In Utero Gene Therapy (IUGT) Using GLOBE Lentiviral Vector Phenotypically Corrects the Heterozygous Humanised Mouse Model and Its Progress Can Be Monitored Using MRI Techniques

In utero gene therapy (IUGT) to the fetal hematopoietic compartment could be used to treat congenital blood disorders such as β-thalassemia. A humanised mouse model of β-thalassemia was used, in which heterozygous animals are anaemic with splenomegaly and extramedullary hematopoiesis. Intrahepatic in utero injections of a β globin-expressing lentiviral vector (GLOBE), were performed in fetuses at E13.5 of gestation. We analysed animals at 12 and 32 weeks of age, for vector copy number in bone marrow, peripheral blood liver and spleen and we performed integration site analysis. Compared to noninjected heterozygous animals IUGT normalised blood haemoglobin levels and spleen weight. Integration site analysis showed polyclonality. The left ventricular ejection fraction measured using magnetic resonance imaging (MRI) in treated heterozygous animals was similar to that of normal non-β-thalassemic mice but significantly higher than untreated heterozygous thalassemia mice suggesting that IUGT ameliorated poor cardiac function. GLOBE LV-mediated IUGT normalised the haematological and anatomical phenotype in a heterozygous humanised model of β-thalassemia.

www.nature.com/scientificreports www.nature.com/scientificreports/ utero, and many countries have efficient screening and prenatal diagnosis programs 2 that may lead to a termination of pregnancy (TOP) if an affected fetus is found. Prenatal diagnosis can detect an affected fetus as early as ten weeks of gestation 3 .
In TM, ineffective erythropoiesis results in expanded marrow cavities that impinge on healthy bone resulting in osteoporosis and osteonecrosis. These can distort the cranium, and of facial and long bones. Symptoms include erythropoiesis in extramedullary hematopoietic sites, hepatosplenomegaly, and, in some cases, extramedullary mass. Without transfusion support, 85% of patients with severe homozygous or compound heterozygous β-thalassemia will die by five years of age from severe anemia 4 . This is a problem in the countries where the cost or blood transfusion is prohibitively expensive and often unavailable.
Fetal therapy of an affected fetus would avoid termination of pregnancy, early death or life-threatening complications for patients with no compatible donor. This would have an impact where α-and β-thalassemia are most prevalent, even where termination of pregnancy may not be available, and blood transfusions are prohibitively expensive 5 .
Possible curative options that show promise are (i) the autologous correction of β-thalassaemia patient HSCs through gene augmentation studies mediated by β-globin expressing lentiviral vectors and (ii) reactivation of the endogenous HBG through lentiviral-mediated gene silencing of HBG silencer genes (e.g. BCL11A) 6 .
Various groups have been trying to construct an ideal gene delivery vehicle, which will be able to correct the disease either in utero or postnatally [7][8][9][10][11] . We choose the GLOBE 12,13 lentiviral vector for an in utero therapy approach in a novel humanised model of β-thalassemia [14][15][16][17] . The GLOBE LV, which is now in phase I/II clinical trial (NCT02453477) was previously shown to correct β-thalassaemia in murine (Miccio et al.) 13 and human (Roselli et al.) 12 haematopoietic progenitor cells. In the study of Miccio et al., the authors showed an in vivo selection of genetically modified erythroblasts in β-thalassaemia mice 13 with 15-30% of maturing erythroblasts being sufficient for correction and correction of human HP cells respectively. Recently A.A. Thompson and colleagues showed correction of transfusion-depended β-thalassaemia and reduction or elimination of transfusion requirements using ex vivo gene therapy with LentiGlobin BB305 vector 18 .
In this study we used the humanised mouse model of β-thalassemia which had been generated by replacing the mouse adult αand β-globin genes with adult human α-globin genes and a human fetal to adult haemoglobin-switching cassette 16,17,19 . Two mouse models were used, one with functional HBB (B383: hBThal-Control) and one with non-functional HBB (B382: hBThal). The B382 with a human fetal to adult haemoglobin-switching cassette (non-functional HBB) kept in a homozygous human alpha globin and heterozygous human fetal to adult Hb-switching cassette. The homozygous animals from the later mouse strain die within two weeks if left untreated. This provides a similar temporal onset of the disease as in humans (early years) if left untreated. Animals become severely anaemic after birth and die within approximately 2-3 weeks of age, similar to the affected human neonate. However, animals can be rescued from lethal anaemia by weekly blood transfusions [20][21][22] . Heterozygous animals show features of thalassemia intermedia such as splenomegaly, extramedullary hematopoiesis, anaemia, and anisocytosis.
In this study, we show the phenotypic (haematological and anatomical) correction of the heterozygous animals using in utero gene therapy (IUGT) approach through prenatal intrahepatic injection of the GLOBE LV.

