Alan Flake: I’m Alan Flake, I’m a pediatric surgeon at The Children’s Hospital of Philadelphia. And today’s CME presentation is on what I consider to be the future of fetal therapy, which is fetal stem cell and gene therapy.
Why am I excited about stem cell and gene therapy? It’s because there’s been a convergence of technologies over the past two decades, including, high-throughput molecular analysis; the human genome project. And these ultimately should allow us to make an early gestational diagnosis of most human genetic abnormalities early in gestation. If you can diagnosis an abnormality early in gestation it always raises the question of fetal therapy.
And if there are biological advantages or efficiency that can be achieved in the fetus safely, versus postnatal treatment, then it would make sense to treat not only diseases that affect the fetus before birth, but also postnatal diseases that can be anticipated before birth. That really changes the whole paradigm of fetal treatment that expands the opportunity to treat the fetus.
This slide just shows a cartoon of the future potential for prenatal diagnosis. You can anticipate genetic mass screening so that whole individuals are screened early in gestation for the majority of human genetic abnormalities. Once again, achieving an early gestation diagnosis and the potential to prenatally treat and anticipate a postnatal disease.
So what are these biological advantages and efficiencies that might make the strong rational for treatment of the fetus with stem cell or gene therapy versus the postnatal individual? The most important is immunologic tolerance, which I’ll talk about in detail. Another is stem cell biology in utero, the frequency of stem cell is much higher and stem cells go through predictable migrations to form tissue compartments that can be taken advantage of with respect to stem cell engraftment or gene transfer. Finally the small size of the fetus allows you to transplant a very large cell dose, or gene vector dose, on a per kilogram basis.
These advantages are literally a once in a lifetime opportunity, they disappear after early gestation, and therefore we’re trying to take advantage of normal developmental events to achieve a therapeutic result by either stem cell engraftment or gene transfer.
Fetal tolerance is a phenomenon that we all experience early in gestation that allows us to identify self-antigens and not react to self. This process is composed of two events:
The first is a positive selection event, which is related to recognition of pre-lymphocyte of self MHC, that’s major histocompatibility complex.
The second step is a negative selection step, which is associated with high-affinity recognition of self-antigen in association with MHC. And these cells are deleted as they would be reacted to self. And so what you’re left with is a population of lymphocytes which react to pathogens and others that don't react to yourself. And the whole rational behind in utero stem cell transplantation is to try to fool the immune system into accepting a foreign antigen, or a group of antigens. And you can do this by transplantation of either an allogeneic hematopoietic stem cell or a Xenogeneic hematopoietic stem cell. And with this you can potentially fool the immune system into processing another individual’s antigens as self and deleting those self-reactive lymphocytes.
The end result of that process is to create, from the perspective of the immune system, an identical twin donor. So these foreign cells from another individual are processed by the fetal thymus cells that are reactive to all of the antigens on those cells are deleted and you end up being completely tolerant to that donor as you would be to an identical twin. We all know that if you have an identical twin as an organ donor, or a cellular transplant donor, this can be done with very little toxicity, no immunosuppression, and you can basically avoid all of the complications associated with current methods of postnatal bone marrow transplant or organ transplants.
Just a reminder of the patient that we're talking about, it’s an approximately 35 g fetus as opposed to a 70 kg adult. So this provides huge efficiencies with respect to the number of cells that you transplant, or the number of gene vector particles required.
If you think about this, there are three strategies that make sense that are shown in this slide. One is to replace the abnormal stem cell compartment, like a bone marrow transplant prenatally. We’re not quite good enough to replace the entire compartment as of yet, and I’ll talk about the reasons for that subsequently. A more promising and imminent approach is to induce tolerance in utero, and then to perform a non-toxic, non-myeloablative cellular transplant or organ transplant after birth. And then the third approach is to target the fetal stem cells with gene therapy, which I’ll talk about last.
So this started many years ago. We initiated this research in the fetal lamb model. And we transplanted fetal liver from hematopoietic stem cells, from an early gestational Type A hemoglobin fetus into an early gestational type B hemoglobin fetus. This was very successful with engraftment of the allogeneic stem cells from a chimerism —that is they had blood elements of both the donor and the recipient. And with this system you can even engraft xenogeneic stem cells, meaning from another species. We were able to engraft human stem cells in the sheep model.
This generated a lot of excitement and led to many inappropriately designed and applied experiments around the world on human patients with various immunologic disorders. And predictably, most of this was unsuccessful. It led to either a state of microchimerism , or the levels of engraftment were so low that it could only be detected by molecular methods and were not associated with tolerance, or no detectable engraftment.
