A sad fact of life for cell biologists is that they can’t expand cells in culture without inducing mutations. Nor, for that matter, can the cells in our bodies grow without doing the same. Under an assault of environmental chemicals and radiation, our cells accumulate mutations. These “somatic mutations” impair tissue integrity and, in the worst cases, give rise to cancers. At least in the body we have an immune system doing its best to track down and remove the deviants. Cells in a culture dish have no such assistance. Worse, the conditions in culture encourage mutation. Imagine that, by chance, a single cell among ten million in a culture dish acquires a genetic change—a mutation—that improves its growth characteristics. As the cell population is expanded, the rogue cells will quickly outgrow their benign neighbors and, all things being equal, will eventually take over the culture. Imagine that this was a preparation of iPS cells taken from a patient with the intention of generating pigment epithelial cells to transplant back into the patient’s retina. Would those epithelial cells be safe? Since their generation involves first growing billions of iPS cells then billions of pigment epithelial cells, all from a single skin fibroblast, there are more than enough cellular generations for mutations to arise over and over again. Quite simply, by the end of this expansion process, the cells will have accumulated some number of mutations. Worse, the number and identity of these variations will themselves vary: each time cells go through this process, the outcome will be different. And though in some cases, the mutations will be recognizably threatening because they have been seen in cancers before, many will be hard to assess.
To add to the complexity, the mutations can be on a variety of scales. They can be tiny, involving literally a single base pair in the chromosomal DNA. Though small, these “point mutations” can have a large effect because they mess with the genetic code, generating proteins with altered structure and disrupted function. But mutations can be much larger than this, involving tens, hundreds, or even millions of base pairs. Sometimes whole chunks of chromosome are lost, or turned back to front so they don’t read properly. Sometimes bits of chromosome get duplicated so that cells carry an extra copy, and often the genes on that duplicated piece are active, thereby increasing the dose of that gene in the cell. These bigger mutations are called “copy number variations” and are often difficult to assess because they frequently involve many genes either duplicated or deleted, and so have complex functional outcomes. Cells in culture are known to accumulate the smaller point mutations, but they are also susceptible to the bigger copy number variations.
Cell biologists take steps to mitigate this risk but are unable to remove it entirely. If the cultured cells are stressed by overcrowding or lack of nutrients, then the growth of variant cells is encouraged. So culture conditions are monitored carefully and kept within precise limits. Nonethelss, mutations will still emerge, and studies have confirmed that human pluripotent cells are not spared these pressures, and moreover that some pretty unsavory mutations can arise.
P53 in a case in point. It is an important cancer control gene. Molecular biologists term it a “tumor suppressor” because it acts in multiple ways to restrict the emergence of cancers. It can activate DNA repair mechanisms, mending the damage caused by chemicals or radiation. It can prevent tumor cells from accelerating aound the cell cycle, and can kill them if they persist. So, having an intact P53 is an important property for a cell. Of concern then that P53 mutations can accumulate in human pluripotent cells in culture. In a recent report, Florian Merkle and colleagues looked at 140 different human hES cell lines, including twenty-six that had been intended for clinical use, and found that five of these lines carried P53 mutations.1 Moreover, the mutations were of a type called “dominant negative,” a type particularly associated with human cancers.
In one sense, mutations in P53 and other cancer genes are relatively easy to deal with, if not prevent. It is not difficult to design assays that would screen out mutations in P53 and other cancer genes. Such assays are now available, and surely all responsible cell therapists in the future will employ them. If a batch of cells has accumulated P53 mutations, you can simply discard them and start again—costly and irritating, but probably relatively infrequent judging from Merkle’s numbers. The trick for cell manufacturers will be to improve their culture technology to minimize the mutation rate, and to refine their in-process screening to detect variants quickly.
More difficult is dealing with the preponderance of mutations accumulated in genes with no tumor association. Many researchers, driven by the understandable desire to know more about their cultures, are beginning to sequence the whole genome of cells. With “next-generation sequencing,” this is relatively cheap and quick. Unsurprisingly, sequencing reveals that many cultures carry genetic variants.2 Some will be somatic, originating in the donor, but many will have arisen in culture. Most will be in genes whose function is poorly understood, and few will have any known association with cancer.
