Although remarkable progress in the correction of murine models of β-thalassemia and sickle cell anemia has been reported with the use of lentiviral globin vectors, translation of these results to large animals is essential to address many issues aimed at the maximization of both safety and the likelihood of ultimate clinical success. For human clinical trials to follow, the development of maximally safe vector systems that result in β-globin expression sufficient to correct the phenotype is needed. Our strategy for the development of a model to achieve these goals involves several steps including lentiviral vector optimization for nonhuman primate use, evaluation of stem cell transduction efficiency through in vivo
transplantation, maximization of both vector-directed globin gene expression and safety of maximally performing vector systems, and, finally, determination of the minimal degree of host conditioning required for adequate engraftment of genetically modified cells (Tisdale and Sadelain, 2001
; Persons and Tisdale, 2004
Indeed, one of the most difficult issues remains the choice of vector systems. To achieve lifelong clinical efficacy in most disorders of the blood, permanent integration of the vector in hematopoietic stem cells is required. In addition, for disorders of globin synthesis, gene expression must be erythroid specific, high level, and sustained. Lentiviral vectors have inherent advantages in this regard, with promising results obtained in murine models (Rivella and Sadelain, 2002
) and verified by several groups (Pawliuk et al.
; Levasseur et al.
; Persons et al.
). In the nonhuman primate model, lentiviral transduction of HSCs with an SIV-based system resulted in moderate levels of in vivo
gene transfer and expression in vivo
(Hanawa et al.
). Although encouraging, we sought to develop vectors based on HIV-1 given the longer duration of use, and the accruing experience with these vectors in human clinical trials (Dropulic and June, 2006
). However, when HIV-1-based vectors have been used to transduce rhesus HSCs, poor gene transfer rates were observed owing at least in part to a species-specific block due to a saturable intracellular inhibitor (Besnier et al.
; Cowan et al.
). Here we developed a modified HIV-1 vector bearing an alternative cyclophilin-binding domain (Kootstra et al.
) and demonstrated avoidance of this block in nonhuman primate HSCs in vitro
. The modified vector performed well in vitro
, with high transduction rates achieved in both human and nonhuman primate CD34+
cells when assessed by the percentage of colonies containing integrated vector by PCR.
To determine the percentage of cells expressing human β-globin, a specific erythroid assay was developed on the basis of small differences between rhesus and human β-globin at the amino acid level. High homology exists between human and rhesus globins, as they are more than 95% identical at both the cDNA and amino acid sequence levels (HomoloGene database at http://www.ncbi.nlm.nih.gov/homologene
). The monoclonal antibody (H2, kindly provided by L. Stanker) used in this assay recognizes the human specific β-125 amino acid site where one amino acid change between rhesus and human exists (Stanker et al.
). To assay erythroid cells, a modified (Wojda et al.
) Fibach two-phase erythroid culture system (Fibach et al.
) was developed from which vigorous erythroid output was obtained. This culture system allows the growth of erythroid cells from transduced CD34+
cells that are subsequently permeabilized and assayed with the H2 antibody. Using this technique, we were able to demonstrate by simple flow cytometry an equally high percentage of cells expressing human β-globin after transduction. Although no large-animal models for the thalassemias or the hemoglobinopathies exist, important information regarding the amount of human β-globin expressed per vector copy should ultimately be measurable in this context.
With these tools in place, two rhesus macaques underwent in vivo
transplantation with TNS9.3-transduced autologous PB CD34+
cells. In one animal, human β-globin expression at approximately 5% was demonstrated by flow cytometry at early time points. This level of human β-globin expression was confirmed by RNase protection assay (data not shown) and similar levels were estimated at the DNA level by genomic Southern blotting. Furthermore, expression of human β-globin was confirmed by mass spectrometry. Although high human β-globin gene transfer rates in rhesus CD34+
cells and the efficient production of human β-globin could be demonstrated in vitro
, the moderate level of human β-globin production in vivo
was limited to only a short period posttransplantation, requiring PCR for detection long term. Indeed, PCR optimization for such low levels observed during long-term follow-up may have resulted in an underestimation of the early marking. Factors such as hematopoietic progenitor cell preparation, transduction method, and cytokine combination were all similar to former protocols that our group has already demonstrated elsewhere to be sufficient for sustained, moderate levels of in vivo
gene transfer (Kiem et al.
; Wu et al.
). In addition, the transduction methods were modeled after those used with SIV in the rhesus macaque (Hanawa et al.