Results
In utero gene therapy corrects haemoglobin in a heterozygous humanised mouse model of thalassemia at 12 weeks' post-IUGT. Results of surgery and postnatal outcome. In order to identify the true fetal and dam survival after in utero gene therapy delivery we analysed the fetuses at E18 after IUGT in two dams. In a representative dam, where all ten fetuses were injected, at post-mortem, a total of 3 fetuses were identified as alive, while the rest of the injected fetuses suffered with fetal demise. In another dam where all six fetuses were injected, three were identified as alive at E18. This gives a survival rate after injections of 37%. This was similar to the survival rate of 36% from IUGT to post-mortem analysis at either 12 or 32 weeks postnatally, suggesting that the loss was due to the in utero injection itself.
Examination of demised fetuses revealed that they probably died immediately after the injection since the fetuses were barely seen at the E18 time-point having been resorbed over the previous 4.5 days. Indeed, in some cases, only the placenta was visible when the dam was sacrificed. The fetuses which were collected at E18 were also genotyped to identify the presence of the homozygote, heterozygote and wild-type animals. The proportion of homozygous pups in uninjected litters at E18 was 25% from n = 3 dams.
Dam survival after IUGT was 100% (n = 9). All fetuses were injected in all litters n = 67 total injected; survival from IUGT to cross-fostering was n = 25. Overall pup survival to post-mortem analysis was 37% for all experiments. There was no difference in the weight of all groups at 12 Weeks (Untreated thalassaemia Control 28.79 ± 2.08 vs Humanised Control 28.09 ± 2.12 vs IUGT 28.74 ± 1.41, p = 0.82, One Way ANOVA).
The animals were cross-fostered to CD1 time mated dams which delivered one day earlier. This was done to avoid maternal cannibalism and also to prevent the passage of maternal antibody against the transgenic protein in the milk. Genotyping at three weeks after birth identified thalassaemia animals all of which were heterozygous. There were no homozygous pups identified by genotyping suggesting that they had demised either in utero, after the intrahepatic injection, or just before cross-fostering as all live cross-fostered pups survived to genotyping. Wild-type animals were excluded from further analysis once identified.
Phenotype results. At 12 weeks of age, the peripheral blood haemoglobin concentration of treated heterozygous pups was not significantly different to that of Humanised non-thalassemia control animals (11 ± 0.21 vs 11 ± 0.22, p > 0.99, One Way ANOVA, Fig. 1B), and was significantly higher than the haemoglobin concentration of untreated heterozygous control pups (8.4 ± 0.17, p = 0.0001). The relative expression of the human beta-globin gene in the peripheral blood confirmed that the normalised haemoglobin concentration was due to the increased www.nature.com/scientificreports www.nature.com/scientificreports/ expression of the transgene (n = 10, p = 0.039, One Way ANOVA, Kruskal-Wallis test, a Two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli, Fig. 1D). At 12 weeks of age, the relative gene expression of human gamma-globin was lower in GLOBE, IUGT, treated compared to untreated heterozygous thalassemia pups (0.11 ± 0.024 versus 1.0 ± 0, n = 10, p = 0.0002, one-way ANOVA, Kruskal-Wallis test, a Two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli, Fig. 1E). This confirmed that in treated heterozygous thalassemia pups, the human gamma to beta-globin switching cassette successfully switched human gamma-globin to beta and subsequently fetal to adult haemoglobin. i.e. there was no hereditary persistence of fetal haemoglobin which could have been responsible for the higher Hb. This production of transgenic beta-haemoglobin was also verified at a protein level using high-performance liquid chromatography (HPLC) where the presence of a human beta chain peak was seen in the treated heterozygous thalassemia pups 12 weeks after injection (Fig. 1F). Data from HPLC were quantified and show an increase in the expression of human beta-globin chain at a protein level (Fig. 1B). The alpha globin in the treated animals was also higher than untreated and the level of gamma-globin was not affected by IUGT.
Heterozygous humanised thalassemia animals show evidence of splenomegaly, extramedullary and abnormal full blood count indices. In IUGT 12 Weeks animals the spleen weight was significantly less (0.12 ± 0.0065) than www.nature.com/scientificreports www.nature.com/scientificreports/ non-treated animals (0.31 ± 0.017, n = 8, p < 0.0001, One Way ANOVA), and similar to Non-thalassemia control animal 20 (Fig. 2A). The haematological indices of IUGT mice, Red Blood Cell Count (Fig. 2B) and Hematocrit (Fig. 2C) were also corrected at 12 weeks to the levels of the Humanised Non-Thalassemia Control animals. The level of extramedullary hematopoiesis (ineffective erythropoiesis), indicated by the ratio of erythroid progenitor cells (CD71+Ter119+) outside the bone marrow was also significantly lower in IUGT 12 Weeks animals (p = 0.00075, One Way ANOVA) compared to untreated thalassemia controls and not statistically different from www.nature.com/scientificreports www.nature.com/scientificreports/ Humanised Non-Thalassemia Controls (p = 0.1, One Way ANOVA) 21 (Fig. 2D). The reduction of extramedullary hematopoiesis was also confirmed by H&E staining of spleen and liver slides (Fig. 2E). These results are comparable to previous studies done on adult thalassemia mice 10,12,22 . Use of MRI techniques to monitor disease progression in animals treated with IUGT. In our second study, the in-utero gene transfer was performed as before, but the animals were sacrificed at 32 weeks of age having had MRI assessment two weeks before post-mortem. The animals were also genotyped, and wild-type animals were excluded from the analysis. The hemoglobin levels in the IUGT 32 Weeks group were also significantly higher (9.91 ± 0.36, n = 8) than that of the uninjected thalassemia control animals (8.42 ± 0.167, n = 8, p = 0.0053) but not significantly different from the humanised non-thalassemia control (10.87 ± 0.21, n = 8, p = 0.1, One Way ANOVA, Bonferroni's Multiple comparisons test, Fig. 3A). The weight of the spleen was also significantly lower (0.2 ± 0.03, n = 8) than that of the uninjected group (0.309 ± 0.017, n = 8 p = 0.006), though significantly different from humanised non-thalassemia control group (0.1 ± 0.0031, n = 8, p = 0.01, Fig. 3B).