What to do with these data? There is a strong argument to say that this information is of limited value. Since every cell in every human body carries a number of harmless mutations, there’s no reason to expect that those in transplanted cells are any less likely to be benign. The usual way to deal with the cancer risk is to run “tumorigenicity assays.” These come in two forms: animal studies and culture assays. Culture assays assess whether the cells demonstrate particular growth characteristics such as being able to grow in the absence of a supporting substrate. Such “anchorage-independent” growth is a feature of cancer cells and thereby a means of identifying cells with a tumorigenic potential. More reliable, though less palatable, are animal study assays, in which an experimenter asks whether the cells can form a tumor in an animal, usually some unfortunate mouse. Strains of mice are now available that have a constitutively compromised immune system. Since these animals are particularly inept at combating cancer, the failure of a population of cells to form a tumor in such mice is accepted as pretty good evidence that they pose a low risk in human patients. Most of the therapies we have considered so far would have been subjected to such assays.
So, tumorigenicity assays have face validity. They seem to show whether cells are or are not tumorigenic. The genetic tests, however, have no such validity. Discarding cells with P53 mutations would seem to make sense from a precautionary perspective. You might also argue that if the plan is to use the cells to make, say, pigment epithelium, you might want to screen important retinal genes for mutations, again just from a precautionary perspective. But that specific application apart, the genetic information adds little to the risk analysis. If you don’t know what a given gene actually does, how do you know whether a mutation in that gene is dangerous? A future complication may be that as we learn more about genes and gene function so the pressure to over-interpret this sort of data will grow.
Returning to the Japanese autologous iPS cell trial, the reason given for the hold on the treatment of the second patient was that his iPS cells had accumulated genetic mutations in culture, and it was not considered safe to proceed.3 Of particular concern were three copy number variations—deletions—that had arisen in the culture. One of the genes involved was indeed a gene previously shown to be associated with cancer.4 Consequently, even though tumorigenicity assays were negative—the cells did not form tumors in immune-compromised mice—the three big genetic deletions was considered to be too big a risk and the trial was put on hold.
We can sympathize with those tasked with reviewing this trial, who, reports suggest, came under considerable pressure from many quarters in what was an enormously high-profile case in Japan.5 The precautionary principle is clearly paramount with such a crucial, novel therapy. Nonetheless, future decisions will need to be more evidence-based if innovative therapies are to proceed.
We can sympathize more with the patient himself. This blind, elderly man had a skin biopsy taken to grow an autologous iPS cell line, with the expectation that, after a long wait, he would receive the retinal graft he needed. Instead, after waiting even longer while experts pondered the genetic results, he was finally told the operation could not proceed. If, as suggested, he was offered enrollment on the study’s other trial, the one employing allogenic grafts—someone else’s cells rather than his own—he might well have asked why he’d had to wait all this time for his own cells to come through if these other cells could do the job. The answer is a complex mix of risk assessment, scientific uncertainty, and pharmaceutical economics, but it starts with the problem of immunogenicity.
The observation is well-established that tissue cannot simply be transplanted from one individual to another without the recipient mounting a rejection response to the foreign material, hence the careful tissue typing that accompanies blood transfusions and organ donations. The biology underlying this response is multifaceted, but is underpinned by the capacity of cells of the immune system to recognize invading cells as foreign by the proteins they express, and display these foreign proteins in such a way that the intruding cells can be attacked and killed. This immune response evolved, of course, to protect us against invading organisms, such as bacteria or viruses, but is unfortunately incapable of distinguishing dangerous invaders from benign therapeutic transplants.
Apart from a brief discussion in chapter 5, I’ve managed to ignore tissue rejection in this narrative so far for a very simple reason. The brain (and the retina) is an immune-privileged site. Evading immune surveillance, it can tolerate foreign antigens in a way most of the body cannot. For scientists trying to develop replacement cells for pancreas to treat diabetes, for example, immune rejection is a serious problem. The allogeneic approach we considered above for the retina would fail disastrously in the pancreas because iPS cell derivatives from another person’s cells would be very quickly attacked and rejected.
That said, the immune privilege is only relative. In some of the trials we’ve considered, the patients would still have received immuno-suppression, without which they might well have rejected the transplanted cells. This is problematic because immunosuppressant drugs tend to be toxic and not well tolerated by patients. Often patients simply stop taking them because they so hate the side effects, clearly not an optimal situation.
