Experimental support for IUHCT

Fleischman and Mintz administered transplacental injection of donor BM cells at E11 into fetal mice with a stem cell deficiency in the absence of c-kit . These early studies were directed toward questions in stem cell biology rather than IUHCT as a therapeutic approach; nevertheless, they established the basic principle that high levels of engraftment could be achieved when a competitive advantage exists for donor cells. In these studies, the degree of erythroid replacement correlated with the degree of underlying anemia, with complete early replacement by donor erythroid cells in lethally anemic homozygous mice. Blazar et al. extended this to lineage deficiency by demonstrating only lymphoid reconstitution (split chimerism) in the mouse severe combined immunodeficiency (SCID) model, which has a T-cell proliferation and survival defect . These studies reaffirm the importance of host cell competition, whether at the level of the stem cell, or lineage progenitor in limiting engraftment after IUHCT.

Until relatively recently, the majority of studies of IUHCT in normal animal models including the goat , dog , primate , and mouse demonstrated minimal or no detectable engraftment. The exception is the ovine model. Early experiments in the ovine model achieved an allogeneic engraftment level of up to 30% after a single injection of fetal liver-derived donor cells . In general, however, results obtained in the sheep model have not been translated into human clinical application, thus limiting its value as a preclinical model.

With increasing experimental and clinical experiences, it became clear that there were significant barriers to successful engraftment after IUHCT, and that the fetal hematopoietic environment posed very different challenges than those encountered in postnatal BM transplantation . In order to systematically examine the relative importance of these barriers and develop strategies to overcome them, the development of the hematopoietically normal allogeneic murine model was critical. The normal mouse initially proved very difficult to engraft, making it a legitimate model for investigation of barriers to engraftment .

Ultimately, the murine model has become more relevant and useful with the achievement of “macrochimerism,” that is, easily measured donor chimerism of ≥1%. This was achieved primarily through intraperitoneal delivery of higher doses of cells or fetal cells. With the generation of macrochimeric mice, it became evident that chimerism levels >1–2% resulted in the consistent association of donor-specific tolerance across full MHC barriers. A mechanistic analysis of tolerance in chimeric mice supported a primary mechanism of deletion of donor-reactive lymphocytes, although deletion was not complete implicating the coexistence of peripheral tolerance mechanisms . However, despite this evidence and the presumptive preimmune status of the fetus, approximately only 30% of the recipients of allogeneic cells were chimeric suggesting the presence of an immune barrier following IUHCT. The development of the intravascular injection model via the vitelline vein allowed delivery of much larger doses of cells and led to our reexamination of immune tolerance. By performing early tracking of donor cells as well as long-term assessment of donor chimerism, we were able to document that 100% of allogeneic and congenic recipients maintained high levels of engraftment up to 3 weeks after IUHCT. However, between 3 and 5 weeks, 70% of allogeneic animals lost their engraftment, whereas 100% of congenic animals remained chimeric . We confirmed the presence of an adaptive cellular and humoral alloresponse that is quantitatively higher in non-chimeric versus chimeric animals. The pivotal observation explaining this inconsistency in the mouse model was that if transplanted pups were placed with foster mothers that had not received IUHCT, 100% of the recipients maintained their chimerism . Maternal alloimmunization by IUHCT resulted in transfer of alloantibodies to the pup via breast milk, inducing an adaptive alloimmune response in the pup with a subsequent loss of chimerism. Notably, the study confirmed that in the absence of a maternal immune response either via foster nursing or through the use of maternal donor cells, engraftment and tolerance were uniformly present via a mechanism of partial deletion of donor-reactive T cells and the induction of a potent T-regulatory cell response. In a separate study using the murine intraperitoneal model of IUHCT and fetal liver-derived donor cells, maternal fetal T-cell trafficking was implicated (albeit indirectly) in the loss of chimerism . Obviously, murine placentation, maternal–fetal trafficking of antibodies and cells, and the time course of events after IUHCT are considerably different in mice when compared to large animal models or during human pregnancy. Nevertheless, it raises the question of whether maternal immunization is an issue in large animal models and clinical circumstances, and whether it is a limitation to engraftment after IUHCT. This critical question remains to be answered for IUHCT. Until this question is addressed, any clinical application of IUHCT should use maternal cells to prevent maternal immune response. Additional recent studies in the murine model provide further information on the significance of donor cell trafficking and mechanisms of donor antigen presentation in the thymus, resulting in bidirectional deletion of donor- and host-reactive cells and generation of T-regulatory cells suppressive of any residual alloresponse that occurs in the periphery . These studies further support the potential of IUHCT to achieve true donor-specific tolerance under appropriate circumstances.

