activity of these cells.

In vivo, you need to look at a model where these cells will survive. Can they survive the delivery? I think it was mentioned in Jonathan's talk. And, also, are there going to be immune responses to these products?

I can tell you that this is not necessarily an easy feat for our programs. I think it took us about a year to a year and a half to establish a robust system where we could get long-term graft survival, and

long-term meaning greater than nine months and to a year. So I think it's important to selecting a model so that if you're looking at these cells, that you can see that the cells, in fact, survive.

An important feature, also, is looking at the phenotype, looking at the phenotype in vitro and looking how does that translate in vivo. Did they maintain the same phenotype? Do they mature over time? Are they proliferative when they're in vivo in their environment?

Also, then, looking at the activity in disease models, looking for clinical efficacy, looking for their histology. Can you find evidence of these cells and are they having any particular effect?

For instance, in this particular case, we're looking at a spinal cord-injured animal and we're looking in brown for a human nuclear antibody in EC4, area chrome cyanine, which is Stan's myelin. You can see in animals that are either injected with vehicle or no injection for that matter, you can see these large cavities that tend to develop in the spine cord-injured animals.

But that in animals injected with GRNOPC1, the oligodendroglial progenitor cells, you can see brown, human cells that are filling the cavity, and that if you look very, very closely, and here's a high magnification of the center of that cavity, you can now see that there are myelinated fibers traversing that lesion site.

So it's important to look at clinical efficacy, histologically what cells are doing. What are the doses and how does that translate into the human equivalent dose? What is your delivery site in volume? What kind of volume of cells are you going to have to inject in order to meet your targeted efficacy? What kind of implications does that have for the number of injections that you might need to give? And then finally the timing of the treatment in relationship to the disease process or the injury process.

Next question is where do the cells go? That has both safety and efficacy implications. You need to look at the sensitivity of your assays to detect particular cells. Do the cells go to the site for intended activity? Are there sufficient numbers of cells at the site in order to elicit efficacy? Is there distribution outside of the target site? And is there migration at the local site? And looking at extended periods over time.

This is just some examples from one of our biodistribution studies where we looked by PCR for looking at tissues throughout the body and also by immunohistochemistry to look at cells within the spinal cord. And here we looked at, in this case, the distance from the most proximal, the most distal cells in the spinal cord as a function of time and dose.

So in this particular case we gave animals either 2.4 x 10⁵ or 2.4 x 10⁶ cells, in alternating 2 x 7⁵, 2 x 7⁶. And we looked at two days, 14 days, and 180 days post transplant. And you can see there that these cells do migrate, and by six months post transplant we can see that there is about a 16 to 17 millimeter distance between the most caudal and the most rostral cells injected into the spinal cord.

For the toxicology studies, one of the big considerations is where do the cells go so that you can start looking for toxic activities. Again, some considerations are looking at the dose of the product, looking at what your tox model is. Again, we've looked at, in our particular case we looked at a spinal cord-injured rat for many of our toxilogical considerations just because we want to see what the cells do in that scenario that probably is the closest model where we can get good long-term cell survival to see whether these cells have any toxilogical effects.

You really need to look at the feasibility of the model. Can you get long term survival of your whatever model you're choosing? The duration of the studies and looking at the duration of the human cell survival, obviously, it's really important to demonstrate that your product, if it's required to be there for a long term, that it is there for a long term.

For our particular program we looked at the toxicity of the delivery. We looked at animal survival, clinical observations, any evidence for systemic toxicity, hematological coagulation parameters, clinical chemistries in detail in both macropathology and micropathology to look to see are there any adverse events associated with the injection of the cells. And then also on some very, very, very -- activities that are specific to our product, in this case the evidence for allodynia.

In the four tumorigenicity studies, again, I think we've heard a lot about that this morning already. But, obviously, you're looking for not only teratomas, but also evidence of ectopic tissue. Obviously looking at the local injection site and distal sites to see, again, where the migration patterns are of the cells. Do you find any evidence of teratomas?

There are some challenges to these studies because in an ideal world you would love to be able to inject the human clinical dose in these models. And if you're looking at, again, as I think Melissa said, that if you're putting these cells in any animal model, it's a xenograft and you want to look where you're going to look at the most boast out of the cell survival is optimized. So, again, looking at long-term cell survival.

You'd like to look at large numbers of animals, not just one or two. You'd like to look at large numbers of animals so you can boost your statistics.

You want to mimic the human clinical setting as best as possible. It's not always feasible, but you want to take that into consideration at least when designing the tumorigenicity studies.

There are large animal models. For my perspective at this point, the large animal models for looking at tumorigenicity studies can be quite a challenge due to, in most cases, the absence of immunocompromised animals, large animals. It requires very long and very laborious xenografted immunosuppression, which I think can be a challenge for looking at the tumorigenicity of any embryonic stem cell-based therapy.

There's also some suggestions, that have been suggested to me anyways, about looking at homologous embryonic stem cell systems. Looking at, for instance, taking your protocol, apply it to a mouse embryonic stem cell, put that into a mouse and see, in fact, do you get teratomas, do you get tumors?

Unfortunately, the protocols that we use for differentiating human embryonic stem cells just aren't applicable to the fine detail that we need for our human embryonic stem cells. And so what you would see with a mouse embryonic stem cells are just not necessarily representative of what you would see with humans.

There are some lessons learned, that -- through our studies -- that we have learned through, again, through the execution of the studies, and that there are many important factors in teratoma formation. For instance, we're looking at embryonic stem cell number. In this case, the more cells you put in, the more changes of having a teratoma.

In this case, we've looked at, we've taken our product and deliberately spiked it with undifferentiated embryonic stem cells. We held the total number of cells constant at 2 x 10⁶ and injected them into the spinal cord, in this case of immunocompromised mice, and assessed for teratoma formation 12 months after.

So what we did was a graded number ranging from 50 percent undifferentiated cells all the way down to actually zero percent undifferentiated cells. What you can see here is that there is this curve that anywhere below one percent we actually did not see any evidence for teratoma formation in the spinal cord and in the spinal cord of these animals. Actually, in this particular experiment, we also did not see any at five percent.

Five percent seems to be about the threshold at which we start seeing teratomas. So this is the undifferentiated cell number in this context of -- our product is something that is important for establishing the formation of teratomas in these animals.

We also know that the site of implantation is important. We find, for instance, in the spinal cord, that there's over a order of magnitude grave sensitivity for teratoma formation than when looking at an intramuscular injection. So the site of delivery of your product is very important.

And, also, the cell aggregation state so that clump cells tend to form teratomas at least more quickly and with a slightly higher frequency than cells that are single cells.

Other studies to take into consideration are allogenicity studies. Is immunosuppression required? And, if so, what is that duration of immunosuppression? I think that most common practice or common logic would suggest that these cells are going to be rejected. But there's snippets of information out there that suggest that they might not be rigorously rejected and that they might in fact reduce the required reduced immunosuppression or reduced irration duration of immunosuppression.

Some of the challenges, again, are that this is very hard to measure, very hard to measure in an animal model because, again, the embryonic stem cell-based products are xenografts in all animals requiring extra rigor to get their survival. There's also, it's very difficult to look at the allogenicity of maturing cells in vivo in these animal models, again, because they are, in this case by definition, a xenograft model.

Humanized models are challenging, and challenging to get the tissue, and challenging with respect to determining allogenicity because it's not necessarily clear that all elements of the immune system are in fact recapitulated in these models with the type of activity that you might expect to see in a human. And so in many cases we're left with in vitro analyses.

Some of the approaches is to look at, again, the allogenicity in vitro; looking at trying to recapitulate in an inflammatory site in in vitro culture; looking at removing immunosuppression in your xenograft model, see if cells survive; and looking for ways to see if you can monitor in the clinical setting engraftment, monitor injection. And, more importantly, what are the effects of rejection? Are there any adverse events in case these cells do, in fact, reject?

A couple of other things I'd like to discuss, and that's the cells for the delivery. Obviously, you have to decide whether it's going to be a local delivery or a systemic delivery. What is going to be the site? Is it going to be an intraoperative delivery? What is going to be the rate of delivery? Is it a one-time administration? Is it multiple times of administration? What's the skill set required for the delivery? And a delivery device, do you need a delivery device?

For our application, again, we, in fact, did design a, what we call a syringe positioning device, because we're looking at injecting these cells over a period of a minute or two in the spinal cord, and one of the things we wanted to do was to make sure that we could stabilize the syringe for neurosurgeons during the actual delivery of this particular product. And, obviously, there's training that's required.