For this reason, the autologous approach is attractive. If you are receiving your own cells, then your immune system is unlikely to see them as foreign. Moreover, this is an avenue opened up by iPSC technology. Clearly, an ES-based therapy has to be allogeneic, but in principle a personalized iPS line could be generated for every patient.
Except it couldn’t. The cost of autologous iPS production is entirely prohibitive. Each therapeutic line would take the best part of a year to produce: the reprogramming of patients cells; the differentiation of the desired derivative; the safety testing on multiple different lines (because you cannot be sure in advance, which individual iPS clone will be suitable); then the production run to give the final therapeutic product. The cost has been estimated to be roughly $1 million per patient. Further innovations will reduce this figure, but surely never to the level of affordability required for broad use.
The Japanese study wanted to start with an autologous therapy to give themselves the best chance of success with this fledgling iPS cell technology. So the single patient who received the autologous retinal graft is currently unique and may remain that way. In the future, probably only billionaires need apply.
Which brings us back to the allogeneic approach and the question of how immune rejection might be overcome. Several approaches are currently under investigation, and which will work and which not is not yet clear. Tissue compatibility hangs on a complex set of cellular mechanisms, but central are a set of proteins that comprise the major histocompatibility complex (MHC). Although these proteins are also complex, each having many variable forms, just two sets of genes need concern us here: MHC class I and MHC class II. Between them, they mediate three pivotal cell interactions.
First, essentially all cells express MHC class I proteins on their surface. There are many variants of class I proteins, and which is expressed is determined genetically. So while every cell in my body expresses the same class I proteins (since they are all genetically identical), my class I proteins wil be different from yours. There is a type of immune cell called a killer T cell, whose job it is to cruise around the body, like some gestapo agent, checking everyone’s class I identity. If it spots a cell carrying the wrong credentials, it induces cell death. A useful cell to have on your side if your objective is genetic purity.
The second category of interaction involves another assassin—the Natural Killer Cell. This cell seeks a different type of deviant. It identifies cells that are expressing no class I protein at all. This happens in tumor cells, which sometimes lose their class I expression, and may therefore escape the attention of other immune mechanisms.
The third category is more complex and involves MHC class II proteins. Class II proteins are not expressed by all cells, but by cells such as macrophages, whose job is to track down foreign cells or cellular debris. Such material is engulfed by the macrophage and digested. It then performs a strange operation. It presents small pieces of the foreign proteins it has ingested on its surface in conjunction with MHC Class II proteins. This antigen presentation attracts naïve T cells—T cells not yet dedicated to a specific immune target. As a consequence of this interaction with the antigen presenting cells, the T cells are primed to seek out material carrying those same antigens—other similar invaders, in other words—and attack those foreign cells.
These multiple categories of seek-and-destroy weapons present a formidable arsenal for transplanted allogeneic cells to overcome. How might stem cell science protect transplants from this attack?
There are several strategies currently in play to overcome this problem, though none has yet definitively proven its worth. The most obvious tactic is to do what hematologists have always done: seek a match. Clearly, the best match is your own tissue, the autologous option, but in the absence of that, an iPS line can be employed that closely matches your own MHC profile. The problem is: the numbers. Imagine we had to have a line to suit everyone. There are estimated to be more than 16,000 MHC variants occurring in combination. There would have to be thousands of iPS lines in order to cover each combination. Then from each of those would have to be derived the particular cell type required, be it retinal epithelium, dopaminergic neurons, or oligodendrocyte precursor cells. Each one of those lines would potentially be considered by regulators as a different medicinal product, and need to be tested as such. Clearly, this could ultimately run to millions of cell lines and is clearly not a manageable approach.
