So here's a glioma cell transfected with a leucine rich protein. It has all those amide protons on the left. This is a control plasma. We turn it on and we see only our tumor lighting up.

What is the beauty of the system? It's that we can design a family of reporter genes with all different resonance frequency and we can saturate specific frequencies so we can do multicolor imaging, sort of analog to fluorescence where you have all these different wave lengths. I think it will be really important in the future to be able, just like immunocytohistochemistry to look at multiple cell populations simultaneously specifically to look at these interactions, also when we co-graft multiple cells.

So this is a summary of the techniques that are currently available. It's a very crowded slide. I am sure they will be posted online so you can maybe look at it later. I think the three that are most closely to becoming clinically important when we start to use human embryonic stem cell therapies in humans are listed here: positron emission tomography; SPECT or gamma camera imaging for systemic injection; and also MRI as you note that it's primarily for correct delivery and local engraftment, less for systemic. The resolution are here. The number of cells, it's possible to see single cells at the right resolution. I should also say that CT and ultrasound and there are things happening right now with contrast agent that we'll hear more about.

If you're interested, some of this work is covered in deeper context in this review article that came out last month, if you want to look that up, with a lot of references.

So, in conclusion, the iron oxide label cells and compatible catheters, and, again, the few centers set up as of yet because the special catheters allows MR-guided realtime targeted delivery and, essentially, accurate cell delivery and it can have various application. And then after we do that, then reporter genes are needed to serve as a beacon for cell survival, uncontrolled cell proliferation, as well as cell differentiation.

And I think that at the end it's probably going to be a combination of techniques, so why not put in a PET reporter gene in a cell and at the same time label it with iron oxide so we can do both delivery and follow it. I think there's an unrecognized potential for interventional radiologists who will do this in their practice or in their academic setting. That it is something that imaging people always talk about they want to track cells to see where they go, to follow. Yes, but more important is to deliver them in realtime using specific interventional instruments and catheters and devices.

So I do think that needs to be an integrated part also with discussions with the stem cell people is how they want to deliver these cells and where because of those issue.

I should acknowledge a lot of people, also people I borrowed slides, and NIH funding. We have worked with a company in Baltimore with mesenchymal stem cells. We have islet cell protocol, got a lot of islets from Justin Diabetes Center.

And I leave it here. Thank you.

(Applause.)

CHAIR URBA: So we have time for a couple of quick questions.

DR. SNYDER: Jeff, which of the markers that you used to mark cells are already being used in patients and are approved for patients?

DR. BULTE: Yes. So the only one that approved is indium-oxine and they are the ones that have been used in patients just recently the thymidine kinase, the reporter gene from the herpes simplex virus and the Theradex label.

DR. SNYDER: And those could be applied to stem cells in your opinion?

DR. BULTE: Yes. There's no difference. Well, you get a practical issue is that embryonic stem cells may be harder to transfect. You know, there are other issues there. They grow in embryo bodies. Do we get the label. But, principally, yes, there's no difference, right.

DR. TAYLOR: So a number of these MRI studies with Theradex we and others have done and shown that as cells don't survive the label is taken up by macrophages and other cells and actually doesn't accurately reflect the cells you transplanted. So I'd like your comments on that in terms of clinical relevance.

And then, two, that's clearly not a quantitative measurement at this point. So can you talk about what the limitations might be clinically then in terms of how we over or under estimate effects if cell number is really critical to some of the deleterious side effects?

DR. BULTE: Yes. I've heard the question many times before. There's a limitation by labeling cells with a fluorescent dye, a lipophilic dye, Theradex indium-oxine it leaches when a cell dye is taken up. In case of the Theradex, the contrast itself disappears quickly since the macrophage is biodegraded. But the Theradex application is really the correct delivery in realtime, initially.

After that, we don't know if the cells are dead or alive because the contrast is going to stay there. Another problem is the label disappears quickly if cells start to proliferate uncontrollably. So the holy grail is an MRI reporter gene. It's the holy grail. And several centers have been working on that, including our artificial approach. People use ferreting. We are not there yet, so that's one way. Yes, so that's a potential artifact.

Your second question was about -- what was your second question?

DR. TAYLOR: How to do with the lack of quantitation?

DR. BULTE: Yes, yes. So the MRI is not quantitative. Reason being it's very simple, we do not a priori if our cells are clustered in groups or if they homogeneously disperse. That affects the MRI contrast differently and that's the problem of that. So the way to do this is indium-oxine or PET bioluminescent imaging, you know semi-quantitative.

What does that mean? We do not really know the depth where the cells are, so we have to correct for the attenuation of the signal. We have to make an estimate so it's semi-quantitative.

Also, by the way, indium oxine leaches out, binds transferring. You get a lot of artifacts, liver uptake, spots that the cells actually are not there. Each technique has it's problem. I'm just hoping in ten years we have a good MRI reporter gene that's safe. The reporter genes we are using are artificial, so we're creating things that don't exist. It's a whole other issue we have looked at and everything looks fine, but I don't know.

CHAIR URBA: Dr. Chien?

DR. CHIEN: I was going to ask you, this is a two-part question. One is, what's the minimum resolution that you can get? I mean how many cells can you pick up? I know it's obviously not a single cell level detection. So what is the minimum mass of cells that you can actually detect, 100,000?

DR. BULTE: Can you put it on four again, sir, on the visual? Real quick I can also tell you. Actually I should point it out because it's a very important question. It's in this column.

The sensitivity, the number of cells, for the reporter genes, bioluminescent imaging, and our artificial , and you go five to ten thousand cells, which you have not talked about them. PET, actually, I have asked the experts. I've not gotten an e-mail yet, but I think it's a low number, perhaps 100 or so. The MRI part with the side, that's a tricky one. And, again, it's not an easy answer. It depends on the field strength. The higher the field strength the better. These particles use contrast. It depends on the voxel size, the resolution.

I would say clinically we can see as much as a 1,000 cells in the lymph nodes in animal systems. We can use bigger magnetic particles, but we can see single cells in vivo at high field in an animal, but this will be clinically translated.

DR. CHIEN: Yes, that's what I thought. And the other question I had is, don't you think you could, like many other surgical protocols, because a lot of this is kind of surgery interventional, are you going to optimize the entire delivery protocol in animals without having to have all these sophisticated imaging things because then you can use, you know, sort of realtime, single- cell-level-almost analysis with lacZ reporters and optimize that and then just go into humans with an optimized protocol that you don't have to demand realtime feedback of the delivery of the cells while the operator is delivering it on the table?

DR. BULTE: Am I allowed to disagree with you?

DR. CHIEN: I didn't say that was the case. I just asked what you thought.

DR. BULTE: Yes. Reporter genes are not realtimed because you get the substrate and it takes a little while, you know, to accumulate in that case. I think in the case of myocardial infarct, if somebody is poking around at the area of the infarct, at the moment within seconds at least it will be known in realtime and the patient is dying, you know, and is claustrophobic and is in the system, I think it is very important to do that fast, not to see afterwards if the cells were injected at the right place.

Same with the lymph nodes, you're suggesting it's not important to do in realtime, but --

DR. CHIEN: I'm saying is is that once you work out the protocol, okay, so for example, for direct injection in a specific location in the heart for example, that I don't know that you necessarily -- you need readout to know where you're at in the heart, but you may not necessarily have to get realtime disposition of the cells in the heart because you've already optimized the protocol for delivery.

DR. BULTE: But if you don't know where you've put the cells, you want to find --

DR. CHIEN: Well, no. You can figure out where you put the cells by electrophysiological feedback, like NOGA and things like that. You don't need to see it, right?

DR. BULTE: Okay.

DR. CHIEN: Okay. Anyway, I don't want to argue with it.

CHAIR URBA: Dr. Firpo?

DR. FIRPO: You know you mentioned that a suicide gene in, again, acyclovir selection and there's a couple of papers out on that now, more than just the one you talked about. But are you aware of any studies where people have done selection to kill the tumor and then allowed the mouse to live after that to see if it comes back?

DR. BULTE: Following treatments? Okay. No, I'm not aware. I don't know how long the -- you know, if it kills all the cells. From the signal it looks like they're all dead, but if a few cells remain. I think it's a matter of dose. I've heard if you inject small numbers of these cells that they may not form tumors. It's just like injecting a subcutaneous tumor. If you inject 100 cells, you don't get a tumor in a nude mouse, but you have to give maybe 100,000 cells. I don't know.