So the last thing I'd like to talk about are the design of the clinical trials. And the design of the clinical trials, again, there's the most important consideration is patient safety. And it's really important to have consideration and interactions with a plethora of different experts in order to design these clinical trials. We really need to, in this case we've looked at, having a steering committee, a data monitoring committee, an outcomes committee so that we can define how these trials and what are we going to be considering for safety and efficacy, in this case, looking with our neurosurgeons, who are the group that are going to be delivering these cells; the radiologists who will be, in fact, monitoring these cells and the progression of these cells, and, hopefully, looking for and not seeing, but looking to see if there are any adverse events at the local sites.

Obviously, with the investigators and the ESCRO committees and various committees to look at -- and IRBs who are going to be reviewing the protocol. So the idea here, again, is to look at the protocol, the logistics of the trial, minimize potential risks. How can this kind of therapy actually be established and maintained in the normal standard of care for your particular patient?

Our proposed phase one clinical trial would be an open label trial that would be looking at treating patients with subacute, functionally complete T3 to T12 thoracic lesions, injuries, transplanting anywhere from seven to 14 days with a temporary immunosuppression. Again, the primary endpoints are safety.

Secondary endpoints are looking at the feasibility of the multiple outcome measures that we're anticipating in the trial. We're proposing frequent short-term and

long-term MRIs, again, to look at the site to see if we're seeing any adverse or positive effects at the site, and looking at a variety of different immunological monitoring.

So I think I will conclude there just to say that there are numerous considerations in developing an embryonic stem cell-based therapy. Some of the questions are actually common to all cell therapies. Some are more specific than others to human embryonic stem cells. That the specific

non-clinical study designs really need to be designed with their clinical application in mind. That the clinical trials designs do require intradisciplinary input and they need to be designed with frequent monitoring involved.

Thank you.

(Applause.)

CHAIR URBA: Thank you. Any questions for Dr. Lebkowski? Dr. Allen?

DR. ALLEN: You mentioned that in the context of long-term safety testing and particularly large animals that it was potentially possible to generate ES lines from these animals, but that your primary concern were that they're manufacture essentially differently, the process differently. I mean I think that this is something that we struggle with with any cell-based therapy.

So I guess my question is, other than that fact, are there any concerns? I mean, if you can demonstrate that your differentiated ES cell, large animal let's say, but essentially animal ES cell functionally behaves similarly, has phenotypic features similar to those of the natural cell, is there any reason not to test that in a large animal model?

DR. LEBKOWSKI: I think there's no reason not to test it in an animal model. But I would add one more criteria, that we have to show that the composition is equivalent, and we need to have the right markers to show that the composition is equivalent.

DR. ALLEN: But, essentially, if the things like, for example, spiking experiments, you could potentially do spiking experiments and confirm that what you see in a mouse with a human ES cell you do or do not see with, for example, a dog-to-dog or primate-to-primate.

DR. LEBKOWSKI: Yes. That's feasible.

DR. ALLEN: Okay. Thank you.

DR. LEBKOWSKI: As long as, you know, the question is is how well do the undifferentiated human embryonic stem cells survive in your large animal model because that could be a question too.

CHAIR URBA: Dr. Calos?

DR. CALOS: Jane, on the teratoma issue, you're scoring teratomas and are you using that strictly to mean benign, or are you also scoring malignant tumors?

DR. LEBKOWSKI: When we have looked at these studies, they're benign teratomas, okay, so they're well differentiated.

DR. CALOS: So what is considered the issue with ES cells if they're always benign teratomas? Do you also get malignant tumors?

DR. LEBKOWSKI: In our experience, if you take enough undifferentiated cells and you put them into a very vulnerable site, okay, you can see cells that I wouldn't necessarily call them malignant, okay, but they will look undifferentiated, okay. So we haven't characterized them as to whether they are malignant or not, just haven't looked at that. But the question is, you can still see cells that look undifferentiated. Again, that's putting lots of undifferentiated cells into a very vulnerable site, okay, for instance, the brain.

The second things is the issue with benign teratomas are, you know, what are the consequences of that? Do they continue to grow uncontrollably? And can you remove it? If you can't remove it, what happens? Can you remove it? You know, what are the consequences of that? So those are the issues that we're dealing with.

DR. GERSON: You showed us a variety of data of cell dose, but I didn't see data on dose effect or dose range. Could you give us some sense about how you approached that?

DR. LEBKOWSKI: Dose range for teratoma formation or dose range --

DR. GERSON: No, you showed us percentage of cells in a fixed dose. And in the migration study, you showed us essentially no dose effect. You showed us a time effect.

DR. LEBKOWSKI: Yes.

DR. GERSON: Is there a dose effect and are you anticipating studying dose effect?

DR. LEBKOWSKI: Yes. We can see dose effects on teratoma formation even when we look at not spiking but using just plain, undifferentiated cells and putting them into animals. We can see a dose effect on teratoma formation.

With regards to any toxicological parameter, for instances, even allodynia or anything like that, we have not seen. We've looked at two doses in many of our studies and have not seen any effect of that.

DR. GERSON: Is there a therapeutic dose effect?

DR. LEBKOWSKI: The answer is no. We have not seen any therapeutic dose effects yet, but we might need to actually go down in dose in order to see that.

DR. TAYLOR: Along those lines, you just talked about if you put lots of cells in a vulnerable site you're likely to see, potentially, a tumor, be it benign or otherwise. And there are data in the literature that suggests small numbers of undifferentiated cells can form malignant tumors depending on the site.

We've also heard from you and a number of other presenters that we don't have a good sense of where cells migrate, how cells migrate. You have data that cells do migrate. So how do you account for and begin to control whether or not cells are reaching "a vulnerable site" and how many cells are? And then, finally, in terms of the homologous cell studies you were mentioning, there is continually a mixture of differentiated and undifferentiated cells present, so how do you begin to account for that in this broader context of tumorigenicity?

DR. LEBKOWSKI: Okay. So let me be a little bit more specific. For instance, if we take and try to inject two million undifferentiated cells into a rodent spinal cord, intact rodent spinal cord, you will see cells that not only have benign teratomas, well differentiated teratomas, but you will see some cells that look still undifferentiated, and that's the target site that we're looking at.

I think the question is, from your biodistribution studies, you need to look at where, for instance, the cells go and to see whether there are vulnerable sites that can, in fact, be -- that are populated by cells or in potentially undifferentiated cells. So, for instance, in the biodistribution sites, I didn't show you all the data, but we've looked at not only where the cells go upon injection into animal that has a spinal cord injury, not only looked at the spinal cord, but we've looked at essentially every tissue in the body by quantitative PCR to look for, are there cells present, and they aren't.

DR. TAYLOR: Let me ask a followup on that. In terms of a location like spinal cord, which is geometrically constrained, and brain, which is geometrically constrained, and heart, likewise, have you noticed adverse effects by injecting even small numbers of undifferentiated cells when there is the potential for compression being an issue, for example, even with a benign tumor?

DR. LEBKOWSKI: Yes, yes, I mean if you have enough undifferentiated cells in your population. So like, for instance, in the data that I showed, if you have 100,000 undifferentiated cells in the population that we deliberately spiked into, for instance, our candidate population, OPC1. We can see teratomas in there, okay. And if we've titrated down, we don't see those.

DR. TAYLOR: The argument is that teratomas are benign. I guess my question is, in some situations a tumor is a geometrically unfavorable condition, be it benign or otherwise, have you accounted for that, how do you account for that, and --

DR. LEBKOWSKI: Well, we're looking for any kind of adverse events. And the question is, do we see a benign or any other kinds of tissue in there that would be a problem?

CHAIR URBA: Dr. Chamberlain?

DR. CHAMBERLAIN: I'm curious. In a human clinical trial where you're delivering to a relatively inaccessible site like the spinal cord, how do you monitor cell survival? And is there any consideration given to tracer studies or putting in marker genes like luciferase or anything like that?

DR. LEBKOWSKI: We have not looked at putting in any marker genes in there, but your question is quite relevant. It is going to be difficult. We are going to be proposing to do some experimental assays to look for whether the cells are there or not, but it is, actually, a challenge, and looking for the appropriate marker or the appropriate tracer that can be used to monitor those cells, that can be used clinically.

