Author contributions: A.T., S.A., A.D., M.K., C.B., A.C., and O.N.: collection and assembly of data, data analysis and interpretation; Z.K.: manuscript reading; C.v.K. and S.F.: financial support, manuscript reading; E.P., S.C., M.S., and C.E.: data analysis and interpretation, participation in manuscript writing; S.C. and M.S.: administrative support; P.L.: provision of patient, administrative support, financial support, data analysis and interpretation, manuscript writing, final approval of manuscript; L.M.-C.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing.
CEA, Institute of Emerging Diseases and Innovative Therapies (iMETI), Fontenay aux Roses, France
CEA, Institute of Emerging Diseases and Innovative Therapies (iMETI), Fontenay aux Roses, France
INSERM U962 and University Paris Sud 11
Correspondence: Leïla Maouche-Chrétien, Ph.D., CEA/INSERM U962, Institut des Maladies Emergentes et des Thérapies Innovantes, 18 route du panorama, 92,265 Fontenay aux Roses, France. Telephone: +33–146-5471-76; Fax: +33–146-5474-99; e-mail: firstname.lastname@example.org,
A patient with βE/β0-thalassemia major was converted to transfusion-independence 4.5 years ago by lentiviral gene transfer in hematopoietic stem cells while showing a myeloid-biased cell clone. Induced pluripotent stem cells (iPSCs) are a potential alternative source of hematopoietic stem cells. If fetal to adult globin class, switching does not occur in vivo in iPSC-derived erythroid cells, β-globin gene transfer would be unnecessary. To investigate both vector integration skewing and the potential use of iPSCs for the treatment of thalassemia, we derived iPSCs from the thalassemia gene therapy patient and compared iPSC-derived hematopoietic cells to their natural isogenic somatic counterparts. In NSG immunodeficient mice, embryonic to fetal and a partial fetal to adult globin class switching were observed, indicating that the gene transfer is likely necessary for iPSC-based therapy of the β-hemoglobinopathies. Lentivector integration occurred in regions of low and high genotoxicity. Surprisingly, common integration sites (CIS) were identified across those iPSCs and cells retrieved from isogenic and nonisogenic gene therapy patients with β-thalassemia and adrenoleukodystrophy, respectively. This suggests that CIS observed in the absence of overt tumorigenesis result from nonrandom lentiviral integration rather than oncogenic in vivo selection. These findings bring the use of iPSCs closer to practicality and further clarify our interpretation of genome-wide lentivector integration. Stem Cells2013;31:1785-1794
In contrast to somatic hematopoietic stem cells (HSCs), induced pluripotent stem cells (iPSCs) can be readily propagated before inducing their differentiation into hematopoietic progenitors with multilineage potential. Gene correction [1, 2] or addition in iPSCs followed by selection of cell clones bearing vectors at sites of chromosomal integration with low genotoxic potential  offer a potentially attractive strategy for clinical development. Yet, many hurdles and unknowns remain. In addition to the potential tumorigenic risk, which might be decreased by the use of nonintegrative vectors, we currently lack methods for generating HSCs capable of robust long-term reconstitution of transplanted hosts from human iPSCs . Another issue relevant to β-hemoglobinopathies is whether persistence of fetal hemoglobin (HbF) or embryonic-fetal-adult globin class switching will occur in vivo after transplantation with human iPSC-derived hematopoietic cells .
Clinical gene therapy trials using transduced CD34+ cells have brought new questions that include the unclear biological meaning of “common integration sites (CIS)”. The concept of CIS was initially coined to describe recurrent retroviral integrations at similar genomic sites in mice injected with replication-competent retroviruses. These integration sites (IS) could be identified because they activated adjacent proto-oncogenes, causing the subsequent generation of tumors . The concept of CIS has been recently extended to cases of recurrent retroviral/lentiviral integrations at certain loci in a given host and/or between individuals, even in the absence of any evidence of a resulting malignancy. Because of a lack of baseline information about the normal range in numbers and sizes of clones that sustain blood cell outputs lifelong, the biological meaning and potential genotoxic risk associated with clones bearing CIS remains unclear .
The primary goal of this study was to bring together in vivo data from the only gene therapy patients transplanted at the time with lentiviral vectors (LVs), one with β-thalassemia  and two with adrenoleukodystrophy , with parallel data obtained from iPSCs generated from the same β-thalassemia patient, to shed light on both the genotoxic potential of lentivector integration and globin gene expression pattern in iPSC-derived erythroid cells (endogenous and transferred globin genes). To this end, we generated Thal-iPSCs 3.5 years ago from the first β-thalassemia patient (transfusion-dependent βE/β0-thalassemia major) who was transplanted with globin LV-transduced autologous CD34+ cells on June 7, 2007 . This patient who became transfusion-independent in 2008 showed a semidominant myeloid-biased cell clone bearing a globin-LV within the HMGA2 gene.
