The promise of embryonic stem cells (ESCs) in transforming regenerative medicine and disease therapies has long captured the fantasy of scientists and the general public. Yet progress toward these goals has been mired by ethical issues surrounding the derivation of ESCs, which results in the destruction of human embryos. In the years following the discovery of ESCs, other biological phenomenon similarly captured scientists' imagination. They witnessed cases of in vivo cellular de-differentiation, the reversion of differentiated cell types to an earlier, mitotic precursor state (Echeverri et al.,2001; Brawley and Matunis,2004). If only they could wave a magic wand and similarly revert somatic cells to a pluripotent state, this new induced cell type could potentially sidestep controversy and replace ESCs.
In an experiment of intelligent alchemy, Yamanaka's group successfully pulled off the magic trick. By misexpressing combinations of ESC maintenance factors in mouse fibroblasts, they hit on a magic elixir of four factors Oct3/4, Sox2, Klf4, and c-Myc (OSKM), that seemingly turn back time (Takahashi and Yamanaka,2006). Somatic cells transform to an ES-cell like state, called induced pluripotent stem cells (iPSCs). It was later shown that similarly derived iPSCs can generate fertile adult mice, demonstrating that they are fully pluripotent (Boland et al.,2009). Human iPSCs also pass the most stringent pluripotency assay for human cells: they can differentiate into teratomas bearing all three germ layers (Yu et al.,2007). When cultured under specific conditions and/or with additional factors, iPSCs can morph into hundreds of specific cell types ranging from neural cells to pancreatic insulin-producing cells (Inoue and Yamanaka,2011).
The seemingly supernatural properties of iPSCs open up a new realm of possibilities for stem cell research and medicine. Given the original motivation for creating iPSCs, initially there was considerable excitement over the possibility that patient-derived iPSCs could be a viable alternative to ESCs in regenerative medicine applications. In a typical scenario, iPSCs would be derived from diseased individuals and genetic lesions repaired with gene targeting techniques (Hanna et al.,2007). Next, the cells would be differentiated to the appropriate cell type, and reintroduced to the patient. However, there are technological and phenotypical hurdles that must be overcome before this fairy tale can become real. This primer discusses those hurdles, iPSC basics, and alternative translational applications that promise to revolutionize medical science.
Upon creation of the promising new cell type, it was immediately apparent that iPSCs were a poisoned apple. They were produced using retrovirus to introduce the OSKM quartet into cultured fibroblasts, a methodology that introduces two main problems. First, the resulting pluripotent cells are tumorigenic, preventing their use in human therepeutics (Okita et al.,2007). This is due to introduction of the oncogene c-Myc, and viral vector induced insertional mutagenesis of the target cell genome. Second, iPSC production is a slow and inefficient process. After OSKM factor delivery, it takes between 1 and 3 weeks for iPSC clones to appear, and potentially weeks longer to differentiate cells to various specialized cell types. What's more, at 10 iPSC colonies per 50,000 transduced cells, reprogramming efficiency is low using the original protocol. These shortcomings hamper high throughput generation of iPSC clones, and increase their production cost, limiting their overall translational utility. Immediately the search was on for ways to make iPSCs less menacing.
Researchers quickly determined that there is more than one recipe for the magic potion that rejuvenates cells. The essential ingredients are factors that can activate pluripotency genes, and facilitate chromatin states associated with pluripotency. These include Oc3/4 and Sox2, which can be replaced by ESC-specific factors Lin-28 and Nanog, or the orphan nuclear receptor Esrrb (Feng et al.,2009). Furthermore, although the c-Myc oncogene significantly enhances the reprogramming process, it is optional. More recently, it was shown that introduction of the micro RNA cluster miR302/367, in the presence of the histone deacetylase (HDAC) inhibitor valproic acid (VPA), speeds iPSC clone production by a few days (Anokye-Danso et al.,2011). The miRNA cluster, a direct target of Oct4 and Sox2, is predicted to inhibit hundreds of mRNA targets including regulators of chromatin remodeling and cell proliferation. Therefore, custom brews can eliminate hazards associated with oncogene overexpression, and accelerate the rate of reprogramming.