The one exception to that is that patients with severe immunodeficiency syndrome. This is a special circumstance where the donor cells have a survival and proliferative advantage over the host cells, and you’re able to achieve engraftment relatively easily. Unfortunately that’s not true with most other… our target diseases, and we had to go back to the drawing board and try to understand what the barriers to engraftment were that prevented clinical success with this approach.
The primary model that we developed was the mouse system. And it had many obvious advantages shown on the slide with respect to trying to define the requirements for engraftment and the barriers to engraftment after in utero stem cell transplantation.
On a stage-for-stage basis, the mouse model was quite good actually. If you look at the timing of when we would anticipate performing this clinically, it would be prior to 14 weeks’ gestation. During this period you have only fetal liver based hematopoiesis, and this is prior to the [thymic] processing of [cell vantagin] that I discussed earlier.
In the mouse model we can do transplants at around 14 days’ gestation, and this corresponds stage for stage through the stage of human development when we would like to perform these transplants, with purely fetal hematopoiesis and prior to thymic processing of antigen. With this system we initially, like the human, achieved only microchimerism; meaning very, very low levels of engraftment.
We achieved these in the intraperitoneal mouse model in which we inject cells at 14 days’ gestation just behind the fetal liver. The maximum dose of cells was only 5 micro-liters, and the maximum number was around 5 million.
And with this system we were able to only achieve microchimerism. I call these the dark days because for about eight years of research, we were only able to achieve microchimerism. It’s very hard to analyze this with respect to trying to determine the barriers to engraftment.
The cell changed, however, with the use of the intravascular approach. This video shows a micro [unclear] injection in the vein in a 14 day gestation fetal mouse. You can delivery approximately 30 micro-liters of cells with this approach, about six times the dose that you can delivery with intraperitoneal transplants.
And this really changed the entire fixture because we were now able to inject very large numbers of cells, take advantage of that efficiency advantage of the small size of the fetus. These cells were engrafted in the fetal liver. The green cells shown here are the donor cells, the EFP cells. And they were initially engrafted in the liver, and then migrate to the spleen, the bone marrow, and the thymus just like it normally occurs during normal development.
These mice that were engrafted were entirely tolerant of skin grafts or any other test of tolerance across full MHC barriers, without any need for immunosuppression or other treatment.
So in the mirroring model at least, the competitive barrier to in utero transplant could be overcome by the intravascular delivery of very high doses of adult derived bone marrow cells. And that will be a recurring theme in this talk.
And with that you can really overcome a competitive barrier and achieve levels of chimerism, shown in the lower line here, of a size 20 to 30% with a single in utero injection in this completely mismatched mouse model.
So the other potential barrier to engraftment is immunologic. Now the fetal tolerance is a well described entity. It was unclear whether we could achieve fetal tolerance by an injection of cells at the time point that we were attempting to perform these transplants.
And in fact, with the intravascular injection technique, it allowed us to track animals that had received high doses of cells. The finding was intriguing in that all of the animals were initially engrafted at birth and chimeric, but only 30% of the animals maintain their chimerism in the long term, and 70% lost their chimerism within four to five weeks of life.
This is very concerning to us obviously. Suggesting an alloresponse, or an immune rejection of the cells.
In fact if you looked at the frequency of alloreactive T-cells in chimeric versus non-chimeric pups, after in utero transplants, you can see that the chimeric animals actually had a much lower level of alloreactive T-cells in the negative, naïve controls, supporting deletion as should occur with fetal tolerance However, the non-chimeric animals actually had elevated levels of alloreactive T-cells that were in the range that you see with intentional immunization of animals.
And so this was clearly an adaptive immune response, and suggested that fetuses were in fact, at least the majority of fetuses, were in fact not tolerant to donor antigen.
This just shows the corresponding humoral immune response, again, supporting adaptive immune response against the donor cells.
So there were two possibilities here. One was a depressing possibility for me, suggesting that in fact the fetus was not tolerant, that Medawar was wrong and didn’t deserve the Noble Prize for his discovery of fetal tolerance, and that this whole endeavor was misguided.
The second possibility, which fortunately was correct, was that the maternal immune system was trigging an immune response in the neonatal host.
This slide shows the maternal cellular alloresponse after a neuro-transplantation. And without dwelling on the histograms there, if you just look at this lower graph on the right side, you can see that the frequency of alloreactive cells after in utero transplantation is actually higher, on the far right column there, than those animals that we had even intentionally immunized. And so in utero transplantation in and of itself induces a strong maternal alloresponse, at least in the mouse, to donor antigen.