There was a difference in animal weight in the IUGT 32 Weeks group which was higher (38.03 ± 2.93, n = 6, p = 0.01, One Way ANOVA, Bonferroni's Multiple comparisons test) than the uninjected Thalassaemia (28.97 ± 1.4, n = 9) and the age-matched humanised Non-Thalassaemia control (28.46 ± 1.53, n = 8) (Fig. 3C). This might be due to the viral integration to Socs6, which was observed during the site integration studies. www.nature.com/scientificreports www.nature.com/scientificreports/ A preclinical MRI assessment was used to investigate, in vivo, the disease progress after in utero gene therapy as described 23 . Spleen volume was initially measured using MRI, and this was then calculated as an organ volume to mass ratio for a more accurate representation of organ volume.
Encouraging results have arisen from the finding that the left ventricular ejection fraction (%) was increased in the IUGT 32 Weeks group (70.85 ± 1.58, n = 6) compared to the uninjected thalassaemia group (50.40 ± 2.83, n = 6, p = 0.0024) but was not different to the Humanised non-thalassaemia control group (68.00 ± 0.41, n = 6, p = 0.34, One Way ANOVA, Bonferroni's Multiple comparisons test, Fig. 3F). We hypothesised that the iron accumulation in control non-injected group caused a reduction in the ejection fraction, which did not occur in the IUGT treated animals that had improved cardiac function 24-26 . Monitoring Iron accumulation in thalassemia animals using MRI T2* techniques. The IUGT 32 Weeks animals were scanned using MRI T2 and T2* relaxation times.
Tissue iron can be detected indirectly by measuring, using MRI, the relaxation times of hydrogen nuclei affected by ferritin and hemosiderin iron. The proton relaxations times are shortened as a result of the iron, specifically T2. The presence of blood can spoil the myocardium and blur its borders. This can be addressed by using T2* which excludes the signal from blood, using multiecho images in late diastole 27 .
The T2 and T2* relaxation times in the spleen of the IUGT group were shortened compared to Humanised Non-Thalassaemia Control. No difference was seen between the IUGT group and the Thalassemia control, which was unexpected (Fig. 4D,G). This might be because of the increased breaking down of erythrocytes and subsequently iron accumulation in the spleens of the thalassaemia animals (IUGT and Thalassemia Control). The excess iron accumulation in the thalassaemia animals could be due to dietary iron, and excess accumulation could be specific to the animal model.
Even though the T2* in IUGT 32 Weeks was similar to uninjected Thalassemia Control group, representative Pearl stains from an animal with successful correction of the defect (HB:12.5g/dl) showed less iron pigmentation compared to an untreated thalassemia control (Hb:9.0g/dl) (Fig. 4A).
T2 relaxation times in the liver of IUGT 32 Weeks were similar to Non-Thalassaemia Control and different from the uninjected Thalassaemia Control group (Fig. 4C). This was not the case for the T2* relaxation times in the liver which were different to the Non-Thalassaemia Control and similar to the uninjected Thalassaemia Control group (Fig. 4F).
Non Invasive disease progress monitoring is possible with MRI, which corresponds to dry weight measurements 23 . For this experiment T2* was more accurate that T2 in the heart even through T2 had the same trend as T2*. This was also reported by L Jackson et al. 23 .

Site Integration (SI) Studies in IUGT 12 Weeks Recipients.