With a sleight of hand, researchers also improved reprogramming techniques (Ho et al.,2011). Most notable are nonintegrating delivery systems that eliminate the risk of invasion of cellular genomes by exogenous sequences. Induction by miRNAs, rather than overexpressed genes, may also lend itself to nonviral mediated delivery techniques (Anokye-Danso et al.,2011). Reprogramming efficiency is enhanced by several means, including down-regulation of tumor suppressor components, and addition of ERK, GSK, or TGFβ inhibitors (Ho et al.,2011). The former group inhibits apoptosis and/or enhances cell cycle kinetics, while the latter stabilizes Nanog protein levels. Subjecting cells to chromatin-modifying, small molecule compounds such as the histone deacetylase inhibitor valproic acid (VPA) and the DNA methyltransferase inhibitor 5-aza-cytidine (AZA), also accelerates clone production (Huangfu et al.,2008; Mikkelsen et al.,2008). The various means of accelerating iPSC production likely work by breaking ill-defined barriers to pluripotency (Hanna et al.,2009).
Akin to transforming a frog into a prince, scientists have determined how to directly reprogram, or transdifferentiate, somatic cells to cell types from alternative lineages. For example, briefly exposing fibroblasts to OSKM factors, then growing them in cell-type specific culture conditions, elicits a direct conversion to neuronal precursor cells (NPCs) or cardiomyocytes (Efe et al.,2011; Kim et al.,2011a). This method differs from other modes of induced transdifferentiation, where overexpressed lineage-specific transcription factors trigger cell-type specific gene programs. Instead, direct reprogramming is thought to first bring cells to a transient, open-chromatin state. This developmentally plastic intermediate functions as a platform from which cell-type specific epigenetic states can be induced. Because direct reprogramming does not require cells to first revert all the way back to a “ground” state (i.e., iPSCs), differentiated cells can be derived relatively quickly and efficiently. Nevertheless, this methodology will never completely replace iPSC production, because the ability to revert cells in time has proven a unique and useful tool (see Future Promise).
Induced Pluripotency Phenotypes
Although the steps of reprogramming remain mysterious close examination has revealed some secrets. The most accessible information comes in the form of observed changes in morphology and gene profiles. After induction, be it by OSKM delivery or other methods, the first signs of reprogramming are proliferation and that cells acquire a characteristic small, roundish appearance (Plath and Lowry,2011). Next, they undergo a mesenchymal to epithelial transition (MET) and begin to cluster tightly. In line with these observations is down-regulation of the somatic and mesenchymal gene programs, and up-regulation of proliferation, epithelial, and some embryonic genes (i.e., AP, SSEA-1; Mikkelsen et al.,2008). A sign that cells have reached a stable, iPSC-state is when they begin to endogenously express pluripotency markers, such as endogenous Nanog, Oct 3/4, and Sox2, and are no longer dependent on introduced reprogramming factors.
Reprogrammed cells also undergo several chromatin modifications. Pluripotency genes, which are suppressed in somatic cells, lose their chromatin repressive marks, such as DNA methylation and histone H3 lysine 9 (K3K9) and H3K27 methylation, and gain the activating mark, H3K4 (Mikkelsen et al.,2008; Plath and Lowry,2011). At the end of the induction process, global gene expression, DNA methylation, and histone tail modification profiles closely resemble ES cell-like patterns.