Materials and Methods
Registration of Clinical Trials
Patient samples were obtained with appropriate authorization from the Agence Nationale de la Sécurité du Médicament et des Produits de Santé (ANSM) of France and signing of informed consent forms. The trials are registered with ANSM and the European Clinical Trials Database (EudraCT).
Thal-iPSC Generation and Characterization
The detailed protocol is described in supporting information.
Transduction of Thal-iPSC with Therapeutic Globin LV
Cells of Thal-iPS clone #4 (Thal-iPSC4), cultured on matrigel, were transduced with a batch of the clinical grade βA(T87Q)-globin LV, used in the thalassemia clinical trial . Five independent transductions were performed at several multiplicities of infection (between one and 50) during 4 hours in the presence of 8 µg/ml of protamine sulfate. At least 2 weeks later, average vector copy numbers were determined by qualitative polymerase chain reaction (qPCR) with primers that amplify the packaging signal Ψ+-gag and the endogenous β-actin gene as described . Cells with an average of 0.9–1.1 LV copies per cell were chosen for subsequent analysis and referred to as Thal-iPSC4G (as a pool). These cells were plated again on Mouse Embryonic Fibroblasts (MEFs) for further amplification.
Hematopoietic Differentiation of Thal-iPSC
Hematopoietic differentiation was induced either by coculture on hematopoietic stromal cells or by embryoid body (EB) strategy. The murine stromal cell line MS5 was maintained in α-Minimum Essential Medium (MEM) (PAA, Piscataway, NJ) supplemented with 10% fetal bovine serum (FBS) . Cells were mitotically inactivated with mitomycin and plated in gelatin-coated six-well culture plates at a concentration of 6 × 105 cells per well. Three wells of Thal-iPSC colonies were harvested by enzymatic (collagenase IV) and mechanical (Cell lifter) dissociation. Cells were seeded on MS5 cells, and the medium was changed to a basic hematopoietic differentiation medium (basic HDM) made of KnockOut Dulbecco's Modified Eagle Medium (KO-DMEM) (Life Technologies, Grand Island, NY) supplemented with 20% FBS (STEMCELL Technologies, Vancouver, Canada), 1 mM l-glutamine, 0.1 mM nonessential amino acids (Life Technologies), 0.1 mM β-mercaptoethanol, 50 µg/ml ascorbic acid, and 200 µg/ml human holo-transferrin (Sigma-Aldrich, St. Louis, MO), without added cytokines. The medium was changed every 3 days.
For EB formation, undifferentiated Thal-iPSC at confluence were treated with collagenase IV. The colonies were collected and transferred to low-attachment dishes (Corning, Tewksbury, MA) in basic HDM. The medium was changed the next day and later every 3 days, using basic HDM medium supplemented with 50 ng/ml of Bone morphogenetic protein 4 (BMP4) and a mixture of hematopoietic cytokines: 300 ng/ml Stem cell factor (SCF), 300 ng/ml FMS-like tyrosine kinase 3 (FLT3), 10 ng/ml interleukin [IL]-3, 10 ng/ml IL-6 and 50 ng/ml Granulocyte colony-stimulating factor (G-CSF)). The cytokine cocktail is identical to that frequently used to support hematopoietic differentiation of human ESCs and iPSCs [11-13].
At different time points, single-cell suspensions were prepared from differentiated Thal-iPSC. The cocultures were treated with 1 mg/ml of collagenase B (Roche Applied Science, Indianapolis, IN) for 1 hour at 37°C, washed with phosphate-buffered saline (PBS), and then treated with 0.25% trypsin/EDTA for 8 minutes at 37°C. EBs were dissociated using collagenase B for 2 hours followed by a 10-minute incubation at 37°C with enzyme-free cell dissociation buffer (Life Technologies, Grand Island, NY). Single cell suspensions were obtained by gentle pipetting and filtering through a 70-µm mesh size cell strainer (BD, San Diego, CA).
Colony forming unit (CFU) assays were performed by plating 200,000 cells into MethoCult H4434 (StemCell Technologies, Vancouver, Canada) according to the manufacturer's instructions. Burst forming unit-erythroid (BFU-E) colonies were picked, washed with PBS, and frozen at −80°C for DNA extraction and HPLC analysis.
Hemoglobins (Hbs) from individual BFU-E colonies were separated by ion-exchange HPLC on a PolyCAT A column (PolyLC Inc.), as described .
For FACS analysis, cells were resuspended to a density of 1–2 × 105 cells per 50 µl in PBS containing 0.5% bovine serum albumin and 2 mM EDTA (PBE) and reacted with specific mouse anti-human antibodies from BD or eBioscience. Live cells were analyzed using a FACSCanto II flow cytometer equipped with FACSDiva software (BD).
