines exerted their toxicity by inhibiting the oxygen consumption rate (OCR) in PSCs. No or only minimal and reversible effects on OCR, sarcomeric integrity, DNA stability, apoptosis rate, ROS levels or beating frequency were observed in PSC-CMs, although effects on human PSC-CMs seemed to be more deleterious at higher SM-concentrations. Teratoma formation from SM6-treated murine PSC-CMs was abolished or delayed compared to untreated cells. We conclude that salicylic diamines represent promising compounds for PSC removal and enrichment of CMs without the need for other selection strategies.


Index
page 32 Table S2  page 33   Table S3  page 34   Table S4 page 35 Supplementary References page 36 murine PSCs after 72 h of treatment. It has been indicated before that PluriSIn #1 has decreased cytotoxicity on murine PSCs compared to human PSCs (Ben-David et al., 2013).
Since PluriSIn #1 was identified to inhibit SCD1 in human PSCs (Ben-David et al., 2013) a differential regulation of SCD expression in mouse may be responsible for those variations.
Indeed, four isoforms of SCD have been identified in mouse (SCD1, SCD2, SCD3, SCD4) while only two have been characterized in human (SCD1, SCD5) (Paton and Ntambi, 2009), but it is not known whether functional differences exist between these isoforms. It is also worth noting that the selective cytotoxicity of PluriSIn #1 was not consistent with human PSCs as well since Kropp and coworkers were not able to reproduce the PSC eliminating effect in experiments with confluent hPSC monolayers which they observed with five different cell lines (Kropp et al., 2015). Varying cell culture conditions like medium composition and involvement of feeder cells seem to repress the potency of PluriSIn #1 (Ben- David and Benvenisty, 2014;Ben-David et al., 2013) which make the achievement of best possible effects challenging.

SM structure-function relationship
Dose-response analyses performed with a newly synthesized batch of SM2, SM6 and SM8 corroborated initial findings and validated their toxicity towards miPSCs ( Supplementary   Supplementary Fig. S3a). Since SM6 possessed the lowest IC 50 value for PIG-AT25 miPSCs, additional dose-response analyses were performed with SM6 stereoisomers (Supplementary Fig. S3b) and ten SM6-based chemical structures C1-C10 ( Supplementary   Fig. S3c) in order to elucidate the structure-function relationship. Comparison of SM6 stereoisomers revealed slightly but not significantly higher toxicity of the initially tested R,R isomer compared to the S,S or the S,R isomers indicating that its activity is not affected by its spatial orientation (Supplementary Fig. S3b). Moreover, none of the ten tested SM6-based chemical structures exerted a significant toxicity against miPSCs (Supplementary Fig. S3c), suggesting that the intact SM6 molecule and not its metabolic derivatives that could be formed within cells is required for its PSC-toxic activity.

Effect of SM6 on action potential (AP) properties of purified cardiac clusters
Electrophysiological analysis revealed same beating frequencies of SM6-treated and nontreated clusters (p>0.05; Supplementary Fig. S6a). Furthermore, action potential (AP) amplitude and maximal diastolic potential (MDP) were similar to control clusters as well (p>0.05; Supplementary Fig. S6b,c). Significant differences were observed in the AP durations (APDs) which increased when the clusters were treated with 5 or 10 µM of SM6 ( Supplementary Fig. S6d). However, the ratios of APD50/APD90 were not significantly compromised ( Supplementary Fig. S6e). Maximal upstroke velocities (Vmax) of APs slightly but not significantly decreased when clusters were treated with 5 and 10 µM of SM6 compared to non-treated clusters (Supplementary Fig. S6f). Taken together, overlays of APs derived from SM6-treated and non-treated cardiac clusters illustrate that SM6 had only minor electrophysiological effects on APs after 48 h of treatment (Supplementary Fig. S6g).