It just depends on how many cells survive. I guess you could do experiments by dosing 10, 50, 100, 200 and see if they form tumors that the same scenario will apply, that the number of surviving cells could again form tumors. So it depends on the specific setting.

DR. GOLDMAN: Jeff, to just follow up on Dr. Chien's question. So the MR resolution is going to depend upon the cellular density, of course per voxel. So just as a base level of maximal resolution, how many cells are required per technique, whether by proton or by polyamide, cells per voxel can be detected let's say in a clinical 3T magnet?

DR. BULTE: Yes. So for clinical 3T magnets I think it's fairly safe to say for MRI within a voxel somewhere between 500 and 1,000 cells, and we're talking here 500 micrometer resolution in each direction, like a cube. The MI probe is less sensitive. Currently we can only really do it very well in animal magnets of higher fields. The sensitivity is about somewhere around 10,000 cells I think at this point.

The most sensitive tracer are these iron oxides. They're the most, so that's going to be the limit.

CHAIR URBA: We'll take two more questions. Dr. Woo and then Dr. Weir.

DR. WOO: Yes. We heard from previous presentations that the formation of teratomas is those dependent of the undifferentiated embryonic stem cells, and the threshold may be around 100,000 plus/minus. And yet in one of your slides, I think it's the first slide of section four, you show a teratoma forming in a mouse and you indicated you only need one cell for causing trouble. So I'm kind of confused as to what is it that is really needed to form a teratoma.

DR. BULTE: No. There's a misunderstanding and I understand your misunderstanding. That slide was with one bone marrow stem cell that reconstitutes the entire bone marrow to see the power that when you have one cell that it starts proliferating. Eventually you can see that. So that one cell was the bone marrow reconstitution experiment.

The other one that I followed up was the teratoma slide. So I understand the confusion, but they are separate studies.

DR. WEIR: You've given us a wonderful look into the future. But I wanted to ask about just conventional radiology techniques as far as monitoring an inject site, for example, for teratoma formation. Just how sensitive do you think it could be if you were looking at the spinal cord or if you were looking at some other site as far as getting a clue that there was a teratoma being formed?

DR. BULTE: Yes. So currently the way it will be done right now is purely anatomical. Right? We get a soft tissue mass or perhaps a skin or keratonin, whatever. So they'll anatomical at that time. It's just like, in general, in tumor formations, the whole issue. By the time a tumor is detected anatomically with MRI, it's already too big. So you want to have more sensitive methods to detect it in its very early stage.

So how many cells you need in order for it to detect it? I think it's the same issue as the sensitivity of these cells perhaps. So I think the bottom line is, is the sensitivity equal or higher than the number of cells that are needed to form a teratoma. So if you can detect fewer cells than are needed to form a teratoma, I think that's a good thing. So at that point you can maybe see it earlier than the cells are able to form a tumor, something like that.

CHAIR URBA: Thank you.

It's time to move onto the public hearing part of the meeting. I'd like to share this announcement before we start.

Both the Food and Drug Administration and the public believe in a transparent process for information gathering and decision making. To ensure such transparency at the open public hearing session of the advisory committee meeting, FDA believes that it is important to understand the context of an individual's presentation.

For this reason FDA encourages you, the open public hearing speaker, at the beginning of your written or oral statement to advise the committee of any financial relationship that may have with any company or any group that is likely to be impacted by the topic of this meeting.

For example, the financial information may include the company's or a group's payment of your travel, lodging, or other expenses in connection with your attendance at the meeting. Likewise, FDA encourages you at the beginning of your statement to advise the committee if you do not have any such financial relationships. If you choose not to address this issue of financial relationships at the beginning of your statement, it will not preclude you from speaking.

Our first speaker is Ms. Amy Comstock Rick from Parkinson's Action Network.

MS. RICK: Thank you and good afternoon. My name is Amy Rick and I am actually here in my capacity as president of the Coalition for the Advancement of Medical Research, which is a non-paying position by the way. But I also serve as CEO of the Parkinson's Action Network. Aside from that I cannot think of any conflict of interest that I have. I drove my own car from my home this morning.

The Coalition for the Advancement of Medical Research is a coalition that was formed in 2001 as a direct response to the President's policy restricting federal funding for embryonic stem cell research, as you all know, for lines that were derived after August 9^th, 2001. The coalition is comprised of patient groups, individual research institutions, some which are represented here, as well as associations of researchers. We have over 100 members.

Our mission remains fairly consistent, which is to promote regenerative medicine with a prime focus on the lifting of the President's policy on the restrictions for federal funding.

As I'm sure you all know, human embryonic stem cell research has been quite a focus for patient groups over the years. With the legislation that has gone through Congress, as well as the scientific breakthroughs that do get a fair amount of media attention, disease groups, Parkinson's, spinal cord injury, diabetes, many cancer, many, many others, the patient advocacy community as been quite focused on the progress. In fact, to the point that a few years ago when the legislation was a hot topic in Congress, we were fearful that the patient community was at a place where, if the legislation passed, that they would be expecting FDA approval and treatments immediately.

I think we are not at that place any more. I find that you have a very educated patient or affliction spinal cord injury population who understand in a very sophisticated way the issues surrounding human embryonic stem cell research, both potential as well as risk.

It is quite enlightening to the patient community that, in spite of the President's restrictions laid down in 2001, that the science has moved forward, not as quickly as it would have but for the restrictions, but both with private funding as well as the emergence of -- merging a fair amount of state funding the science has moved forward and it is inspiring to patients and we commend the FDA for actually having the advisory committee meeting to address some of the issue that we hope you will be facing in the coming months and years as you begin to see applications for approval for clinical testing using human embryonic stem cells.

As you all know, we actually anticipate even hopefully more eligibility for federal funding in the coming months and years as the three main presidential candidates are all from the U.S. Senate. So we happen to have on record that all three of them voted twice in favor of the stem cell research enhancement act which would have allowed federal funding for human embryonic stem cell research on lines derived, would have lifted the President's restriction if they were left over embryos from IVF clinics that otherwise would have been discarded and a few other limitations. So given the fact that we have this record, we do anticipate that in 2009 more research will, in fact, be funded using federal funds.

It is in that context of hopeful anticipation for this field that I want to raise two cautions to this committee. One would be our request on behalf of CAMR and the patient communities that in spite of the high visibility and great amount of controversy that there has been around human embryonic stem cell research that you not put an extra layer of risk averseness or safety requirements simply because the nature of the visibility or the controversy on the issue.

It is the risk benefit analysis, if you will, which I understand is extremely complicated on all these issue, not unique to stem cell research, and is not formulaic in any way, but the fact that there's external controversy I would plead that you do not, as scientists, allow external controversy in any way to interfere with your analysis.

And the second caution that I would raise is, in fact, not as directly related to stem cell research. But in an article in Bloomberg News this week, actually about this meeting, Dr. Robert Lanza was quoted as saying, in this field there can be no risk whatsoever. Now, I know that all of us know you can't take that literally because there's risk in everything in life. But I ask that you -- for the diseases that we're talking about in this room this morning, Parkinson's, spinal cord injury, juvenile diabetes, cancer, if you could, these decisions cannot be made in a vacuum. We are talking about risks, but you're talking about risks as you know balanced against the life of living with a chronic, progressive disease like Parkinson's. You're talking about cancer, juvenile diabetes. There is the ever present, horrible risk of living and dying a miserable death with one of these diseases or with this injury, and if I ask, as you always do, to keep that in mind as you're assessing the risk of some of the very serious questions that we heard about this morning, scientific questions.

Thank you.

CHAIR URBA: Thank you very much.

Next we'll hear from Dr. Chris Airriess, California Stem Cell, Incorporated.

DR. AIRRIESS: Dr. Urba, committee, thank you for the opportunity to speak today.

First off, I am speaking on behalf of a private company, California Stem Cell, and I'm an employee of that company. We are actively developing therapies and are engaged in preclinical development currently in spinal muscular atrophy and ALS, as well as spinal cord injury, so some of the diseases of the previous speaker has just brought up.

The stem cell research field is currently at a turning point. Research findings enabling the scalable, current, good manufacturing practice production of human cell populations at extremely high purity move the therapeutic potential of stem cell derived treatments from the real of hope to that of practical application.