Transplantation in NSG Mice
Unsorted cells or CD34+ and/or CD45+ sorted cells from the differentiation cultures were injected intrafemorally into 7- to 10-week-old, sublethally irradiated (315 cGy 137Cs γ-rays) NSG (Nonobese diabetic/severe combined immunodeficient IL-2 receptor γ-chain-null [NOD.Cγ- Prkdcscid Il2rγtm1Wjl/SzJ]) mice at doses of approximately 2 or 0.2 × 106 cells per mouse, respectively. After 3 and 8 weeks, sequential femoral bone marrow (BM) aspirates were obtained and then, after 10–12 weeks, the mice were killed. BM cells obtained from individual mice were stained with anti-human CD45, CD15, CD33, CD66b, CD19, CD20, CD34 antibodies (all from BD) and glycophorin A (10F7MN, from P. Lansdorp, Terry Fox Laboratory, Vancouver, BC, Canada) after blocking with 10% human serum and anti-mouse FcR antibody (STEMCELL Technologies). To try to enhance the output of mature human erythroid cells, 100 units of human recombinant erythropoietin (EPO)were injected intraperitoneally into each transplanted mouse 24 and 48 hours before harvesting BM cells for the week 3 analyses. Stained cells were analyzed using a FACSCalibur (BD, San Diego, CA) with gates established from analyses of BM cells aspirated from control mice (not transplanted) and subjected to the same antibody staining protocol. To obtain isolate populations for further analyses, specific subsets were sorted using a BD FACSAria II. NSG mice, originally obtained from the Jackson Laboratory (Bar Harbor, MN), were subsequently bred and maintained in the animal facility of the British Columbia Cancer Research Centre. All procedures were carried out according to protocols approved by the University of British Columbia Animal Care Committee.
PCR and RT-qPCR
Total RNA was extracted from bulk cells using the PureLink RNA Mini Kit (Life Technologies) and from individual colonies or sorted CD45–GlyA+ erythroid cells produced in vivo in NSG mice using an ArrayPure Nano-scale RNA purification kit (Epicenter Biotechnologies, Madison, WI). After DNase I treatment, reverse transcription was performed using SuperScript III (invitrogen). Genomic DNA was isolated with a NucleoSpin Blood Kit (Macherey-Nagel, Düren, Germany). All primers used in real time (RT)-PCR, RT-qPCR, and genomic PCR analyses are shown in supporting information Table S1. Detection of the βA(T87Q)-globin LV in individual hematopoietic colonies was carried out by PCR with specific primers (LGF10, LGR11) and (Epo24, Epo26) as described . Quantification of the vector in genomic DNA was performed as described . TaqMan probes were obtained from Applied Biosystems ready to use or were synthetized by Eurogentec (supporting information Table S1). qPCR assays were carried out in duplicate or triplicate with the ABI PRISM 7300 detection system (Applied Biosystems, Grand Island, NY). A recombinant DNA plasmid containing the five human globin genes (ζ, α, ε, γ, β) subcloned in tandem was used to generate standard curves for quantification. For all standard curves, the correlation coefficients (R2) were above 0.99. The relative fold-changes in expression were calculated using the ΔΔCt method.
LIS Analysis and Statistics
Genome-wide mapping of lentivirus integration sites (LISs) in Thal-hiPSC4G was performed by high-throughput 454 DNA pyrosequencing after Linear Amplification Mediated - PCR (LAM-PCR) using two restriction enzymes (Tsp509I and NlaIII). To track LIS heterogeneity over time, genomic DNA was prepared from these Thal-iPSC4G pools after 10, 23, 36, 53, and 66 population doublings and from individual BFU-Es. CIS identification and analysis were accomplished by the aid of the methods and programs (idsincisdet, CISUNIFc, CISLENTIc, coinc1) described in Abel et al. .
The lentiviral control dataset is an in silico dataset based on computer simulations. We made use of the program CISLENTIc for our analysis to compare the observed numbers of CIS with the exprected values under a lentiviral distribution. We applied “conditional analysis”, which implies that simulations are performed chromosome-wise by distributing the observed or expected number of IS in the gene coding regions and their complement, respectively, on each chromosome. Inside of each region, the distribution is uniform. “Conditional analysis” means that it is conditional both on the observed number of LIS on the chromosomes and on the observable model parameters. In this conditional analysis, the number of LIS is attributed to each chromosome (under H0) to gene or nongene regions.
The different steps of the study are summarized in supporting information Fig. S1.