Elimination of PSCs in murine ESC-derived cardiac clusters with SM6
The morphology of murine ESC-derived cardiac clusters after SM6 treatment was similar to those of DMSO or puromycin-treated controls (Supplementary Fig. S7b). Interestingly, even though pre-purification was performed with double the concentration of puromycin (4 µg/ml) between day 9 and day 14 compared to that used in experiments with the PIG-AT25 miPSCs (compare with the scheme in Fig. 5a Fig. S7c,d).
Comparing the relative cell yield in ESC-derived day 16 clusters, the number of cells generated in all treated groups was lower than in the DMSO control group ( Supplementary   Fig. S7e). However, the cell yield in the 8 µg/ml puromycin-treated group was lower (55.9±7.3% of the DMSO control) than in SM6-treated clusters (on average 72.9±8.9% of the DMSO group). As determined by flow cytometric analysis, the loss of cells in the drug treated clusters was due to a decrease in the GFP -/cTnTpopulation of non-CMs. While 38% of cells in DMSO control group were non-CMs this fraction was reduced to 12% in puromycin group and to 31%, 12% and 8% in 1 µM, 5 µM and 10 µM SM6-treated groups, respectively ( Supplementary Fig. S7f). Interestingly, a population of GFP -/cTnT + CMs in ESC-derived clusters was 1.7-fold larger in SM6-treated than in puromycin-treated clusters ( Supplementary Fig. S7f S7g) and staining for the PSC marker Oct4 showed the presence of pluripotent cells after treatment with 1 µM and 5 µM of SM6 but not after treatment with 10 µM of SM6 ( Supplementary Fig. S7h).
Taken together, by using SM6 in large-scale suspension cultures of mESCs ( Supplementary   Fig. S7) and miPSCs (main Figures 5 and 6), we demonstrated that it was possible to reduce the number of contaminating PSCs in murine ESC-and iPSC-derived cardiac cell aggregates by more than 96-99%. This efficiency was comparable to that of the antibiotic-based genetic selection method and did not compromise the CM viability or their functional properties.
Furthermore, even when used alone, SM6 was capable of eliminating PSCs in differentiating EB cultures to the same extent as the puromycin-based genetic selection method. Moreover, the yield of CMs in SM6-treated EB cultures was higher than that obtained with puromycin because SM6, but not puromycin, preserved the subpopulation of cTnT-positive CMs in which the transgenic MHC-promoter-driven selection markers were not expressed. This puromycin-sensitive and GFP-negative CM subpopulation most likely represents ventricular CMs because the MHC promoter in the developing rodent heart is predominantly active in the atrium whereas in ventricles this isoform is activated only after birth (England et al., 2013). This is also in agreement with Kolossov et al. who found that the MHC promoter in mESC-CMs was active only in pacemaker and atrial but not in ventricular CMs (Kolossov et al., 2006). Thus, in murine system, the small molecule based elimination of PSCs has significant advantage over genetic selection because it preserves more heterogeneous CM population independently of subtype and results in higher CM yields than the genetic approach.
At 70-80% of confluence, hiPSCs were passaged as small aggregates every 4-5 days at 1:20 splitting ratio after dissociation with 0.48 mM Versene solution (Life Technologies). Medium was changed every day.

Cardiac differentiation of murine iPSCs and ESCs
The differentiation of transgenic mPSCs was initiated in mass cultures by inoculating one hiPSC-derived CMs (hiPSC-CMs) were kindly provided by Axiogenesis (Cologne, Germany) and cultured following the manufacturer's recommendations.

Cardiac differentiation of human iPSCs
Prior to cardiac differentiation the NP0040-8 hiPSC line was cultured in E8 medium in 6-well plates coated with 10 μg/cm 2 Matrigel Matrix (hESC-qualified, Corning, Cat. # 734-1440) in a humidified incubator at 37°C and 5% CO 2 . The E8 medium was supplemented with 5 μM of Rho Kinase (ROCK) inhibitor Y27632 (AdooQ Bioscience, Cat. # A11001-5) for the first two days after each passage. Once the hiPSC-confluency had reached about 80%, the medium was discarded, cells washed with Dulbecco's phosphate buffered saline without calcium and magnesium (DPBS -/-) and hiPSC colonies were dissociated into single cells by adding 1 mL of 0.48 mM Versene into each well of a 6-well plate and incubating at 37°C for 7 minutes.
After counting, 70,000 cells/cm 2 were seeded onto Matrigel-coated 6-well plates and maintained in 2 mL of E8 medium supplemented with 5 μM ROCK inhibitor for two days.
After two days, fresh E8 medium without ROCK inhibitor was exchanged and cells were cultured for additional two days.

Dissociation of hiPSC-CMs
At the specified day of differentiation, the culture medium was washed with DBPS -/and CMs were dissociated by using 0.05% Trypsin-EDTA (Thermo Fisher Scientific, Cat. # 25300-054) for 25 minutes at 37°C. The Trypsin was inhibited with 0.5% bovine serum albumin in DMEM:F12 and the dissociated cells were passed through an EASYstrainer filter with 40 μm pore size. Cell number was determined in an automatic cell counter and CMs were used for downstream analyses.
Procedures for the synthesis of compounds C8 and C9 can be found in the reference of Berkessel et al. (Berkessel et al., 2007).
The nicotinoyl and iso-nicotinoyl hydrazides PluriSIn #1, P1, P3, P5 and P7 were prepared by standard coupling of the corresponding hydrazines with acid chlorides in the presence of triethylamine. The same holds for the benzoyl hydrazine P4. The PluriSIn #1 derivatives P2, P6 and P8 were prepared by reductive alkylation of iso-nicotinoylhydrazide with benzaldehyde, cyclohexanone and cycloheptanone, respectively, in the presence of sodium borohydride. Synthesized PluriSIn #1 and derivatives were applied on mPSCs and compared to PluriSIn #1 purchased from Sigma-Aldrich (catalogue number SML0682). For all compounds (PluriSIn#1, P1-8), the analytical and spectroscopic data were in agreement with the structures shown in Supplementary Fig. S1a, and confirmed the purity of the materials synthesized.
All substances were diluted in DMSO at the final stock concentration of 50 mM (PluriSIn #1 and its derivatives P1-8) or 20 mM (all 16 SM compounds SM1-SM16). For mPSCcytotoxicity assays, serial dilutions of DMSO stock solutions were prepared with cell culture medium.

