Methods to analyze the clonality of an adverse event in preclinical or clinical retroviral stem cell gene therapy protocols are needed. We analyzed the progeny of retrovirally transduced human peripheral blood progenitor cells (PBPCs) after transplantation and engraftment in immune-deficient mice. The integration site of the provirus serves as a unique tag of the individual transduced PBPC. A plasmid library of junctions between proviral and human genomic DNA was generated. We were able to detect individual transduced cell clones that amounted to 0.14%–0.0001% of chimeric bone marrow cells. This is the first report in which the contribution of individual marrow-repopulating cells to human hematopoiesis is directly quantified.
Recently, two cases of insertional mutagenesis have been described following retroviral gene transfer to human CD34+-selected bone marrow (BM) cells . Methods to monitor and quantify the contribution of individual retrovirally transduced marrow-repopulating cells to hematopoietic engraftment with high sensitivity and reproducibility would allow the recognition of aberrant proliferation at an early stage. For example, patients with leukemia having normal blood counts after high-dose therapy and stem cell transplantation are monitored for minimal residual disease. An increase in the clonal size of leukemic cells by one log, as detected with real-time polymerase chain reaction (PCR), can predict relapse . In the field of retrovirus-mediated gene therapy, quantitative  and semiquantitative protocols  to monitor all transduced cells have been applied. Additionally, a quantitative competitive PCR protocol to quantify individual transduced clones [1,5] has been described, whereas highly accurate protocols using real-time PCR to quantify individual hematopoietic clones have not been reported yet.
We previously established a transplantation model for human retrovirally transduced peripheral blood progenitor cells (PBPCs) in nonobese diabetic/severe combined immunodeficient (NOD/SCID) mice . The retroviral vector used for transduction contains the human multidrug resistance 1 gene (MDR1) as a transgene. The retroviral vector—through its integration into the genomic DNA—serves as a unique tag of individually transduced cells and their progeny. We characterized retroviral integration sites in human marrow-repopulating cells .
To determine unique junctions between proviral and human genomic DNA, we used a ligation-mediated (LM)-PCR technique, cloned the LM-PCR products into plasmids, thus generating a plasmid library, and sequenced all the obtained proviral-genomic DNA junctions . The plasmids obtained in the cloning step were used later for quantitation of individual hematopoietic clones using real-time PCR technique [8–10].
Our results allow us to draw conclusions on the clone size with high sensitivity and specificity, suggesting that a plethora of individual hematopoietic cells are simultaneously contributing to human hematopoietic BM engraftment. The quantitative analysis of individual transduced cell clones also offers a means to analyze the clonality of an adverse event in preclinical  or clinical [4, 12, 13] stem cell gene therapy protocols.
Materials and Methods
Selection of CD34+ Cells
Material from three healthy donors was used in this study. Informed consent was obtained from all donors before CD34+ cell collection. CD34+ cells were prepared as previously described . Briefly, CD34+ cells were isolated from frozen PBPC samples by magnetic microbead selection using the CliniMACS system (Miltenyi Biotech, Bergisch Gladbach, Germany) according to the manufacturer's description.
Retroviral vector stocks were produced and stored as described . Retroviral transduction was performed as described . In brief, CD34+ cells were prestimulated for 16–20 hours at a density of 1 × 106 cells/ml X-VIVO-10 medium, supplemented with interleukin (IL)-3 (20 ng/ml), IL-6 (10 ng/ml), stem cell factor (50 ng/ml), Flt3-ligand (100 ng/ml; CellSystems, St. Katharinen, Germany), and thrombopoietin (20 ng/ml; R & D Systems, Wiesbaden, Germany). After prestimulation, cells were exposed over 3 consecutive days to a retroviral supernatant containing the hybrid vector SF91m , which is based on the Friend mink cell focus-forming/murine embryonic stem cell virus and carries the human MDR1 gene. Twenty-four hours after the last infection period, cells were harvested.
Twenty-four hours before transplantation, female NOD/SCID mice were conditioned by sublethal irradiation with a total dose of 3 Gy. A total of 3–5 × 106 bulk transduced human CD34+-selected PBPCs were transplanted intravenously per mouse, as previously described . Starting on the day of transplantation, animals received 2 μg of human IL-3 (Strathmann Biotec AG, Hannover, Germany) and 4 μg of human G-CSF (Amgen GmbH, Munich, Germany) three times per week subcutaneously. Additionally, all mice were treated with anti-asialo GM1 (Wako Chemicals, Neuss, Germany) on days 0, 5, and 11. All animal studies were approved by the appropriate authorizing bodies.
NOD/SCID Mouse Reconstitution Assay
Mice (n = 7) were killed by cervical dislocation 6–8 weeks after transplantation. Engraftment of human cells isolated from mouse BM and presence and expression of the MDR1 transgene were evaluated as previously described [3,6]. Engraftment was measured by quantifying human CD45 antigen-expressing cells. Control mice received transplants of untransduced human cells.
Real-time PCR was used to quantify the presence of proviral sequences (data not shown). The difference between the threshold cycles of the proviral MDR1 gene and the human erythropoietin receptor gene was used to quantify the percentage of MDR1-transduced human cells with reference to a standard curve over a 5-log range . Rhodamine-123 efflux analysis served to detect the expression of the functional transcript—a plasma membrane P-glycoprotein—of the MDR1 transgene  in engrafted human CD45+ cells.