Generation and Characterization of iPSCs
We generated an adherent CD34−CD73+CD90+CD105+ mesenchymal stem cell (MSC)-like population in vitro from BM CD34− cells harvested from a gene therapy thalassemia patient , and verified the absence of a preintegrated βA(T87Q)-globin LV in the genomic DNA of these MSCs by PCR analysis (data not shown). We then transduced these MSCs with retroviral vectors expressing OCT4, SOX2, KLF4, and cMYC and obtained several iPSC clones. These expressed the anticipated pluripotency markers and showed silenced expression of all four transferred retroviral vectors (Fig. 1A and 1B). All iPSC clones injected into immunocompromised mice produced teratomas containing tissues of all three germ layers (Fig. 1C). One clone, referred to as Thal-iPSC4 was chosen for further studies. Its karyotype performed with G-/R-banding at Passage 17 was normal and that of a male.
To investigate the hematopoietic potential of Thal-iPSC4, we generated EBs in the presence of hematopoietic cytokines and 12–24 days later, assayed the cells for hematopoietic colony-forming cells. Both lineage-restricted and bipotent myelo-erythroid progenitors were detected in these assays (Fig. 1D). Thal-iPSC4 cocultured on MS5 stromal cells in absence of supplementary cytokines also differentiated into hematopoietic cells as indicated by a CD45+CD34+ phenotype (Fig. 1E). This cell population increased over time during the coculture from an average of 3% ± 2% at day 12 to reach 3.8% ± 1.3% and 6.7% ± 2.1% at days 19 and 22, respectively. During the same time period, the CD34+CD45− cell population was 8.0% ± 3.9%, 10.5% ± 1.7%, and 10.4% ± 0.4% at days 12, 19, and 22, respectively.
Differential Expression and Regulation of Endogenous Globin Genes in Thal-iPSC Versus Blood-Derived BFU-Es
To characterize the expression of the various endogenous globin genes in erythroid cells generated in vitro from Thal-iPSC4 derived BFU-Es, we first examined globin transcript levels by RT-qPCR. For comparison, we applied the same analysis to BFU-E colonies obtained (i) from cryopreserved blood cells of the same βE/β0-thalassemia (Thal) patient before the gene therapy procedure was performed, (ii) from the blood of a healthy adult, and (iii) from a sample of normal cord blood (CB). Peripheral blood contains approximately 100 BFU-Es/ml .
We found that ratios of α- and β-like globin mRNAs were similar in the progeny of BFU-Es obtained directly from the Thal patient or from the normal CB or adult blood samples (Fig. 2A). In contrast, the Thal-iPSC4-derived BFU-Es produced erythroid progeny that displayed a phenotype with strong embryonic and fetal features. This phenotype included the production of ζ-globin mRNA at a level that represented 4% of the combined α-like globin mRNAs, which stands in contrast to the barely detectable levels of this globin mRNA found in any of the other three samples. Similarly, in Thal-iPSC4-derived cells, β-like globin mRNA ratios, γ/(ε+γ+β) and ε/(ε+γ+β) reached 47% and 52%, respectively, with β/(ε+γ+β) at only 1% (Fig. 2A).
To quantify α-/β-like Hb from the same cells, we performed HPLC analysis on individually plucked colonies. The results showed that the Thal-iPSC4 cells produced mainly fetal (HbF) and embryonic Hbs in variable proportions (Fig. 2B and 2C), but no adult abnormal HbE (Fig. 2B). On average, HbF levels constituted 65% of the total Hb, although with a substantial degree of variability (range 36–100%) (Fig. 2B and 2C). Cells dissociated from iPSCs-derived BFU-E colonies were phenotyped by flow cytometric analysis and found to express the human erythroid-specific GpA antigen (supporting information Fig. S2).
To further investigate the regulation of endogenous globin gene expression in the erythroid cells derived from Thal-iPSC4, we examined the levels of BCL11A-XL (ultra-long), BCL11A-L (long) and SOX6 mRNAs. The choice of these transcripts was based on the known importance of these factors for adult-fetal globin class switching. We found that the first two were expressed at 20-fold and sevenfold lower levels, respectively, in the thalassemia patient's BFU-E-derived cells as compared to the cells derived from a normal adult donor, whereas SOX6 mRNA levels were the same (Fig. 3A). In the cells produced by CB BFU-E, these transcripts were also reduced relative to the adult cells (by sevenfold, 20-fold, and sixfold, respectively). However, all of these transcripts were much more profoundly reduced in the cells produced by Thal-iPSC4-derived BFU-Es (333-fold, 200-fold, and 20-fold, respectively, (Fig. 3A). Interestingly, the β/γ globin mRNAs ratios in the four groups of erythroid cells were significantly correlated (p < .05) with the levels of BCL11A-L and SOX6 gene expression (Fig. 3B).