Genetic and Epigenetic Aberrations
Though iPSCs resemble ESCs in many ways, it was revealed that behind the facade of the lovely maiden lies an ugly witch. Originally, it was assumed that transplantation of host-derived iPSCs would escape immune-rejection by the host recipient, a problem faced by ESCs. However, we now know that autologous iPSCs suffer the same fate (Zhao et al.,2011). At this stage it remains to be tested whether fully differentiated iPSC progeny, or iPSCs derived by alternative methodologies, will be able to slip past the host immune surveillance system. iPS and ES cells also suffer from genetic instability that can cause aneuploidy, copy number variation, and protein coding point mutations (Gore et al.,2011; Hussein et al.,2011; Laurent et al.,2011; Mayshar et al.,2011). Many of these aberrations are likely byproducts of selection for robust, rapidly dividing cells in cell culture.
Moreover, the reprogramming process introduces its own set of problems. Comparisons between iPS, ES, and somatic cells show that iPSCs harbor differentially methylated regions (DMRs; Lister et al.,2011). Some of these DMRs are signs of epigenetic memory, a failure to fully reprogram the somatic methylation pattern. Other DMRs are newly acquired aberrant methylation patterns, many of which occur in regions identified as hotspots for epigenetic mutations, and non-CG methylation that can stretch over 1 MB in length. Some DMRs are transmitted to differentiated progeny, and separate reports indicate that epigenetic memory can hamper iPSC differentiation potential (Kim et al.,2011b). These findings cast a shadow on the potential for using iPSCs in the clinic.
Despite a blemished reputation, iPSCs may still get their fairy tale ending. One application that has already met success is the “disease in a dish” model. iPSCs derived from patients with genetic diseases are differentiated into affected cell types. As the cells mature, they acquire disease characteristics in vitro. This procedure gives scientists ringside seats to the blow-by-blow action of disease development. iPSCs have already been used to model genetic diseases such as spinal muscular atrophy (SMA), Rett syndrome (RTT), and Hutchinson-Gilford Progeria syndrome (HGPS), yielding new insights into disease etiologies (Ebert et al.,2009; Marchetto et al.,2010; œLiu et al.,2011; Zhang et al.,2011). The iPSC lines yield useful data despite the fact that they likely have incurred mutations. This finding suggests that iPSCs are well-suited for this application as long as proper controls are in place.
Another application that may place iPSCs in good graces is using them as a platform to test disease therapies. For example, iPSCs can be used in high throughput drug screens to test therapeutic value and toxicity (Inoue and Yamanaka,2011). In addition, panels of iPSCs representing diverse genomes can be used to test the efficacy of a certain drug across the population. A natural extension of this technology is toward personalized medicine, where the response of a patient's cells to specific drugs can help physicians choose the one that will work best for that individual. At present some of these approaches are too costly, an obstacle that may be overcome if production of iPSCs and their differentiated progeny are streamlined.
At a more basic level, iPSCs can also be used to model normal development. The cells can be differentiated into rare cell types or those that are otherwise hard to isolate, observe, or manipulate in vivo. From there, much can be learned about their functionality, and cell intrinsic or extrinsic developmental molecular mechanisms.
Whether iPSCs will be able to fulfill their much-hyped role in regenerative medicine is still unclear. Scientists must first create ways to monitor genetic stability or avoid those issues altogether, reliably weed out undifferentiated cells that could form teratomas, and resolve immune rejection issues (Lowry and Quan,2010; Zhao et al.,2011). How will the story end? Since 2007, there has been a 100-fold increase in the number of iPSC papers published annually. At this rate, it won't be long before we find out.
A CONVERSATION WITH THE EXPERTS
Here we feature interviews with Clive Svendsen, PhD, and William Lowry, PhD (Fig. 1). Svendsen discusses the limitations and potential of iPSCs in cellular therepeutics. Lowry compares the usefulness of ES and iPS cells for modeling development.
Interview with Clive Svendsen
Developmental Dynamics: What is your lab's research focus?
Clive Svendsen: We are interested in induced pluripotent stem cell (iPSC) technology from two different angles. One is disease modeling, and the other is the potential for treating diseases by means of stem cell transplantation.
