Can induced pluripotent stem cells be used to generate useful Parkinson's disease cell models?

Dr. Bill Langston sat down with  Dr. Michael Rogan and Dr. Holli Kawadler to discuss the future of induced pluripotent stem cells (iPS cells), and their use in Parkinson's disease modeling.

BL: The discovery was made back in 2007 (Takahashi et al, 2007) and the concept is to insert certain genes into adult cells and allow those cells to revert back into so-called “induced pluripotent cells” or “stem-cell like” cells. It was first done by taking skin cells and growing fibroblasts, but in theory you could do this with other cell lines. It’s been done with ketarinocytes from human hair – you just pluck out human hairs and there are enough cells to get a cell culture going – and it’s also been done with hepatocytes (Aasen et al, 2008; Aoi et al, 2008).

There are four different genes that drive these adult cells into stem cell-like cells and then are a whole series of characteristics they are supposed to meet, including injecting them into muscle in mice where they form teratomas and basically develop into all three cell types [endoderm, mesoderm, ectoderm]. This proves it’s a stem cell and can differentiate into different types.

MR: Why were fibroblasts used first?

BL: I think it was just the easiest and most obvious. Skin biopsies are quite easy to obtain now; they come with little kits, any physician can do them. And, fibroblasts grow like crazy and are very hardy. But conceivably we may see a day when people routinely go to the doctor and pluck a few hairs and they can generate new tissues out of it, so it’s quite an interesting concept.

HK: How satisfied are you that iPS cells are equivalent to natural stem cells?

BL: You’re never 100% in science, but I personally am 90-95%. There are a number of papers out showing that there are very similar to authentic stem cells. There’s even some recent evidence that telomeric length (the aging clock in cells) reverts to the length one typically sees in embryonic stem cells (Marion et al, 2009). Quite a number of labs have replicated iPS cell work now; these cells are clearly capable of differentiating into multiple tissue types and look very “stem cell-like”. It may turn out they’re not identical, but they’re awfully awfully close, and I think for many of the uses we’re hoping to put them to they’re going to be quite adequate.

HK: In terms of generating PD cell models from Parkinson’s patients, what about the epigenetic modifications that may be lost in the process?

BL: That’s a good question, and it does appear that some of those epigenetic features are lost. Now that could be a good thing, or a bad thing. It might be for certain purposes, starting fresh is a good thing. If on the other hand, you are trying to model the disease, you might want to have all of those modifications. And, I would say that’s one reason I’m less optimistic that cells derived from idiopathic PD patients are going to be useful for modeling.

Let’s back up a little bit: let’s assume you follow this process and you get very nice iPS cells from your patient. The next step would be to drive them forward into a cell type of interest, and for the last 50 years we’ve been focused on nigrostriatal dopaminergic neurons. It’s very obvious that there are other cell types that are of just as much interest in PD – but let’s stay with the dopaminergic nigrostriatal system for the moment since there’s so much work on that system and it does cause the classic motor symptoms of PD.

There’s pretty good technology now for inducing the differentiation of dopaminergic neurons, a lot of funded by the Fox Foundation  in the early days when we were so excited about stem cells for transplantation purposes. [PDOR Note: Dr. Langston was the Founding Chief Scientific Advisor for MJFF]. The first grant programs we ever launched were to get people to develop dopaminergic neurons from embryonic cell lines, mostly in animal work back then. And now there are a lot of good formulas out there; a lot of labs have successfully taken embryonic stem cells into dopaminergic cell lines. While that was originally done for transplantation, it really paved the way for modeling.

So the next question is: will these cells from PD patients develop a parkinsonian phenotype? I think the fact that we are losing a lot of those epigenetic factors, plus we still don’t know why those patients got Parkinson’s in the first place, makes me less optimistic you’ll get a disease phenotype in cells from idiopathic patients. Now, if you’re using them for neurotransplantation, you don’t want a disease phenotype. I think when iPS cell technology first came out, everybody thought, “Oh wow, transplantation. We’re going to use this to cure the disease!” As you know, we’ve been trying to cure Parkinson’s with fetal cells for 20 years without much success. And while someday that may be a use for iPS cells, and they may solve ethical issues and immunorejection issues, I think we are a long way away from these cells being used for transplantation to cure the motor features of the disease.