Expression of βA(T87Q)-Globin in the Erythroid Progeny of LV-Transduced Thal-iPSCs
We then transduced Thal-iPSC4 with the βA(T87Q)-globin LV to generate Thal-iPSC4G (as a pool). When subjected to our differentiation protocol, these cells generated hematopoietic progenitors (BFU-Es, CFU-Gs, CFU-Ms, and CFU-GMs) as readily as the parental Thal-iPSC4 (supporting information Fig. S3A). PCR on genomic DNA of individual BFU-E-derived colonies showed that approximately 70% of them contained an integrated globin LV (supporting information Fig. S3B). All of these transduced colonies expressed βA(T87Q)-globin in addition to HbF, but HbE was not detected (Fig. 4A; supporting information Fig. S3D). Nonerythroid but globin LV-sequence containing colonies (CFU-GM.44G+) derived from the Thal-iPSC4G cells did not express the βA(T87Q)-globin gene (supporting information Fig. S3C), confirming that expression of this transgene is erythroid specific.
The levels of HbAT87Q in the erythroid colonies obtained from Thal-iPSC4G cells ranged from 25–75% (average of 43%) of the total Hbs and from 40–85% (average of 60%) of the combined HbAT87Q and HbF, as quantified by HPLC (Fig. 4B; supporting information Fig. S3D). 70% of the Hb expressed in erythroid cells derived from BFU-Es of normal adults was HbA (with 30% HbF), also with substantial variation in levels (data not shown). Similarly, BFU-Es obtained from the Thal patient before gene therapy produced erythroid cells in which 66% of the total Hb was HbE (supporting information Fig. S3E), whereas after gene therapy these values were converted to 30% HbAT87Q and 30% HbE . Expression level of HbAT87Q did not increase with the vector copy number (Fig. 4B). Together, these data indicate that a high level expression of HbAT87Q was obtained when Thal-iPSC4G cells were induced to differentiate in vitro into Hb-producing erythroid cells.
Characterization of the Hematopoietic Progeny of Thal-iPSC-Derived Cells Regenerated in NSG Mice
To determine whether Thal-iPSC and Thal-iPSC4G cells would produce erythroid progeny in vivo and to characterize the cells obtained, we subjected these to our MS5 hematopoietic induction coculture protocol for 15–22 days without cytokines before transplanting cells into NSG mice. Because there was no preexisting evidence as to what would be the optimal cell subset to repopulate NSG mice, three iPSCs-derived hematopoietic cell populations were transplanted intrafemorally into sublethally irradiated NSG mice, as follows: (1) CD34+ cells isolated by FACS, (2) CD34+CD45+ isolated by FACS, or (3) unsorted cells having undergone hematopoietic differentiation.
Analysis by cytofluorometry of unsorted differentiated cells around Day 19, showed 10% hemato/endothelial CD34+ bipotent cells, 4% CD34+CD45+ hematopoietic cells and 1.2% endothelial VEGFR2+CD31+ cells. The cells were negative for the erythroid specific marker GpA, and we did not look at megakaryocytic markers (e.g., CD41) (Fig. 1E and data not shown). Two NSG cohorts (Groups A and B) with a total of 24 mice were transplanted, plus three positive (CB CD34+ cells) and two negative (no injection) controls (supporting information Table S2A and S2B). Thal-iPSCs transduced with the β-AT87Q-globin lentivector were used for Group A, whereas untransduced Thal-iPSC were used for Group B. Sequential BM aspirates were obtained and analyzed 3–12 weeks later.
Human GpA+CD45− erythroblasts were detected in eight (plus the two positive control) mice at frequencies up to 133%, for the better mouse, of the lowest positive control. Human myeloid CD15/66b/33+ cells of the granulocyte–macrophage lineage and CD19/20+ cells of the B-lineage populations were detected in several mice 3 weeks post-transplantation (Fig. 5Aa and 5Ab; supporting information Table S2A). After 8–12 weeks, transient human hematopoietic populations of small size were detected whereas no human cells were found thereafter (Fig. 5Ac; supporting information Table S2B). Three mice showed GpA+ cells at both three and 12 weeks post transplantation. No mouse was positive for myeloid CD15/66b/33+ or B-lineage CD19/20+ populations at both 3 and 12 weeks (supporting information Table S2A and S2B).
Using an anti-human CD3, we were unable to detect T-cells at 3 or 12 weeks post-transplantation. Of note is that CD3 cells represented only 1–5% of total hCD45+ cells in our positive controls, and this is in agreement with published reports . We did not look for NK cells. In a third NSG cohort (Group C in supporting information Table S2C), mice were injected with 2 × 106 unsorted, 2 × 105 CD34+ or 2 ×105 CD34+CD45+ sorted cells derived from iPSC transduced with the globin gene therapy vector. Four mice were weakly positive for GpA cells at 3 weeks, and no GpA cells were detected at 8 or 12 weeks. A large number of human CD45+ cells was detected in most mice 3 weeks post transplantation, but a very small number of such cells and in fewer mice were detected at 8 weeks, with only two mice being positive for CD15/66b/33 12 weeks post transplantation. Injection of sorted versus unsorted cells did not improve the hematopoietic engraftment in this experiment, as previously observed in mice of Group B (supporting information Table S2B and S2C).