DR. GOLDMAN: It's a fault, but to the point with regards to benign teratoma versus malignant teratocarcinoma, that's essentially a histologic definition based on degree of anaplasia. But within the nervous system, any neurologist or neuroncologist or neurosurgeon will tell you there's no such thing as benign tumor within the brain or spinal cord. Anyway, that's a comment.

But I wanted to ask essentially the same question that I did of Dr. Dinsmore. Are you sorting up front or is the assumption that the population you're starting with is essentially purified with regards to aligoprogenitors?

DR. LEBKOWSKI: We do not do any sorting by any flow cytometry events or anything. It's strictly driven by the differentiation protocol.

DR. GOLDMAN: So this thing segues into the spiking experiment. I'm just wondering about that. If I understood that graft correctly, when you get to roughly five percent spiking, you start to see approximately ten percent teratoma generation?

DR. LEBKOWSKI: Yes.

DR. GOLDMAN: That's out of two million. So that means that your threshold for teratoma generation is effective 100,000 undifferentiated ES?

DR. LEBKOWSKI: In this model system, that's correct.

DR. GOLDMAN: So that's what I'm getting at. The model systems, in the typical -- in an uninjured model, the cell number's much lower than that, well, of course, because of teratoma?

DR. LEBKOWSKI: Yes, that was a uninjured model.

DR. GOLDMAN: That wasn't an uninjured model.

DR. LEBKOWSKI: On an injured model, yes.

DR. GOLDMAN: So then I'm wondering then, is it --

DR. LEBKOWSKI: We've done both.

DR. GOLDMAN: Is the cell survival less than you think it is or are there paracrine effects within the population such that the undifferentiated are being suppressed by differentiated? Because I'd be very surprised that it takes 100,000 cells to get a teratoma or at least any incidence of teratoma.

DR. LEBKOWSKI: You know, all I can say is what we deliver. We have monitored for human cell survival in all of the animal models for all of the tumorigenicity and for all the toxicology studies, and we see them there. We see them in good numbers.

Can I say that, you know, 100 percent of the cells that I have delivered are there? I can't say that, you know. There's just not the tools to be able to answer that question.

CHAIR URBA: Two last questions.

DR. CHAPPELL: I'd like to follow up on the previous discussion. You originally presented the results that we've been discussing in terms of percent.

DR. LEBKOWSKI: Yes.

DR. CHAPPELL: So five percent or more undifferentiated cells give rise to detectable teratoma levels. But we've converted those to 100,000, which is five percent of two million.

DR. LEBKOWSKI: Yes.

DR. CHAPPELL: So I'm wondering whether it's the absolute number that you think is more relevant. So would it be 100,000 with 1,900,000? When your experiment, would that be equal to 100,000 alone, or 100,000 plus five million? Or do you think it's the percent that's more relevant? Or do you have any evidence to answer that?

DR. LEBKOWSKI: We have evidence to say that it's probably the cell number. We've done some studies, but not as many as in -- not as many animals. Looking at just undifferentiated cells alone injected, graded numbers of undifferentiated cells in the spinal cord versus spiking them into, for instance, the product, and we see about the same thresholds. So that is data suggested that it's probably not necessarily what's in there with it's buddies, okay.

On the other hand, I am a little cautious about making that conclusive remark yet because we do know that there are states of aggregation. Other things can influence how these cells and how they form teratomas.

CHAIR URBA: Last question.

DR. FRIEDLANDER: So in assessing risk benefit, one likes to think there is always some benefit to balance the risk here and we didn't talk a lot about that. So I'm just wondering, when you were injecting cells like these oligodendroglial progenitor cells, I presume you're hoping these are going to differentiate into glial cells or oligodendra sites, myeline damaged nerves. But presumably there's a secondary or paracrine effect of these cells also. How do you -- I mean it's fine in the tissue culture dish to look at this, but how do you do this in vivo? How do you assess these sorts of issues?

DR. LEBKOWSKI: I mean, one of the things that we've looked at extensively is looking at their effects on behavioral activities, for instance, the BBB score, but also looking at histological. I just showed one little picture. But looking histologically at what these spinal cord injury sites look like, okay, and seeing sparing of tissue and looking at reconstitution or myelinated fibers going through these sites, so it's a combination of factors.

CHAIR URBA: Okay. We've reached the time for the morning break. I'd like to thank this morning's speakers for their great presentations, for staying on time. I'd like to remind the committee members that during the break we ought to not discuss the topics that will be held this afternoon because we all want to hear everyone's comments. And we will reconvene here at 11:20 a.m. Thank you.

(Whereupon, the foregoing matter

went off the record at 11:08 a.m.

and went back on the record at

11:25 a.m.)

CHAIR URBA: So we're ready to begin the second part of the morning session. I'd like to start with our next speaker, Dr. Isacson, Director, Center for Neuroregeneration Research, McLean Hospital/Harvard Medical School.

DR. ISACSON: Don't want to deprive the member of coffee.

I want to thank as an academic and invited guest, the FDA and CBER and CGT for this terrific exploratory meeting which I think is unique to us as a country and gives us an opportunity to really air the issues. And the briefing document that was described by Dr. Bauer to begin with I think is a very good start.

So what I would describe to you today is essentially a set of experiments and experiences over the last 15 years in which the desired cell type is actually one for potential use in Parkinson's disease. And I will speak to you mainly as a scientist, but also as someone who has participated in clinical trials.

I wanted to start by showing my opinion from current and previous work about this field, in particular because I saw in a previous slide the word pharmacology show up, and I do think that when we are discussing stem cell therapy we really are discussing live cell therapy and they are really, specifically issues when you use investigational new cells as opposed to the standard IND. And, obviously, we already talked about what is the reasonable way to evaluate stem cell derived cell therapy compared to the risks inherent in such therapies.

And, finally, or almost finally, I want to discuss with you the ability in each of these models, be they muscle or diabetes or neurological disorders, how we generate a desired cell type just like we would do in pharmacology for a desired substance, and how we define that stability. And then to reiterate that the animal model or models that we use, how are they going to tell us both about benefit and safety data as the previous speakers also brought up.

Now, in Parkinson's disease cell therapy has been considered for a number of years. The reason being that in the mid brain, the base of the brain of a Parkinson patient, the cell that's most vulnerable and actually produces the signs and symptoms of Parkinsonism died. They are known as dopaminergic neurons.

And in a paradigm in which the basic idea is that since drug therapy fails over time, the replacement of such neurons producing dopamine, here you see a patient from our trial in Canada that participated in, which here you have the patient prior to transplantation with a PET signal for fluorodopa and here you have three years after transplantation to one side that restoration of signal indicates that the cells produced dopamine.

And they do so and they grow so in a very normal phenotypic way. In these transplants here are actually four or five tracts and they grow very specifically to create this dark substance, which are dopamine terminals, in the unplanted side. There are very few fibers left and the threshold for diseased is when the patient had lost about 75 percent of their normal terminals.

And these cells grow in what we call phenotypic manner and restore a terminal network. This is very exploratory therapy, but it has some up side. In particular we found that patients typically recover over a period of two to three to four years when the terminals grow into place that the desired cell type, here in red, is the cell type that dies in the diseased. So as of early this week a paper came out, in fact three papers, out of the four groups in the world that are actively pursuing this work, three in nature one and one is ongoing, in which, in our case, these transplants survived for up to 14 years without any signs of Parkinson's disease pathology. Two other studies showing signs of some very minor protein aggregation. But the overall message I believe from the four studies are that the majority of patients who get these cells, very phenotypically specific cells, have them active for a long period of time which gives you a rationale for pursuing a better cell source than fetal cells.

So I hope you saw the title of my slide here, which is that these are non-stem cell derived fetal neurons. But, in fact, those are the very one cells we desire to get from the stem cells.

So the next question then is, what kind of animals would you use here to gain recent information about safety and benefits? And in the use of animal models, as we heard previously, the combination of both your desired clinical method and safety is the best choice I believe. Certainly models that inform us that are disease relevant, and I will show you examples of this, and also models that provide safety data at the same time. And even though it's more difficult I believe to present data for clinical trials, the combinations of such models frequently give you a more comprehensive view.

As we discussed previously, in all our cases we've had to choose animal models both of allo- and xenografted combinations, and the immunosuppression methods have been applied I would say actually with some reasonable data outcome. I feel that those experiments have given us a lot of information that weren't too influenced by the host donor immunology. So as we heard from one previous speaker, I think these are issues that may be manageable.