The real interest we have is quite the opposite: we want to get a disease phenotype in these cells. The real excitement is to use these cell lines to study disease mechanism and use them  for drug discovery. In that case, if we don’t get a phenotype it will be disappointing.

Our strategy is to go to patients with known genetic forms. If you’re looking for the chance to develop disease models, it makes sense to start with patients who have a known cause, and that known cause is in their DNA. It’s not epigenetic, it’s genetic. Our idea is to develop cell lines from patients who (A) have the disease and (B) we know the cause of the disease (say a LRRK2 mutation as an example). It’s a patient with the disease phenotype, with the cause known, and the cause is known to be in the DNA so presumably it would go with those cell lines. Our idea, along with a number of other groups, is to use iPS cell technology to develop new ways to study the disease. Obviously, it’s just not possible to biopsy the substantia nigra and grow those cells to find drugs that stop the disease or to study mechanism, and this has been a huge bottleneck. Using iPS cells we may be able to get cells with authentic Parkinson’s in a dish to study underlying disease mechanisms and for drug development. That’s where the real excitement is.

The concept is to find these cases with a known genetic cause, develop cell lines, take them back to iPS cells, and then forward to dopaminergic neurons. Hopefully one or more of these genetic forms will develop a phenotype or a signal that we can use to study disease mechanism and eventually it would be wonderful to develop high-throughput screening that can be used for drug discovery. I like to argue that most of our current models are “best guess” models (MPTP, viral injection, synuclein injection).  Developing iPS cells from people with genetic forms of PD will  in fact come  as close to actually having the disease itself in a system that we can use experimentally that we have come to so far, assuming that we are able to get a disease phenotype.

MR: What are the limitations of the ‘disease in a dish’ model, given that it lacks much of the circuit connections and complex environment  of  the brain?

BL: You could do it several ways. You could develop pure cultures – that would be the simplest because your experimental work wouldn’t be confounded, but I think there’s a fairly good chance we’re going to need to do a lot more. You could do co-cultures with glia, or conceivably with other neuronal cell types, and I think there’s a good chance we’re going to need to stress these cells, whether it’s oxidative stress or neurotoxins. It may turn out that just growing them isn’t going to be enough and we may have to do a lot of different things, including co-cultures and more complex culture systems, and/or various types of stress.  It may take quite a bit of work to get these cells to cough up a phenotype.

MR: What’s your favorite vision of a phenotype-in-a-dish?

BL: The dream would be Lewy bodies or Lewy neurites, but I don’t think we’ll be that lucky. But if we got even substantial synuclein aggregates and it was clear that those cells were not functioning normally.  It would be nice to see some functional deficit in these cells compared to controls from normal subjects or spontaneous PD patients. But, a morphologic signal would be the best.

MR: Do you think the search for phenotype is going to focus on alpha-synuclein?

BL: That’s what we’re going to do, but there are many other things we may end up looking for: other markers of oxidative stress, mitochondrial dysfunction, functional tyrosine hydroxylase, uptake sites, dopamine release. The more complex the phenotype, the harder it’s going to be to adapt it to high throughput screening, of course.  Sometimes a simple aggregation where you can put some kind of marker on it might be much better, but those are all the challenges we’re going to face.

MR: Are there assays that you could use for alpha-synuclein without killing the culture?

BL: There might be. If we get some type of aggregation, say a beta-pleated sheet that fluoresces, then yes. That type of marker would be a dream for the high-throughput screening. Otherwise we may have to start tagging or get something that would affect cell viability. The big drug companies have spent decades developing high throughput screens to study their systems, so we’re hopefully going to collaborate with some people who have experience with those steps. We have a lot of candidate compounds we’d like to test – we’ve developed a high-throughput screening system in vitro using a fluorescent system where you can measure synuclein aggregation.  Synuclein will spontaneously form fibrils over about 60 hours in a test tube, and we’ve been working to develop a high throughput screening method with a number of compounds. We’ve screened about 4000 compounds and found a number of compounds that prevent and even reverse aggregation in a test tube. So we’ve got some very good candidates. Even if we don’t have high throughput screening, we have a number of lead candidates we would like to test the old fashioned way if we had a model like that, doing the experiments in culture one at a time. And I’m sure many of groups have similar compounds they’d love to look at.