Vector integration studies were done in the animals which were sacrificed at 12 weeks. The integration site analysis was done on bone marrow and liver. Vector Integration Studies in IUGT 12 Weeks animals showed no similar integration sites (Fig. 5). The mean vector copy number (VCN) in the IUGT 12 Weeks group, in the liver, was 0.093 ± 0.14, n = 10 and the corresponding integration sites were 60.56 ± 30.73 (Fig. 6G). The VCN in the bone marrow was 0.0059 ± 0.0014, n = 10 and the corresponding integration sites were 9.25 ± 4.27 (Fig. 6H). No similar integration sites were found during the vector integration site analysis. Cumulatively the number of integration sites in the liver was 504 and in the bone marrow 51. The SI study highlighted the importance of investigating the specific vector in each biological application (IUGT) before human application. Even though oncogenesis through viral genome integration is intrinsically less likely in terminally differentiated cells, this is not the case for rapidly dividing cells of the haematopoietic lineage 28 . Themis and colleagues found a high incidence of hepatocellular carcinoma in mice receiving intravenous vitelline vein injection of the Equine Immune Anaemia Virus (EIAV) SMART 2hFIX, SMART 2Z, and SMART 3NZ lentiviral vectors in utero, but no oncogenesis was observed in HIV-1 based vector injections using the same route 29 . Thus the particular vector type may be of importance in the development of oncogenesis. Further studies suggest that the fetal mouse may be a sensitive genotoxicity model that exposes particular lentiviral-associated mutagenesis resulting in liver oncogenesis 30 . In this study, we observed integration to Peg12 gene which is associated with tumorigenesis but no hepatocellular carcinomas were observed at any time point. We did observe that IUGT animals were significantly heavier than non-thalassemic controls at 32 weeks. A possible explanation is integration to the gene SOCS-6, which was observed during the site integration studies. Socs-6 is involved in IGF-1 signalling and integration to the its site might have caused overexpression of the gene resulting to a significant weight difference 31,32 . Also, SI was observed in genes which are related to the development and function of the nervous system and also spermatogenesis. A future repeat study should include an assessment of the neural development of the offspring and also their reproductive potential, especially in males.
www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ Vectors need to be assessed for safety and shown to have low oncogenic potential, in the actual biological system since the integration profile of the same vector may differ in different cell types, ex vivo and in vivo and during different developmental stages, before gene therapy application 33 . The safety in such system needs to be addressed 34 .

Discussion
In this study, we show successful phenotypic correction of thalassaemia intermedia in a humanised mouse model using an in-utero gene therapy approach. Normalisation of haemoglobin, spleen size and a reduction in the level of extramedullary haematopoiesis were the main findings of this preclinical study. We proved that there was a successful fetal to adult haemoglobin switch in the treated mice, as this was necessary to demonstrate that the improved phenotype was due to the IUGT. In vivo mouse, MRI was used to monitor the disease 23 and showed improved cardiac function in the IUGT treated pups, even at 32 weeks after birth. We hypothesise that this was mediated through reduced iron accumulation in the heart.
Fetal application of gene therapy in mouse models of congenital disease such as haemophilia A 35 and B 36 , congenital blindness 37 , Crigler-Najjar type 1 syndrome 38 and Pompe disease (glycogen storage disease type II) [35][36][37][38][39] have previously shown phenotypic correction of the condition. Fetal gene therapy has also been found to prevent the development of fetal structural anomalies. For example transient transduction of the periderm via intra-amniotic delivery of adenoviral vector encoding transforming growth factor β3 (TGFβ3) prevents cleft palate in a mouse model of disease 40 . For obstetric conditions that affect the fetus, maternal uterine artery injection of adenovirus containing the vascular endothelial growth factor (VEGF) gene improves fetal growth in growth restricted sheep and guinea pigs 41,42 . With regard to thalassaemia specifically a study by May et al., 43 showed that www.nature.com/scientificreports www.nature.com/scientificreports/ gene transfer using a LV that contained proximal and distal regulatory elements with a large segment of the LCR (TNS9) could correct β-thalassaemia in mice, with an average increase in Hb by 3-4g/dL per vector copy and improvement of haematological indices. Similar studies have been published using other thalassaemia mouse models but with variable results 8,[43][44][45][46] .
The target cell niche in this study was the expanding population of fetal liver haematopoietic stem cells. There are intrinsic differences between fetal and adult HSCs properties especially fetal liver HSCs which can give rise to thymocytes, especially Tregs which are responsible for promoting tolerance to self and potentially foreign antigen 47 . The fetal to adult switch in these HSC properties occurs between week one and two after birth in mice. During fetal life, the HSC pools in the fetal liver are actively cycling and rapidly expanding whereas, once the HSCs reach the fetal bone marrow, they become quiescent. This property of fetal HSC pools in the liver and their potential of thymic repopulation may be an added advantage for intrahepatic IUGT 48 which is easily able www.nature.com/scientificreports www.nature.com/scientificreports/ to direct gene transfer to the fetal liver through direct injection. Fetal liver HSCs are also capable of repopulating five times more effectively long term than short term 49 . Any manipulation of liver HSCs should have a long-term effect, ideally permanent, especially if these cells migrate to the thymus 47 . Intrahepatic fetal gene transfer to haematopoietic progenitors has also been demonstrated in primates where the maximum effect was achieved using lentiviral vectors 50 .