But perhaps it could be approached stepwise. First, although there are many MHC class I variants, there are only three dominant genetic loci, called “HLA-A,” “HLA-B,” and “HLA-DRB1.”6 So long as these three loci are matched, then organ transplant studies suggest that rejection can at least be attenuated.7 There is, however, still a problem: most of us are heterozygous at these loci, that is, we’ll have inherited a different variant from each parent. So most carry two genetic variants at each locus. This can be overcome by only using iPS cell lines whose donors are homozygous—where both maternal and paternal genes are identical. So, instead of having to match two variants per genetic locus, only one match is needed. Suddenly the numbers start to look fractionally more manageable. If the most common homozygous combination were used to make an iPS cell line, that line would potentially fit with 14.5 percent of the Caucasian population. The second most common combination would add a further 6.5 percent. So, by generating just the right two homozygous iPS cell lines, 20 percent of the Caucasian population become potentially matched recipients. Moving forward from there, however, with real ethnically mixed populations and decreasingly common variants, we meet a serious problem of limiting returns. You need 17 iPS lines to cover 50 percent of the European population, and this set doesn’t travel well. Only 13 of the 20 most common European combinations appear in the 50 most common among Hispanic populations. This number drops to 8 for African-Americans, and just 3 for Asians. So again, the numbers quickly become unmanageable.
The other logistical problem with this approach is finding donors who are homozygous at these genetic loci. Marc Peschanski and colleagues have calculated that to find just a single individual carrying the most common Caucasian variant (the one that would fit 14.5 percent of the population), they would need to screen roughly 180 individuals.8 To get one of each of the ten most common, they would need to screen 11,000 people. And it gets progressively worse from there on.
Despite this difficulty, projects are under way to develop “haplobanks,” repositories of iPS cell lines suitable for allogeneic transplantation. Most notably, the Center for iPS Cell Research and Application (CiRA) in Kyoto, Japan, is pursuing this alternative,9 and an international collaboration, the Global Alliance for induced Pluripotent Stem Cell Therapies (GaiT), is seeking to coordinate these activities.10
Quite apart from logistical problems, the haplobank strategy has two serious issues: cost and comparability. The cost problem is obvious: who’s going to make the substantial financial investment required to establish and maintain these banks? Haplobanks have been compared to public umbilical cord blood banks such as exist in many countries, including the United States and the United Kingdom. But the difference is that cord blood banks have demonstrable public health utility, whereas haplobanks will need to overcome some major problems before such utility could be demonstrated.
The conundrum is this. The iPS cells are the starting material for a cell therapy product, rather than the product itself. The actual products are the retinal pigment epithelium cells, the dopaminergic neurons, or the oligodendrocyte progenitor cells derived from the iPS cells. Like all medicines, allogeneic cell therapy products have to be tested for quality, safety, and efficacy. Each product must complete preclinical and clinical evaluation. Thus each therapeutic cell line made from an individual iPS cell line would be a new cell therapy product. So a hypothetical haplobank of, say, twenty iPS cell lines, used to generate twenty retinal pigment epithelium cell lines would be deemed by regulators to have generated twenty distinct medicinal products, each one of which would need to go through the entire preclinical and clinical testing process. And if those same twenty iPS cell lines were then used to make, say, twenty oligodendrocyte progenitor cell lines, that would be another twenty distinct medicinal products.
Were this really required, it would threaten the entire project. The estimated cost to develop a single medicinal product has been estimated to be $2 billion.11 Is any company seriously going to run this process twenty times (or even just twice) for what is effectively just a single medicine? Certainly, there would be economies of scale in running the processes in parallel, and each success would reduce both the risk and the cost. Nonetheless, this is surely not a viable prospect.
This is where comparability might help. Comparability is a well-established regulatory solution to a medicine manufacturing problem. Pharmaceutical companies frequently have to make a changes in their manufacturing process. Sometimes, the manufacturer can no longer find a supplier for a particular reagent required for the production of the drug, or a superior process is discovered for making the drug. Sometimes, the company just wants to transfer production to a new facility. Consistency is clearly important: every batch of a drug must be identical. So how can the company be sure that the revised product is unchanged despite the change in the process?
The answer is that regulators will ask for comparability studies to be conducted. The regulators and the manufacturer will have agreed on the medicine’s “critical quality attributes” (CQAs), that is, the key properties that assure the quality, safety, and efficacy of the drug. The regulators will ask the manufacturer to demonstrate that the new version of the drug has the same CQAs as the original version, assuring confidence that consistency has been maintained.
One proposition is that the same concept of comparability could be applied to the range of iPS cell–derived products. Thereby, all twenty retinal pigment epithelium cell products derived from the twenty iPS cell lines could be considered comparable if one line had been tested and approved as a novel cell therapy product, and the other nineteen could be shown to have the same CQAs as the approved line and were therefore comparable. Instead of running the entire process twenty times, you run it just once, and show all the others are essentially the same.