This alloresponse was also present in the form of allo-corrected antibodies. And you can see that the timing of the antibody response corresponds almost exactly to the timing of loss on engraftment in our 70% of injected animals.
And the real clinching experiment here was if we took the pups away from the injected mother, who was presumably immunized, and put them on a naïve foster mother for breastmilk feeding, then 100% of the animals maintain their chimerism, and none lost their chimerism. Whereas in contrast, in our injected mothers, we had the usual 30/70 split.
So this suggested that the mother actually induced an adaptive immune response in the pups by transfer of alloreactive antibody via breastmilk to those pups.
And this was proven by taking the serum of immunized moms, giving it orally to the pups that were foster-fed, and once again, inducing a 100% loss of chimerism with maternal serum alone.
We also showed in these animals that the other side of the self-tolerance equation, the regulatory side, also corresponded to maintenance of tolerance, and that animals that maintain their chimerism had a higher and more potent T-regulatory suppressive response than those that lost their chimerism.
So the upshot of those experiments was, in the absence of maternal immune response, 100% of recipients engrafted across full MHC barriers, that tolerance occurs by mechanisms of deletion and generation of T-regulatory cells, just like it’s supposed to, and that the average levels of chimerism in this completely mismatched model were in the 20 to 30% range, which would be therapeutic for many of the target diseases that this applies to.
Importantly, from a clinical perspective, this means that if this applies in humans, which we’re unsure of, but even if it does, we can use maternal cells as a source of cells and proceed potentially with clinical trials.
So the immunologic barrier really won’t exist with the planned approach clinically.
So what about the strategy of prenatal tolerance induction followed by postnatal cellular organ transplantation?
We’ve actually shown this in three different systems, all of which are minimally n non-myeloablative, non-toxic methods of enhancing engraftment after birth. These include general lymphocyte infusion, low-dose total body radiation, and a single dose of busulfan.
And with these prepared regimens we’ve been able to achieve near complete replacement of the post-hematopoietic system. And combined with fostering the pups we were able to do this in approximately 100% of animals that we transplant.
An example of this is shown here where we have a regimen of total body irradiation of the chimeric pups after in utero transplantation. These are very, very low doses of total body irradiation, not even adequate for causing thrombocytopenia. And then after they received their radiation, they received a second transplant from the same donor strain of T-cells depleted bone marrow cells.
And with that you see a dose-dependent increase in the levels of engraftment to once again near complete replacement of the host hematopoietic system by donor cells. This is done with no host toxicity, and no graft versus host disease.
If you apply this to hemoglobinopathy, which is probably the primary clinical target for this therapy, we’ve done this in both human sickle mice, and thal mice. And if you look at these two graphs, what’s interesting about them is that the animals that have donor chimerism with the myeloidchimerism in the bone marrow, approximately 20 to 30% have near complete replacement of the peripheral red cell compartment. And you get amplification of the peripheral of engraftment and the peripheral red cell compartment because of the prolonged half-life of normal cells relative to diseased cells. So the diseased cells are cleared from the circulation and the normal cells essentially make up 100% or close to it, of circulating red cells.
So if you apply this to these models, you can correct the phenotype of these hemoglobinopathies, both sickle cell and thal. And this is supportive evidence that this can be used clinically for these diseases.
So the upshot from these experiments was basically that tolerance can be used as a platform for non-toxic postnatal strategies for cellular transplants. You can get near complete or complete donor chimerism across full MHC barriers without significant myeloablation or immunosuppression. And hemoglobinopathies, because of this amplification of red cell expression in the recipients, our attractive initial target disorders for clinical translation respondents.
So we all know that mice are not men, and we need a better pre-clinical model that more closely resembles the human for pre-clinical testing of this therapy.
We chose the canine model because historically it’s been difficult to engraft, like humans. And that suggests that the same barriers are present in a canine as in human. They’re outbred model obviously, like humans. And they’ve been used as the primary pre-clinical model for bone marrow transplantation, supporting that they have a similar hematopoietic profile and a similar threshold for graft versus host disease to humans.
Finally there are disease models in the canines, such as the leukocyte adhesion deficiency model, which are genetically and phenotypically very similar to humans.
Our initial experience with in utero transplant in the canine model, after developing a model that was similar to the methodology we would use in human, was somewhat disappointing. There were very minimal levels of initial chimerism, 0.1 to 2.1% of microchimerism. The good news was we saw no graft versus host disease, but the bad news was that this was very inconsistently associated with tolerance. And we were unable to consistently enhance chimerism with non-myeloablative preparation in these animals. So basically it wasn’t good enough for clinical translation and we had to go back to the drawing board similar to what we did in the mouse.