Experimental support for IUHCT

Fleischman and Mintz administered transplacental injection of donor BM cells at E11 into fetal mice with a stem cell deficiency in the absence of c-kit . These early studies were directed toward questions in stem cell biology rather than IUHCT as a therapeutic approach; nevertheless, they established the basic principle that high levels of engraftment could be achieved when a competitive advantage exists for donor cells. In these studies, the degree of erythroid replacement correlated with the degree of underlying anemia, with complete early replacement by donor erythroid cells in lethally anemic homozygous mice. Blazar et al. extended this to lineage deficiency by demonstrating only lymphoid reconstitution (split chimerism) in the mouse severe combined immunodeficiency (SCID) model, which has a T-cell proliferation and survival defect . These studies reaffirm the importance of host cell competition, whether at the level of the stem cell, or lineage progenitor in limiting engraftment after IUHCT.

Until relatively recently, the majority of studies of IUHCT in normal animal models including the goat , dog , primate , and mouse demonstrated minimal or no detectable engraftment. The exception is the ovine model. Early experiments in the ovine model achieved an allogeneic engraftment level of up to 30% after a single injection of fetal liver-derived donor cells . In general, however, results obtained in the sheep model have not been translated into human clinical application, thus limiting its value as a preclinical model.

With increasing experimental and clinical experiences, it became clear that there were significant barriers to successful engraftment after IUHCT, and that the fetal hematopoietic environment posed very different challenges than those encountered in postnatal BM transplantation . In order to systematically examine the relative importance of these barriers and develop strategies to overcome them, the development of the hematopoietically normal allogeneic murine model was critical. The normal mouse initially proved very difficult to engraft, making it a legitimate model for investigation of barriers to engraftment .

Ultimately, the murine model has become more relevant and useful with the achievement of “macrochimerism,” that is, easily measured donor chimerism of ≥1%. This was achieved primarily through intraperitoneal delivery of higher doses of cells or fetal cells. With the generation of macrochimeric mice, it became evident that chimerism levels >1–2% resulted in the consistent association of donor-specific tolerance across full MHC barriers. A mechanistic analysis of tolerance in chimeric mice supported a primary mechanism of deletion of donor-reactive lymphocytes, although deletion was not complete implicating the coexistence of peripheral tolerance mechanisms . However, despite this evidence and the presumptive preimmune status of the fetus, approximately only 30% of the recipients of allogeneic cells were chimeric suggesting the presence of an immune barrier following IUHCT. The development of the intravascular injection model via the vitelline vein allowed delivery of much larger doses of cells and led to our reexamination of immune tolerance. By performing early tracking of donor cells as well as long-term assessment of donor chimerism, we were able to document that 100% of allogeneic and congenic recipients maintained high levels of engraftment up to 3 weeks after IUHCT. However, between 3 and 5 weeks, 70% of allogeneic animals lost their engraftment, whereas 100% of congenic animals remained chimeric . We confirmed the presence of an adaptive cellular and humoral alloresponse that is quantitatively higher in non-chimeric versus chimeric animals. The pivotal observation explaining this inconsistency in the mouse model was that if transplanted pups were placed with foster mothers that had not received IUHCT, 100% of the recipients maintained their chimerism . Maternal alloimmunization by IUHCT resulted in transfer of alloantibodies to the pup via breast milk, inducing an adaptive alloimmune response in the pup with a subsequent loss of chimerism. Notably, the study confirmed that in the absence of a maternal immune response either via foster nursing or through the use of maternal donor cells, engraftment and tolerance were uniformly present via a mechanism of partial deletion of donor-reactive T cells and the induction of a potent T-regulatory cell response. In a separate study using the murine intraperitoneal model of IUHCT and fetal liver-derived donor cells, maternal fetal T-cell trafficking was implicated (albeit indirectly) in the loss of chimerism . Obviously, murine placentation, maternal–fetal trafficking of antibodies and cells, and the time course of events after IUHCT are considerably different in mice when compared to large animal models or during human pregnancy. Nevertheless, it raises the question of whether maternal immunization is an issue in large animal models and clinical circumstances, and whether it is a limitation to engraftment after IUHCT. This critical question remains to be answered for IUHCT. Until this question is addressed, any clinical application of IUHCT should use maternal cells to prevent maternal immune response. Additional recent studies in the murine model provide further information on the significance of donor cell trafficking and mechanisms of donor antigen presentation in the thymus, resulting in bidirectional deletion of donor- and host-reactive cells and generation of T-regulatory cells suppressive of any residual alloresponse that occurs in the periphery . These studies further support the potential of IUHCT to achieve true donor-specific tolerance under appropriate circumstances.