Correction of haemoglobin was achieved with a low VCN of 0.093 in the liver. This is similar to the findings of Miccio et al. 13 where correction of haemoglobin was achieved with an HSC VCN of <1 (0.2-0.9) 13 in adult mice using ex vivo transduction of CD34+ HSCs before reinfusion in an autologous setting. The liver VCN in this study is 5-10 times lower than the HSC VCN that Miccio et al. observed 13 . The VCN in the liver was expressed in cells from the whole liver cell population and not only on the HSCs. There is a marked discrepancy between the VCN in the BM and that in the liver. This can be explained by the fact, that possibly, a large percentage of non-hematopoietic (off-target) cells were transduced following the direct intrahepatic lentiviral injection. Other authors have demonstrated that non-haematopoietic cells are transduced in an intrahepatic IUGT. The developmental stage of each organ, in this case the liver, could determine the susceptibility to in utero gene transfer and transgene expression 51 . The efficiency and distribution of transduction, is also highly depended on the type of vector used, as well as, the route of administration 52 . This could raise safety concerns, even though in a Cynomolgus Macaque model no tumours were observed, 6 years post IUGT, with an Adeno Associated viral vector. The authors, in this non-human primate study, addressed the importance for life long surveillance 53 . Thus, the VCN in the HSCs residing in the liver will be even lower. The use of direct in utero gene therapy, took advantage of the high proliferation rate of the fetal HSCs which resulted in an adequate number of transduced daughter cells that were able to correct the phenotypic Hb defect. It was not possible to do individual sorting of HSCs from each transduced mouse due to the minimal amount of tissue collected and the scope of analysis initially planned such as EMH, histology and HPLC. Thus, we were unable to calculate the VCN in the HSC population only and instead present data on the VCN for the whole heterogeneous population of hepatocytes and whole bone marrow. It was unexpected that a level of gene correction (VCN) in the BM of 0.5% resulted in such a substantial boost in hemoglobin levels in the periphery. It is possible that the HSCs might have had a greater percentage of initial correction since the fetal liver is mainly made of HSC progenitors. Alternatively this result may not be true as we concentrated on analysis of the whole bone marrow rather than just the BM HSCs and our VCN result may have been lowered due to dilution. Further studies are needed to confirm this effect. Previous studies have found low genotoxicity of lentiviral vector integration in mice that are prone to developing tumours 33 . Nevertheless, it is an important safety consideration to keep the VCN as low as possible whilst maintaining the efficacy of the gene transfer.
The humanised mouse model of beta thalassaemia 14,15,17,19 that we used is unique in that the fetal mice produce human gamma-globin (human fetal haemoglobin) in utero 15 , and have a human gamma to beta switching cassette which results in the death of homozygous animals within two weeks of birth, as the production of defective beta globin takes over. Homozygous animals can be rescued postnatally using regular blood transfusions 16 . This model allows human beta-globin producing vectors to be tested in utero 12,13,54,55 , and since the in utero phenotype is less severe than of other thalassaemia mouse models, eg th3, it allows the evaluation of IUGT to rescue the mouse model in utero and post-natally 45,56 .
In the first cohort (12 Weeks), where the animals were analysed at 12 weeks after birth, we observed normalisation of the phenotype with functional correction of anaemia, excessive extramedullary haematopoiesis and the other abnormal haematological indices. Blood films and H&E were also examined by an expert (University of Cambridge, Veterinary College). There was no evidence of malignancy and no abnormal looking red blood cells in the corrected (Hb > 10.0g/dl) animals (Fig. 6D). On the contrary in the untreated animals (Hb < 9.0g/dl), we observed the presence of moderate anisocytosis, polychromatic cells and a mixed population of normocytes, schistocytes, crenated cells and acanthocytes 21 (Fig. 6E).The presence of normal blood films in the treated animals compared to untreated was a confirmatory result. To confirm the presence of normal Hb we used real-time PCR and high-performance liquid chromatography to demonstrate beta-globin gene expression with downregulation of the gamma-globin gene as well as the additional peak from the beta globin chain in HPLC. These results excluded the hereditary persistence of fetal haemoglobin which could have contributed to the phenotypic correction of the overall haemoglobin.
In the long term 32-week postnatal group, the phenotypic correction appeared to be less effective as in the group analysed at 12 weeks after birth suggesting that the effect may reduce over time. This could be due to transcriptional silencing of the transgene 57 or even due to epigenetic changes leading to deactivation of the beta-globin gene as the pups grew 58,59 . This was evident in the lower Hb concentration at the later time point (9.91 ± 0.36g/dl at 32 weeks' vs 11 ± 0.22g/dl at 12 weeks), although the haemoglobin concentration at 32 weeks post-delivery was still significantly higher than that of age-matched controls. Histological analysis of the animals showed a correlation of spleen hyperplasia with iron deposits in the spleen, moderate anisocytosis and anaemia.
In the long-term group, MRI was used as a novel method of monitoring in utero gene therapy functional success (left ventricular ejection fraction, spleen volume, iron overload) [60][61][62][63] . We found that MRI can be used to monitor the disease progress, using T2 or T2* without sacrificing the animal. T2* is currently used clinically for the non-invasive diagnosis of iron overload in both heart and liver. It is also used for the assessment of the thalassemia major patient prior to stem cell transplantation 64 . This study confirms that this can be applied in small animal studies which can be used for the non-invasive assessment of new therapies 23 .