Immunocytochemistry (ICC)
For ICC analyses, miPSC-CMs were plated on fibronectin-coated multi-well plates (110 5 cells/cm 2 ) and treated with 8 µg/ml puromycin for two more days as described above. CM   Analysis was performed using FCS express 6 (De Novo Software, Glendale, CA).
Cells were disrupted by pipetting with the yellow pipette tip 10 times. The homogenate was sonicated twice for 5 sec per cycle in order to completely disrupt the cells and to shred the DNA. Samples were incubated on ice for 15 minutes to generate a whole cell lysate (WCL).
Control WCLs of human HEK293, HT29 and COS9 cells were prepared in the same way.
Lysates were then centrifuged at 10,000 xg for 10 min at 4°C and the supernatant was collected. Protein concentration was determined by Pierce BCA Protein Assay kit (Thermo Fisher Scientific), aliquots were snap-frozen in liquid nitrogen and then stored at -80°C until use.

Subcellular fractionation
For fractionation into nuclear, mitochondrial and cytosolic fractions cells cultured in 6 cm plates were collected after 8 h of treatment with SM6-or 0.05% DMSO, washed with DPBS +/+ and resuspended in 500 μl of ice-cold subcellular fractionation buffer containing 250 mM sucrose, 20 mM HEPES (pH 7.4), 10 mM KCl, 2 mM MgCl 2 , 1 mM EDTA, 1 mM EGTA, 1 mM DTT and 1x Protease Inhibitor Cocktail, cOmplete Mini (Roche). Cells were disrupted by passing cell suspension ten times through a 27G needle using 1 mL syringe and lysates subsequently incubated on ice for 20 min. Samples were then centrifuged at 720 x g for 5 min at 4°C to collect nuclei in the pellet. The supernatant containing cytoplasm, membranes and mitochondria was transferred into a fresh tube on ice and centrifuged at 10,000 x g for 5 min at 4°C to pellet the mitochondrial fraction. The supernatant containing the cytoplasm and membrane fraction was aliquoted into a new tubes sitting on ice. The nuclear pellet from previous step was washed with 500 μl of subcellular fractionation buffer and passed through a 25G needle ten times. After centrifugation at 720 x g for 10 min at 4°C the supernatant was discarded and the pellet containing nuclei was resuspended in 300 μl of Tris-buffered saline (TBS) containing 0.1% SDS, sonicated briefly to shear genomic DNA and homogenize the lysate. The mitochondrial pellet was processed as described for the nuclear pellet to obtain mitochondrial lysate in TBS/0.1% SDS. Protein concentration in samples was determined using Pierce BCA Protein Assay kit (Thermo Fisher Scientific).

For immunoblotting, protein samples (14 μg/lane) and the SeeBlue Plus2 Pre-Stained
Standard (Life Technologies) were loaded onto 12% separating polyacrylamide gel. After electrophoresis they were transferred onto PVDF membrane using the iBlot system (Life Technologies). After blocking for 60 min at room temperature in buffer containing 0.1 % Tween 20 and 0.5 % milk powder in TBS, pH 7.5, the membrane was incubated overnight at 4°C with primary antibodies against p53, cytosolic marker α-tubulin, or mitochondrial marker voltage-dependent anion channel (VDAC) (see Table S4). After washing with TBS, membrane was incubated for 1 h at room temperature with fluorescently conjugated secondary anti-rabbit IgG or anti-mouse IgG polyclonal antibodies in 0,5% milk in TBS (see

Reactive oxygen species (ROS)
The ROS levels in miPSC-CMs after treatment with various compounds was determined by

Action potential (AP) recordings
APs were measured in spontaneously beating cardiac clusters from day 16 of differentiation using sharp glass microelectrodes with a resistance of 20-50 MΩ when filled with 3 M KCl (World Precision Instrument, Sarasota, USA). The intracellular recordings were performed as described before (Halbach et al., 2006) and signals were acquired with a SEC-10LX amplifier (npi electronic, Tamm, Germany) and Pulse software (HEKA, Lambrecht/Pfalz, Germany).
Data was analyzed with Mini Analysis Program (Synaptosoft, Fort Lee, USA).

Teratoma histology
Mice were sacrificed and hind limbs were dissected, fixed with 4% PFA overnight at 4°C),   Table S2 and Table S3 for exact IC 50 values for three human and three murine PSC lines, respectively. Figure S3. mPSC-cytotoxic effects of SM6 and its structurally related compounds and stereoisomers. (a) Relative viabilities of PIG-AT25 miPSCs exposed for 48 h to SM2, SM6 and SM8 derived from two different batches of molecule synthesis and (b) to SM6 stereoisomers (n=4). Differentially colored symbols indicate different batches (in a) or different stereoisomers (in b). SM6 is in an R,R conformation. (c) Relative viabilities of PIG-AT25 miPSCs exposed for 48 h to SM6 and ten structurally related compounds C1-C10 (n=4).       Abbreviations: CPVT1 -Catecholaminergic polymorphic ventricular tachycardia, type 1;

Supplementary Tables
PBMC -peripheral blood mononuclear cells.