We're continuing the Spinal Muscular Atrophy (SMA) work from our Nature paper (Ebert et al.,2009). We show that motor neurons die in the SMA line, and we see similar results in lines from three other SMA patients. Now we are exploring, for example, which cell death pathway is activated. We also just got a $2 million grant from the California Institute of Regenerative Medicine (CIRM) to develop high content screening assays to look for novel small molecules to slow down the cell death in our cell model for SMA.
A few years ago we were funded by NIH to explore iPSC models for Huntington's disease. Most other diseases that have been modeled in iPS cells so far, including SMA, have been autosomal recessive diseases. The reason is that it's very straightforward to understand the biology, because cells homozygous for the recessive allele have a simple knockout dysfunction. HD is caused by a CAG repeat, has a very definitive molecular mechanism, and is autosomal dominant. Jamie Thomson and I generated three HD lines before I left University of Wisconsin-Madison, and have been distributing them to a consortium of five different researchers to see if we can come up with any kind of abnormality in those cells, particularly neuronal dysfunction of some sort. We have some evidence that there are changes in those cells and we are just now confirming what those are.
Dev Dyn: What nudged you into the iPS cell field?
CS: We had been playing around with transcription factors for many, many years and reprogramming cells in our own way. For example, we were trying to get dopamine neurons from neural stem cells by over-expressing transcription factors. When Yamanaka's 2006 paper came out everybody was thinking, “Well, can you do that with human cells?” Jamie Thomson was working on it, and my lab was next to his. It was a natural extension for us to get involved.
Our goal was to use these cells to model human diseases. Jamie agreed SMA was a good disease to model and he prioritized making those lines even though there were one hundred different things he could have done instead. Having that interaction with him got me into the field, and now we are 100% invested. I have taken the technology here to Cedars-Sinai, and I'm introducing it to other physicians and researchers.
Dev Dyn: Which papers have most impacted your research?
CS: The three main papers (Takahashi and Yamanaka,2006; Yu et al.,2007; Park et al.,2008) were the ones that got me into the field. The field has evolved from there.
Jamie had a paper approximately 2 years later showing that he could derive human iPS cells using nonintegrating, episomal EBNA1 vectors (Yu et al.,2009). When that paper came out, I was excited because I didn't like integration. We still use that method for making lines. To be frank, I don't care anymore whether cells are reprogrammed using small molecules, RNA, or soup, as long as it is a nonintegrating method. All those papers, and those that show increased efficiency and so on, enriches what you can do. But as long as a line expands exponentially and is immortal, who cares?
The three Nature papers recently documented mutations in iPS cell lines that were not present in their parent fibroblasts, demonstrating that the process of making the lines can lead to changes in the cell's DNA (Gore et al.,2011; Hussein et al.,2011; Lister et al.,2011). However, while these papers are interesting, there was no effort to do the same study with other stem cell lines. I think if you deep sequence any cell type that has been expanded in culture you will find a nightmare of problems, including some new mutations that may be carcinogenic. So, the real question is, are mutations occurring more frequently in iPS cells or ES cells than in any other cell type that divides in culture? I don't think we've addressed that efficiently.
Dev Dyn: How are iPS cells useful for understanding disease development?
CS: The best analogy I can give is a traffic accident. You are the police officer who arrives on the scene only to find that the cyclist is on the road, the truck driver is in the ditch, and somebody has died in a car. You only have the endpoint to look at. What you'd really like to know is how it happened. The guy came over the bridge, the sun got in his eyes, he was lighting a cigarette, and then the cyclist pulled out . . . Three things happened at once to cause that accident. You would never have known that by looking at the position of the truck, the bike, and everything else. You just know that they are involved.