The common theme here is that we were doing interperennial transplants in dogs and getting inadequate results. So we went back and we looked at both the ontogeny of the immune system and the effect of delivering the cells via intravascular versus interperennial approaches.
We used a florescent dye to track the cells, and we developed a technique of an intracardiac injection to deliver the cells by an intravascular move.
When we looked at dog development, immunologic development, we identified at the timing of the expansion of double positive cells in the thymus, which is predictive of a pre-thymic phase of immune processing, and that timing was around 40 days.
And we looked at the mode of administration of cells, we saw the same thing that we saw in the mouse that basically intravascular was far superior to interperennial.
These panels here show the upper two panels of the interperennial approach in 24 to 48 hours. You can see almost no cells in the fetal liver and pool cells in the perennial cavity. Whereas with the intravascular approach in the lower two panels, you can see the high levels of circulating cells and very high levels of fetal liver engraftment after the intravascular delivery of the cells.
So this was applied in our next series of in utero transplants in dogs. And this figure shows the initial engraftment levels after inter-cardiac injections at 40 days’ gestation. And you can see that this was a really breakthrough in that we had very high levels of chimerism in most of the dogs. The average was over 10%, and 40% of the dogs were over the 20% threshold, which is required for successful treatment of sickle cell disease.
This just shows the chimerism over time in the dogs. It’s important to note that none of the dogs lost their chimerism. They were initially engrafted and the chimerism levels were actually quite stable for up to two years, which is as long as the study lasted.
We were able to demonstrate associated tolerance in these dogs. These were four animals that were, received renal transplants from their maternal donor at various time points. And these dogs had various levels of chimerism, shown by the arrows for each of the recipients.
And what we found was that the three dogs with the higher levels of chimerism were all completely tolerant to their kidney grafts; controls acutely rejected their grafts. These were animals that received IUHCT but had microchimeric engraftment. And then there was one animal that had a low grade chronic rejection of the kidney who had engraftment chimerism level of 7%. So overall these animals were, by very strict criteria of renal transplantation, tolerant to their donor animal.
So in the nine model currently, we can achieve microchimerism, or levels of chimerism associated with tolerance in over 90% of our recipients. Stable without a single loss of…instance of loss of chimerism. The chimerism levels are high enough to achieve boosting, or to allow organ grafts, and we’ve seen no GVHD or other detrimental effects with T-cell doses that we used.
So this really supports the potential for clinical application. The slide simply shows the threshold for GVHD, and we don’t see GVHD until we provide much higher levels of T-cells than we need to use to engraft the donor cells. So this is a good safety margin and supports a reconsideration in clinical trials for this endeavor.
We plan on a trial for sickle cell disease following regulatory approval, and if it works for sickle cell disease, then it should work for thalassemia, which is the most common genetic disorder in the world. If it works for hemoglobinopathies, it should definitely work for most of the immunodeficiency disorders and some of the storage diseases. So we’re optimistic about the future of in utero stem cell therapy.
The other thing I’d like to discuss today quickly is something called fetal gene therapy. And this is where we target fetal stem cells using viral bacteria’s to achieve gene transfer.
The rationale for this is similar, but not exactly like the rationale for stem cell therapy. The rationale is primarily that you can target stem cell populations in various compartments in the fetus at different stages of gestation. So at some stage of gestation, every tissue compartment will have stem cells that are exposed, if you will, for fetal transfer. And I’ll elaborate on that.
So this slide simply shows our experimental model, which is once again the [Miran] model. We use a very high resolution ultrasound system that allows us to do intra-amniotic injections at E8 or 9. That’s as early as the blastocyst stage of fetal development. We can to intracardiac, or intravascular injections at E10. And inject almost any compartment we wish to in the fetus.
This is the paradigm experiment for exposure of a specific tissue stem cells, a specific time in gestation. This shows skin development in the fetal mouse, which is identical to that in human. You start off with one layer which contains all of your [anasen] stem cells for all skin derived structures. These include your sweat glands, your hair follicles, your breast and stem cells. This is exposed to the amniotic fluid, and a simple intra-amniotic injection at E8 or E9. It can achieve a very high level of transduction and G transfer into these cells, which if you use an integrating virus like lentiviral vector, will persist for the lifetime of the animals.
And the lower panel shows a green florescent protein expression in black and white after animals have been injected at E8 or E9 at very high levels. These are adult animals in these figures. At E10 we get less expression, and that’s due to formation of the [paraderm], which obscures access to the fetal stem cells. And then after E11, we see no long-term gene transfer to stem cells because of the paraderm formation, also the stratification of the layers of the skin, which cover up the stem cells and deny access to them.