Human erythroid cells were isolated from NSG mouse BM by fluorescence activated cell sorting of cells stained with an antibody specific for human GpA. We extracted RNA from two positive mice (AI7R and BA1N) from Groups A and B and quantified globin transcripts using probes that recognize specifically endogenous human βA, β0, and βE-globin cDNAs but not the βA(T87Q)-globin cDNA. From this analysis, we found no ζ-globin mRNA in the Thal-iPSC-derived human GpA+ cells produced in the mice α/(ζ+α) = 100% (Fig. 5B). Evidence of a switch from an embryonic to a fetal Hb program was also reflected in the expression pattern of β-like genes ε/(ε+γ+β) = 0% and γ/(ε+γ+β) = 91%. A trend toward an adult program was also evident from level of β expression β/(ε+γ+β) = 9% (Fig. 5B). By comparison, GpA+ cells regenerated in NSG mice transplanted with CD34+ CB cells also showed no evidence of embryonic (ζ and ε) globin gene expression and showed 68% β/(ε+γ+β) (Fig. 5B).
Ranking of Globin LV Integration According to Genotoxicity Potential
It was also of interest to compare the LV integration sites (LISs) obtained in the Thal-iPSC4G cells with those detected in the transplanted patient. Five independent iPSC transduction experiments were performed. Two of them were chosen for further analysis because qPCR showed that cells from these two transduction experiments had an average of 0.9-1.1 LV copies per cell, which is similar to what was obtained in the β-thalassemia gene therapy patient. Genome-wide mapping of LIS in Thal-hiPSC4G was performed by high-throughput 454 DNA pyrosequencing after Linear Amplification Mediated - PCR (LAM-PCR) using two restriction enzymes (Tsp509I and NlaIII). We identified a total of 1,074 unique (LISs) in all of the Thal-iPSC4G cells examined, of which 21% (227 LISs) were assigned to regions of low (“safe harbors”) and 79% (847 LISs) to regions of high genotoxic potential. The distinction between regions of low or high genotoxic potential was made on the basis of criteria previously published , with the difference that we used the cancer gene list of the Memorial Sloan Kettering Cancer Center . In the latter group, 213 LISs were located in or near (±300Kb) cancer genes - a frequency that is significantly higher than expected from LV experimental control data set (p < .001, chi-square test). In contrast, only 5% (54) of the LISs were located in or near tumor suppressor genes (±300kb), and this was not a significant association.
No LIS was detected in ultraconserved regions, but 25% (271) of LISs were identified in or near (±300Kb) sequences that encode miRNAs. Notably, after 66 population doublings (PDs), only a few LISs (n = 13) were detected in the Thal-iPSC4G cultures (Table 1). However, there was no significant change in the distribution of LISs in regions of high or low genotoxicity potential throughout Thal-iPSC4G culture. PDs of iPSC lines were estimated by sampling cell numbers throughout serial passages. Our estimated PDs are in good agreement with calculations based on published population doubling times for ES and iPSC (i.e., approximately 32–38 hours) [20, 21].
Table 1. Distribution of βA(T87Q)-globin lentiviral vector integration sites with low or high genotoxic potential among transduced Thal-iPSC4G, according to published criteria  and the list of the 12 left genes at PDL 66 with low or high genotoxic potential
PDLs of Thal-iPSC4G
Low genotoxic potential (%)
High genotoxic potential (%)
NRD1: Tumor suppressor gene; 22248 bp downstream of MIR761; NDST4: Cancer gene; Stability, Oncogene; 275213 bp upstream of MIR1973; TAF2: In gene, Intron 18; XPO7: In gene, Intron 1; SFRS15: In gene, Intron 18; UBQLN1: In gene, Intron 1; 268606 bp downstream of MIR7-1; TSC22D1: In gene, Intron 1; ZC3H3: Tumor suppressor gene; Stability; PFTK1: Cancer gene; TNF_alpha_NF_kB; MACC1: 284363 bp upstream of MIR3146; PDGFRA: In gene, Intron 1; SOST: 16824 bp downstream of gene; 292022 bp downstream of MIR2117
Abbreviations: IS, integration sites; PDL, population doubling level
Total unique IS
In addition to the analysis of undifferentiated Thal-iPSC4G, we determined the distribution of LISs in 24 individual BFU-Es derived from Thal-iPSC4G that had undergone 23 or 36 population doublings. We found that 13 BFU-Es colonies of the 24 contained a single copy of the βA(T87Q)-globin LV, and for one of these (8%), the LISs satisfied the criteria for lower genotoxicity potential (supporting information Table S3).