However, the last point there, for those of you in the back, it says engraftment, implantation and transplantation procedures. Typically, in my field, which is an academic field not driven by industry yet, a cell concentration dose, delivery and implantation sites don't sound very exciting to a scientific paper, but they are extremely important to the outcome for the patients and the rodent models and primate models we view. So I hope to give you some examples of those.

Now, what about an overall understanding of function? The key problem in our field of Parkinson's disease was that people basically didn't believe you could use cell therapy when we started this work in the early '80s. And it took a long time before the field realized that, in fact, it was a dopamine neuron specific behavioral effect. And I've used a summary slide here in which the behavioral syndrome is on the left side of this y-axis, which your graph actually rotates because had lost dopamine neurons on one side of the brain.

And then you have here weeks

post-transplantation, and I believe these experiments illustrate an important point. If you do a xenogeneic combination of putting human fetal dopamine neurons, these are

post-mitotic, very young, but they will not divide after this point, our desired cell type will grow in the brain, takes about 20 weeks, maybe sometimes 16, in which a rat receiving the same cell dose essentially as a allogeneic rat-to-rat combination recover. If you put in, say, a primate fetal VM which has a faster gestational growth, the same dopamine neuron post-mitotic will make the animal recover faster. A pig has 115 days gestation, also does it over eight weeks. And if you take a mouse dopamine neuron, it takes about four to six weeks.

If you put in, and this is work we originally used to show the principle that you could get the desired phenotype from ES cell using what's known as a default pathway, we implanted ES cells at very low concentration where ectoderm dominates and we got dopamine neurons, and now we face shifted this by 14 days because that's the number of days it takes for dopamine neurons to be produced from embryonic stem cells. So if you follow the reasoning here, you can predict the behavior recovery depending on your donor cell age and growth patterns. So that's sort of proof of principle.

Another example I would like to share with you is that it's very different than pharmacology, and I like that it's different, meaning cells have -- the reason we use cells many times is because they have feedback control of releasing the desired substance, here dopamine. So the dose response curve with dopamine cells here from survival of 100 cells, where you see survival in the rat, up to 10,000 cells, the behavior response plateaus even though you increase your dose from 100 to 10,000 surviving cells.

The reason being, and you can find this in articles, that the cells will shut down their own release of dopamine as an adaptation to the host brain. So that's why we are pursuing this particular cell therapy rather than more drug design.

Now, what about stem cells? So the first work we worked with actually involved Dr. Dinsmore, who was then collaborating with us, was control experimenting, which we were actually producing another cell type. But in the control experiment, which was scary to most post docs in the group, we actually implanted pure ES cells.

And the remarkable finding was that not only did we find teratoma, which is in the center here, but also large number of neurons of the desired cell type that we have been looking for previously for about 15 years. So, obviously, we were very attuned to this finding, and over the next couple of years we used the finding to define a way to obtain a better outcome that did not include teratomas.

But, actually, very pertinent to the discussion this morning, we did the cell dose experiments that we had questions about. When we injected, and this is using the same concentration but actually increasing cell dose, 50,000 mouse ES cells into a rat brain generate smooth muscles, a little bit of neurectoderm, but also endothelial cells and skin. If you reduce the dose somewhat, you get towards a more neurectodermal neuronal graft, and, finally, at very low levels, this was actually based on early work that was done many, many years ago based on neuronal induction paradigm, if you don't tell an ES cell what to do, it becomes a neuron. We could in some cases obtain pure neuronal grafts that were of this desired dopamine neuron.

And, moreover, we used this finding to test the functional hypothesis. Could these dopamine neurons function? Indeed, they did. They had very, exactly the desired cell type and they could grow exactly into the phenotype with the markers we had. And implanting them into rats and later into primates we could show by PET scanning, post mortem that these dopamine neurons functioned exactly the way we wanted to. Using FMRI and PET, they activate blood flow the same way that the fetal cells had in early clinical trials that we had done on patients. So the functional issues, which obviously is my interest, had been accomplished.

But we also had to pursue very fundamental questions as had the previous speakers. For example, one example which goes against the common notion that ES cell depends on implantation site, and perhaps suggests that some of it depends on survival in the implantation site, was that we performed in those early experiments implantation of this very low dose of ES cell either into the brain or the kidney capsule and they are virtually impossible to distinguish, meaning certainly for teratoma formation, that was dependent on cell dose not location. So the old experiments using kidney capsule as a location was a fairly good way to test these tumors.

Now, obviously, the next step was to use embryonic stem cells, and we pursued this, and many times in collaboration, but the obvious problem for us was actually two-fold, that human cells grow into bigger tissues, as we heard previously. I think we need to be plain about this. Tumors are not unicells grow into larger bodies than do mouse cells and that is a problem.

Normally ES cells, and I want to quote Mahendra Rao is in the room, unless you have a teratoma, you don't have an ES cell. And I think he put it best of all, namely, teratoma is a desired outcome to prove that you have a real ES cell, namely, just all the tissues are produced in this benign, I should say at least in teratology, as a childhood tumor where you find all the cell types. So this is the normal outcome of a normal ES cell.

But more interestingly in a recent paper by Steve Goldman, who is on the committee and in this room, showed that when you use human tissue, if you have a cell that proliferates, say you know that our brains are very large and that's primarily because our cortex grows much larger than other species, and if you have such cell in your cell population it will grow to the size or try to grow to the size of a human cortex, which makes it very problematic for animal models. And so one needs to consider these issues I think very plainly and biologically.

A more challenging problem that we have already discussed is when an ES cell becomes transformed. And in my view and my experience it has to do primarily with passaging of cells. So the more a cell is allowed to or will divide, the higher the likelihood obviously that some kind of event will occur that will perhaps favor division.

Obviously, for those of you who are not biologists in the room, it's all the in vitro cell culture favor cells that divide. So there is the selection pressure to have cells transform. So passaging needs to be known and, obviously, studied very carefully.

And, as mentioned previously, subtle effects may occur. And teratocarcinoma is a very thing to me than a teratoma consequently. That would be uncontrollable growth, meaning the cells will not stop dividing even though you differentiate them. So that's a different type of cell.

Another issue there I think the biologists can help you understand is the cancer biologist's concept of de-differentiation. During passaging sometimes a cell will lose it's control of division and proliferation and basically become like an ES cell. In fact, that is the concept used for induced pluripotency, which is the current work on making skin cells into ES-like cells.

But for today's discussion let's focus on the human ES cell. And we performed early, this is now seven years ago, the equivalent study that we did in the mouse using uninduced, the pure embryonic stems as implantation human cells into rats. And, as expected, with a very low dose of cells here we ranged, again, on the order of a few hundred to a thousand.

We typically, here's a rat brain, right side/left side, and here's the red outline is of graft. These grafts contain not only sometimes teratoma, another tissue type, but frequently these neural tissues. These are called neurectoderms. Here we have human antigens, we have in green here, and they form these structures that are the beginning of the nervous system.

But they also at nine weeks produced the desired cell type, the dopamine neuron. So we knew then that it was worthwhile to try to work with these cells to get out of this mass of cells the desired type.

How did we do this? Well, actually, by this time other groups had developed new techniques using standard cell tissue procedures, and, in particular, Ron McKie and Lauren Studer's groups developed ways in which they try to simulate the process of the Day 0 cell, which is the embryonic stem cell, this induction of the rosette that I just showed you, the green rosettes, and then as the cell matures, we have to expand those precursors into large numbers called neuroprecursors, and finally, in the last few days, we hope that they turn into the desired cell type.

They look typically like these balls. But the reality is, and that's why I don't like to show the schematic, and this is a population of cells. It contains all this beautiful green dopamine neurons, but also in the center of the ball other cells that may have escaped pattern.

This became very obvious to us. And even though a number of groups, and I apologize for the details of this scheme here, but essentially, again, Day 0 to 45 using all sorts of factors that scientists believe are necessary to grow the cell, sometimes it actually spontaneously differentiates, we can stimulate these rosettes, shown as gray spots here, we actually have to cut them out of the dish. This is how crude this technology is.

And then propagate them again until we get large masses of those, and eventually we get some dopamine neurons that, thankfully when we transplanted them, and this is now also worked on by my colleagues in other universities, they restore function in the dopamine-depleted rat model at the predicted rate of recovery. So as we showed in the first slide of function, they don't immediately create benefit to the animal, but at the rate in which they would grow into the host brain.