We did not observe any rescued homozygous pups in the genotyped litters; indeed, given the dam numbers and from previous studies in this specific humanised mouse model, we would have expected to have at least 5-6 (25% of the total) homozygous pups if they were corrected. The only homozygote pup had to be culled at 14 days of age, due to severe illness, and haemoglobin measurement showed severe anaemia (3g/dL). All live cross-fostered pups survived to genotyping. There are a number of possible reasons for this low survival of www.nature.com/scientificreports www.nature.com/scientificreports/ homozygote injected pups. It is possible that the treated homozygous pups demised in utero after the intrahepatic surgery or in the first hours after birth before they were cross-fostered due to a higher susceptibility to miscarriage or perinatal loss. Pups were usually born overnight and immediate cross-fostering was not possible as there was not always staff available to transfer them to the foster mother. It is also possible that homozygote rescue was incomplete rendering the born homozygote newborn pups more susceptible to being rejected or attacked and eaten by their mother. This has been reported in other studies where fetal intervention occurred 65 . Future studies using a higher titre of the vector will be conducted to determine whether this improves survival in the homozygous pups.
In this study, we show that targeting the specific stem cell niche 66,67 early enough in gestation with gene therapy could be critical for fetal gene transfer, especially if most of the haematopoietic stem cells are within the compartment, in this case, the liver. Alternative routes such as intraperitoneal and intravascular injection need investigation to achieve a phenotypic cure, especially rescuing the homozygous animals. Boelig et al. found that the intravenous route was most successful for achieving rapid diffuse donor cell population of the fetal liver after in utero hematopoietic cell transplantation in mice 68 . Whether this achieves the same effect for fetal gene therapy remains to be seen. Importantly for clinical translation, the tissues were studied by an expert, and no animals were found to have malignancy, which is in keeping with other studies using this stable GLOBE vector 13,18 . This can also be confirmed that GLOBE vector is currently used in Phase II clinical trial, using autologous stem cell/ gene therapy approach in adult patients 69 . The number of vector integration sites corresponded to the vector copy number, even though this was low. This could be a sign that if higher VCN is achieved, the level of haemoglobin could also rise.
The integration site analysis revealed integration to be associated with the neural system, stem cells, spermatogenesis and also with an oncogene. This must be carefully studied since the potential of tumor formation was shown to be high in a fetal context when using certain lentiviral vectors 29 . This also supports the use of an ex-vivo gene therapy approach using amniotic fluid stem cells or other highly efficient HSC progenitors 70 .
One of the benefits of perinatal treatment lies in the potential to limit organ damage through early intervention. Also, applying a therapy to the fetus, where stem cell proliferation is high, results in a higher number of transduced cells, which leads to a better therapeutic effect 71 . The fetus also offers a size advantage, allowing a higher vector-to-target cell ratio. Certain organs that are challenging to target after birth may be more easily accessible in the fetus due to their developmental stages or relative immaturity [72][73][74][75] . The fetal epidermis is an example of this, as it undergoes remodelling by programmed cell death and is replaced by mature keratinocytes. The epidermis forms a thick barrier to gene transfer following birth 76 but could be better targeted in utero 77 .
One of the main obstacles of postnatal gene therapy is the development of an immune response against the transgenic protein or the vector 78 . This is of particular importance when gene therapy aims to correct a genetic disease, which completely lacks a gene product. It is also possible that some patients may have pre-existing antibodies to the viral vector that will inhibit long-term expression of the transgenic protein; this will limit therapeutic efficacy and prevent repeated vector administration. For example, pre-existing neutralising antibodies against adeno-associated virus (AAV) serotype 2 have been shown to interfere with AAV2 vector-mediated factor IX (FIX) gene transfer to the liver [78][79][80][81] . Delivering foreign protein to the fetus can provide an advantage of immune tolerance during fetal life, a concept first suggested more than 60 years ago 82,83 . Induction of tolerance relies on the introduction and expression of the foreign protein early in gestation before the immune system is fully developed. The protein needs to remain at a detectable level within the fetus and presented to the thymus at the correct time [84][85][86][87][88] . In a postnatal treatment this can be achieved by ex-vivo correction of autologous HSCs, which will prevent an immune response, since the immune system will recognise the autologous, corrected cells as self. This can be demonstrated in a study, by A Thompson and colleagues. Using the BB305 vector, it reduced or eliminated the requirement for blood transfusions in 22 patients with severe β-thalassemia. No complications or reactions to the vector were reported 18 .