At California Stem Cell we have conducted extensive safety testing along the lines that have been discussed here this morning on our human embryonic stem cell lines, as well as the high purity, differentiated human cell products of these lines. Studies such as these help to minimize the risks of potential therapies to prospective patients.

Until the technology for safe and scalable generation of patient specifics outlines has proven, compassion compels us now to use existing technologies to develop therapy as addressing devastating and currently untreatable human disorders.

With appropriate safety testing and careful administration of safe and effective immunosuppressive regimes, emerging therapeutics based on current human embryonic stem cell technologies are an immediate and viable solution for treatment of the widest variety of such conditions.

We've been highly impressed thus far with the dedication and insights of the team of Mercedes Serabian and her colleagues here at CBER. Two items in particular that we feel will be conducive to the efficient development of stem cell-based therapies are continued opportunities for early interaction with the FDA through the pre pre-IND process. This has been very beneficial to us so far.

We've got a lot of valuable feedback and we encourage the continuation of this process. And we would like to see clarity, which I'm sure is coming, on the FDA's requirements for preclinical efficacy in safety for stem cell therapies in the form of a formal guidance document.

Again, I thank you all again for the opportunity to speak today, and we also thank our key partners, the ALS Therapy Development Institute, Families With Spinal Muscular Atrophy, Johns Hopkins University, the University of California, Irvine.

CHAIR URBA: Thank you.

Now, notice of this meeting was made available to the public and anyone wishing to speak was asked to register prior to the meeting. However, we have a few moments of additional time if anyone else in the audience wishes to address the committee at this time.

If not, we'll adjourn for lunch until 2:05. Thank you.

(Whereupon, the foregoing matter

went off the record at 1:02 p.m.

and went back on the record at

2:07 p.m.)

A-F-T-E-R-N-O-O-N S-E-S-S-I-O-N

2:07 p.m.

CHAIR URBA: If everyone could please take their seats, we'll begin the afternoon session.

Okay. So we'll get started with the afternoon session, which, if you remember Dr. Bauer's presentation this morning, was to address three rather broad questions. And the first question up for discussion is on the slide that's before us. And I guess just to set the stage, I will read it.

Inappropriate Differentiation and Tumorigenicity, and what we're being asked to consider and discuss are:

Criteria for selection of clinically relevant animal species or models that support engraftment of the administered human embryonic stem cells, for example, optimal strategies for evaluating potential host rejection of administered stem cell-derived products?

What may be the optimal site of implantation in the animals in order to obtain meaningful test results?

What sorts of study durations are required?

And what is the most appropriate dosing method, that is, absolute undifferentiated human embryonic stem cell number versus percentage of undifferentiated stem cells present in the product to extrapolate cell doses tested in animals to plan the clinical dose?

So that's where we'll start. And Dr. Goldman, if you'd like to kick off the discussion?

DR. GOLDMAN: Sure. So trying to break that down a bit more operationally, both tumorigenicity and inappropriate differentiation can be looked at as functions of model and disease environment, of course of site, site of implantation especially for nervous system targets, function of the survival and the study duration, the cell dose, of course how that cell dose is calculated, whether before or after transplantation or as a function of both. And then, of course, both tumorigenicity and differentiation state have to be looked at as a function of the immune state as well whether we're dealing with immunocompetent patients, I mean suppressed patients, or as far as disease models are concerned, immunocompetent, suppressed or deficients.

So essentially we're looking at a combinatorial function of all those variables and we need to establish an algorithm for being able to apply these as uniform criteria across cell types, and at least in some reasonable fashion across disease models.

So I'll start just discussing at least the issue of tumorigenicity from the standpoint of just presenting a couple of questions for the committee, and then of course looking at differentiation or inappropriate differentiation from the same standpoint.

Tumorigenicity strikes me with the most important issue, at least in my own mind from what we heard this morning, is what constitutes a tumor, how to define it? Of course there was already some debate, if you will, in terms of whether a tumor could be benign, whether a histologic benigness connoted physiologic outcome benign nature. The issue of whether infiltration and to what degree was a measure of tumorigenicity and to what extent that precluded the use of ES or ES derivatives.

And then really in a more fundamental level, how looking at histologic tissues can we define a tumor? Should we look at the division rate, the survival rates over time, the expansion rates of the population over time? Do we need to assess that from the standpoint of the proliferation or turnover rates of need of cells in the organ into which the cells are being transplanted? Or as we looking rather for an absolute absence, as the case may be, of proliferation or undifferentiated expansion? What kind of markers can be used to define anaplasia? What kind of markers can be used to define undifferentiated expansion?

Now, of course, we heard the ES markers used as indices of the persistence of undifferentiated cells in grafts. But worrying about more from the standpoint also of the things not discussed. What happens as ES derivatives are implanted? And if those derivatives are still at the progenitor state and undergoing expansion themselves, what allows us to define whether the undifferentiated expansion of already committed progenitors, when does that become a tumor?

There are instances, some mentioned earlier, some others in literature, of undifferentiated neuroepithelial expansion of ES-derived neuroepithelial cells. This may be a problem in a variety of organ systems. So it's not just a question of undifferentiated ES persisting in a graft, but also of their mitotically competent derivatives. At what level do we need to exercise control over their expansion?

At what level do we permit the implantation or introduction of any persistent, undifferentiated ES cells, or, as the case may be, still mitotically competent derivatives thereof? And what kind of markers can we use to define the existence of these cells? The cancer literature, of course, has a number of markers defined phospho-Akt survive in a variety of proteins that can be used for identification purposes, but they tend to correspond to markers of anaplasia or uncontrolled expansion.

Cells that are no longer controlled by normal cell cycle checkpoints. The issue is going to be, I think, for this field as progenitors are transplanted and then undergo persistent expansion that will not necessarily express markers of anaplastic transformation. At what point can we define that as normal expansion versus uncontrollable?

So those are all questions, but they really come down to the point of, what can we tolerate in terms of implantation of potentially undifferentiated or partially differentiated cells that are still capable of expansion? So I would broaden the issue behind just ES cells and behind just

ES-derived teratomas or teratocarcinomas.

So that segues into the issue of differentiation and to my mind it's very much a parallel question. What constitutes inappropriate differentiation?

And so it's essentially by definition. Ectopic differentiation of a functional, mature phenotype in an area in which that phenotype would not normally be present or in numbers in which that phenotype would normally not be represented, would represent inappropriate differentiation.

But we don't yet have criteria by which to establish whether an inappropriately differentiated pool is dangerous. Under what circumstances inappropriate differentiation may be relatively harmless? In what cases it may be beneficial? There are no general rules where this is concerned and these outcomes may depend very much on the disease state and the disease target.

In my own field, if we put glial progenitors into say a demyelinating lesion and we're looking oligodendrocytic differentiation and we see astrocytic, well that would be inappropriate. If we go into let's say a stoke bed and pick up astrocytic differentiation, well, that would be beneficial. So this is very much a function of the disease target as well as the cell type that's being implanted.

Then even appropriate differentiation into a set of phenotypes that is correct, if you will, for the organ may still not yield an appropriate functional outcome, and so the necessity becomes to match, essentially, the quantitative representation of phenotypes generated from progenitors that may be capable of giving rise to multiple phenotypes.

For islets, for islet progenitors derived from ES, it's one thing to look for beta cells as an appropriate cellular target, but one may expect these progenitors to potentially give rise let's say to alpha or delta cells, but potentially antagonizing the effects of the beta cells. For example, in the examples mentioned earlier of dopanergic production, well those same progenitors as derived from ES can also give rise to serotonergic and gabergic cells which in vivo/in vitro can potentially antagonize within the steroid of some of the effects of the dopanergic neurons.

So even when we have ES that are giving rise to the cell types of interests and even the precise representation of cells that would normally be derived from those progenitors, unless the proportions are correct that we're going to need, we may see if not dangerous, then at least counterproductive effects that would potentially diminish ultimately efficacy, as well as in some cases potentially presenting safety issues as well.

And so we need to define and characterize the state of differentiation that we want cells to be implanted at, what types of lineage potential they have at that point, and it becomes a big of a ying yang in that by the time we have cells that are sufficiently mature to yield the cell type of interest with the highest possible fidelity, in other words the purest possible population of the cell type of interest, well by that point we're far enough down the lineage and the cells may be post mitotic, they may not tolerate the engraftment well, and so we may not have a viable, engraftable cell population.