These are just other images of adult animals that were injected in utero that have high expression of GFE transgene.
And this shows that we’ve transfused the stem cells. These figures show a punch biopsy in the center of these larger biopsies of skin. And you can see the green fluorescent cells that are migrating toward the wound to aid in healing of the wound.
The figure on the upper left shows the baseline stem cells coming in from the periphery of the wound. And the figure on the lower right hand corner shows stem cells trending in from individual hair follicles, showing that we transfused the bald cells in the hair follicles, which is the other population of stem cells and skin.
What can you do with this? You can potentially treat a disease like epidermolysis bullosa with a simple intra-amniotic injection of gene vector. We’ve done this in the analogous mouse model. And by replacing the β3 sub-unit of laminin five in a knockout mouse,
We’ve been able to correct phenotype of the skin in these mice both at an electron microscopic level and at the level of preservation of skin grafts.
The CNS is another target for in utero gene therapy, and also illustrates the principle of exposure of stem cell population. So prior to closure of the neuro canal, you have the neuro [exederm], which is exposed to amniotic fluid in a simple intra-amniotic injection [unclear] vector, will transduce all three cell types in the CNS and all areas of the CNS at a very high level. And this includes even extra CNS neuro structures related to the neuro crest, for instance. And you see peripheral nerves and dorsal root ganglion for instance, transduced by a simple trans-amniotic injection.
This just shows the very high levels of neuro transduction that you get of three types of neuro cells.
Finally this is a 2-year-old mouse brain that is still glowing green. This is permanent expression of the transgene and has obvious potential application for a variety of storage disorders and neuro degenerative disorders.
Muscular dystrophies are another potential target for in utero gene therapy. Very difficult to access stem cells in muscle after birth because of the large muscle mass, its extra vascular location. And the other problem is of course the immune response if you do transduce muscle cells in the adult, you can have an immune response against the dystrophin protein, which is deficient in muscular dystrophy.
With muscular dystrophy you have an advantage for in utero treatment because you get muscle destruction of the abnormal muscle fibers, and over time if you can replace the stem cell compartment of a fetus with 10% normal stem cells, you can replace the muscle compartment with normal muscle because it is selectively preserved versus those that are diseased, which are destroyed.
This is an illustration of the potential of in utero gene therapy for muscular dystrophy. This is an E14 injection via the intravascular approach I showed you earlier using AAV vector. And you can see here that you have almost 100% transduction, and GFP expression muscle fibers, and all of the anatomic locations, including the diaphragm and intercosta musculature that are effected in muscular dystrophy.
This panel simply illustrates that we are transducing the stem cells in muscle, and these are the satellite cells that you can transduce and genetically correct in 15% of the satellite cells and muscle tissue. Then you can potentially cure the disorder. And we were able to do that with a single in utero injection of vector.
Another example is the liver. You can think of the number of genetic disorders that would be amendable to genetic correction that are liver based. And these include hemophilias, Wilson’s disease and many others. With an intracardiac injection at E10 of lentiviral vector, we can achieve a very high level or permanent [hematopoietic] expression of the transgene.
And we’ve applied this to Wilson’s disease in the mouse and have been able to correct phenotype of Wilson’s disease with a single in utero injection of the appropriate vector.
And what about anatomic abnormalities? Can you correct an anatomic abnormality by fetal gene therapy? These images show cleft lip palate, one of the most common anomalies on a worldwide basis and very devastating abnormality in third-world countries and certainly one that requires a great deal of effort to correct in the developed world.
Although we don’t know the genetic cause of cleft lip and palate in a human, it’s likely related to the TGF-β3 pathway. And the TGF-β1 knockout mouse is essentially 100% penetrance of cleft palate as shown in these images.
If you provide transient expressing of TGF-β3 at the appropriate place, which is the palatal shelf, which is accessible by intra-amniotic injection using an antiviral vector that expresses TGF-β3, then you can essentially correct the cleft pallet in these mice.
And this shows correction of an anatomic defect using in utero gene transfer with a simple intra-amniotic injection of a viral vector.
So prenatal gene therapy is very promising, but is a little farther away, I think, than stem cell therapy because of the issues related to safety of viral vectors. There are technologic barriers to gene delivery that must be overcome. There are many biological questions that remain to be answered, and we have to satisfy societal and ethical concerns related to germ line transduction; Insertion mutagenesis and other potential complications. I have no doubt that within the next 10 years to 15 years, gene vector technology will be relatively safe and specific, and we can start considering treatment of the fetus for the biological and efficiency advantages that apply.