MR: Earlier you brought up the fact that PD involves more than just dopamine neurons.  What would you propose as a  suite of cell types that would constitute  a good model for PD?

BL: That’s a great question. The one I’m most interested in is the olfactory bulb. The question is which cell type is it in the olfactory bulb. I’m still not quite clear on that. There’s increasing evidence, started by Heiko Braak, that the olfactory bulb is one of the very first areas affected in this disease, and it’s consistently and fairly significantly affected. That may be a system we’d want to look at much earlier as a feature than the nigrostriatal system. If you’re thinking neuroprotection, that would be a fascinating one. The trouble is that it’s not the Th-positive cells that are actually affected, as I’m told by John Duda (Kranick and Duda, 2008), one would need to sort out what cell type in there is affected.

Another would be the peripheral autonomic system: the superior cervical ganglion is affected; the enteric enervation of the gut, the colon is affected. These are cell types that are generally affected much earlier. The irony here is that the transplantation field, and I think the Fox Foundation which was a real part of this, got really focused on dopaminergic neurons early on because we thought we were going to transplant dopaminergic neurons into the adult brain. The technology and the cocktails are very well developed to differentiate stem cells and embryonic stem cells into dopamine neuronsand will be very useful for this work. The question is what other cell types has that been done with? I know cardiomyocytes have been done, but I’m not sure in the nervous system what groundwork’s been done to really transform stem cells into the other cell types.  My dream would be a set of cells from some of these systems that we think are hit the earliest: the cardiac sympathetic enervation may be earlier than even brain – we don’t know that yet; work from the Hawaiian Heart Study found that mid-life constipation has a four-fold increase in risk for PD, and we think now that may be just early involvement of synuclein pathology in the enervation of the gut.  And synuclein is just a marker, we don’t know if it’s the cause, but it’s certainly a very good marker.  Some of those would be really interesting to think about for the future.

The great hope is that iPS cells will yield something that is much closer to being a real authentic model of Parkinson’s than anything we’ve had to date, which are really just “best guess” attempts to model the disease. I think that’s where the excitement is and the big unknown, for those who are interested in modeling, mechanisms, or drug development, is whether we are going to get a phenotype.  For neurotransplantation, the last thing you want to see is Lewy bodies – it’s a very different direction.

Dr. Lorenz Studer discusses his recent work using patient-derived iPS cells as models of pathogenesis and treatment of familial dysautonomia, and the issues surrounding using this technology for Parkinson's disease modeling and drug discovery. - HK

HK: Can you summarize your findings?

LS: The key point is that we were able to model a neurological disorder using iPS cells for which there are no good alternative models (animal models not predictive). While previous iPSC studies (3 key papers todate) have been able to generate patient specific cells of interest, our study manages to go beyond simply generating the cells and identify three disease related phenotypes in a dish (defect in splicing, defect in neurogenesis / peripheral neuron birth, and a defect in cell motility. These in vitro surrogate markers of disease help to understand how a mutation in a gene with a rather generic function (controlling mRNA "elongation") can affect rather specifically the peripheral nervous system. The results of reduced expression of ASCL1 was particularly interesting, as knockout mice for that gene lack autonomic neurons completely, which is the neuron type most affected in familial dysautonomia. Finally, we took all the drugs/compounds that based on the literature have been proposed at some point in time as being candidate therapeutic drugs, and found that one of them indeed was able to reverse splicing and neurogenesis defect. It also gave some insights into the length and timing of treatment regimen that may be particularly effective in combating FD symptoms.

HK: What aspects of familial dysautonomia make it a good candidate for iPSC modeling?

LS: It is a disease with a clear genetic origin, early onset and affecting a tissue type for which we can model differentiation in vitro. Mouse models have not been successful at modeling the disease stating the need for alternative approaches. The splicing defect (exon skipping of IKBKAP) was an in vitro surrogate that we expected to see in vitro while the ability to model the neurogenesis and cell motility aspects of the disorder was unclear at the beginning of the project. The question is obviously whether diseases with a less obvious genetic origin and diseases that are late onset can be modeled using the same approach (see below)

HK: Previous work has shown derivation of PD patient-specific dopamine neurons from iPS cells (Soldner et al., 2009). What are the potential differences in modeling developmental disorders versus neurodegenerative disorders such as Parkinson’s disease? Are there differences in studying disease pathogenesis and/or drug discovery?