Clinical application of IUHCT

The clinical experience with IUHCT worldwide was previously reviewed and will not be reviewed here. In summary, the experience is discouraging with the exception of SCID, which has been successfully treated by IUHCT in a number of centers . However, SCID is a unique disorder that facilitates donor T-cell survival and proliferation, and the engraftment achieved has only been documented to reconstitute the T-cell lineage (split chimerism). Thus, it can be stated that IUHCT has not been clinically successful in establishing engraftment in a hematopoietically competitive recipient. In the current era, however, the consistent achievement of therapeutically significant levels of donor cell engraftment in the normal murine model and, more importantly, the canine model suggests that the competitive barrier in human fetuses can be overcome. Toward that end, recently, a meeting of interested translational and basic scientists and clinicians was convened, where a consensus statement was developed describing guidelines for clinical trials of in utero stem cell transplantation . The regulatory process and planning is currently under way to support a multicenter trial of IUHCT for the hemoglobinopathies.

Methods of gene transfer

There are a number of gene transfer methods to choose from depending upon one’s goals and the circumstances that apply. Nonviral DNA delivery methods have been proposed as safer alternatives to viral vectors . However, these methods have generally been too inefficient or impractical for most in vivo applications. In contrast to nonviral methods, gene transfer by viral vectors is a relatively efficient and extremely versatile approach. Ideally, viral vectors harness the viral infection pathway but avoid the subsequent replicative expression of viral genes that causes toxicity. Because the vector is ultimately responsible for the transfer of genes to the fetus, the choice of vector is of utmost importance in fetal gene therapy. While a complete discussion of viral vector technology is beyond the scope of this review, a specific vector type is usually chosen based on a number of considerations such as tissue tropism, packaging capacity, the ability to integrate into host genomic DNA, immunogenicity, and the ability of the investigator to obtain or manufacture the vector.