On the other hand, the still mitotically competent cells that may have much greater efficacy, as well as survivability upon transplantation, may be those that potentially may give rise still to undesired phenotypes that may still be capable of uncontrolled expansion. And so this is the, essentially, dilemma I think that we all face is establishing what is the appropriate stage of differentiation for transplantation and how essentially enriched or purified do those populations have to be at the time of transplantation? So I'll leave those as essentially entry points for discussion.

CHAIR URBA: Do you have a couple of comments on what you would do preclinically to identify and answer those questions?

DR. GOLDMAN: Well, then it's a question of disease target and, specifically, the cell type of interest. I think these answer are going to be, to the extent that answers can be derived for any, but I think that conceptions are going to be driven by exactly what disease target and exactly what cell types are being used.

So, for example, my own target of interest, the hypermyelinating disorders. This is a set of disorders where glial progenitors, as Jane was discussing before for example, can be productively used. However, here's little ability to control the oligodendrocytic versus astrocytic differentiation of these cells. We're still learning what the rules are.

But we know that by the time the cells are oligodendrocytes, oligodendrocytes at least primate and human, oligodendrocytes are every bit as post mitotic as neurons, and so they become very difficult to engraft. And someone in order to have a working strategy has to be able to give the cells are progenitors.

And so then the issues of persistent expansion and potentially unregulatable differentiation come to the fore. And the imperative then becomes matching up the disease target to the phenotypic potential and likely in vivo activity and behavior of those cells. That can be productively done with some disease targets, not so much with others. So I don't know that there's a one size fits all answer to these.

CHAIR URBA: But if you were going to take the oligo cells to trial, you would take the earlier stage of differentiation, which is more susceptible to some of the problems we talked about. What preclinical things would you want to see have done before you actually went and did that clinical trial?

DR. GOLDMAN: So I would want to see a definition of the expansion kinetics of that cell population over time in vivo in animal models, in animals models that severally replicate the disease target of interest. The net population expansion from the standpoint of how many cells do you have at given points of time after the initial transplantation, taking out for very long periods of time essentially for the experimental models we typically use, mice and rats, the life of the animals.

I would want to see what the mitotic rate, the fraction of cells in cell cycle was as a function of time. And what point the mitotic index of the implanted cell population fell to that of the background, essentially the native cell population of the same phenotype in vivo, in other words of the host cycle kinetics, at what point do they actually match up.

At that point I would want to make sure that there was no heterotrophic migration of the cells into in this case non-white matter areas, in other words that we weren't seeing heterotrophic foci oligodendrocytes in areas of gray matter for example. I would want to see that there was no overt anaplastic transformation of any of the cells at any point. And that there was no evidence of heterotrophic differentiation to non-glial phenotypes, much less non-neural.

I think all of those are critical safety end points, and that in the final analysis that the overall cell number was in no way perturbed by virtue of the graft, and that the final representation of phenotypes at least was analogous to that of the native tissue that one was trying to either repair or replace. So that, in effect, we're looking for the establishment over long periods of time of histologic and, therefore, physiologic normalcy.

CHAIR URBA: Two other questions I think that tie in with all the other questions and then we'll let other people comment.

Is the mouse and rat model, is that as far as you go, or do you think you've got to establish it --

DR. GOLDMAN: Depends on the disease target. So, for example, the congenital hypomyelinating disorders, the only models that exist, by and large, are mice. And yet there's no reason in terms of the known biology, of course we only know what we know and we don't know what we don't, but in terms of the known biology, there's no reason why those models wouldn't be reflective of the congential hymyelinations of humans, of children.

On the other hand, if one is looking at, for example, models of adult either traumatic injury or stroke, the issue then becomes what the kinds of sizes involved, what the kinds of unique features of the primate and human vascular supply to regions that are being challenged, that one, I believe, does need larger animal, and preferably primate models because there are primate and human specific features of the anatomy, and in some cases of the cell biology, that require large animal, or as the case may be, primate modeling. So, again, it depends on the disease target that one is approaching I believe.

CHAIR URBA: Is there an acceptable rate of tumor formation or does it have to be zero?

DR. GOLDMAN: I would say zero, and just getting back to the point that we raised earlier, at least in the spinal cord and brain, there is absolutely no such thing as a benign tumor. It doesn't matter what they look like histologically.

CHAIR URBA: Would the acceptable rate of tumor formation be target specific too, or would you generalize that to all the things you heard about today?

DR. GOLDMAN: This, of course, is all personal opinion. But with few exceptions, the disease targets of regenerative medicine are not ones that, to my mind, should permit the genesis of tumors, of cancers from implanted cells. When one is talking about biologic therapeutics or any therapeutics in the setting of diseases with short life spans, of course the bar becomes smaller.

So, for example, one may tolerate the potential development of lymphoma or leukemia decades out in the setting of chemotherapy, radiation therapy of say a child with leukemia, because you're looking at extending the life span tremendously and you essentially are faced with an all or none situation. Most of the disease targets that I think we're all looking at in regenerative medicine are a bit different in terms of these are disease that often are chronic, they often are involving tissue loss over long periods of time, and so looking at that from the standpoint of playing out essentially the cost benefit over time, from my own standpoint I don't see the risk of tumorigenesis as being really justified.

Again, I'm speaking with very broad strokes, but in general terms I don't think it's justified. There are examples, such as Huntington's disease where one may think in terms of inducing endogenous progenitor cells, where the patient may have a relatively short life span and will potentially be using a strategy that may increase the risk of brain tumorigenesis over time. But then just as the example before with childhood leukemias, if we're looking at potentially extending the life span of that individual considerably beyond what it would normally be, then the risk of late stage tumor formation becomes I think justified.

But it's for disease targets where the life span is not significantly curtailed or is curtailed but not potentially to the point where the appearance of tumorigenesis would necessarily be where it's need to be, that's where I don't think it's justifiable.

CHAIR URBA: Thank you. Dr. Taylor?

DR. TAYLOR: Thank you.

In thinking about optimal preclinical study design I think you made an excellent point. I think the first thing is that the preclinical studies have to be clinically relevant. So they have to be done in animal models that reflect the disease state at which we're looking and they should reflect the questions that we're likely to see clinically with regard to migration, whether route of administration matters more than site of delivery, and I think we're really asking the question, what it really comes down to, what can cells do versus what they actually do, and so we need to design preclinical, in vitro studies to examine both the potential of the cells as well as in vivo studies that show not only the potential, but what actually happens given the population that you ultimately are going to use.

I wonder if, perhaps, given the preclinical data in mice at least suggest site of undifferentiated cell administration really has an impact on tumorigenicity, and given that we heard this morning that vulnerable sites exist, perhaps site localization studies should be recommended to include direct administration of the final cell product to some of those vulnerable sites, most likely vulnerable sites to see if, in fact, tumorigenesis or some other adverse effect is going to be an issue.

We've also heard that dosing matters, cell number matters, and I think back to some of the angiogenesis assays and tumorigenesis assays that we all did when we were looking at angiogenic cell types, and there's a standard subcutaneous tumorigenesis assay that we all do to determine if cells are angiogenic. It seems to me that we can develop similar sorts of assays for the uncontrolled cell proliferation potential of these cells.

And then, finally, I wonder if -- I think we need to be careful not to add an extra burden on embryonic stem cells, but at the same time recognize that given the fact that we don't yet know to what degree minimally differentiated cells have adverse events, we need to develop strict definitions of what potency and release criteria are going to be, and I'll speak to that more when we get to the next question.

I guess there is one other thing that I would like to say and that is patient criteria have to fit in here somewhere. It may be that an age of a patient matters, just like disease matters. Age may matter, sex may matter, co-morbidities may matter, and so I think as we begin to ask some of these questions about the both positive and negative potentials of these cells, we have to consider the context in which the cells are going to be delivered.

DR. SNYDER: I just wanted to just briefly reiterate a point that Steve started to make and Doris began to make, too, and that's that in answer to your question as to what should our cell population be that we put in there and how do we know. It really so much depends on having really faithful animal models that reflect the real disease. And that in turn goes hand in hand with a better understanding of what the pathophysiology of the diseases are we want to treat, and that, in and of itself, is turning out to be a moving target.

