Our screening approach provides a new platform technology for using hESCs and purified populations of their differentiated neural derivatives to rapidly screen and identify compounds that exert specific effects on these cell types. This screening approach relies on the observable phenotype of cell death coupled with gene expression analysis to identify pathways of cell-type specific drug activity. To extend its utility, this approach can also provide clues to the molecular mechanisms that participate in stage-specific cytotoxic effects of candidate drugs. We had reasoned that because of fundamental differences in cell cycle and growth factor dependence, there would likely be drugs that were specific to one cell type versus another. Indeed, as expected in our primary screen we identified nine such compounds. Of these initial 9 candidates, 6 compounds demonstrated dose responsive toxicity exclusively in NSC populations. Interestingly, the compounds amiodarone HCL and selamectin had the most dramatic ameliorating effect on NSC survival (). It was surprising to us that none of these compounds were in the expected classes of anti cancer or anti-proliferative agents but instead included anti-parasitic and antiarrhythmic drugs.
We chose to further investigate one of these drugs, amiodarone HCl, which specifically killed NSCs but not dopaminergic neurons differentiated from NSCs. Amiodarone has for decades achieved clinical status as an effective class III antiarrhythmic drug in cardiac patients 
. Importantly, because it is already approved for clinical use, amiodarone HCL may have clinical applications in cell replacement therapies by selectively removing only the unwanted undifferentiated NSCs during the pre-transplant period.
In order to confirm that the cytotoxic effect seen in the amiodarone HCL-treated NSCs was specific to the activity of the drug, we first sought to determine which cellular pathways were affected in the amiodarone HCL susceptible NSC population relative to unaffected dopaminergic neurons receiving the same treatment (Fig. S2
). The GSEA data revealed amiodarone HCL treated samples had significantly reduced expression of factors involved in ion channel activity. Amiodarone is known to specifically block ion channels, which suggests that the effect seen in the drug treated samples is specific to amiodarone HCL activity. To further test this, we reasoned that populations of cells with a greater basal expression of ion channel activity mediators would be most susceptible to drug treatment. Indeed, microarray data confirmed that amiodarone HCL-susceptible NSCs have significantly increased base-line expression of certain ion channels (, SLC2A1 and CLC1A). It is tantalizing to speculate that amiodarone HCl might also be toxic to other stem cell populations that demonstrate increased ion channel expression relative to their differentiated derivatives, including mesenchymal stem cells (MSCs) and endothelial precursor cells 
, thus expanding the utility of the automated screening assay described here.
Amiodarone has been shown to exert its cytotoxic effect via
a TNF-related signaling pathway that includes caspase-8 mediated apoptosis 
. Thus, we next wanted to determine whether our assay could detect subtle changes in TNF activity in samples treated with amiodarone HCL. Notably, downstream members of the TNFR2 pathway were significantly augmented in the amiodarone HCL-treated NSC population (Fig. S2
). TNFR2 belongs to a class of membrane glycoprotein receptors that specifically bind TNFα. TNFR1 is expressed on most cell types, while TNFR2 expression is restricted to endothelial, hematopoietic and some neuronal populations 
. TNFα is a potent pro-inflammatory cytokine with two biologically active forms that are either soluble (solTNF) or membrane bound (tmTNF), and TNFR2 is preferentially activated by tmTNF 
. It was initially thought that TNFα-mediated signaling downstream of TNFR1 results in apoptosis, while those downstream of TNFR2 induce proliferation 
. Additional work, however, revealed that in collaboration with TNFR1, TNFα can act upon TNFR2 through a ligand passing mechanism and trigger apoptosis 
These published reports in their aggregate support that TNFR2 can lower the threshold of bioavailable TNFα needed to cause apoptosis through TNFR1 thus amplifying extrinsic cell death pathways. In fact, short term treatment of patients with amiodarone leads to a significant decrease in the patient's serum TNFα concentrations while paradoxically the amiodarone toxicity is exerted through TNF-mediated apoptotic pathways 
. These observations are explained by the fact that amiodarone HCL up regulates TNFR2, and TNFR2 is more dependent on ligation with tmTNF than solTNF. To test this model, we treated amiodarone HCL-susceptible NSCs with solTNF. If amiodarone HCL toxicity is mediated through TNFR2, and TNFR2 is not sensitive to solTNF, then addition of solTNFα should not be cytotoxic to the NSCs. Indeed, three doses of solTNFα (0.1 µM, 1 µM and 10 µM) were tested in NSC culture for 48 hours and no increase in cell death relative to untreated cultures was observed (). This supports published reports that the addition of solTNFα to NSC cultures actually induces proliferation and differentiation 
. Since TNFα is such a potent inducer of apoptosis through TNFR1 death domain signaling, and amiodarone treatment results in the down regulation of TNFα with concomitant upregulation in TNFR2 signaling in NSC alone, it is possible that amiodarone selectively kills NSCs by lowering the threshold of TNFα required to trigger apoptosis in NSCs via
upregulation of TNFR2 pathways in NSCs and not dopaminergic neurons.
Our results support our primary goal of identifying a previously approved drug that may allow us to deplete mitotic NSCs from an otherwise differentiated population of dopaminergic neurons, thus ensuring their safety for use in transplantation. Importantly, this automated screening assay allowed us to interrogate some of the specific molecular mechanisms that may be responsible for the targeted cytotoxic effect amiodarone HCL had on NSCs and not cells differentiated from NSCs. While we do not purport to know the molecular mechanisms by which amiodarone HCL leads to the toxicity we observed in NSCs, it is notable that the results of our automated screening, including GSEA and microarray analysis, are all consistent with published literature that implicates the roles of ion channels and TNFα signaling in amiodarone-mediated cytotoxicity. This suggests that our automatic screening assay is specifically measuring the effect amiodarone HCL has on different populations of cells. Our methodology can also be easily expanded to other screens in the neural system. For example, we note that purified populations of motor neurons and oligodendrocytes are now readily available from hESCs and our screening strategy can likely be extended to these cell populations as well.
In conclusion, we describe a method using hESCs and their differentiated neural derivatives that permits the rapid screening of clinically approved drugs for compounds that can be safely used to selectively deplete progenitor cells from a differentiated cell product. Importantly, this approach is adaptable for use in a Chemistry, Manufacture and Control drug screening protocol and may have applications in identifying lineage specific reagents, thus providing additional evidence for the utility of stem cells in screening and discovery paradigms.