But the reality, which I'd like to share with you, is that this is a fairly big graft in a rat host. So in a human brain it wouldn't be so detrimental. But these cells were relatively rare, and even though many of them had the right cell type, there are also cell types that we didn't recognize immediately but later found were derivatives of dopamine neurons that were not in the mid brain but in other places of the brain, not necessarily harmful, but not exactly the functional cell type.

And I think this is typical for the stage of research of current embryonic stem cell work. We are trying to get a cell type. But, in fact, as pointed out wonderfully by a previous speaker, we actually have to use in vitro in the dish methods to do what nature does normally.

So these in vitro cell culture models then became our work for the next five years. And to bullet some of those discoveries, the questions that we always ask now, are the changes produced by cell culture and passaging monitored and understood? Do we understand what happens to the cell over time? And also, obviously, if you grow the cells for 40 days or sometimes for six months, do you really know whether those functional cells you have are the typical ones that you want for heart islets or brain? Are they as optimal as say a drug design would be? What criteria do we use?

And finally on the safety, to what extent can we remove the unpatterned cells? It's a very pragmatic question. If you know you have the desired cell type, how do you get rid of the other ones? And that's the last point.

And I will show you some experiments, but I wanted to share with you how it looks like in the dish. If you look into the microscope at we called Passage Zero, Day 9 of a human embryonic stem cell culture this is how it looks like. It's not well ordered lines of crop cells, but rather a series of aggregated cells that, depending on concentration when you plate them, some of those will form into these rosette structures that we know will turn into dopamine neurons.

Over time we can grow them, and we actually have to manually cut around those and passage them further. And we do so using methods that are very standard in the field and we start at zero and around 42, sometimes 47 days.

But I wanted to share with you a discovery or an observation we have in a paper published by Jan Pruszak last year that at Day 35, when you think you have patterned the cell, actually in some cases you have these SSEA-4, which is the unpatterned ES cells in green, the precursors to dopamine in red, and also some of the neurons desired here in blue. This is a population of cells.

And the obvious idea that follows, perhaps I should show you the explanation I believe in first, is that at that time, 1, 2, 3, 4, 5, et cetera in your dish, some cells will differentiate according to their normal, natural schedule, others will be remain on pattern because stem cells tend to talk to each other at certain densities. And I know there was interesting question from the committee earlier. These things also matter.

If a cell is very close, if an ES cell is very close to its neighbor, it may exchange what we call cytokines and other patterning factors to tell the other cells what to do so that's why I showed you the dish before. A cell C can remain in this dish for 35 days without really seeing the differentiation factor you want.

So if this is the reality, what do you do about it? Well, there are examples of how to reduce this risk, and I believe that science eventually will solve this problem and I'll give you a few examples. That actually was illustrated previously.

When you look at the embryonic antigen 4 and you grow the cells over time, in certain instances you can grow, here you see days in vitro 42, that this green antigen, here studied by flow cytometry, is decreased as the cell population ages. So less than 0.5 percent in our best cases, and here's another example of a marker that disappears in the population.

Now, as we heard previously, well, is that sufficient? Is 50,000 human cells or 100,000 going to be too big a risk in these trials? So obviously, as suggested, one can perhaps limit the procedure to a more differentiated proliferating cell as a precursor of dopamine neuron.

Here is an example where we labeled the precursor using a marker and then connecting it to a fluorescent protein. So when the cell expressed this marker, it was also green, fluorescent green. And in that way, when we implanted the cells, instead of getting these big tumors here, we managed to eliminate virtually all the tumors. So this is technically and scientifically quite possible.

Now, also, as my last data slide, I want to show you something that we actually published last Friday in stem cells by the author Hedlund et al. And where we actually labeled by a green fluorescent protein the very dopamine when it expressed a factor that's unique to this cell that dyes in Parkinson's disease, and using flow cytometry we transplanted only neurons and, quite surprisingly, they didn't apparently need any neighbors because we had surviving cells that then had eliminated completely or virtually completely all other cells types. So even though these are very challenging issues, I believe there are technical and scientific ways around this.

So, in my experience, the factors that come into play are the animals models, and there are reasonable ways of choosing those including species and cross species. The animal species in new suppression methods are known and can be addressed in an effective way. If I may add one thing is that you actually, if you have a teratoma that's not growing any more, it will not have mitotic cells in it. So there is actually a very good way of addressing, I think some member of the committee said this, you can actually look at earlier time points, and, if your graft is not growing, that means that all the stem cells are depleted, meaning they had gone down their natural developmental path. So, I think, again, this can be understood and dealt with.

However, the engraftment implantation transplantation procedures I do think belong not only in the laboratory but in industry. The number of variables you have to test here are demanding, but I think they perhaps should be the same way as we deal with pharmacology.

Potential tumorigenicity we have talked about. Stem cell derived cell identity and stability to me means that you really want to decide cell type there for the patient you are treating. I think that's an absolutely given and that there should be sufficient criteria from the proposing clinical group or company to prove that and that they have the desired cell type because the benefit must be present to the patient.

And, finally, in summary then, my point is that I think these models can provide both benefit and safety data and should be carried out before clinical investigations.

Thank you very much.

(Applause.)

CHAIR URBA: Dr. Taylor?

DR. TAYLOR: Hi. That was some elegant work. I have a couple of questions.

So most of the abilities to sort that you used were based on labeling with something like GFP. And, clearly, that's not clinically applicable.

DR. ISACSON: That was just an example. In fact, we have other methods including cell surface antigens and so I didn't have time to go through them. But I would say that there are at least two or three antigens, again, depending on the specificity of the antibody for each stage. So we actually have a paper that's in preparation showing what we call a code, so each cell type, like in hematology. As for hematology, there is a way for obtaining a cell surface marker code that is useful.

DR. TAYLOR: Sure.

DR. ISACSON: So I don't think that's a necessary requirement in the future, but an example that you can identify and isolate the cell type.

DR. TAYLOR: And so are you saying you have codes for various stages of differentiation then?

DR. ISACSON: Exactly.

DR. FRIEDLANDER: I have three very short questions for you.

In your cultures where you saw SSEA-4 positive cells, did you see them associated with any other specific cell type that might help you predict the future?

DR. ISACSON: I don't think so, but I would defer to someone who looked at more cultures than I do, but I have not seen any association. Except when you saw that neural structure, the rosette formation, that actually does have a biology in it. So cells are born in the center of that crater and then migrate through and out.

DR. FRIEDLANDER: The second question is, did you ever look for SSEA-4 in your tumors?

DR. ISACSON: Yes. That's the one thing I tried to say at the end there. We started doing this actually in the mouse teratoma, and by 21 days, I remember this very specifically, by 21 days in mouse teratoma, we no longer see any growth, so that the teratology, the embryological childhood tumor is formed. Whereas with human cells, the SSEA-4 typically will go beyond several months. And, obviously, that actually sometimes leads that the animal is lost as is stated in the briefing document. The animal dies from a tumor. But you can use that marker effectively to say whether there is still undifferentiating cells because they will self renew.

DR. FRIEDLANDER: Correct.

DR. ISACSON: But the point biologically is that the cell will tend to try to fully differentiate. There's nothing that says that an SSEA-4 cell necessarily will maintain it's proliferation because I've seen that in other animal species stem cells.

DR. FRIEDLANDER: So that leads me into the third question, which is, in your primary cultures that you talked about first, how did you measure the viability of those cells and do those cells actually have to be viable or proliferating to actually have a positive effect in the graft?

DR. ISACSON: Absolutely. I hope that that was a premise that went through. If you transplant dead cells, you get absolutely nothing. Otherwise, we wouldn't be doing this work quite frankly.

But let's see if I can answer the first part of your question. Yes, so it's quite easy to do that. You have live viability and dye exclusion tests, so typically 80, 90 percent of the cells are alive at the implantation.

But I want to point out something that was mentioned previously. Somebody called cataclysmic death. When you implant cells into the adult brain, which I think is the desired host, there's a lot of cell death in the first few days.

In fact, in the developing brain you have something called developmental cell death. Fifty percent of your brain has been sculptured from a larger number of cells. But I would say in most cases up to 90 percent of the cells will die in the first few days. So that's an important thing will live cell therapy that the scientists and the clinicians and the companies understand that they need to go through those paradigms and simulate it in a reasonable way.

DR. SNYDER: I just had a brief, two-part question. In addition to looking for the gross evidence of teratoma, did you ever see any evidence of non-neural cell types in the brain that may not have been deforming but would have only been picked up by immunocytochemistry, like smooth muscle or early evidence of cartilage, cell types that are inappropriate to the brain?