IUGT was delivered at embryonic day (E13.5) just before the time of HSC fetal bone marrow colonisation in mice, which occurs around E16-E17 89 , and which corresponds to an equivalent of 10 to 12 weeks of gestation in a human fetus 90 . Fetal gene transfer to the haematopoietic compartment 91 may offer a third option to couples with a prenatal diagnosis of a fetal congenital blood disorder, especially in the beta haemoglobinopathies 92 . For human therapy, non-invasive prenatal diagnostic techniques are being developed using maternal blood to diagnose fetal thalassaemia from 8-10 weeks of gestation 93 . This would allow enough time to deliver gene therapy to the human fetus using ultrasound-guided intrahepatic injection. Using this methodology, no ex-vivo expansion and correction of the appropriate haematopoietic progenitors is required. More animal studies are required with the aim to rescue the homozygous mouse model. If this is successful, treatment could even be applied in low-income countries where the palliative cost of homozygous patients is unsustainable for the local health systems [94][95][96] . In order to reach the clinic, a gradual progression from small animal disease models, such as the mouse, followed by studies in larger animals, for example sheep to test the feasibility and safety of the delivery method will be necessary. A reproductive toxicology study would then be required, which might be performed either in rabbits or possibly in non-human primates before establishing a fist-in-human clinical trial of fetal gene therapy 73,97 .

Methods
Animals, animal care, and procedures. Humanised CA mice (α2α1/α2α1, γ(HPFH117) β 0 /+, Heterozygous Thalassemia) and the HbA mice (α2α1/α2α1, hγβ A /hγβ A , Humanised Non-Thalassemia Controls) were obtained from Dr Thomas Ryan, the University of Alabama at Birmingham U.S.A. 15,19 . Experimental protocols were performed following the U.K Home Office Regulations and Guidelines for the Operations of Animals (Scientific Procedures) Act (1986) and approved by the local UCL animal welfare ethics committee. Mice were housed in single cages after plugging. The dams were given wet food after the laparotomy and closely monitored www.nature.com/scientificreports www.nature.com/scientificreports/ for the first seven days after the procedure. The animal experiments were done according to the project license PPL No: 70/7408 and Animals [Scientific Procedures] Act 1986 and the NC3Rs ARRIVE guidelines 2013. The operator was personal licensee PIL 70/23796. Lentiviral vector production. GLOBE Lentiviral vectors were produced as described using standard methods 13,98 . Compared to other transcription designs, GLOBE vector has only HS2 and HS3 which have the crucial LCR enhancer and chromatin-opening functions 99,100 . GLOBE-LV is a minimal LCR-β-globin transcription unit containing a 2.7 kb fragment encompassing LCR elements HS2 and HS3, linked to a fully functional mini-β-globin gene with 265 bp of 5′ and 300 bp of 3′ flanking sequences; intron 2 features an internal deletion, reducing its size to 257 bp.
Uninjected control litters. Genotyping of control litters (n = 3) was performed at E18 to determine the proportion of heterozygous and homozygous pups.
In utero injections. IUGT was performed in timed-mated pregnant heterozygous thalassemia mice as described 38,101 with minor modifications. On E13.5 day of gestation, the animals were shaved in the abdominal area; the skin was cleaned using ChloraPrep ® and under 5% isoflurane (VetTech Solutions Ltd, UK) general anaesthesia, a midline laparotomy was performed, and the uterine horns exposed under sterile conditions. 20 μl (10 7 VP/ml) of Globe LV viral particles were injected through the myometrium into the hepatic region of each fetus using a 33-gauge needle (Hamilton, Switzerland). The uterus was returned to the maternal peritoneal cavity, the abdomen was closed using an absorbable vicryl 6-0 suture (Ethicon Inc., USA) and analgesia (Marcain, 0.25% AstraZeneca) was administrated. Mice were recovered in a warm cage (28 °C) overnight. After spontaneous delivery at E20-E21 day of gestation, treated pups were cross-fostered to CD1 dams by day 2-3 after birth, to avoid maternal cannibalism and maternal antibody response to the transgenic protein [102][103][104] . Pups were genotyped at three weeks of age; earlier genotyping was precluded by the Home Office Project license. The pups were bred until scheduled post-mortem examination at 12 and 32 weeks postnatal.
Post-mortem tissue harvest. At post-mortem examination, blood (0.5-1 ml) was collected by cardiac puncture under terminal anaesthesia, induced by injecting 0.8 ml of 1.25% Tribromoethanol (Avertin Sigma-Aldrich, UK) solution. The spleen and liver were placed in 1.5 ml Safe-Lock microtubes (Eppendorf, UK). For bone marrow collection, the femur and tibia were separated from the body and muscle was removed off the bones. The epiphysis was removed from both ends and placed in 1.5 ml Safe-Lock microtubes (Eppendorf, UK). All the tissues were snap frozen in liquid nitrogen and stored at −80 °C.
Isolation of bone marrow. The bone marrow was harvested by inserting a syringe needle (Terumo, myjector 27G, 0.4 x 12 mm, Belgium) into one end of the bone and flushing 1 ml of Phosphate Buffer Saline (1x PBS, pH 7.2) (Sigma, Life Sciences, USA) repeatedly through the bone. The bone marrow aspirates were collected into 1.5 ml Eppendorf tubes and centrifuged at 300xg for 5 minutes to pellet the cells.