Now go to the iPS cell model for Spinal Muscular Atrophy. Before the SMA model, you could look at a diseased patient's spinal cord and see that most motor neurons were gone, but there was no information telling how it happened. I think the beauty of iPSCs is that you can use them to replay the disease over and over. You can start from the beginning when the cells are normal, as they are on our model. At 4 weeks, there is very little loss of motor neurons. Then between 4 and 8 weeks, they die in front of you in the dish. You are watching the accident. That, to me, is the power of modeling, as long as you have a good phenotype—as long as an accident occurs. If an accident doesn't happen, then you are in trouble.
Dev Dyn: However there are certain limitations. You are missing surrounding tissues or other factors that might be involved in disease progression.
CS: Exactly. It is an artificial system. The only real advantage is that it's human. You really need to combine models. The whole-mouse model is beautiful because it is the whole organism with the blood and immune systems and everything else. Of course, the downside is that it's a mouse and the biology is different. Quite often, that confounds interpretation. The best model is a real human being with the disease. Anything else is an approximation, and you have to go along with the weaknesses and strengths of each model.
iPSCs add a “human flavor” that we haven't had before. It's not perfect because, as you mentioned, it doesn't have the architecture of a living human being. On the other hand, the closer we get to approximating a real person, the more trouble we get into with the IRB or with ethics groups. We could grow a mini-brain in a dish, but if it's thinking, we're in trouble.
Another real problem is with adult-onset diseases. When you reprogram cells, they kind of go back to the beginning. Theoretically, you would have to wait 40 years before you would get Parkinson's or Huntington's disease, and we don't have that kind of timeframe in our experiments.
In fact we don't see an overt phenotype with our HD lines. We bring out a phenotype by stressing them with a toxic signal. But then you have to wonder whether you are reproducing what actually happens in the disease, or whether you are artificially simulating a pathway that isn't part of the native disease? The drug companies are interested in validating molecules in these models. I think they would be happier with a more “natural” degeneration of the cells because then they can be fairly sure that that is as close as you can get to mimicking a disease.
Dev Dyn: What are some of the other limitations of iPS cells in therapeutics?
CS: I think there are many challenges. There are a lot of data coming out on unreliability of the lines in terms of genetic instability. When you expand cells in culture they change, so the mutations build up.
I am also really worried about controls. I don't think we know what a good control line is. I can pick what looks like a young, healthy student, a 19-year-old who has no disease whatsoever, and say, “There is my control Parkinson's line for my Parkinson's study.” In 20 years, that student could get Parkinson's disease, as a percentage of us will. It is even worse for Alzheimer's. Fifty percent of everybody over the age of 80 gets Alzheimer's disease now. So, you've got to imagine that 50% of your so-called controls will get Alzheimer's. If you combine all the different diseases that we know about, I contest that we don't have a really good control line. We are all going to die of something.
Dev Dyn: Do you foresee that genetic instability will be something that can be controlled in the future, or at least monitored?
CS: It can be monitored; I am not sure about controlled. Even though some of our SMA lines are abnormal genetically, they still show the same phenotype. In other words, it may not be so crucial if your phenotype is strong enough that it can shine out above genetic instability. But if the phenotype is very subtle, a genetic abnormality could throw the whole model out the window.
Dev Dyn: Do you see a future for iPS cells in personalized medicine?
CS: A huge future. I think cardiology is probably the prime area where they will be used in personalized medicine. One of the biggest problems in drug development is that neurological drugs often have off-target effects in the heart, causing arrhythmias. I think screening compounds on beating human cardiomyocytes and, for instance, looking at receptor profiles, will be cheap and quick, and very important for early detection of problems. Normally, you can't do that until you get to phase II clinical trials.
Extending that paradigm, what I predict is that in 20–30 years' time, if not sooner with all of our iPSC lines, is that we will be able to screen several different drugs on cardiomyocytes derived from your own iPS cells. Then we can find out which one works best with the least side effects for you. When I think of personalized medicine, that's the kind of future I'm thinking of, rather than genomics.