You know five or six years ago, for example, for ALS, the slam dunk answer would have been, well, we just need motor neurons. Now as our sophistication about the nature of ALS is growing, we're starting to realize, well, sure, motor neurons are important, but maybe we actually do need astrocytes, which means that perhaps we need to somehow figure out how to put in a mixed population of cells that'll be necessary.

This kind of thing I think even extends to our view of embryonic stem cells, and this is a point that Willie and I started talking about. On the fact of it, to say I'm going to use embryonic stem cells in a neurologic disease, it would be a deal breaker to say, oh, and by the way, I'm also getting some mesoderm, I'm getting some vasculoendothelium, and I'm getting some smooth muscle. Ordinarily, you would say, well, I guess you can't use your cells.

Well, with a broader view of how we're starting to appreciate the injury niche, we realize the injury niche is neural elements, and also vascular elements, and some extracellular matrix. It may be exactly what we want to be able to reconstruct a niche that you have a cell that safety gives you some elements and then some support elements, and Willie talked a little bit about the cross talk.

So I guess the point comes back to, even as cell biologists, we still have to go back to really good, faithful, animal models. And that's independent of whether it's a larger animal or a small animal. We just need models that at least model the path of physiology and then be prepared to revise our notion of what pathophysiology is and what you really want to do to fix something that's broken.

CHAIR URBA: Yes. I think your summary was brilliant. I think you've really summed it up. The difficulty, of course, there are so many questions simultaneously. It's trying to solve a complex puzzle. So let me just suggest parsing this out into the cell types because each cell, when you examine it, interrogate it, there are different questions. So ES cell, the intermediate, which would be the progenitor committed but can still have renewal capability in the differentiated cell.

And so in the ES cell, what's interesting about this is is that your starting material, which would be kind of like let's think about this like a monoclonal antibody, and then you want to purify the soup and eventually get a purified humanized antibody, in this case the definition of your optimal study material is actually a cell that will form a tumor and so that's very different. So you have to have criteria to ensure that the cell is what you want to begin with.

So, obviously, first thing, there has to be criteria that it is an ES cell that meets criteria of interest. That can be done by transcriptional profiling. It can be done by -- and other issues. I'm not sure. I was concerned about the karyotypic abnormalities, but I think that can be sorted out.

The other thing that I was quite impressed with is how little, how few assays there are for monitoring for teratoma. I mean if we, and I say we meaning the whole field, we're all interested in our diseases, heart disease, neurologic disease, et cetera. But what we're not doing is focusing on a core barrier where, if we don't get over it, we're not going to be able to do anything. And that is having better assays for teratoma. There's just no question.

I look at this and I say, the idea that you're going to take an ES cell, up it into a nude rat, and have any kind of confidence that this or a derivative, however you're purifying that, be it ibolometric or by FAC sorting, okay, choose anything. There's a kind of a gradient there. I think it's not stringent enough and there are things that you can do. So here's a couple of things and I'm surprised to hear that people haven't done this.

First of all, you can do xenotransplantation without rejection. Take a human cell, put it into a pig, you just have to do it at a window where tolerance is not fully established. And so you could do it in a very late fetal stage in pigs or in

non-human primates.

You could give it systemically. You could put it into the fetal circulation, and if those cells, if they were tagged and did not produce tumors anywhere after about three months or four months, and they were put in very large quantities, I would have confidence that that cell preparation was as good as we could tell non-tumorigenic.

The difficulty of using a rat or a mouse is, okay, the good news, it's not being rejected; the bad news is all the competence factors, receptors, growth factors, et cetera, won't even tickle appropriately the human ES cell. Many times they are variant enough that they won't even activate the downstream pathway. And all of those effects, including the niche itself, are not going to be operative and so you wouldn't have that confidence. So I think that's one issue.

The second issue is we need better surrogates. We need, if you will, a cholesterol for teratoma. This sounds very cardiological. Sorry. But tumors secrete factors. There could be signatures. They could be unique. If nothing else, in a pig you could detect a human antigen for sure if you did this correctly. I say it as a group. I'm not saying --

We just have not been aggressive enough in developing these assays that we need. What I don't think we should do is just settle. So I am a very strong advocate for embryonic stem cell research.

On the other hand, I think we have to let science take us there and we have not be aggressive enough in developing these assays. That's very clear to me. And until we do and have confidence in them, and so what we need to do is we need to take ES cells and then differentiate them and see when do they lose their tumorigenic capability and what level.

So you can say, well, we won't tolerate any tumors. Okay. Well, I can understand that. But on the other hand we expose ourselves to mutagenic stimuli every day. We get X-rays. Okay. We know there's a defined risk, but we accept it. Okay. You manage it. I think that's what this is.

We have to figure out what is a tolerable risk for whatever dose of whatever the number of cells, or whatever. But until we get quantitative with that, we are just guessing and I don't think guessing is good enough.

The other point is is the niche, which I'm sensitized to, is that, of course, you know I'm a cardiologist and we've -- not me, but our field has been putting cells into the heart for some time with quite mixed and ambiguous results. But one of the things I think is going to be clear depending upon the organ system and the disease is that there's going to be a different mix of benefit ratio. It has to be very carefully examined so that the benefit's always -- and I think you were saying, that the benefit's always got to be in favor of the patient. Okay, that's point one.

Point two, there are going to be certain areas where getting the cells back out is not going to be trivial and there has to be some thought as to what you're going to do if things go wrong. And that's one of the things I think that I share your view on this CNS as being difficult there because, if things go wrong, it's hard to undo them. And in other cases it might be a little bit easier. So that would be one issue.

Now, the other thought was progenitors. So one thing to think about is, I didn't hear this at all today, is that we do know that, and these factors and pathways are being flushed out every day, that ES cells, including human ES cells, can be directed to go in a certain direction in response to specific queues. Some of these are well defined factors. Some of them are used every day. They don't have to be a gene. They don't have to be gene therapy.

But if you can get a cell, an ES cell, to move in a certain direction, it may move away from being a teratoma. So you might be able to design protocols to move it so it has zero potential to be teratoma and still has the potential to be your cell. Maybe not the cell that your neighbor wants, but maybe the cell that you want. And I think figuring those first steps out and standardizing, so you want mesoderm, you hit everything with this; you want ectoderm, you hit everything with this; you want endoderm, you hit everything with this, and I think that you can flush this out to take those steps.

Now, this is not a product, but somebody has to take those steps because that could eliminate from a finite to perhaps darn near zero because you've moved away from the gold standard of ES cells, which is the ability to make a tumor. So I think was one step.

The other things is is there's clear evidence in humans that you can implant cells that are progenitors into an organ, into a living human being and it won't cause cancer and that is myoblast therapy for heart failure. Well, it didn't work in the Magic trial. I know it's still a thought. I know you're aware of some of this work. They didn't get tumors because the cell has a limited capability of replication.

So I think just the fact that it divides doesn't necessarily mean it's going to be a tumor. It might be good. You might get an amplification signal. So I think each cell has to be looked at very, very carefully and I think putting some attention on getting the intermediate and characterizing it rather than the ES cell itself, because if you can catch the intermediate and purify it and characterize it, you've got it. You know what you have and you know what you don't have.

It's not an ES cell, and you can reproducibly say these are the criteria that we are going to establish that you have that intermediate. It has to fulfill these criteria then you go forward with it.

The last one with the differentiated cell type, what I was amazed at is is that the criteria for what people would accept for a cell that they would be thinking about, maybe not yet, but think about putting in a human being, it's clear that the technology for isolating this cell has not purified it to a level of homogeneity where you could be sure that it was a cell, even the cell you wanted. It doesn't mean it didn't have a few cells in there that you want, but that 90 percent or 95 percent of the cells were the cells you want. So the one criteria is you need to figure out is the cell you want.

The other is is that it could be a variant of the cell you want. So, for example, there are many different types of heart cells. You have a pacemaker cell. You don't want to put a pacemaker cell in your ventricle if you want to contract. It's not going to be a good thing, you know, you have electrical confusion.

I think these sorts of issues, it looked like a heart cell, expressed myocin, but it had spontaneous pacemaking capability. We didn't know that. There may be an advantage of having a more homogeneous starting material that will go in a certain direction rather than thinking that all the cells that express these four markers are all the same cell. I think that could be dangerous.

