everal challenges must be overcome before successful implementation of iPSC-based drug screening and pathway discovery can be achieved (). The most critical issues are whether the relevant disease phenotypes can be faithfully reproduced in vitro
and, if so, whether they can accurately predict disease behaviour in vivo
. Despite promising studies suggesting that certain features of familial dysautonomia, SMA and RETT syndrome can be generated using iPSC-derived neural cells, other neurological disorders such as Parkinson's disease seem more difficult to model so far20
. Three main factors may influence the amenability of diseases to in vitro
modelling: the onset of disease in patients, the cell-autonomous nature of the disorder and the complexity of the underlying genetic defects. For example, evidence from animal models and clinical data indicate that familial dysautonomia, SMA and RETT syndrome manifest early in life, may have a strong cell-autonomous component and are caused by mutations in single genes, whereas Parkinson's disease generally occurs later in life and is caused by environmental and complex genetic factors. However, it is still unclear which of these three elements most strongly influences our ability to generate the relevant disease phenotype in vitro
. It is possible that a disease such as autism, which involves complex genetics but manifests early in life, could still be modelled with an appropriate iPSC-derived cell type. Many diseases with the greatest societal impact are polygenic and highly influenced by environment (for example, congestive heart failure, Alzheimer's disease, diabetes, sudden cardiac death, emphysema and Parkinson's disease). It remains to be seen whether their key phenotypes can be reproduced in vitro
If the aetiology for disease development is known or suspected, there may be ways to introduce the causal agent into purified iPSC-derived cells to induce or accelerate the manifestation of disease phenotypes (see ). For example, in amyotrophic lateral sclerosis (ALS), superoxide dismutase (SOD) mutations affect the function of glial cells surrounding motor neurons. Studies have shown that co-culture of human ESC-derived motor neurons with glial cells carrying the mutation induces neuronal death48,49
. An in vitro
disease model could therefore potentially be generated that uses glial cells and motor neurons derived from iPSCs from an ALS patient in a similar co-culture system. Another example is Duchenne muscular dystrophy (DMD); the skeletal muscle phenotype of this disease is thought to be due to both the presence of dystrophin mutations and cumulative mechanical stretch injury from muscle use50
. Thus, mechanical stress (or catecholamine stimulation) may need to be applied to iPSC-derived skeletal muscle to appropriately model this disease in vitro
. For other complex diseases, exposure of relevant chemical agents or toxins to iPSC-derived cells may reveal phenotypes that would otherwise remain undetectable. For example, in one study that generated dopaminergic neurons from iPSCs derived from patients with sporadic cases of Parkinson's disease, no obvious abnormalities could be detected20
. However, in a subsequent study, dopaminergic neurons derived from iPSCs obtained from a single Parkinson's disease patient harbouring a mutation in the leucine-rich repeat kinase 2 gene (LRRK2
) were exposed to oxidative stress and demonstrated increased susceptibility to cell death51
Potential human iPSC models of complex diseases involving environmental factors
Assuming that disease features can be reproduced in vitro
, it is still unclear whether the phenotypes can be used for high throughput small-molecule screening. A major limitation is the lack of robust lineage-specific differentiation protocols that enable researchers to generate sufficient quantities of purified cells of a specific type for large-scale screening applications. Although significant advances have been made to direct the differentiation of ESCs or iPSCs into certain types of neurons52,53
and pancreatic cells41,62
, none of these protocols generates the cell types of interest with > 95% purity. Sorting of these cells from the heterogeneous iPSC mixture to reproduce the disease phenotypes for high-throughput small molecule screening remains a challenge. Improvements in cell-purification strategies (for example, fluorescence activated cell sorting, drug selection, gradient centrifugation and functional marker isolation) may eventually allow us to overcome this barrier. The use of small-molecule screens to identify compounds that can enrich for a cell type of interest has also proved valuable51,63–67
. Despite these challenges, some companies already offer human iPSC (hiPSC)-derived cardiomyocytes in quantities that are suitable for drug discovery and toxicology testing.
The heterogeneity of the maturation stage of the differentiated iPSCs is also a potential limitation. A high-throughput screen aimed at identifying small molecules that improve cardiomyocyte contractility may have a high rate of false-positive and -negative hits if there are well-to-well differences in differentiation state, as mature cardiomyocytes exhibit greater contractility than their immature counterparts. Similar issues may apply to small-molecule screening using hepatocytes or pancreatic β cells if the end-point of analysis is the secretion of specific enzymes or hormones, which strictly depend on the cells’ maturation stages.
Once these barriers to the development of robust in vitro disease models using iPSCs are overcome and small molecules that can reverse the disease phenotype in vitro have been identified, an appropriate animal model will be needed to validate the in vitro screen ‘hits’ in vivo. For candidates that are Food and Drug Administration (FDA)-approved drugs, with known pharmacokinetic and toxicity profiles, no additional animal studies might be needed. For small molecules that have not been previously tested for their pharmacokinetic, toxicity and efficacy profiles, a standard pre-clinical evaluation of these molecules in vivo will still be required. It is likely that a number of small-molecule candidates identified from such screens might show efficacy only in the artificial conditions of an in vitro assay. Thus, large-animal models of disease would be essential to help eliminate these candidates with insufficient biological efficacy or enhanced toxicity in vivo. Investment of research resources to create reliable animal disease models should thus be a significant priority if we are to realize the full potential of therapeutic drug screening efforts using disease-specific human iPSCs.