Dev Dyn: What sort of limitations do you think should be placed on the use of iPS cells in therapeutics?
CS: The biggest challenges surround issues of patient confidentiality. Who will have control over all that information we're going to get on all these so-called controls? Someone donated a skin cell on a Wednesday afternoon because it sounded like a good thing to do. When we spit him out as a noncontrol because he has a deletion that will give him Parkinson's in 5 years, who is going to give him the bad news? It's going to be perhaps one of the hardest issues we are going to face. These cells that you are “playing” with and distributing to your colleagues are ethically charged because there is a living, breathing patient associated with them.
We think embryonic stem cells are tough to deal with ethically. But if you think about it, the good side of embryonic stem cells is the embryo was destroyed. There is no patient linked to those cells.
Dev Dyn: Do you still see a place for embryonic stem (ES) cells in therapeutics?
CS: Yes. One reason is that I think the challenges that lie ahead for iPS cells are underestimated. Second, we're still working out subtle differences between the two cell types. We are still trying to untangle the variability in ES cells from the variability in iPS cells. iPS cells are going to be epigenetically scarred and may have differences in the number of mutations due to the reprogramming process. Whenever we do an experiment, we run an embryonic stem cell line alongside to normalize against.
Dev Dyn: ES cells are the gold standard.
CS: Kind of. I hate that term in a way because everything we do is an artifact. There is no “normal” in this game. The ES cells themselves are an abnormality. They are very transient and don't expand indefinitely in the embryo. You are comparing one tissue artifact with another.
Dev Dyn: What are some exciting ideas that are emerging in the field?
CS: One is partial reprogramming. That you don't have to push the cells all the way back in time. You can go a little bit back and you can go forward. I think we have a lot to learn about how a skin cell can go directly into a neuron, for instance. Does it go from one to the other along a direct pathway, or does it have to go to back to some sort of precursor state and then re-differentiate?
Second, we always thought iPS cells injected back into the same animal would not reject. However, a recent paper surprised the field showing that iPS cells do reject in mice (Zhao et al.,2011). While extremely interesting, in this study teratoma assays were used to look at rejection. For the field of neurology we are interested in brain transplantation—and the brain is immune privileged which may dampen any rejection iPS cell derived neural tissue may illicit. It's a fascinating area of science to keep close watch on.
My feeling is the most lasting effect of this whole technology will be on understanding how aging occurs. When I saw the paper from Kevin Eggan on ALS it kind of blew my mind (Dimos et al.,2008). They took fibroblasts from an 82-year-old patient—you would think skin cells would be the last thing you would want to take. They are wrinkled and old and there is no regeneration, and yet they got those to be like an embryonic stem cell again. It's got to tell you something about aging.
Along those same lines, the papers on progeria, or premature aging, were also fascinating (œLiu et al.,2011; Zhang et al.,2011). They took fibroblasts from a progeria patient who was 9 years old but looked 90. The fibroblasts looked 90, as well. After they reprogrammed them, they behaved and looked just like a normal pluripotent cell. But then, when they made them differentiate, they could replay the disease. They rapidly became old in the culture dish. That was another fascinating exploration of aging, which is where I think a lot of interest will be in the future.
Interview with William Lowry
Developmental Dynamics: What is your lab's research focus?
William Lowry: I come from a background thinking about ectodermal development, particularly from the perspective of the epidermis and hair follicle. My postdoctoral training was in the identification, characterization, and manipulation of adult hair follicle stem cells. Part of our lab is now looking at how stem cells versus other cells in the skin serve as cancer cells of origin for skin cancer, squamous cell carcinoma. We also think a lot about how during the process, adult stem cells undergo proliferation, de-differentiation, and an epithelial–mesenchymal transition (EMT)-type process. A lot of these processes appear to have analogs in the reprogramming process for iPS cells. So, we also think about whether there are parallels between the generation of cancer and the reprogramming of somatic cells to a more primitive state.