And that's it. That's all I have to say.

CHAIR URBA: Take your niche argument one step further. The mouse and rat don't provide the other factors, but the pig does? Would a non-human primate provide more and better and are you arguing --

DR. CHIEN: No. I think value and primate would be better. I was just concerned that people would be upset that I was advocating giving human ES cells in utero to non-human primates. But I think if that's what it takes to have complete confidence that you're ready to do something as potentially revolutionary and as exciting and as valuable as this, I think it's worth it. And you do it in very high dose.

It would be kind of like the Ames Assay, you know, you just really put a huge dose of cells systemically into the fetal circulation. They would be genetically tagged lacZ, something else you could see very quickly, and then you see did any tumors form in the progeny over three months.

CHAIR URBA: Dr. Weir?

DR. WEIR: Ken's brought up all kinds of points that need further discussion. But I did want to get back to specificity of the animal models and how valuable they are, and Evan and you talked about that, and particularly in relation to diabetes. Because as far as dosing and characterization of what you want to put in, it seems to me that diabetes is actually rather unique in that regard because we have so much experience with just experimental islet transplantation.

We know what the beta cell mass should be. We have very defined markers as to what a differentiated beta cell is. So you can put the human embryonic stem cell precursors in and see exactly what they turn into. And you can put them in various transplant sites, as in fact De Novo Company has published recently, and get very good characterization which should allow you to make very good guesses as far as what you should implant into humans. In other words, what beta cell mass do you need?

And, in addition, the whole relationship between the beta cells and the non-beta cells I think can be defined in animal graft sites in a way that it'll have a lot of predictive value. The issue of malignancy is really a different issue, which I think you're never off the hook with that in terms of safety. But, at any rate, I just want to make that specific point about diabetes.

CHAIR URBA: Dr. Allen?

DR. ALLEN: I'd just like to address the issue of animals. I think I'm a little surprised. I mean we are worrying about niches and the exact, you know, replicating the disease. And I guess my bigger concern with all this discussion so far is the ultimate in almost all of the experiments we're talking is xeno experiments, the xeno transplants. I've heard nothing today that explains to me why we would not have on the insistence a strong suggestion towards doing allogeneic experiments within species. So, as a veterinarian, I mean I have a potential interest in this because, obviously, my patients could potentially benefit.

But there is at least technologically no particular reason why one couldn't, for example, do a dog or a pig allogeneic experiment. I find it strange that we not consider, if we were looking at

cell-based therapies, we would not wish to assess, we make decisions on the efficacy, potential efficacy of treatments by looking not at what a human cell would do in a goat, but what a goat cell would do in a goat. And it seems strange that we would suddenly abandon that simply, as I said, for expediency.

So much as I don't want to see unnecessary animal experiments being done, I think if a model exists in a large animal that can be tested, for example, a spinal cord injury model, if it exists and it can recapitulate the phenotype of the disease in humans or the injury in humans, that it would be appropriate to at least explore that.

If it doesn't exist, then I think doing a mouse xeno, potentially, experiment is appropriate, but I find it sort of strange that the concept of the two animal rule, not only are we not requiring the necessary two animals, but we also are accepting xeno data and then we're talking about tumors.

DR. CHIEN: No. I don't necessarily disagree with you. But there's one important point here, and that is that, as we've already heard, human ES cells and mouse ES cells are quite different. And I would suspect that pig ES cells and all creatures great and small, if you were fortunate enough to get their ES cells, are going to be slightly different. And the markers are sure as heck going to be different. There will be some that will be conserved.

So you can do all of this and still not answer the safety issue because you didn't do it on the actual material. So my concern was really safety, not efficacy. I'm not arguing you with efficacy. If you want to get there, that's certainly important. But I think for safety, you have to have a way in an animal to test the human material that's going to go into the human being and you're going to have to have that nailed as well as humanly possible.

DR. SALOMON: So there's a lot going on here. The comment I'd make first would be on the animal models.

I think if we're trying to give advice to the FDA on how they might put together a guidance document for sponsors, I think that's why we're here, I think we're going to have to be a little bit more specific about these animal model recommendations.

I think that we first have to ask ourselves, and I think you've picked up on that theme, in your comments is, what is the animal model supposed to be telling us? First of all, sure, it's easy to say that we want to have these animal models that have integrity with respect to human diseases. But the fact is is that most animal models are poor approximations of human diseases, diabetes, the NOD mouse, inflammatory bowel disease, multiple sclerosis with myelin immunizations, and I could go on.

Secondly, immunosuppression, we're going to mouse models and test immunosuppression? I can give seven days of cyclosporine to a kidney transplant and get long-term tolerance. Now, you want me to try that in a human patient? Doesn't work.

Xenotransplantation, talk about complicating life dramatically. I've sat on this committee and others talking about xenotransplantation for years. So I just want to point out the fact that what a sponsor should do with the animal model is answer specific questions in incremental fashion that relates to what they want to do in a human patient. And the idea of expecting sponsors to concoct these really complicated, multifactorial animal models before we'll let them do a human clinical trial, that part I think is wrong.

DR. CHAPPELL: I wanted to bring up two points when discussing phase one trials in question three. But Dr. Chien already started along that road, but maybe we can keep these in the backs of our minds.

The first one is the issue of what incidents of toxicities we are going to tolerate. And I warn you against saying you'll tolerate zero toxicities only, even the most severe ones, because you can do sample site calculations and realize you need -- no finite sample size will demonstrate that. And by saying you want zero toxicities in early phase, what you're doing is you'll maybe see zero toxicities by luck and then in a randomized phase, big clinical trial later on you'll probably find some of those if they're at all common and then wonder why you weren't warned in advance.

Cancer research has gone down that road to great confusion. They define acceptable toxicities in terms of one toxicity out of a certain number in a clinical trial, but those numbers varied and so nobody really knows what the percentages are in cancer vaguely under a three, but it's very vague and there's tremendous confusion.

And I urge you, even though it may seem artificial, to say, to tell us statisticians and others that design clinical trials what percentage would be acceptable, arbitrary as it may be, five percent teratomas, three percent teratomas, et cetera. I mean things work out much more neatly and are much more broadly interpretable if you actually have a number even though you may not want to give us one.

The second issue is surrogate end points. We're placing tremendous demands on these phase one trials, first of all in terms of toxicities. Right? We're not going to tolerate a fifth of people having grade three or toxicities as in cancer. And so we need some surrogate for toxicities as Dr. Chien pointed out.

But, also, the comments so far seem to indicate that we want some evidence of efficacy in a phase one clinical trial, and usually I think of surrogates as the enemy for phase three trial. But for phase one we're going to need surrogates of efficacy and toxicity I think because we won't have the sample size to determine rates of clinically important relevant efficacy and toxicity events.

So you'll hear from me again in the third question on those.

CHAIR URBA: Dr. Gunter.

DR. GUNTER: Well, I think the discussion has been very good and I have to say I agree with most of the points that have been made. I just wanted to pitch in regarding the suggesting about allogeneic animal model, and I understood those to be a model where you would be using embryonic stem cells or derivatives thereof not from the product you're going to be giving to people. And I can understand theoretically why that's important.

But, in the end, I'm afraid concerned sponsors are going to spend a lot of time developing a product that they never intend to give to people and end up trying to justify comparability between that product and what they actually want to give to people. And based on what I've seen about abilities to characterize these cellular products, I think it's going to be hard to establish that comparability, and so data you get from an allogeneic animal model might be very difficult to interpret. So I'm just concerned about that approach as useful. That's my main point.

I would also suggest that maybe we shouldn't expect our animal models to solve -- one animal model to solve every problem. I understand the need for animal models to establish efficacy, and I think in this field we do need to show some evidence of efficacy prior to phase one. But an animal model for efficacy doesn't necessarily have to be the same one you use for safety.

I think the FDA put together a very good background package for us and there I saw pretty good evidence that immunocompromised rodents were actually quite a sensitive vehicle for testing tumorigenicity, and there's some numbers that I have in some of these references. But it's fairly impressive how these models can detect very low numbers of tumorigenic cells.

I agree with the comment that we need to define what a tumor is. There's certainly a need for standardization in the field and that would help all sponsors if they knew what kinds of assays the FDA was looking for.