Unfortunately, this proposal involves some wishful thinking. The first problem would be simply that comparability is not normally applied to starting materials. If a process starts from different materials, it would usually be deemed to be an altogether different process. A compliant regulator might allow this one to slip by, but other issues might be more sticky. For example, regulators are unlikely to be happy with the current level of variance between iPS cell lines. Currently, even two lines from the same donor can vary considerably in their growth and differentiation properties. The first challenge to comparability, therefore, will be to develop reprogramming strategies that are sufficiently robust and reproducible to generate cell lines that could conceivably be considered comparable. Beyond that, how can there be certainty that two cell lines will behave equivalently once injected into a patient, especially since they have been purposely chosen to be distinct in precisely the genes that determine how the body will react to the cell transplant?
Beyond these technical questions, there are conceptual problems enough to keep bioethicists engaged for a long while. Fairly obviously, for two products to be considered comparable, they would need to target the same patients. If a pair of cell lines have been specifically designed to be different precisely so that they can target two distinct patient groups—in this case, patient groups defined by their genetics—then surely they can’t be considered comparable.
Even if the comparability concept is accepted by regulators, it may not get the haplobank approach over the line. Demonstrating comparability itself is a challenge, as Christopher Bravery has pointed out.12 If the mode of action of a therapy isn’t known (as is the case with many cell therapies), then how can it be satisfactorily demonstrated that two lines are functionally equivalent. Moreover, comparability does not completely alleviate the cost problem. Three batches of a product usually need to be analyzed for comparability to be demonstrated. This is itself a substantial undertaking. Then, even following registration, problems remain. All drugs, particularly new drugs, are subject to intense regulatory scrutiny. Regulators monitor therapies carefully to see if any adverse outcomes emerge once the therapies start to be prescribed by doctors out in the real world. If just one “comparable” product runs into a problem, how should regulators view the whole combined set of therapies? This could be particularly complicated since each haplotype cell line will be taken up at different rates, some genetic variants being much more common than others. Thus batch size and penetration rate for each variant are likely to differ widely.
As things currently stand, most of these issues remain unaddressed. Nonetheless, considerable sums of money are being invested in strategies that require comparability to be ultimately workable. Regulators, who contrary to the opinion of some generally do like to say “yes,” are currently keeping relatively quiet. They may not be able to maintain that silence for too much longer.
Is there a cleverer way out of this predicament? Couldn’t we design a pluripotent cell that could act as a universal donor, suitable for all patients, regardless of their MHC profile? Well, gene editing provides one way this might be done. What if we simply took a single iPS lines and knocked out its MHC class I genes? The immune system would no longer see the cells as foreign, and that one line could be used for everyone. Except, as we’ve already seen, the immune system has evolved a strategy to counter this tactic, precisely because this trick is adoted by some cancer cells to evade surveillance. This is where the natural killer cells come in. Not only must a cell avoid showing the wrong class I markers; it must show the right ones to elude these killer cells.
So we need to be clever: we need to create a marker to keep the natural killers at bay without alerting the killer T cells, which are seeking out the wrong credentials. This is an area of intense research, and although no one has the final answer yet, an exciting recent approach is that of Germán Gornalusse and colleagues.13 They engineered a decoy MHC into their iPS cells. First, they used gene editing to knock out a gene called “beta-2-microglobulin” (B2M). Normally, B2M is required to partner the class I proteins on the cell surface. So by removing this gene, they were able to ensure that no native class I genes were expressed by the iPS cells. This kept the cells out of the clutches of the killer T cells, but what about the natural killers?
To fool them, the Gornalusse team engineered an artificial gene into the cells, composed of two parts: B2M coupled to a minor class 1 gene (HLA-E) that doesn’t vary very much between individuals. On its own, HLA-E isn’t sufficient to elicit a killer T cell response, but its appearance on the cell surface in conjunction with the engineered B2M is enough to elude the natural killers. By coupling the two partners together—B2M and HLA-E—the Gornalusse team ensured that the B2M couldn’t partner any of the more immunogenic class I proteins, so these engineered iPS cells had the best of both worlds: evading the killer T’s, while placating the Natural Killers.