The other half of the lab is thinking more about using pluripotent cells, both embryonic stem (ES) cells or iPS cells, to model human development. It's clear that you can make lots of different types of cells from ES and iPS cells. Two things aren't so clear yet. First, how similar are the progeny of ES cells and iPS cells to each other? Second, how similar are either of those to cells that you find in tissue? We make lots of different kinds of cells from ES and iPS cells and then compare them directly to what we think are their natural counterparts in tissue. We typically find significant differences between the in vitro-derived and the in vivo-derived cells. So that's what we're trying to figure out now: how to make the cells that you generate in vitro more like those you find in vivo.
Dev Dyn: What do you think that these cells can teach us about normal development?
WL: Most of our work is coming to the conclusion that the derivatives that you get from the human ES and iPS cells more closely represent cell types found early during development than late in development. We're trying to narrow that window by trying to birthdate progeny of the cells by comparing them wherever we can to tissue-derived cells. The problem is coming by the tissue-derived cells. But it appears as though the earlier we go, the more similar they become. A working hypothesis is that they represent cells found early during development, but we don't know what “early” means, necessarily. Early in development could be 2 weeks, it could be 16 weeks, it could be 9 months. It all depends on your frame of reference. We're trying to narrow the window as best we can.
Dev Dyn: What are you hoping to learn from those cells that you can't learn from cells in vivo?
WL: If the cells that we generate accurately represent something that's found in vivo, for one thing that would be an interesting finding from a basic science perspective. Two, it would give us a model system where we could use that time window during human development to understand what's going on in early embryos, which you wouldn't otherwise be able to study. You wouldn't be able to do gain- and loss-of-function on human cells at day X of development, but maybe we can do that with ES and iPS cells. Three, if you know not only the identity of the cell type you're making, but knowledge about the functional maturity or the stage of development that it represents, it's going to be much more helpful for when you plan some kind of cell-based therapy. If you're going to put these cells into patients, it would be nice to know what stage of development they represent. Are they cells that would be found in that adult, or a newborn, or a very early fetus?
Dev Dyn: Can you talk about the limitations of using an in vitro system to model developmental processes?
WL: The ES and iPS cells not only are developing in vitro but they are developing from a starting point which is poorly defined. We don't have precise understanding of what the human ES and iPS cells represent compared with what's found in vivo. Are they inner cell mass cells? Are they epiblast cells? Maybe there is some kind of in-between cell type or artifact that isn't found in vivo?. That not only should influence how you differentiate the cells, but also the identity of the cells that you expect to get. If you're starting with something natural, you expect to get something natural. If you're starting with something unnatural, then who knows? So, not only is it in vitro, but it's coming from a poorly defined source.
Also, with in vitro differentiation you're talking about two dimensions, whereas in vivo you are talking about three dimensions, which changes everything. You don't have all the surrounding cells there providing whatever cues they provide. What scientists try to do instead is use the literature to see what growth factors, compounds, or conditions represent the situation in vivo. But really we don't have a complete knowledge of what goes on in vivo in any species, much less humans. So, it's difficult to say what the best differentiation conditions are because we really don't know what's going on accurately in vivo anyway.
Dev Dyn: What would be the advantages of IPS cells over ES cells for the type of research that you are doing?
WL: iPS cells are just much easier to come by. It's pretty trivial to make 50 iPS cell lines. But the success rate of making 50 embryonic stem cell lines, generating ES cells from an embryo that has been frozen and thawed, is not very high. In addition, procuring those human embryos for the purpose of making ES cells is very difficult.
The problem is that there is a wide variation across the iPS cell lines, and so it's not clear how to even categorize them as good, or OK, or bad cell lines. Some cell lines tend to make one particular type of differentiated cell versus another. Some are more prone to spontaneous differentiation than others. Some are more prone to incurring more genomic aberrations than others. There is just a wide variety right now, and we don't have very good methods for distinguishing the more reliable lines.