DR. ISACSON: Well, I showed you a slide, right? Did you see that?

DR. SNYDER: I did at the low dose. You did not.

DR. ISACSON: At 50,000 cells.

DR. SNYDER: Okay. So you never saw inappropriate cell types, not in teratoma formation, but just forming there like vascular endothelium, things of that sort.

DR. ISACSON: No, that was what that slide showed. If you have 50,000 cells, you get all the tissues in a teratoma. Wasn't that clear?

Okay. But let me ask you a question another way. So, yes, teratoma is exactly what teratomas are. And you make me think of another point which is that some neural cells, including glial cells, will migrate away from the site of implantation.

So if you have a glial cell, and this pertains to some of the previous questions, a glial cell is a microglia. It has to be. It is on the non-neuronal tissue. Neurons will remain within gray matter and never move if it's a fetal cell. Progenitor cells in glials that will migrate along white matter bundles in the brain quite extensively, and we have published this and others have, so there is actually an importance in understanding what they do and where they may end up in the brain.

Fortunately, I have never seen any case of demyelination. In fact, some people will argue that maybe they are beneficial to the brain. But they are not the functional cell that we have studied for Parkinson's disease.

DR. CHIEN: I was going to ask you about the one component I guess we haven't talked a lot about is let's say you have the correct cell of interest, getting it to survive in the organ system or tissue you're interested in is a challenge, for example. In most of the heart stem cell therapy or

cell-based therapy that's been done, it's become fairly clear that very few of the cells delivered in any approach and almost any cell survives long term. It's a very small percentage.

And I was wondering in the Nature Medicine issue where your paper just appeared, there were two others that also suggested that the transplanted cells actually either don't survive or they acquire the disease, the graft actually gets the disease. And I was wondering two independent groups, and so that wouldn't necessarily be a vote of confidence that this is going to work. It's a no-brainer let me put it that way.

DR. ISACSON: No, but let's --

DR. CHIEN: I'm trying to liven it up a little bit.

DR. ISACSON: I thought I described the facts more clearly. So all three papers show survival of cell for up to 15 years. That's indisputable. All papers show that most cells, in fact probably 99 percent, because I have looked at the other transplants. We have this as a shared collaborative effort.

When the transplants show what's called protein inclusions, which doesn't mean that the cells don't work by the way, this is also a fact, they do so at a fraction of less than one percent. So my interpretation is actually different than yours then, which is that it actually lends a lot of optimists to this field.

And the fact that -- I mean one of the papers speculates that maybe the transplant gets Parkinson's disease, but there absolutely no evidence that's the same type of reaction as in the patients because the transplant that we have looked at from other groups have have a lot of microglia around them. So it actually could be something that pertains to today's discussion is something that the host may generate as a response to the transplant not unlike what you may get in your heart transplants if you put in the wrong cells in the wrong place.

But I actually think that you said two papers, but they have the same basic tenet. There's a fourth paper coming out that has exactly the same outcome as our paper. I think it basically shows in fact that there are variable outcomes and that the variable outcomes depend on surgery, technique, cell dose, cells that don't belong in the transplant were transferred including angiogenesis.

The paper that had most of these cells had donor tissue that wasn't dissociated as a primary cell culture. So the whole tissue chunk was put in the brain. And we know from science that that contains donor blood vessels which have very high component of antigens such as MHC1, which are, of course, T-cell targets and microglial triggers. So I believe that will be signs and our observations will be explaining these findings.

But the bottom line is that cell therapy, which was stated by all papers if I may add, have functional effect in the patients for at least ten years, which is as well as perhaps heart surgery can do currently, and this is a very prototypical and, as I hope I said, very crude method at present, which will not be comparable I hope, I think to future work on pure cell neural transplants.

DR. CHIEN: Okay. That was transient though, I think, in one of the papers. I just read it.

DR. ISACSON: It's quite different than heart.

DR. GERSON: You, if I could, the preceding presenters have suggested that there is cell loss at the point of the insertion of the cells by whatever method into a tissue. Is that a non event or are there cellular events or host response events due to this cell death that is purported to take place? Is this something that we should give attention to or something that we should ignore?

DR. ISACSON: That's an interesting question and I don't think there is an easy answer to that one. I mean fact of the matter is when you implant cell, 80 percent die. The remaining 20 percent will grow quite well over a 15 year period. Those are the facts.

But I mean it may be important to study that specifically. So if you implant cells that may contain fibroblasts or endothelial precursors, they will have a very massive -- I mean shouldn't exaggerate, but the immune system will look at those very differently than nerve cells. So, again, you know, specificity in experiment is required. Generalizations are somewhat weak.

But I would say that if you as an experimenter know yourself have composition, you can make reasonable predictions. If somebody comes to the FDA and makes a suggestion, I want to transplant these cell types, and they contain a population that you know is going to be highly immunogenic, even though most of the cells may die, they need to specifically show that there isn't an adverse effect. But I wouldn't say generally that it's a huge problem. But in most of the work that we do currently, we spend a lot of time designing the experiment and the therapy to avoid inflammation and try to avoid all of the immune reactions.

DR. SALOMON: I think that I just want to pick up on that too. What bothers me is this idea that we just accept the way you put it that there's going to be 90 percent, 80 percent, different percentages thrown out of cells dying with a transplant.

Number one, I don't think that that's something that you want to accept. I think if you get the efficacy and you set up the protocols in such a way that that's there in the early stages of a study, then that's what you find, and I'm not saying that, per se, I'm biased against that, but as a principle I don't think that's a good principle. You know, we do kidney transplants. We don't expect 90 percent of the kidney to die after the kidney transplant.

I also want to say that I think that there is an easy answer to the question. That if you put a cell transplant in any tissue site and 90 percent of the cells die within a few hours or a few days, that will have immunological consequences both for innate and adaptive immunity.

DR. ISACSON: Yes. So I want to differ with you on one point though, which is important, just like pharmacology. Tissue organ transplants, even though we are part of the same society, American Society of Transplantation you and I, I would say cell therapy, and we have Gordon Weir here, is a very different principle than organ transplants.

Your analogy I don't feel applies because I'm not talking at all about that the transplant itself died. If you put in another organ and the organ dies, that's a complete failure. But in those patients that we have that have been off and on dopa for a decade, I would say maybe 75 percent of initial dopamine population died, but that was sufficient survival that these patients improved.

Of course I agree with you on the principle, that in the future we would like to have 100 percent survival and it may be possible to do that in various ways. I just want to point out that, you know, it's an early young field and we need to know the differences.

DR. TOMFORD: You mentioned the idea of the cells acting individually versus sort of together. What stimulates the transformation of the cells into a tissue? In other words, we're transplanting individual cells. Eventually they're going to form a tissue. Could you comment on what happens there or what your thinking is?

DR. ISACSON: I wish I was able to. No, if I understand your question correctly, for example, brain tissue is formed from the neuroectoderm. So at some point these cells talk to each other and send signals that are translated into transcription factors that then eventually tells the cell what to become, and they are spatially dependent and time dependent. It's more like initiating programs that run for a certain period of time.

To some extent, we can force these programs. And experimentally the lab, because we are not so good a simulating the environment, many scientists in my field have now resumed what they call forced expression of a transcription factor. So you express the factor that's dominant and forcefully drives the program to which cell type you want. So you can do it either by intrinsic genetic mechanism or try to simulate in the environment the time, the dose of those factors that are typically in surrounding cells.

CHAIR URBA: One last question.

DR. FRIEDLANDER: I'd like to come back to that concept of the effects of the transplanted cells on endogenous cells. And I recognize there's a difference between putting a cell which you expect to become a dopaminergic -- producing neuron and cell like a vascular endothelial cell, which may have some sort of paracrine effect on tissue around it.

But should we be thinking a lot more about the staging at which we put these in and what effect they could have on endogenous cells in terms of resuscitating or maintaining diseased cells, which ordinarily would die but perhaps not? So how do you look, in fact, endogenous cells be maintained as opposed to newly differentiated cells with tissue you put in?

DR. ISACSON: I think your question falls in the range of, you know, can you implant a cell that stimulates the process in the host? The answer is probably yes. Can you put in a cell that does something specific to heart and muscle or brain that you know you have lost? And the former yes means that it's almost -- the second time I've used the word, it's more challenging than the latter, meaning when we put in nerve cells in the brain, we get the cells making the rats less Parkinsonian because if we kill the cells, they are Parkinsonian again. So they depend on the factor you put in the cell.