The preclinical data accompanying this study are very encouraging, and the approach will probably end up being tested in the clinic. Still, because genome manipulation has become so easy, further refinements will surely follow. One obvious candidate is the insertion of a “suicide gene” into the cells. This would be the ultimate safety switch, so that if the transplanted cells turned rogue and became tumorogenic, a drug could be administered to the patient that would simply kill all the engrafted cells, and any progeny they had produced.
Dare I end this chapter with the prediction that neural cell therapies are finally poised to deliver? We have been here before. The ultimate success with advanced therapies always seems to be just around the corner. Nonetheless, there are reasons to be optimistic.
First, pluripotent stem cells really have liberated the field from the cell availability and scalability constraints that plagued early studies. Whether the therapies emerging from pluripotent cells prove to be safe and efficacious remains to be seen, but at least now the right cells will be available—reproducible and at scale—so they can be tested conclusively. The therapies may or may not work, but at least we should learn the answer, one way or the other.
A second cause for optimism is the advance in other parallel biotechnologies that are impacting cell therapies, permitting ever cleverer experiments. Studer’s optogenetic study is just one such development. Technologists can add value to their cell products with refinements such as the HLA-engineered cells we considered above. As gene editing becomes increasingly tractable, pluripotent cells and their derivatives can be manipulated in increasingly sophisticated ways. Again, this is an area that is only beginning to be explored, but surely represents the future.
Nonetheless, there remains much to do. Skills beyond those of cell and molecular biology are beginning to be applied to neural cell therapies, yet there is still a long way still to travel before truly regenerative therapies for the brain are finally to emerge. We’ve seen that earlier brain cell replacement approaches were adopted with considerable naïveté: slurried suspensions of stem cells squirted into areas of brain damage. This won’t do any longer. The “retinal patch” has taught us (if we hadn’t noticed already) that design criteria need to be specified in advance, and products manufactured to those precise specifications. Certainly, this will lead to more products, like the retinal patch, a combination of cells and a supportive matrix. The search for the appropriate substrates has been progressing in parallel with stem cell science for some years. Natural decellularized materials were an obvious starting point, but synthetic biodegradable polymers have been combined with stem cell implants for a number of decades. In 2002, for example, Evan Snyder’s group implanted neural stem cells into lesioned cortex on nanoparticles made from polyglycolic acid, and observed extensive growth of neural fibers into the host brain.14 Recently, more elaborate substrates have been devised, like manganese dioxide (MnO2) nanoscaffolds combining biological substrates such as “laminin.”15 Advances in bioprinting hold considerable promise in generating purpose-built tissues and organs, while nothing has yet been printed comes close to the complexity of complexity of true biology. How this forced simplification will impact function is currently unclear.
Moreover, identifying appropriate substrates is just step one. None of the artificial structures have as yet addressed serious clinical limitations, such as those associated with scaling. Any tissue of any reasonable size will require a blood supply. We noted in chapter 10 that while three-dimensional organoids can be grown from pluripotent cells in culture, the absence of a blood supply ultimately leads to metabolic failure. So, in addition to the complex histogenesis required to build the primary brain tissue, there needs to be a concomitant construction of blood vessels, with all their biological and mechanical complexities. Then follows the challenge of linking any newly formed vessels with the host blood stream. These are no simple tasks, and no therapeutic has come close to accomplishing them on a clinical scale.16
Conversely, there are opportunities that have scarcely been pursued. Gene therapies can be used in conjunction with cell therapies. In some nonbrain applications—immunotherapies in particular—this is proving to be highly effective. “Car-T cells,” for example, are cancer therapies in which a patient’s own T cells are removed, engineered to carry a gene that boosts the immune system’s ability to identify and kill cancer cells, then injected back into the patient. The “cells plus genes” approach could be used in different ways. The cells could be engineered to release therapeutic molecules or factors that would improve their own survival and efficacy in the damaged brain. Safe to say, an enormous range of possibilities is starting to emerge.
1. Merkle, F. T., Ghosh, S., Kamitaki, N., Mitchell, J., Avior, Y., Mello, C., et al. “Human Pluripotent Stem Cells Recurrently Acquire and Expand Dominant Negative P53 Mutations,” Nature 29, no. 7653 (2017): 1–11. http://doi.org/10.1038/nature22312.