). A number of other factors could explain the decline in human transgene expression over time. Others have demonstrated more efficient transduction of human peripheral blood CD34+
cells, using the amphotropic envelope in a nonobese diabetic/severe combined immunodeficient (NOD/SCID) mouse transplant model (Hanawa et al.
). We are currently comparing alternative envelopes such as the amphotropic envelope to determine whether this can improve transduction of the more primitive hematopoietic stem cell compartment. However, efficient transduction and engraftment of repopulating cells with VSV-G-pseudotyped lentiviral vectors in the pigtail macaque have been described (Trobridge et al.
), and these promising results are likely due to an escape from host restriction by TRIM5α in this model. In addition, equivalent in vivo
marking levels were demonstrated when oncoretroviral vectors pseudotyped with either the amphotropic or VSV-G envelope were compared (Shi et al.
). HIV-1 vectors encoding marker genes have been more successful in the rhesus model, although a high degree of variation remains present (An et al.
). This discrepancy with our results could relate to the difficulty in achieving high-titer vectors encoding human β-globin along with key regulatory elements. A reduction in absolute HSC numbers could account for the fall in marking derived from the long-term repopulating cell compartment as the gene marking level at 4–6 months or greater posttransplantation represents the “true HSC” marking level (Kim et al.
). Unfortunately, our gene marking levels detected by Q-PCR over 6 months posttransplantation were approximately 0.01% in both animals, certainly insufficient to expect clinical benefit in humans, and potentially indicating that our vector system/transduction method has not achieved efficient transduction to LT-HSCs.
The restriction to infection by HIV in rhesus cells was shown to be only partially overcome by modification of the cyclophilin-binding domain (Kahl et al.
). Indeed, this modification was insufficient in our hands to achieve high-level engraftment by HSCs transduced with our HIV-1-based vector. The early peak in gene transfer observed thus likely represents a burst in marking derived from successfully transduced committed progenitors but continued restriction of HIV-1 from the HSC compartment. Among the restriction factors identified to date, tripartite motif-5 isoform α (TRIM5α) and apo
atalytic polypeptide 3G (APOBEC3G) have both been shown to restrict HIV-1 infection and in combination may potentially decrease transduction efficiency of HIV-1-based viral vectors (Sakuma et al.
). Rhesus TRIM5α binds to the HIV-1 capsid (CA) to initiate E3 ubiquitin ligase-mediated degradation and prevent viral replication (Stremlau et al.
), whereas SIV can escape rhesus TRIM5α. Our current work is focused on the development of a lentiviral vector system that can perform with equal efficiency in the human and rhesus nonhuman primate setting. Preliminary results using a newly developed vector system combining components of both HIV with SIV capsid to overcome the restriction demonstrate high-level in vivo
marking with a vector encoding green fluorescent protein (GFP) (Uchida et al.
For human clinical trials to proceed, one must also consider the risk of insertional mutagenesis. Certainly the theoretical risk of insertional oncogenesis has been documented in the nonhuman primate model (Donahue et al.
), and unfortunately, more recently in human clinical trials of gene therapy for immunodeficiency diseases (Cavazzana-Calvo et al.
; Hacein-Bey-Abina et al.
). Although both of our recipient animals maintained normal blood counts during follow-up, the overall low-level contribution by genetically modified cells does not permit a realistic assessment of the risk of insertional mutagenesis with our current vector system and such analysis will have to await further optimization. Indeed, our attempts at cloning lentiviral integrations were met with difficulty owing to the low level of marking and the competing internal bands that are frequent with our construct. The assessment of integration patterns with our lentiviral system will require long-term, high-level engraftment of genetically modified cells in order to assess the safety of our approach. Lentiviral vector have already shown a propensity to insert into transcriptionally active genes (Schroder et al.
; Wu et al.
; Hematti et al.
) and target cells often undergo multiple integration events, particularly at high multiplicities of infection (Woods et al.
). Whether vectors that direct erythroid-specific expression of the therapeutic gene will result in an integration site pattern reflecting engraftment patterns that differ from those already described remains of great interest, as vectors with such expression patterns may ultimately prove safer than vectors that constitutively express the therapeutic gene.
In summary, we report transient in vivo β-globin production after lentiviral gene transfer to hematopoietic stem cells in the nonhuman primate. Further optimization should ultimately allow us to comprehensively model HIV-1-based globin gene transfer.