Rationale for specific vector application to IUGT

The use of a specific vector for fetal gene transfer is evolving toward newer-generation vectors that provide both relative safety and efficacy. Although first-generation adenoviral vectors were highly efficient with broad tropism, they did not integrate into the host genome and are highly immunogenic. This resulted in short duration of expression in rapidly dividing fetal cells, and significant inflammatory responses when administered after the onset of immunocompetence . Experimental applications for adenoviral vector include any circumstance where only short-term gene expression is desirable, and/or when transgene expression is required within 12–24 h after transduction . Cleft palate is an example, wherein a missing gene is required for a short duration during a critical developmental event, that is, palatal fusion . Adeno-associated virus vectors (AAV) are single-stranded DNA viruses that are gaining in application due to their safety and low immunogenicity . Although they integrate into the genome at low frequency, most of the AAV expression is episomal and therefore limited in rapidly dividing fetal tissues. However, tropism can be influenced by viral serotype and the vectors can be targeted to tissues with low turnover such as skeletal muscle, liver, and the CNS to achieve relatively durable expression. The limitations of AAV are their limited packaging capacity and slow expression profile; that is, transgene expression may take up to two or more weeks. Permanent expression of transgene after IUGT in most tissues requires gene transfer to the stem cell population with an integrating viral vector. Retroviral vectors were the first integrating vectors to be used. However, retroviruses require cell division for transduction, a major limitation for the transduction of relatively quiescent stem cell populations. More recently, lentiviral vectors, including those derived from a replication-incompetent human immunodeficiency virus (HIV), have been introduced and are highly efficient at infecting dividing and nondividing cells with low immunogenicity . Pseudotyping the lentivirus with a specific viral envelope, such as vesicular stomatitis virus protein G (VSVG), improves lentivirus stability and helps target the transgene to specific tissues .

Modes and timing of prenatal gene transfer

In addition to the type and titer of vector used, the developing fetus offers a multitude of variables that profoundly influence the distribution and efficiency of transduction. The primary variables correspond to the mode and timing of vector administration and are again integrally associated with normal developmental anatomy and events. A prime example is the amniotic cavity. This amniotic space is formed at the initial stages of embryogenesis and is theoretically accessible for gene transfer from the developmental stage of formation of the bilaminar embryonic disk throughout the remainder of gestation. The transduction efficiency of injecting the vector into the amniotic space would depend on the ability of the vector to come in contact with the cell, and the tropism and titer of the vector used. However, for each stem cell population or tissue that comes in contact with the amniotic fluid, there is a developmental window of accessibility for specific cell populations . For instance, in the amniotic cavity, macroscopic changes of embryonic body shape, such as folding and closure, may determine the period of direct contact with the amniotic fluid ; differentiation of epithelium, such as formation of the periderm and epithelial stratification in skin, may obscure access to the expanding stem cell or progenitor cell population ; and fetal physiological movements such as breathing and swallowing help extend the distribution of amniotic fluid to internal spaces exposing additional cell populations . In this context, it is not surprising that major differences in distribution of transduction are observed with intra-amniotic injection at different developmental stages. For instance, skin stem cells that give rise to all of the skin and skin appendages can be easily and efficiently transduced between E8 and E10 but the formation of periderm makes them inaccessible thereafter . Similar examples of this principle have been described for the neuroectoderm , lung epithelium , and other tissues.

A primary determinant of the distribution of transduction is the developmental compartment that the vector is delivered into. For instance, the distribution of transduction after intra-amniotic administration (ectodermal and neuroectodermal) is markedly different from the administration of vector into the extra-coelomic cavity (heart, kidney, and pancreas) at the same time in gestation. Similarly, intravascular administration via an intracardiac injection at E10 results in a completely different distribution of transduction (hematopoietic, endothelial, osteogenic, liver, and cardiac) from intra-amniotic or extra-coelomic administration. This can be particularly relevant to targeting specific organs and providing transduction specificity in the presence of anatomic barriers. In general, it can be stated that the dependence of stem cell accessibility on the developmental stage is greater for all modes of administration at earlier gestational time points than later; likewise, the efficiency of stem cell transduction is, in general, greater earlier than in the later stages of development. In addition, advantages of prenatal gene transfer such as immunologic tolerance and the vector particle-to-cell number ratio are favored by gene transfer during early developmental stages. Therefore, we believe that the most promising results will be obtained when prenatal gene transfer is applied earlier, rather than later.