DNA Isolation and quantitation. DNA was isolated from bone marrow aspirates using Invitrogen PureLink Genomic DNA Mini kit (Cat No: K1820-01) as per manufacturer's instructions. The concentration of the DNA samples was by the NanoDrop 1000 Spectrophotometer (Thermo Scientific, UK). The quality of the DNA samples was determined by calculating the A260/A280 ratio. PCR for genotyping. Animals were genotyped using primers specific for mouse or human haemoglobin (Table 1). The reactions were done in 0.2ml polypropylene PCR tubes (Corning) in a Thermo Scientific thermal cycler using MyRedTaq PCR mix (Bioline).
RNA isolation and quantitation. RNA was extracted from blood and bone marrow samples of mice using the Bioline Isolate II RNA Mini kit (BIO-52072) manufacturer's instructions. RNA from the human samples was extracted using TRIzol reagent (Invitrogen) per manufacturer's instructions. RNA samples were quantitated using the Nanodrop 1000 spectrophotometer (Thermo Scientific, UK) and purity was ascertained by A260/280 ratio (>1.9). qpcR using taqman ® probes. Reverse Transcription: cDNA was synthesised by using Bioline Sensifast cDNA synthesis kit (Bio-65053). Taqman ® oligonucleotide probes have been used in the qPCR experiments (Table 1)  The reactions were run on the Applied Biosystems Step-One Plus Real-Time PCR system machine. Results were analysed using the Life Technologies Step-One software.
www.nature.com/scientificreports www.nature.com/scientificreports/ Determination of RNA expression levels by qPCR. Expression levels of human β and γ-globin chains (target genes) were assessed in the blood and bone marrow samples of treated and untreated mice at 12 weeks. The relative quantity of each target gene studied was calculated by the ΔΔCt method.
Assessment of vector copy number by qPCR. Viral titers were quantified by determination of integrated viral copy number (VCN) in cell genomic DNA or unintegrated proviral vector genomes in the cytoplasm by a TaqMan ® real-time qPCR method. The qPCR reactions were set-up in 96-well reaction plates (MicrAmpTM Optical; Applied Biosystems, USA) and analysed using Butler's late reverse transcriptase (LRT) primer/probe set 105 .
The endogenous titin was used to determine absolute numbers of genomes present in each qPCR reaction using a primer/probe set. Titin, also called connectin, is a protein kinase involved in the structural assembly of microfilaments in muscle and during chromosome segregation and cell division in non-muscle cells. Standard curves were generated using serial dilutions (10 6 copies to 10 1 copies) of the GLOBE and the Titin plasmids (PGK plasmid) 106  Primers and probes. The primers already described in the literature 19 were used for the study. The oligonucleotides were custom-made from Eurofins MWG Operon, Germany (Table 1).

Haematological indices.
Peripheral blood was collected from anaesthetised mice using a heparinised syringe into BD micro trainer EDTA collection tubes (Fisher Scientific). Red blood cell counts and Hematocrit values were measured. Hb concentrations were determined by Hemocue Hb 201+ system. Full blood counts and Serum Protein Electrophoresis (SPE) analysis for all the samples were performed by Central Diagnostics facility at the University of Cambridge. Animal and Spleen weight was also measured using a high precision balance (Kern, UK).
In vivo MRI. In vivo imaging, spleen volume, cardiac function and relaxometry were performed as described in L. Jackson et al. 23 , at 30 weeks of age using a 9.4T MR system (Agilent Technologies, Santa Clara, USA) equipped with 1000 mT/m gradient inserts and a 39mm volume resonator RF coil (RAPID Biomedical, Rimpar, Germany). A small animal physiological monitoring system (SA Instruments, Stony Brook, NY) was used to record the ECG trace, respiration rate, and internal temperature. Animals were anaesthetized with a mixture of isoflurane and oxygen with physiological measurements used to maintain the depth of anaesthesia 23 .
Cardiac function. Cardiac function was assessed with a gradient echo cine MRI sequence with a temporal resolution of 5.2 ms, an in-plane spatial resolution of 117 μm and a slice thickness of 1 mm 23 . The left ventricular blood pool was segmented at systole and diastole using Segment v1.8 R0462 107 and the corresponding volumes used to calculate left ventricular ejection fraction (EF), stroke volume (SV) and end systolic/diastolic volumes (ESV/EDV) 62 .
Site integration studies. SI was performed by the Nationales Centrum für Tumorerkrankungen (NCT) Heidelberg, German Cancer Research Center (DKFZ), Department of Translational Oncology, Germany.
Statistical analysis. All the haematological indices in the different groups (IUGT 32 Weeks, IUGT 12 Weeks, age-matched Humanised Non-Thalassemia Control, age-matched Uninjected Thalassemia Control group) were compared using one-way ANOVA with Bonferroni's multiple comparison tests. For the qPCR data, statistical significance was determined by One Way ANOVA and Kruskal-Wallis test. To achieve a statistically meaningful result, for each group we aimed to have at least six mice pups in each group, born by more than three different dams, in order to have more than three independent biological replicates Graph Pad Prism Version 7.0a for MAC, Graph Pad Software, LA Jolla California, USA was used for statistical analyses and graph drawing.