Generic LV Common Integration Sites
We then compared the distribution of the 1,074 LISs obtained in the Thal-iPSC4G cells with that found in the blood of the isogenic thalassemia gene therapy patient (357 LIS) . To this comparison, we added the LIS dataset (4,859 LIS) obtained from blood samples of the two first human patients included in the hematopoietic gene therapy trial for adrenoleukodystrophy (ALD) . LIS of both Thal-iPSC4G cells and the ALD-Trial were identified by LAM-PCR and the enzymes Tsp509I, NlaIII or HpyCH4IV. For LIS detection in the β-Thalassemia trial, we made use of LM-PCR and the enzymes MseI and NlaIII. CIS were detected and classified as follows: 2 LISs within 30 kb for second order, 3 LISs within 50 kb for third order, 4 LISs within 100 kb for fourth order, and n LISs within 200 kb for the nth order .
Computer simulations based on assumed LV distribution (CISLENTIc , established that CISs of order ≥3 were unlikely to have occurred by chance alone and there were 176-fold, 1,879-fold, 11,471-fold, 134,783-fold, and 1,420,000-fold more CIS of third, fourth, fifth, sixth, and seventh orders, respectively, than expected for the 1,074 LIS of the Thal-iPSC4G dataset. In the iPSC dataset, we found that 18.06% of all experimental LISs are involved in CIS (194 observed LISs; 36.62 expected LIS; p < 0.00001). In the Thal gene therapy patient blood dataset, we observed that 15.78% of all experimental LISs are involved in CIS (83 observed LISs; 9.15 expected LISs; p < 0.00001). We then ran a “MegaBase (MB)-scale analysis”, based on kernel convolution, on the three datasets and observed a good concordance between overlapping peaks (regions of integration) and our CIS definition in the Kb scale (supporting information Fig. S4). Interestingly, most of the CIS of order ≥3 were common between ALD and β-Thal samples (Thal-iPSC4G and/or Thal patient, Table 2). This indicates that certain genomic regions were preferred targets for LV integration in all three studies, in which the vector DNA insert was not the same and a common selection pressure could not be invoked as an explanation, unless one surmizes that in vitro selection during iPSC culture mimicks in vivo selection after cell transplantation. Notably, none of the 13 LISs remaining after extensive Thal-iPSC4G culture was involved in a CIS.
Table 2. The 50 most prominent common integration site clusters
Cluster size (bp)]
Nb IS in cluster
% LIS within genes
Next RefSeq genes
This list was determined on the basis of fulfillment of the following criteria: common integration site (CIS) must be formed by LIS from at least two studies (ALD, Thal patient, Thal-hiPSC4G) and should be of sixth order or greater (fifth order if present within a larger CIS cluster). ALD represents two adrenoleukodystrophy patients enrolled in the lentiviral gene therapy clinical trial . Thal represents thalassemia patient enrolled in the lentiviral gene therapy clinical trial , iPSC represents Thal-iPSC4G. Also see Supporting Information Figure 3.
We then determined the “50 most important CIS” from the combined dataset (Table 2). These 50 CISs are grouped as follows: 31 are shared by the three datasets, six are only shared by ALD and Thal patient, 13 are only shared by ALD and Thal-iPSC4G, and none are shared only by Thal patient and Thal-iPSC4G. Use of the Ingenuity Pathway Analysis (IPA) program to examine the CIS clusters that were found to be shared between the cells in the two clinical trials (six CIS clusters affecting 17 genes) revealed that 10 of the 17 genes are associated with the top functions “Cancer, Hematological Disease, Immunological Disease”. In addition, seven of the 17 LISs were found within HLA-gene regions and were observed only in datasets obtained from ALD and Thal patients but not in any of the 2 Thal-iPSC4G transduction experiments.
Discussion and Conclusions
Although exploration of the potential future use of iPSCs for the gene therapy of the β-hemoglobinopathies has been initiated [2, 3], we aimed here at sheding light on certain key remaining issues by comparing relevant outcome parameters of recent clinical trials with quantifiable observations in iPSCs.
We observed that the use of human mesenchymal stem cells (from BM CD34- cells) together with valproic acid treatment was highly conducive to iPSC formation. Thal-iPSCs generated from the gene therapy thalassemia patient were readily induced to differentiate into hematopoietic cells in vitro using conventional protocols. Analysis of endogenous Hbs produced in vitro showed high levels of ε- and γ-globin expression and barely detectable endogenous β-globin expression. Complete embryonic-to-fetal as well as a clear trend toward fetal-to-adult globin class switching occurred after injection of Thal-iPSC derivatives into NSG mice. It is thus unlikely that switching to the adult β-globin expression phenotype would not occur in vivo after transplantation in humans, making β-globin gene addition or correction necessary.
Injection into NSG mice of purified (CD36+CD235a+CD71+) erythroid cells produced in vitro from human iPSCs resulted in complete terminal maturation . However, human cells were no longer detected after 4 days in transplanted mice. This is clearly distinct from this study where more primitive hematopoietic cells were injected, resulting in human GpA and β-globin expression ≥3 weeks after transplantation together with B-cell formation.