The potential for gene editing in prenatal gene therapy

The recent development of gene-editing tools has the potential to revolutionize gene therapy with major implications for future prenatal gene therapy strategies. The ability to modify a chosen sequence in its native locus offers remarkable advantages over current gene therapy technology; if this is adequately specific and efficient, many of the concerns cited subsequently can be eliminated completely. A corrected gene operating under its normal regulatory and transcriptional elements would potentially eliminate concerns related to random gene insertion, over- or under-expression, and germ-line alteration (except for the better). While gene-editing technology has been available for some time in the form of zinc finger nucleases (ZFN) or TAL effector nuclease (TALEN), the efficiency was low (ZFN) and the nuclease design and assembly requirements were challenging (ZFN) or extremely labor intensive (TALEN). In 2012, Jinek et al. published a ground-breaking study demonstrating that an immune mechanism in bacteria that uses RNA-guided endonucleases could provide a straightforward approach for constructing customizable nucleases for gene targeting . This technology (clustered regularly interspaced short palindromic repeats – CRISPR + Cas endonuclease complex = CRISPR/Cas) has proven highly efficient and could be applied to correct genetic abnormalities in autologous cells (i.e., amniotic fluid stem cells), or to even correct the percentage of endogenous stem cells in situ by in vivo administration. Considerable research is required to reduce off-target events, but this technology has far-reaching implications for all aspects of gene therapy.

Adverse effects and safety concerns

Prior to the application of in utero gene therapy in humans, a number of safety concerns must be addressed. While the risks of postnatal gene therapy have been recognized and extensively discussed, specific risks may be higher for the fetus than for the postnatal subject. The risks that cause the most concern include insertional mutagenesis, disruption of normal organ development, and germ-line transmission .

Insertional mutagenesis is a major concern with integrating viral vectors and is the subject of intense investigation. The clinical observation of four cases of T-cell leukemia, diagnosed 31–68 months after retroviral-mediated gene therapy for X-linked severe combined immune deficiency (SCID-XI), led to a temporary halt of gene therapy trials using a retroviral vector. Lymphocyte analysis revealed that insertional mutagenesis had occurred in all four cases . In the only study documenting oncogenesis after prenatal gene transfer, Themis et al. reported a high incidence of postnatal liver tumors in mice after prenatal injection with an early form of third-generation equine infectious anemia virus (EAIV) vectors with self-inactivating (SIN) configuration but not when using a similar vector with an HIV backbone . It remains unclear whether insertional mutagenesis led to tumor formation, but it demonstrates that the fetus may be particularly sensitive to certain vectors.

The effects of prenatal gene transfer on organ development need to be considered with any prenatal gene transfer strategy. Certainly, strategies involving expression of growth factors, transcription factors, or other regulatory molecules have significant potential to alter normal organ development, particularly early in gestation. A notable example is our recent finding that interstitial expression of fibroblast growth factor 10 in the developing rat lung results in cystic adenomatous malformations . In addition, direct toxicity of either the virus or inappropriately regulated transgene would have great potential to impact organ growth.

Germ-line transmission is a safety concern as well as a bioethical issue. The goal of in utero gene therapy is to modify somatic cells, but undesired gene transfer to germ-line cells is possible. In the human fetus, the primordial germ cells are compartmentalized in the gonads at 7 weeks of gestation . The germ line should only be accessible through the vascular system, so targeted gene therapy that is administered after this time period should not affect the germ line. Several groups have directly assessed germ-line transmission after in utero gene transfer. Tran et al. administered intraperitoneal injections of retroviral vector into fetal sheep and examined the germ cells by polymerase chain reaction (PCR) and breeding experiments . Gene transfer to the germ line was not detected. Lee et al. investigated germ cell transfer after intraperitoneal, intrapulmonary, and intracardiac administration of HIV-1-derived lentiviral pseudotyped VSV-G vectors in fetal rhesus monkeys . At early postnatal time points, there was no sign of a transgene in the germ cells of monkeys that received intrapulmonary or intracardiac injections. The intraperitoneal approach, however, resulted in gene transfer to a subpopulation of female gonadal cells, but it was not detected in male gonads. Porada et al. also investigated germ cell transmission after intraperitoneal retroviral gene transfer in fatal sheep. Breeding studies were negative, but PCR on the purified sperm from injected rams and immunohistochemistry of sectioned testes showed low-level transduction of germ cells , which was dependent on the gestational age of the injected fetus . These studies suggest that the frequency of germ-line transduction is low and related to gestational age and mode of vector administration.