However, over time you see, we cannot today really measure that. When the cell starts communicating -- and this could be paracrine by the way; I'm not saying for example in ALS, which is a different type of pathology of the Parkinson's disease, even though they're related -- there may be a different way of using glial cells to substitute for cell that participate more actively there in the disease process than I believe they do in Parkinson's disease.

So I'm open. I think we should all be open to innovation and keep a very open mind of the choice the scientists and the clinicians for what reason and how they prove they have efficacy. The most frustrating thing in the field if you're in it is actually when somebody, we had that opportunity, right. You put in a cell, it dies, and the group says, well, the animal got better.

So you ask, well, how did the dying cell provide the fact of that? And now scientifically you're in such a broad scenario, universe of possibilities, that you simply can't move because there is no hypothesis to test. But you would presume there was a substance then, right, released by the cell that you perhaps could develop a drug for. And I think that's perhaps beyond the FDA/CBER premise of what stem cells do.

But since you asked the question, I think a lot of people think about those issues with live cells, and I think it will be a major scientific field in the future. Homeostasis of cells, they talk to each other all the time.

CHAIR URBA: Okay. We've heard all morning about where the cells go and how do we track them. And that leads to our last talk of the morning with Dr. Bulte, who is the Director of Cellular Imaging, Institute for Cell Engineering at Johns Hopkins.

DR. BULTE: Good morning. There are three things that I will talk about, and I've heard some of the issues before that are very, very relevant for clinical translations. The three things: are the cells delivered correctly; where do they go; and what do they become, assuming that they survive?

So during the last 50 years we have made tremendous progress in organ transplantation, be it the pancreas, kidney, together with the other development pioneered by Howard Hughes, the rise of the modern airline industry so that the organs arrive alive. So the purpose of the meeting today is sort of I think looking into a crystal ball what the future will bring, and we don't know.

I do think that one of the things that has not been addressed is perhaps the administration of cell cocktails. As we know, transplanted cells interact with the host, and they work in teams and perhaps they work better with friends they already know, let's say, from the same background.

The majority of examples that I will show do not involve the use of human embryonic stem cells for a variety of reasons. However, the examples I will show, and I will really show a variety of imaging techniques that can be used, they're all applicable to human embryonic stem cells.

I'll start right off. There is no perfect imaging technique. They all have their limitations. So that's kind of unfortunate that we cannot just pick one for all applications.

Now, I'm showing this example here worked on by Steven Goldman and worked on earlier by Evan Snyder, how we can transplant in the brain of animals if it is myelinated disease, and we see here on the right that they shiver less, and you can measure that on a graph, and perhaps the amplitudes can be used as an outcome measure and we could possibly apply our imaging techniques to optimize that.

Now, the reason why I showed this example is that I'm fascinated by the histology shown here that these cells, in this case they're human fetal glial progenitor cells, they're stained here in red, for human nuclear antigen.

The question really is, what are the limits of cell migration and cell proliferation. We have here a mouse brain, if we transplant these cells in humans or in the spinal cords, are they at some point going to stop? Where do they stop? So these are questions we cannot answer by taking biopsies. We have heard immunohistochemistry or PCR. Of course that's being very complicated.

The relevance of imaging, and this is in general for any sort of stem cell specifically for human embryonic stem cells we're talking about, is the imaging, once we know in animals that they work and we can measure, let's say, the outcome measures such as the BBB score, that we correlate that with this accurate cell delivery, the cell migration, and correlate the therapeutic effect and then we can optimize the number of cells, the delivery routes, the timing, and the dose. We know very little about this currently, and, eventually, we can hopefully apply it clinically.

Now listed here, and this list might not all be inclusive, are ten techniques that are currently can be used, and I'm talking about noninvasive imaging. Not all of these can be used clinically, and I will show some examples, and at the end come back to these, what the different advantages and disadvantages are.

The only FDA-approved cell tracker as of today is indium-oxine. It's primarily used for immune cell trafficking. The half life is short. It's 2.8 days. I'm showing here an example of mesenchymal stem cells that have been labeled with indium-oxine. They have been administered IV. And what we do then, we image IFE Spect and the SPECT/CT, so I'm starting right away with two imaging techniques. One only gives the hot spots; the other one, the CT, gives anatomical information, and we fuse and superimpose them so we can interpret the hot spots within its anatomical location and what you see is mesenchymal stem cells are large. They will clog up the lungs.

So you have to be very careful. You don't want to use emboli. You have to inject single cell suspension. And after two days, you can see here a hot spot where we have induced in dogs myocardial infarct persisting for five days.

We have found some benefit. It's very hard to validate that because myocardial infarct is very variable. But what's interesting, that out of 10 or 20 million cells we inject, we can quantify only about 50,000 arrive in the heart, and the rest is all staying in the liver and the spleen being trapped. And apparently a small number may be enough to induce a beneficial effect.

Much of the work that I've started with and I think clinically will be the way that cells are going to be delivered involves stem cells with a magnetic dye, and super paramagnetic iron oxide is a material that's also found in birds that migrate along the geomagnetic field, dolphins, salmons, it's sort of like magnetic colloidal fluid. It's FDA-approved as a liver agent. How does it work? This is before, after injection. These particles IV are taken up by macrophages. They have scavenger receptors. These particles are opsonized with proteins, and then the problem we face in the MRI and you get black spots.

So here is normal liver that contains Kupffer cells, which are the scavenger cell, the macrophage that turn black, and you can see the isointense tissue is a tumor. So macrophages take this up.

Stem cells are non-phagocytic and there are now a number of techniques that introduce these particles into cells most commonly using transfection agents.

So I come back to the three themes. The first one, are the cells delivered correctly? I think that's the most important thing to start out with. For the brain, we have different routes of injection, one is stereotactic parenchymal injection. Such is the case of Parkinson's disease.

The other one is intravenous infusion that I have shown. It's less invasive, but small numbers may reach the target area as compared to parenchymal injection, but that may not always be without risk.

Another method is intraventricular injection. I think this is a good route of administration for diseases that are located throughout the brain, close to the ventricles such as multiple sclerosis, the entire area can be targeted. And, for instance, in cases of stroke, what about intra-arterial infusion on the side of the stroke so we can target ourselves?

Shown here is an example of the latter of cells were injected in the carotid artery on the site of the middle cerebral artery occlusion, and you can see here a moderate graftment on this side with some cells also localizing in the hemisphere. And one of the things we have done is we have monitored the injection process using laser Doppler flow monitoring. You can see here we stop our infusion, the signal goes up and then it slows down again. So if you have cells in the brain, they're going to impede the blood flow.

And in this case here you can see a massive engraftment here on that side. You can see the LDF drop significantly and then we stop the injection. So even if certain administration routes IV where we inject larger cells, we have to be very careful that we are not going to clog up capillaries when we target these areas because actually we can induce mini strokes.

We can with MRI, shown one hemisphere on the bottom, we can look in 3-D where our cells are localized, in this case primarily in one hemisphere as you can see.

Now, not many places are able to do MRI-guided injections. That will be the future to deliver cells. We have catheters that are compatible if the MRI that are steerable, they can be targeted towards the area of interest.

I'll show you an example of work on here with Dara Kraitchman in a myocardial infarct. We can see the MRI. We can inject the paramagnetic tracer gadolinium that is bright. When tissue is -- it goes in but it takes a little bit longer to go out because it resides there. And we can precisely place a catheter near the infarct so we can see. And this is a frame movie. In realtime it's much better. We see this in realtime on the screen with special software and then we can inject our cells right here and we can see that.

Now, mistakes are going to be made by let's say less experienced radiologists and at this point we do not know what the best grafting site is. You can imagine if injected at a myocardial infarct itself it's a very hostile environment. There's low oxygen, low nutrients. Not too far away, perhaps at the border zone it's best.

So if you can see where to locate it, then we can correlate it to our outcome measures and the good thing it's done in realtime. So if the therapy does not work and the patient is being taken out of the magnet, goes home or in intensive care, if it doesn't work it's not the injection itself. It's the cells that are failing, perhaps not enough.

If we don't do this and it doesn't work, we do not know what the problem is. It may be that the injections were done incorrectly.

So I will show an example where it really is important. This is the first clinical study on MR tracking of cells performed by some colleagues with me in the Netherlands, a very large team. And, again, this is not human embryonic stem cells. This is with dendritic cells that are used as cancer vaccines. That area of cell therapy, same as islet cell therapy, has been more developed so we know a lot more about what we can do with that.