LS: I think the approach is particularly suited for the study of disease with a clear genetic origin, and particularly genetic disease with relatively early onset. Examples that fit this category are numerous (several dozen) such as Rett syndrome, Fragile X, neuropathies such as Charcot-Marie Tooth (CMT),  Down Syndrome, types of Leukodystrophies (Niemann Pick etc..), Tay-Sachs disease and many others. Outside the nervous system some of the most common ones that come to mind are Duchenne Muscular dystrophy and cystic fibrosis. Obviously, one key ingredient in the success of the approach is the ability to direct the fate of undifferentiated iPSCs into the cell types affected and to purify the differentiated cell type(s) of interest for comparative studies of disease-specific and control cell lines.

Modeling early onset genetic disorder may be easier, as iPSC derived cells  reflect a young developmental stage, and it is currently difficult to conceive the generation of iPS cell derived cells in culture that correspond to cells living in a 60 or 70 year old person. This could be the reason that the only other disease that showed some promise to date for being modeled in vitro via iPSC approach is SMA, a motoneuron disorder with strong genetic contribution and early onset (Ebert et al., 2009). However, it is too early to say that late onset disorders such as genetic forms of Alzheimer's disease or Parkinson's disease cannot be modeled at all, they may simply need some additional "tricks", such as exposure to environmental stresses mimicking aging process.

Specifically for the case of PD (Soldner et al., 2009), there are a number of issues to consider.

First of all the differentiation protocols for midbrain DA neurons from human ESC/iPSC are still very suboptimal. My lab has worked on this for many years and the protocols used by Soldner et al., where based on some of our old protocols. However, unlike mouse ESC/iPSC derived dopamine neurons that can be generated and purified quite efficiently, human ESC/iPSC derived dopamine neuron have a number of issues best illustrated by their overall poor performance in vivo. For example in the case of the Soldner study, it would be essential to demonstrate that the dopamine neurons indeed express midbrain specific markers at sufficient levels such as FoxA2, Pitx3 to be suitable for modeling PD..

A second issue is that they tried to model idiopathic PD rather than the more rare forms of PD with a clear genetic contribution. It is currently unclear to what extent that will be possible and it may be important to first demonstrate whether or not genetic cases of PD (LRRK2, Parkin, DJ1, a-syn etc) show any disease phenotype in vitro.

Third, we need to be able to better understand aging process to recreate a cells that have at least some features of a 50 or 60 year old neuron Finally, we need to figure out what cells types are needed to best model the disorder. Obviously DA neurons are important but not the only cell type affected and it may be that non-cell autonomous factors play an important contribution to their demise

HK: What are your next steps?

LS: Some of the immediate follow-up studies in FD include performing gene correction by knock-in technology in the homozygous disease lines and testing iPSC lines from carriers that are typically asymptomatic. We are also interested in testing the predictive power of the iPSC approach at even higher resolution, i.e. comparing individual patients with distinct disease onset or distinct disease manifestation (sensory versus autonomic versus non central symptoms), and to test individual responses to kinetin and possibly other compounds in those patient populations. We are also planning to screen larger panels of compounds based on their ability to reverse reversing splicing but also reverse phenotypic changes. Finally, we are pursuing collaborative studies that explore the basic mechanism underlying the tissue specific splicing defect and the role of kinetin in reversing mis-splicing in FD-iPSCs.

Beyond FD, we have started to extend the approach to other more common neuropathies such as Charcot Marie Tooth and a number of other neural crest disorders. We also work on iPS approaches in the CNS including PD.

Gabsang Lee, Eirini P. Papapetrou, Hyesoo Kim, Stuart M. Chambers, Mark J. Tomishima,
Christopher A. Fasano, Yosif M. Ganat, Jayanthi Menon, Fumiko Shimizu, Agnes Viale, Viviane Tabar, Michel Sadelain & Lorenz Studer. Modelling pathogenesis and treatment of familial dysautonomia using patient-specific iPSCs. Nature. Published online Aug 19. dio:10.1038/nature08320

The California Institute of Regenerative Medicine (CIRM) released yesterday a list of its newly funded stem cell research grants. Although none have a clear focus on Parkinson's disease, what is interesting is how many were not focused on embryonic stem cell-based approaches. (The New York Times discuss this in a news item posted yesterday as well.) In fact, several projects focus more on the potential promise of 'adult' stem cell approaches, including the use of induced pluripotent stem cells.