Dev Dyn: What is your approach to dealing with this quality issue?
WL: It's not as though all ES cell lines are identical. It's been argued in a couple of recent papers that essentially iPS cells are across a broader spectrum of quality, by whatever criteria, whereas ES cells are within a more narrow spectrum. The only way around it essentially is to do all of your experiments on multiple lines and look for things that are consistent across them. If you only do your experiments with one ES or one iPS cell line to quantify the efficiency of differentiation to a particular cell type, then you can end up with a very misleading result. With every experiment we try to compare at least three ES and three iPS cell lines. Of course it would be better to do 10 or 20 or 50 lines, but it's difficult.
Dev Dyn: What initially provoked your interest in the stem cell field?
WL: When I started at UCLA, I was surrounded by labs that were working on human ES cells. Talking to them made me wonder if we could study the cell fate decision whereby human pluripotent cells make ectodermal cells, which then make the epidermis and nervous system, in vitro. Are the pathways that have been discovered for frog and mouse conserved in human? Without having ever seen pluripotent stem cells through a microscope, I wrote a grant to figure that out and was lucky enough to get it. It has since become a huge part of my lab.
Dev Dyn: What are some of the papers that have most impacted your research?
WL: I think the most interesting development since the initial Yamanaka finding (Takahashi and Yamanaka,2006) is direct reprogramming, which is really just reprogramming but going from a differentiated cell to a differentiated cell (Vierbuchen et al.,2010; Efe et al.,2011; Kim et al.,2011a). It is really exciting, not only for clinical purposes but also as a cell biology experiment. I think it's fascinating that expression of just a few genes can cause such a cascade that leads to such a dramatic cell fate change. Maybe it's the kind of change and reprogramming that goes on during cancer. I think there are a lot of interesting connections to be made, but it's still so early.
It could also be a useful way to make cells that either haven't been made before or that are very difficult to differentiate from pluripotent cells. For whatever reason, some cell types are made from pluripotent stem cells at 80%, and some are made at 1%. So we're thinking about trying to make the more difficult ones with direct reprogramming. The big question is, what are the reprogramming factors? There are so many differences between two different cell types that it is difficult to then parse through all that and say these are going to be the genes that reprogram this cell type to that cell type. We don't want to screen through a thousand combinations of a hundred genes to figure it out. We are trying to figure out ways of narrowing that a bit.
Dev Dyn: What do you foresee is the future of IPS cell research?
WL: There is obviously going to be a huge push to do anything that can facilitate their movement into the clinic. Here at UCLA, we've built a whole lab designed to make FDA-compatible iPS cells and iPS derivatives for the purpose of some day initiating a clinical trial. As part of that, there are approximately a hundred different steps that you have to go through. We are looking very closely at each step and how do you make that step more efficient, how do you make it cleaner, how do you do it without animal byproducts, how do you do it in such a way that it's repeatable a hundred times over without any differences?
In terms of basic science, it's hard for me to even imagine how many different interesting things are going on with these cells. It opens up so many possibilities that you can make any kind of cell from any other kind of cell, and you can do it all in human, and you can do it relatively easily and relatively cheaply.
Modeling human diseases in a dish will be enormously useful. Again, we don't know how phenotypes in vitro correlate with those in vivo, but at least we have something we can use to study a disease. How else do you study diseases like ALS and SMA and Parkinson's and Alzheimer's? There's no other way to do gain- and loss-of-function in those diseases in humans. Before, all you could do was study the pathology of the disease after it had happened. You couldn't study the etiology of the disease very easily. Now we can do that.
Many thanks to Clive Svendsen and William Lowry for generously sharing their expertise, insights, and time. I am also grateful to Courtney Montgomery for her expert assistance as transcriptionist. Interviews were performed and edited by JCK. The author regrets that due to space constraints she was unable to reference many excellent papers.