And, finally, just a suggestion and this may be way off the wall. But in an earlier meeting we had a very impressive group from the National Toxicology Program who discussed a proposal for testing the safety of retroviruses. It may be reasonable to consider their involvement in this problem given the importance of the field in trying to develop some standardization of tumorigenicity assays.

Thank you.

CHAIR URBA: Do you have a specific response?

DR. ALLEN: Yes. Just one. I think just to be clear, my interest in allogeneic models is, I mean apart from anything else, we're looking at data now that have been done with essentially dosing experiments, spiking experiments, dosing experiments where you increase the number of these potentially less differentiated cells and say there's an increased tumor risk. What I'm saying is that some of that science needs to be done in an allogeneic model rather than xenogeneic.

For example, it may be that if you gave a pig a pig embryonic stem cell line that was actually very, very low or highly manageable risk of tumor, and that it wasn't in fact that dose dependent, unlike when you do a xeno model, so I think it's sometimes -- I'm not suggesting the burden all falls to industry. What industry has to do goes in parallel with what science has to do. But within that framework I think there's an enormous amount of information.

I just find the concept of ignoring a xeno environment strange. I mean I think we have to get information on an allo environment as well and use that to our best. And I think, say, just confirming the fact this risk of teratoma and these other things really exists in other species when it's not a xeno environment I think is really important.

CHAIR URBA: Dr. Taylor?

DR. TAYLOR: I think it's important that we remember, although we're talking about embryonic stem cells here, there are other stem cells that have already been down many of these pathways and that a number of these questions have been answered for some of those in terms of whether or not it's appropriate to do allo or xeno testing. And so I think we need to learn from what's already been done and only make different recommendations if we feel the risk benefit ratio is significantly different.

And in the situation where tumorigenicity comes up, I think that that's going to directly relate to the product involved and how well the product involved has been characterized both in vitro and in vivo. We know that differentiation directly relates to tumorigenicity. That the more differentiated a cell type is and the purer a cell type is with regard to differentiation, the less likely it is to be tumorigenic.

So I certainly think that this can't be dissociated from the cell product and the initial composition. That maybe, in fact, we set more rigorous guidelines for the composition and understanding the product and the potential benefit of that product, then define a few safety assays, tumorigenicity assays that have to be employed, just like they do now for other cell types, and then go forward from there.

DR. SNYDER: Yes. I just had a quick followup on the animal model comment. That certainly when we're working out the biology, at least in neurobiology, we always stay within the species to work out some of these things. And the only time we tend to move into using human cells is, as Kurt and Ken said, when you're actually dealing with a product that might go into humans.

The one part about having good, predictive animal models comes down to the realization that the road of clinical trials is littered with failures that were based on jumping from an unfaithful animal model to humans. Brain tumor therapies come to mind as one glaring example of where the clinical trials simply failed because they were based on poor models of brain tumors and we're only now getting to derive better animal models.

So I think that even though we're all -- certainly animal models for safety is one thing. Animal models for efficacy I think have to better reflect what we think we're going to anticipate seeing in the humans. Otherwise, we'll get into clinical trials that'll simply fail and give the stem cell field just a black eye, I fear.

DR. FRIEDLANDER: I think a lot of very important points have been discussed quite eloquently by most my colleagues. I'd like to just emphasize and go over a few that came up.

First of all, I couldn't agree more that I think it needs to be indication specific. That depending upon the disease you're looking at and the cell source you're looking at, what you're expecting those cells to do, the bar is going to be very different.

So, for example, we might have a much higher tolerance for some side effects in someone who is going to be dead within two months or a year in spite of the treatment than something which is a more chronic disease and might threaten some vision for example. I think the micro environments are very different.

For example, if we take cells from the progenitor cells we work, we put into the eye, they don't proliferate, they do certain things. You put them into a model of brain tumors, they proliferate as he says by Ki-67 staining. So, again, depending upon what you're looking at, I think you can have different expectations and different results.

In terms of the animal models, I agree. I think it's very important you have something that's physiologically relevant, that's disease appropriate. However, there are many instances in which you can best perhaps expect proof of concept. So, for example, rodent eyes, for example, are intrinsically resistant to ischemic damage. Yet this is the vast majority of diseases that blind you do so as a result of abnormal angiogenesis and reflective ischemic conditions. So I think you have to be careful about what kind of a bar you set there in terms of how rigorous that's going to be.

I think in the consideration we discussed a little bit about the size of the animal eye and we can get by by doing rodents and is it really necessary to expand to larger eyes. So let me give you another example. When you inject tiny volumes into tiny eyes, much of it comes back out again. You get reflux. So if you're looking at dosing or you're trying to get a sense about how effective the population is, if three-quarters of the stuff refluxes out, you really need to think about another model. So, again, I think depending upon the disease and the indication you're looking at, larger eye models might be more appropriate.

I think depending where those cells wind up, again, the micro environment, they will spread and proliferate and do things very differently than if they're localized in another area. And, finally, again, I think the eyes is where most of our experience is. If you're proposing putting something

sub-retinally into an eye that's diseased, that already has extensive degeneration as opposed to something which is normal, you're going to see very different results also.

So, again, I think the appropriateness of the model you're using become very, very important.

CHAIR URBA: Dr. Gerson?

DR. GERSON: I'd like to make an observation comment and then a suggestion, specifically in the area of the issue of benign and malignant teratomas.

First of all, in the presentations this morning, I was struck with the interest in sort of understanding the tumorigenicity potential of the ES cells and the differentiated cells. I did not come away from that with the impression that there was a desire to instill tumorigenic ES cells in any setting regardless of the disease characterization.

So I think we would -- I would suggest that we sort of reinforce that. I think that there are ways to inch into this field therapeutically managing risk benefit with an emphasis at the start on safety and require, as best we can, that we not encourage therapeutic trials with any intention of tumorigenic cell infiltration. And there are ways to manage that.

One is to make sure there is no ES phenotype by BCR of gene expression. I think we have some pretty clear data on that. I would implore the field to develop a single, unified approach to a tumorigenicity assay, should be reasonably straightforward. I don't really care what it is, where it is, whether it's right or not because it never will be right. But if there was a standardized assay, so on a relative basis one could assess the risk of tumorigenicity, I think that would be wise.

And I think one can revise downward that stringent approach as there's experience with differentiated ES cells over time, but to start with a less stringent approach in my mind doesn't make any sense.

CHAIR URBA: Dr. Goldman.

DR. GOLDMAN: This is just a followup to Dr. Gerson's point really, and it really picks up from Dr. Chappell's before.

I think we need to distinguish between toxicity accruing to inappropriate differentiation and then disease-dependent toxicities, a la the interaction of those inappropriately differentiated phenotypes in the disease setting versus toxicity accruing to tumorigenicity. I think we can reasonably look at inappropriate differentiation toxicity in the same way we look at any treatment associated risk and simply assess the cost benefit by virtue of that patient's condition or that group of patients' conditions.

But the issue of tumorigenesis from ES I think is more fraught with really fundamental dangers. We don't work in a vacuum and having any malignant cancers generated from inadvertently introduced, still undifferentiated ES in patients would be to my mind disastrous for the field and it would take very few patients suffering cancers from ES transplants to potentially put a really significant damper on the field potentially for years. And so I'm really looking at it from that standpoint as much as the precise balancing of cost benefit in suggesting that we have a higher threshold for potentially accepting toxicity due to tumorigenesis from ES rather than the toxicities where we're

vis-a-vis inappropriate differentiation, which essentially is just what Dr. Gerson just said.

CHAIR URBA: Would you identify yourself?

DR. MCDONALD: Yes. Dr. John McDonald, Johns Hopkins. I really wanted to clarify two issues. One is embryonic stem cells being similar, as other points have mentioned and our approach needs to be similar as well and I think regulating this. And the second one being a comment on primate, on modeling in the use of primates.

So first of all, I mean obviously for political reasons embryonic stem cells are different. But, biologically, every cell that we're starting came from an early embryonic stem cells and progenitor, and the same regulatory issues that would apply to any cell being transplanted, whether it be derived from a late-stage progenitor or a late-stage progenitor derived from ES cells, should also be similar. There's not a great need to be treated dramatically different and that would be a mistake, in my opinion.