Superficially, this seems to be creating a stir in the whole 'value of ES vs. adult stem cell' debate that has fueled political/ethical fights for years. But it strikes me that what is perhaps lost in this is the deeper discussion of the fact that not every disease necessarily will benefit from the same stem cell approach in the same way. So bone marrow-derived stem cell approaches clearly have promise for leukemia and similar adult stem cell approaches seem to be growing in rationale for heart disease. But for CNS degenerative disorders, my sense from speaking to researchers in the field is that ES cells (and maybe iPS cells) are still at the top of investigator's lists as best approaches to take forward for creating replacement neurons (including dopamine neurons for PD). This apparently has to do with the fact that it is easier to get more and better neurons from ES/iPS cells than from other adult stem cell sources.

Is this still the right way to think about the field in this way? Or is the line between value of adult stem cell approaches vs. embryonic stem (or even iPS) cell approaches becoming fuzzier with respect to diseases like PD? Where should the dollars be prioritized if goal at end of day is to generate quality (and safe!) replacement dopamine neurons for transplantation purposes? Do we even have the right benchmarks to use to even make that sort of direct comparison between different stem cell sources?

There are still a number of questions to consider when looking at using iPS in both regenerative medicine and disease modeling, as outlined previously. How similar are ES and iPS cells, can we mitigate the differences, and how much will these differences impact use of these cells? Two new papers this week address the first question (Feng et al., 2010; Hu et al., 2010); both groups directly compare ES and iPS cells looking at differences in differentiation, proliferation, and senecence. Their results boil down to few differences in the functional cells that are produced, but a vast decrease in the efficiency of iPS cells.

Feng et al. directly compared 14 hESC lines with 8 hiPSC lines, the latter generated by lentiviral Oct4, Sox2, Nanog, and Lin28 expresstion or retroviral Oct4, Sox2, Klf4 or c-Myc expression. All iPSC lines could be differentiated into hemangioblasts/blast cells, and hematopoietic and epithelial lineages, though with much lower efficiency than ESC lines (~500 fold). The iPSC-derived cells grew very slowly as compared to the ESC-derived cells, yielding significantly lower cell numbers (>1000 fold) and a higher percentage of senescent cells.

Hu et al. focused on neural differentiation. They compared 12 hiPSC lines (retroviral, lentiviral and non-integrated expression) with 5 hESC lines. All lines were able to differentiate into neuroepithelial cells using the same program, and the iPSC lines could be differentiated into both neurons and glia similarly to the ESC lines. Additionally, both groups could be differentiated to functional spinal motoneurons. However, the iPSC lines showed highly variable efficiency and differentiation, regardless of source or reprogramming method.

Regarding function, iPS cells have already been shown to be truly pluripotent, able to give rise to healthy adult mice via tetraploid complementation (Boland et al., 2009; Kang et al., 2009; Zhao et al., 2009); this work seems to support few functional or developmental differences between ES- or iPS-derived cells. The overwhelming issue remains to be the very low efficiency and variability of iPS cells to expand and differentiate. Hu et al. report this variability in cells of the same lines, derived from the same fibroblast, likely meaning that the cell of origin is not to blame. Is it the technique of reprogramming that triggers a senescent phenotype? Would other reprogramming methods (ie.with small molecules opposed to viral vectors) still result in this low efficiency? Furthermore, how much does this finding impact using iPS-derived neurons therapeutically? If the resulting neurons are functional, does this low efficiency matter (as long as we can generate enough cells)?

Bao-Yang Hu, Jason P. Weick, Junying Yu, Li-Xiang Ma, Xiao-Qing Zhang, James A. Thomson, and Su-Chun Zhang. Neural differentiation of human induced pluripotent stem cells follows developmental principles but with variable potency. PNAS published online before print February 16, 2010, doi:10.1073/pnas.0910012107