So you can see here cells are labeled with iron. They turn blue. We can stain that, inject it in the draining lymph node, that of stage three melanoma patients; lymph node is resected; and the resolution is really very high of MRI. You can see here the cells through vessels going to lymph node from another lymph node.

Now what have you learned from the study? One is yes, in a clinical magnets the patients with few numbers of cells we can see migration of the injected nodes to nearby lymph nodes, and this is validated here with the scintigraphy. So I think that's important, but it's translational from the animal work.

What the real surprise was is that in half the patients the cells were misinjected completely and nobody knew that it happened until the MRI was done. So they're not set up in the Netherlands to do these MR-guided injections. They do it under ultrasound. I think all of those who have seen those images, they're much harder to interpret. They have less resolution.

So here's the lymph nodes where the cells need to be injected in, and dendritic cells need to stimulate T-cells in the medulla. You can see it injected in the subcutaneous fat. So with indium-oxine you can't see that. You see a cloud of radioactivity, but you do not know where you are.

So I think this illustrates perhaps also with spinal cord injections, you want to inject cells in a myocardial infarct or in the striatum. It's important to know where they are. So the second thing is, where do they go?

This is an early example done with oligodendrocyte progenitors, worked on with Ian Duncan. You can see here the injection site. You can see here the cells and they also myelinate sort of like a B-line migrating away we actually sorted about 40mm away from the injection site and it's interesting that the MRI distribution of these cells also correlates to the new myelination that the cells induced. There's no normal myelin in these animals.

We can also do that in vivo and we put cells in the ventricles so we don't put it in the parenchyma. We can see cells moving out of the ventricles into the brain. They myelinate and we can correlate with histology.

Now we can look at 3-D reconstruction. Now this is an example, this is Evan Snyder's cell line, the C17.2 of the cell line that is immortalized. So it continues to proliferate. It does not form tumors. So I think this is benign proliferation. But you can see here, in 3-dimension, you can see those fingerlike projections, how they grow into a glioma, like a glioma also into the brain, and I think it's very hard with histology to really look at these invasive patterns that pop up again. You see here all these fingers coming out where they migrate probably along white matter tracks.

Now we can also, when we do our histology, this is an ex vivo image. So it's noninvasive. We can determine the exact plane of cutting. If we don't know a priority of where our cells are, we may be unlucky and cut the wrong plane. We can do this in any dimension. We can slice it up. We have the computer NCR cells there, that black spot, so we know we can focus in on that area.

The resolution of MR is such that when we look at cancer vaccines we see here cells homing into the medulla. There is some vaccine injected into the food path, so the direct cells, immunosentinel cells, they pick it up and they go in the lymph node and we can see here very small numbers of cells. And this is a lymph node of a mouse so if you know how small these are.

Now, what about quantifying the local number of cells? With MRI it's very difficult in certain cases. It does work in this case. It was very homogenously distributed. So we sampled the number of black pixels on the MRI. We made single cell suspensions off these nodes, and, because of the iron oxide labeling, the cells were magnetic. We could isolate on a magnetic cell separation column and we, in this case, actually can use MRI to quantify cells.

I'm not going to show it. We are looking at immunoadjuvants to see to enhance the trafficking and it's very nice. We can do this over time non-invasively.

What can we learn from islets transplantation studies? MR tracking has also been pursued for a group in Czech Republic and at MGH. One of the things that came out in graft reject allogenic transplant is that the islets that you can see as black spots they disappear because they're being rejected and the iron is dissolved.

So the problem of islet cell transplantation is really the immunosuppression, which is very toxic to the beta cell. We don't know how long the islet survives.

Another approach is to protect these cells in semipermeable alginate capsules so the insulin can go out but the cells are protected from antibodies. And there is now some clinical work ongoing in Moscow with xenografts in patients. One of the questions we have when we label these capsules magnetically, we made magnetocapsules, is, can we monitor capsule rupture so we can see that when we induce this artificially. We can see single capsules, so we possibly want to monitor capsule breakdown. Nobody knows how long these capsules stay intact, and at that point maybe do immunosuppression.

This is, again, an example of how these things can be infused. This is an interventional radiologist, Arepally and Howard Hughes Fellow Brad Barnett, so they can be targeting the portal vein and then we can engraft these capsules in the liver and we can exactly follow the grafting process. We can do that in realtime. I don't know if the liver is the best place, but it's an example how we can monitor the grafting.

So now one of I think the most important question once the cells have delivered correctly, that's the most important thing. But the most important question is, what do they become? And we need a reporter gene for that.

I've heard a lot about cell survival, how many cells survive. A technique that's already mentioned is called bioluminescent imaging that uses the luciferase from the firefly, we get luciferin and photons are being emitted.

And it's interesting some work I'm doing with Doug Kerr is when we started these experiments, now almost two years ago, that we had a very hard time to get good survival of cells, and we found out that immunosuppression, the direction you choose, is a major determinate. And standard cyclosporine that people use from other studies is actually not best. There may be other regiments that enhance survival. So I think it's very important that you can monitor cell survival and use it.

Now, unfortunately, it cannot be used clinical. Why? Light has a limited penetration depth. You can also only do this in white mice. If there's pigment, melanin, it blocks the light, so there are limitations here.

The other thing you can do is you can put the luciferase on the reporter gene. For instance, we have done that on the GFAP when they become astrocytes. So if you don't put it on the promoter and you have glial-restricted precursor cells and the control is kidney cells, they all light up because there's no promoter. But then you put it under GFAP, you can see that only the glial-restricted precursor cells will light because they did then change to astrocytes and not the control cells.

You can do this for many reports in the interest of the neurons, of course. I think the key is you can see when it happens. Because this is noninvasive, you can stick in the mice two times a day if you want. Some gas anesthesia, they can handle that, and you can follow this over time.

Bioluminescent imaging has been developed by Christopher Contag at Stanford University. He has shown a very powerful experiment of injecting only one cell, bone marrow stem cell, following irradiation reconstitution of bone marrow. Only one cell that you can see reconstitutes the whole mouse, and in this case it failed. So this means that this technique is so sensitive that if there's one human embryonic stem cell that will form a teratoma or teratocarcinoma that that can be followed over time.

Do they form teratomas? And I learned something new today. There was an auto teratocarcinoma and I exchanged it here. I forgot to do it in the first one. So the issue is, and I'm showing the example of one cell, let's say people say adult cells are much safer than human embryonic stem cells. People do magnetic cell sorting or perhaps flow cytometry, but I think if there is only one cell that is undifferentiated, that one cell can still cause trouble.

This is work done also at Hopkins. You can see it as a PET imaging of a reporter gene. So this is the herpes simplex virus thymidine kinase that is a enzyme that can trap a probe in a cell that can be detected by positron emission tomographies. You can look here at teratoma formation. I think this is very interesting work done also from Stanford from Joe Wu, that the HSV-TK system developed by Juri Gelovani at Sloan Kettering in '94 had previous been used for gene therapy for gliomas with ganciclovir with the bystander effect. So they have done an experiment to induce teratomas in mice, used in a double fusion and a triple reporter gene. The triple one has the thymidine in kinase so they both have luciferase.

So this is the bioluminescent experiment. So you can see that both tumors grow. They make teratomas. They increase proliferation to get more signal.

Now, at this point, again, ganciclovir is given, right, and only the one with the thymidine kinase, which is a triple one, will act as a suicide gene. So you can see the tumor disappearing in here.

So I think this is so cool that you have one reporter, you can image it, and at the same time you can intervene with the drug if it gets out of hand. So this is the thymidine kinase system.

You can also monitor with positron emission tomography like you see here, the tumor growing and then it goes away with the ganciclovir.

Now, where is this clinically, there are now four clinical studies including islet cell transplantation. As far as I am aware, and I may be wrong, I know of one study with thymidine kinase also performed at Stanford by Sam Gambhir and others. They have checked the T-cells in glioma patients for immunotherapy. You can see here T-cells homing. There are two tumors here and some peripheral artifacts the system.

Now, so these other techniques I think that are clinically now being implemented at this point, in 2008, 2007.

There are a lot of things brewing around the corner. One of those I think is the first time that it's being done. We have cloned an artificial reporter gene not found in nature, based on rational design, and this reporter gene is chock full of amide protons which have a different resonance frequency. We can saturate those protons. They exchange with water. We always image water with MRI. We can give a specific radio frequency pulse that turns that on.