Dose finding with retroviral vectors: correlation of retroviral vector copy numbers in single cells with gene transfer efficiency in a cell population

Retroviral vectors based on murine leukemia viruses (MLVs) are the most commonly used gene transfer system for stable transduction of various target cells,¹  but their random genome insertion²  may lead to functional or structural alterations of cellular genes. This process, called insertional mutagenesis, has been considered a possible limitation of somatic gene therapy for more than 2 decades.^3-5  However, until recently there were no reports of malignant transformation of normal somatic cells related to the insertion of replication-defective retroviral vectors. Indeed, based on the experience of numerous preclinical and clinical studies, the frequency of those events was predicted to be low.^5-9 

Last year we reported the first evidence that a retroviral vector insertion may directly lead to the activation of an oncogene, Evi-1, causing acute myeloid leukemia in a mouse gene-marking model.¹⁰  Later, a lymphoproliferative disease, probably causally linked to retroviral insertions into the oncogene LMO-2, developed in 2 patients in an otherwise highly successful gene therapy trial.^11-13  These observations underscore that insertional mutagenesis represents a real risk factor for somatic gene therapy and that insertional alterations of crucial regulatory genes (eg, oncogenes) are more common than estimated.⁹ 

Given that problems related to insertional mutagenesis are likely dose dependent,⁹  dose finding requires careful investigations in powerful preclinical models. Here we investigated the correlation between retroviral transduction rates and numbers of vector insertions. To address this question we transduced primary hematopoietic CD34⁺ progenitors and cultured K562 cells with an enhanced green fluorescence protein (EGFP)-expressing vector at different multiplicities of infection (MOIs) and analyzed gene transfer (expression, copy numbers) in mass cultures and cell clones derived from them.

Primary CD34⁺ peripheral blood stem cells were isolated using magnetic cell sorting (MACS) technology (Miltenyi Biotec, Bergisch-Gladbach, Germany).^14-16  CD34⁺, K562, and PG13 cells producing the retroviral vector SF11αEGFPrev¹⁷  were cultivated as described.¹⁴  For transduction, K562 or prestimulated CD34⁺ cells¹⁴  were transferred onto 6- or 24-well plates, preloaded up to 8 times with diluted (1:5, group A) or undiluted (groups B-F) supernatant.¹⁴  Twenty-four hours after the first round, cells from some groups were submitted to a second infection on plates preloaded accordingly. MOIs given in Table 1 were calculated based on titers determined as described.¹⁴  It was assumed that subsequent preloading steps are equally as efficient as the first one, although there might be a saturation effect (B.F. and K.K., unpublished data, April 2003). Transduced CD34⁺ and K562 cells were cloned in methyl cellulose (MC) or by limiting dilution.^15,16 

Gene transfer and mean fluorescence were estimated by flow cytometry.^14-17  DNA was isolated using Qiagen (Hilden, Germany) columns or by direct lysis of MC clones.¹⁵  Southern blots and real-time polymerase chain reaction (PCR) were performed as described.^14,16 

Data summarized in Table 1 show the expected linear relationship^17,18  between MOI and vector copy number for primary CD34⁺ (groups A-D) and K562 cells (groups A-C). However, the increase in the number of transduced cells was linear (in relation to the MOI) only up to approximately 30% of gene transfer. Moreover, once maximal transduction was reached, further increasing the number of infectious vector particles per cell did not lead to an elevated number of transduced cells but only to much higher transgene copy numbers per cell (eg, K562 group C vs group D). As a consequence, raising the MOI under these circumstances would not only be useless, it would potentially be dangerous. Instead, to increase transduction rates (up to 68% for CD34⁺ and 93% for K562 cells) a second infection cycle had to be performed (groups E and F for both).

Nonlinear interdependence between transduction efficiency with MLV vectors and the applied MOI (Table 1) is based on an important feature of the original virus: provirus integration occurs during mitosis, after chromosomal duplication.^2,19,20  In the presence of limiting amounts of infectious particles, only 1 of the 2 daughter cells will thus bear a virus copy, resulting in a theoretical maximal gene transfer of 50% with an MOI of 1. However, because the interaction of viral particles and target cells occurs by chance, the statistical use of 1 infectious vector copy per cell results in a gene transfer rate of approximately 30% to 35% with 1.4 to 1.7 vector copies per transduced cell and a Poisson-like distribution of insertion numbers (usually not exceeding 3). Consequently, any linear increase of gene transfer should be accompanied by an exponential increase of insertion numbers. This problem becomes even more complex if the target cells are not equally susceptible to infection at a given time. Indeed, we observed a mean of more than 9 vector copies per transduced K562 cells after 2 infection cycles in cell populations showing approximately 90% transgene expression (Table 1). Similarly, Brenner et al²¹  recently reported average copy numbers of approximately 8 for 90% gene transfer into primary CD34⁺ cells.

To estimate the number of insertions in single-cell-derived clones, transduced cells from all groups were subjected to limiting dilution (K562) or methyl cellulose (CD34⁺) cloning, and 5 randomly chosen EGFP-positive clones per group were further analyzed by transgene-specific, real-time PCR, Southern blot analysis, or both.

As evident from Table 1, data obtained with single-cell-derived clones are generally in good agreement with vector copy numbers in the corresponding mass culture. However, with increased transduction rates, the analysis of single-cell-derived clones revealed strong differences in copy numbers among clones (Table 1; Figure 1). Therefore, transduction at high MOI may strongly influence the outcome of functional studies by elevating transgene expression in mass cultures¹⁷  and creating clones with multiple transgene copies (K562 clone 10/F, more than 31 copies; CD34 clone 12/F, 21 copies), which might be far from normal. Recently, of CD34⁺ mass cultures transduced to 90% (average copy number, 8), only those CD34⁺ cells with limited vector copy number (mean, 2) were capable of repopulating NOD-SCID mice, which probably reflects receptor or postreceptor restrictions (eg, cell-cycle status) to gene transfer in true stem cells.²¹  However, a direct genotoxic impact of multiple vector insertions on cell survival (compare K562 groups C vs D, E vs F) cannot be excluded¹⁸  and requires urgent investigation, considering that preinsertion blocks will soon be overcome (as with lentiviral technology).²² 

Finally, there was a clear positive correlation between transgene expression and insertion frequency in K562 clones (Figure 1), which is consistent with a limited variability of transgene expression depending on the individual insertion site. Indeed, variations in expression levels per single integration were in the range of less than 4 (groups A, B; Figure 1; Table 1). Thus, although dependent on insertion sites, retrovirus-mediated expression may be less variable than observed with many physicochemical transfection methods.²³ 

In conclusion, generating cell clones with multiple vector insertions after high-efficiency retroviral gene transfer may lead to erroneous predictions of transgene efficiency and toxicity. Moreover, those clones might be at a higher risk for malignant transformation because the frequency of insertional mutagenesis may directly correlate with the number of integrated vector copies.^6,7  Consequently, transduction protocols should be evaluated for transgene copy numbers obtained in single cell clones, even if that risk is lower in mature cells or with alternative, safety-improved vector systems. If a target population contains a fraction of cells that is more resistant to retroviral (including lentiviral)²²  transduction, a further increase of the MOI cannot be recommended. Rather, procedures for posttransduction cell sorting and clonal selection in vivo must be elaborated with even stronger emphasis.

We thank Dung Nguyen for expert technical assistance and Francis A. Ayuk for critical reading.