BCL11A and SOX6 expressions were well correlated with γ/β-globin ratios in BFU-Es derived from Thal-iPSC4, indicating that BCL11A and SOX6 control γ and β globin gene expression in this setting. An alternative to globin gene transfer might be the erythroid specific knock down of BCL11A by means of shRNA thereby forcing γ-globin expression and β-globin downregulation. Another approach might be the modulation of nuclear receptor TR2/TR4 activity, which has been shown to play a role in γ-globin gene expression .
Lineage specificity of expression and lack of silencing of the transferred βA(T87Q)-globin gene were observed. The variability in βA(T87Q)-globin expression in BFU-Es per integrated LV copy in vitro was not greater than that observed for endogenous globin genes, thus pointing to an intrinsic variability in globin gene expression . The high expression of the transduced βA(T87Q)-globin achieved is consistent with levels obtained in the gene therapy patient, where steady-state βA(T87Q)-globin expression per copy is estimated to be at least 70% of endogenous levels of a normal human β-globin gene in circulating red blood cells .
Human cell output was sustained but low in transplanted NSG mice, similar to that observed after engraftment of human ES-derived hematopoietic cells in mouse models [26, 27]. In our study, transplantation of CD34+ or CD34+ CD45+ sorted cells did not improve NSG reconstitution as compared to transplantation of unsorted cells. To our knowledge, convincing data on sustained engraftment of human iPSC-derived cells had not been reported before this study.
If and when remaining hurdles to the use of HSCs derived from iPSCs for human patients are surmounted, such as with the help of nonintegrative expression of iPSC-inducing genes, it is envisaged that patient-specific autologous iPSCs will be generated and HSCs obtained by in vitro differentiation. HSCs derived from iPSCs may then be corrected by gene editing (e.g., TALE/Cas9 nucleases) or modified by lentiviral transfer of a therapeutic gene followed by screening for integration in chromosomal areas of lower genotoxic potential. Appropriate pretransplant conditioning of the patient will then be applied before intravenous or intrabone infusion of corrected/modified HSCs at a suitable dose following cell expansion and, if applicable, ex vivo or in vivo selection for corrected/modified cells.
In agreement with a previous report , we found that a substantial fraction of LIS in Thal-iPSC4G (8% here vs. Seven percent in the previous study, using the same criteria) were located in relatively safe areas of the human genome. However, we noticed that LIS heterogeneity diminished rapidly over time in LV transduced iPSCs, indicating the importance of selecting iPSCs with safer LIS soon after LV transduction. We did not detect globin LV integration into the HMGA2 gene of Thal-iPSC4G.
Transduced Thal-iPSCs did not have LISs within the HLA locus, and this contrasts with findings with blood cells from both Thal and ALD gene therapy patients and from CD34+ hematopoietic progenitor cells . This difference may be explained by the fact that HLA genes are not expressed in ES or iPS cells, consistent with the evidence that loci containing expressed or poised genes are the most likely candidates for LV insertion [29, 30].
Our comparative data sets strongly suggest that most identified CIS, in the absence of blatant malignancy and in vivo selection, result from LV integration biases and are not, per se, reliable predictors of genotoxicity and oncogenesis. Caveats include the relative small size of the LIS dataset from the Thal gene therapy patient blood cells, slight differences in methods employed (LAM- vs. LM-PCR), and the possible existence of in vitro selection during extensive culture of LV transduced iPSC, although we did not observe any change in the distribution of LIS over time in vitro between regions of high and low genotoxicity potential. However, CIS formation may, in limited instances, point to benign in vivo selection, when self-limited cell proliferation or relative clonal dominance occurs and dysregulation of adjacent gene expression can be documented. This may be the case of CIS identified within or in the vicinity of HMGA2 [8, 31-34].
We thank C. Courne, B. Gillet-Legrand, M. Granger-Locatelli, M. Pla, M. Chopin, and S. Guenounou for excellent technical contributions, S. Fichelson (INSERM) for providing MS5 cells and I.H. Park (Yale University) for providing the four reprogramming MIG vectors, G. Tachdjian, L. Tosca, and D. Pineau (Béclère Hospital) for Thal-iPSCs Karyotyping. This work was supported by INSERM and ANR Chaire d'Excellence to P.L., a grant to C.E. from the Canadian Stem Cell Network, Grant 01GM1004 (GETHERTAL) to M.S. by Bundesministerium für Bildung und Forschung BMBF; A.T was supported by a fellowship from Mahidol University and an RGJ-PHD scholarship from the Thailand Research Fund, S.A. by the French Research Ministry and the French Society of Hematology (SFH), A.C. by a Croucher Foundation Fellowship, and M.K. by a Studentship from the Canadian Institutes of Health Research and the Michael Smith Foundation for Health Research. A.T. and S.A. contributed equally to this article.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.