2. For a review of studies on gene sequencing of stem cell cultures, see Martin, U. “Genome Stability of Programmed Stem Cell Products,” Advanced Drug Delivery Reviews 120 (2017): 108–117. http://doi.org/10.1016/j.addr.2017.09.004.
3. See Mandai et al., “Autologous Induced Stem-Cell-Derived Retinal Cells.”
4. See Chakradhar, S. “An Eye to the Future: Researchers Debate Best Path for Stem Cell–Derived Therapies,” Nature Medicine 22, no. 2 (2016): 116–119. http://doi.org/10.1038/nm0216-116.
5. Takashima, K., Inoue, Y., Tashiro, S., and Muto, K., “Lessons for Reviewing Clinical Trials Using Induced Pluripotent Stem Cells: Examining the Case of a First-in-Human Trial for Age-Related Macular Degeneration,” Regenerative Medicine 13, no. 2 (2018): 123–128. http://doi.org/10.2217/rme-2017-0130.
6. Note there are two different nomenclatures here: sometimes major histocompatibility complex (MHC) antigens are referred to as “human leukocyte antigens” (HLAs), but both are referring to the same thing.
7. Opelz, G., and Döhler, B., “Pediatric Kidney Transplantation: Analysis of Donor Age, HLA Match, and Posttransplant Non-Hodgkin Lymphoma: A Collaborative Transplant Study Report,” Transplantation 90, no. 3 (2010): 292–297. http://doi.org/10.1097/TP.0b013e3181e46a22.
8. Gourraud, P.-A., Gilson, L., Gilson, Girard, M., and Peschanski, M. “The Role of Human Leukocyte Antigen Matching in the Development of Multiethnic ‘Haplobank’ of Induced Pluripotent Stem Cell Lines,” Stem Cells 30, no. 2 (2012): 180–186. http://doi.org/10.1002/stem.772.
9. See Center for iPS Cell Research and Application (CiRA), Kyoto University, website: http://www.cira.kyoto-u.ac.jp/e/index.html.
10. See Global Alliance for induced Pluripotent Stem Cell Therapies (GAiT), Edinbugh, website: http://www.gait.global/about-gait/.
11. Sullivan, T., “A Tough Road: Cost to Develop One New Drug Is $2.6 Billion; Approval Rate for Drugs Entering Clinical Development Is Less Than 12%,” Policy & Medicine, last updated March 21, 2019. https://www.policymed.com/2014/12/a-tough-road-cost-to-develop-one-new-drug-is-26-billion-approval-rate-for-drugs-entering-clinical-de.html.
12. Bravery, C. A., “Do Human Leukocyte Antigen–Typed Cellular Therapeutics Based on Induced Pluripotent Stem Cells Make Commercial Sense?,” Stem Cells and Development 24, no. 1 (2015): 1–10. http://doi.org/10.1089/scd.2014.0136.
13. Gornalusse, G. G., Hirata, R. K., Funk, S. E., Riolobos, L., Lopes, V. S., Manske, G., et al., “HLA-E-Expressing Pluripotent Stem Cells Escape Allogeneic Responses and Lysis by NK Cells,” Nature Biotechnology 35, no. 8 (2017): 765–772. http://doi.org/10.1038/nbt.3860.
14. Park, K. I., Teng, Y. D., and Snyder, E. Y., “The Injured Brain Interacts Reciprocally with Neural Stem Cells Supported by Scaffolds to Reconstitute Lost Tissue,” Nature Biotechnology 20, no. 11 (2002): 1111–1117. http://doi.org/10.1038/nbt751.
15. Yang, L., Chueng, S.-T. D., Li, Y., Patel, M., Rathnam, C., Dey, G., et al., “A Biodegradable Hybrid Inorganic Nanoscaffold for Advanced Stem Cell Therapy,” Nature Communications 9, no. 1; article no. 3147: 1–14. http://doi.org/10.1038/s41467-018-05599-2.
16. For a more detailed discussion of these issues, see Stevens, K. R., and Murry, C. E., “Human Pluripotent Stem Cell–Derived Engineered Tissues: Clinical Considerations,” Stem Cell 22, no. 3 (2018): 294–297. http://doi.org/10.1016/j.stem.2018.01.015.