But we do need to be aware of the risks in the field, and Dr. Goldman's comment I believe strongly in, too. For example, it would be relatively easy to get approval to do a single patient with a drug, or stem cells, if you're a doctor a or surgeon at a hospital. And I think there we need to be very stringent, because all it takes is one negative. Even whether that patient was going to die from the disease in two months anyways, a negative there, something going awry would cost the field a tremendous amount.

So I think we just need to be careful, as Dr. Goldman mentioned in the cost benefit analysis, just we may -- keep the same stringency across, and the stringency and regulation of ES cells in medical usage should really not be much different than the regulatory approaches that we have with other cellular forms of delivery that have been commonly used over the last 20 years.

My second comment is on modeling. And as many of the speakers have brought up, modeling is an issue, because, it's inherent in the a name; it's just a model. But we've seen over history also the use of primates, even to appropriate, you know, approximate the size of what you're trying to treat, like your comment about the eye. Oftentimes, you know, it's really just a band-aid, and it's not getting us the data we want, and you could figure out from the beginning that it didn't give us the data we want.

So we don't want to impose barriers to progress forward just simply because they've been put there before. But I think, traditionally, a lot of the requirements to go to larger animals hasn't given us the data we want, and predictably wouldn't give us the data that we want, and it sets the bar higher, but it also can set a false barrier.

And the great example of this is the Parkinson's disease transplantation trials, where adverse events were seen with transplantation with dyskinesias and so forth that weren't observed in primates. Well, it turns out that the dosage that they gave in primates was so substantially lower, you know, because they couldn't get enough cells to do the transplantation. Examples like that, you know, I think we can look back and see many of them in our field.

Now, it made everyone feel better to do those primate trials, but it didn't help at all in terms of predicting. In fact, it gave a false sense of security.

CHAIR URBA: Can you go to the microphone, please?

DR. ISACSON: So you're actually right in point, but wrong in fact. So the primate model would not give an L-dopa, so that didn't have any prior dyskinesia. That was the reason they didn't give the appropriate answer.

CHAIR URBA: Did you know that?

DR. ISACSON: Cell dosage was not an issue. They were given a very large transplant, but they didn't have any prior L-dopa exposure, which is what's the difference between the real clinical situation.

DR. MCDONALD: So what was the difference in the dosage -- what was the dosage range used in primates versus the dosage range used in humans?

DR. ISACSON: Similar.

DR. MCDONALD: What were they?

DR. ISACSON: They were three to four transplant per case. But the primate may not --

DR. MCDONALD: How many cells, it's the cells, number of cells? This is kind of getting at the issue of dosage. Is dosage number of transplant sites, is dosage number of cells, is dosage number of surviving cells?

DR. ISACSON: Obviously, all of the above, but the difference was really that -- and we have actually used the primate model with L-dopa, which is now giving us answers. So it's not actually an absolute version of what you said. With L-dopa you get dyskinesia in the primate model, and then you can test -- it's the same side effect of transplants that you see in the patient. So just a point of clarification.

DR. MCDONALD: Well, thanks for clarifying that fact. My point had nothing to do with the dyskinesia; that was just an example. But I think you bringing that up pointed out the issues of just simply dosing across species and size is substantial --

DR. SNYDER: Well it actually brings up more about the modeling. So what Ole was mentioning then, the primate model from which the human trials was launched did not accurately reflect the human population. The human population typically is started on pharmacology first, then gets the transplants.

Now, the monkeys were also started on pharmacology, then got the transplants. Now they do show the dyskinesias, which would have predicted what we would have been seeing in the human population.

DR. MCDONALD: All right. Thanks.

CHAIR URBA: Specific response for that?

DR. CHIEN: Yes. I just wanted to come back, because I just wanted clarify one point, and that is is that we would not be asking, if we were to ask for non-human primate data for human ES cell safety issues about, are the cell preparations you're going in with, do they have any residual cells, or any residual potential in a primate situation in normal, okay, to form a teratoma? To me, that seems not unreasonable. It's not that much different from what you have to do if you're going to have a humanized antibody, and go into the clinic. Okay? I mean, what's more dangerous: monoclonal antibody, or human ES cell-derived products? It's a no-brainer, okay? You should at least do what you're going to do with monoclonal antibodies therapy.

I don't think it's asking -- now to ask to create a whole humanized disease model in the eye, or whatever else it is that you do, and then to get the ES cells from humans, and then to show that they would have efficacy, that's another issue. We're talking about a standardized tox assay. I think this is not asking too much.

CHAIR URBA: Dr. Woo?

DR. WOO: Yes. I thought this morning's presentations were very informative, and this afternoon's discussion is very, very stimulating.

My comments really has to do with the product and the potential tox. From what we hear this morning is that, in some of these trials that are going forward, we're hearing, 95 to 99 percent purity sounds pretty good, and we can go ahead and do the transplant, and so on, and hopefully the other one to five percent of cells is going to be just bystanders, and even if they're inappropriate cells, they may not cause any particular pathology.

And yet, on the other hand, I'm hearing that the cells, these undifferentiated human ES cells, will have some problems in chromosomal instability, and there has been observation of aneuploidy when you do the chromosome spreads, and so on. Well, to me, aneuploidy, that's a hallmark for tumorigenicity. And then we hear there's intrachromosomal deletions, and this chromosome and that. Well, if you can see intrachromosomal deletions in the finite number of cells that you look at in the metaphase spread, and we are talking about transplanting millions and millions and millions of cells out of which one to five percent, we don't know what they are. That, to me, seems like a very -- it's a condition that is conducive to the development of teratomas and teratocarcinomas in the recipients.

So I would really urge the field to spend much more time trying to purify these cells before really going into "translation of medicine" in too fast and too big of a leap. To minimize the formation, like Richard says, you know, you can never minimize everything to zero, but the best we can do within the technical capabilities is certainly something that we should try before we go into transplants.

And then, of course, how do we know that we have minimized the risk? Then I think Ken Chien made a point that I totally support, which is, we need to develop assays to know what we're dealing with, whether it's gene profiling, or whether it's secreted products, or for cholesterol, whatever it is.

(Laughter.)

DR. WOO: And so, finally, it seems to me that every sponsor can potentially be dealing with a human ES cell line that is proprietary, and that means each company is dealing with a product that is completely different from one another. So for every one of those products, it is the responsibility of the sponsor to demonstrate to the FDA that they have purified the products, and they have developed assays to determine that this is the threshold of cells that will not form tumors in the animal model, whatever animal model that is appropriate, and then the maximal dose of the transplant is not going to exceed the number of cells that will form tumors in the animal models. I think this kind of criteria will have to be set before we just jump into human patients and say, let's do it, and hope for the best.

Thank you.

DR. GERSON: I just have a quick concern. There was discussion earlier about animal models in the context of toxicity and efficacy, and I think that's very appropriate. I'm a little bit concerned about the results of a negative animal model that's designed to look at the tumorigenicity of these cells in a specific disease entity because you're much more likely to have negative data without any standardization. So I would argue that there should be, if anything, a standardized model for tumorigenicity so that, on a relative basis across the field, we can assess what's going on.

CHAIR URBA: Dr. Chen, from the FDA, you've been patiently waiting.

DR. CHEN: Okay. I heard there are many good suggestions there, but there are still some points that the panel haven't covered, which is the study duration. I mean, you have animal that you suggest is rodent, or a large animal. Does the study need to be carried out to the extent of the life span of the animal, and how long are you supposed to follow them?

DR. MCDONALD: Okay. So I'll address that issue. But first, I want to make a comment.

You know, so we spend a lot of time talking about tumors, as an example. But just to stimulate some thought related to this question, let's talk about graft versus host disease, and host versus graft disease, which is, in my mind, a much more likely, huge negative outcome than tumorigenicity. And, you know, there are standards for dealing with this.

If you want to look at a cost benefit analysis, anybody who's gone to medical school, and has been on one of those units where they do bone marrow ablation to get rid of a solid tumor, and then replace with other things, just making it through that is an unbelievable negative. And then to only have to succumb with, actually, a fair percentage of people to graft versus host, or host versus graft disease, which can come in many forms, shapes and sizes in terms of severity and onset both in terms of time, you know, being subacute, subchronic, or very long term, and yes, so I think some discussion on how long should begin with looking at how long do we already do.

What is the standard in the FDA now for cell-based therapies? Let's take a look at bone marrow replacement. How did they have to demonstrate in an animal model that things didn't go a negative? And then we can