Ligation-Mediated PCR and Analysis of PCR Products
BM DNA from seven chimeric mice (E15M1, E15M5, E15M6, E15M15, E17M19, E18M22, and E18M24) was analyzed by LM-PCR. The LM-PCR was then performed . Briefly, DNA isolated from chimeric BMs was digested with restriction enzymes BsmAI or PvuII. Proviral-human DNA junctions were marked using a biotinylated long-terminal repeat (LTR)–specific primer followed by enrichment of the biotin-marked fragments using paramagnetic beads. Next, an adapter oligo cassette was blunt-end ligated to the LTR-distant portion of enriched fragments to create binding sites for forward primers. Nested PCR with a total number of 60 amplification cycles was performed on the purified fragments. PCR products were analyzed on agarose gels, and gel blocks were excised and purified by a Gel Extraction Kit (Qiagen). DNA of each excised gel block was cloned into pCR4 plasmid vector (Topo TA Cloning Kit, Invitrogen) according to the manufacturer's instructions. Ten colonies of each cloning reaction were screened for insert length by direct PCR of bacterial colonies with standard vector primers T3 and T7. Cycle sequencing of PCR products obtained by screening was performed using an ABI Prism Genetic Analyzer 310 (Applied Biosystems, Weiterstadt, Germany) according to the manufacturer's instructions. The plasmids containing the LM-PCR product inserts were used to create standard curves for further real-time quantitative PCR experiments. A plasmid integration site library was compiled that consisted of 207 integration sites. The average insert size was 510 bp (range, 127–1,600 bp).
Chromosomal Mapping of Retroviral Integration Sites
For chromosomal mapping, the following criteria had to be met: sequence of interest (soi) flanked by LTR and adapter sequence and unique match (>90% identity) of soi with human genome database. Standard National Center for Biotechnology Information blast program was used to map the soi to human genome. Chromosomal localizations of mapped integration sites were determined using the projectEnsembl Contig Viewer. To confirm mapping of integration sites, we also used the UCSC Genome Browser BLAT program (April 2003 Assembly). The chromosomal mapping of integration sites was considered correct if the same mapping result was obtained using both programs.
Real-Time Quantitative PCR
Reference standard curves were constructed by diluting plasmids containing an LM-PCR product. To test the real-time quantitative PCR method for our purposes, we used a plasmid mixture. Three plasmids each containing an LM-PCR product (E15M5K4, E15M5K5, and E15M5K6) were mixed in a constant amount of 105 copy numbers, while the amount of the plasmid E15M5K4 decreased from 100% to 0.01% and the amount of the other two plasmids was kept in a fraction of 1:1 (e.g., 20% plasmid E15M5K4, 40% plasmid E15M5K5, and 40% plasmid E15M5K6). Unique primers binding to the flanking human DNA were designed for clones E15M5K4, E15M5K5, and E15M5K6 using Primer Express™ 1.0 software (Applied Biosystems, Foster City, CA). For quantitation of the total copy number of all plasmids in the plasmid mixture, a primer and probe set was placed in the backbone of the plasmids. The amount of an individual plasmid was normalized by dividing the copy number of the plasmid with the total copy number. The reactions were performed three times in triplicate using the following reaction mix: 300 nM of final concentration of forward and reverse primers and 200 nM of final concentration of probe were mixed with 3 μl of template volume in a total volume of 30 μl.
Unique reverse primers binding to the flanking human DNA were designed for 17 integration site sequences detected in seven chimeric mouse BM DNA by LM-PCR. A primer and probe set that detects the MDR1 transgene was used to quantify the total copy number of transduced human genomes in the chimeric BM. The reactions were performed three times in duplicate (except samples E15M15 and E18M42, in which only two repeats and one reaction, respectively, were possible because of low DNA amount) using the following reaction mix: 300 nM of final concentration of forward and reverse primers and 200 nM of final concentration of probe was mixed with 1.2–1.5 μg of chimeric BM DNA in a total reaction volume of 30 or 50 μl.
Real-time quantitative PCR was performed on an ABI PRISM® 7700 Sequence Detection System instrument (Applied Biosystems) using the following PCR program: 2 minutes at 50°C followed by 10 minutes at 95°C and 45 cycles of amplification cycles (95°C for 15 seconds and 60°C for 60 seconds).
Our aim was to assess the contribution of individual progenitor cells to human hematopoietic engraftment in the BM of NOD/SCID mice (SCID-repopulating cells [SRCs]). In our experiments, purified CD34+ PBPCs (CD34+ >87.9%) were retrovirally transduced and transplanted in a total of seven mice. The proportion of human CD45+ leukocytes in the chimeric NOD/SCID BM amounted to a median of 13% (range, 7%–42%). The proviral MDR1-DNA was found in a median of 23% (range, 13%–31%) of human cells, and a median of 10% (range, 4%–13%) of the human leukocytes expressed the MDR1 transgene (Table 1). The granularity of the leukocytes shows that both myeloid and lymphoid human cell populations harbored the transgene (Fig. 1). Individual hematopoietic clones were identified by their unique retroviral integration sites into the genomic DNA.
Table Table 1.. Data of retrovirally transduced peripheral blood progenitor cells (PBPCs) in nonobese diabetic/severe combined immunodeficient mouse bone marrow
Human PBPCs were transduced with the retroviral vector SF91m carrying the human multidrug resistance 1 gene transgene (MDR1). Integration site sequences identified by LM-PCR were mapped to human genome and localized on human chromosomes using both the NCBI BLAST and the UCSC BLAT programs.
aPercent of all nucleated chimeric mouse bone marrow cells.
bPercent of all human CD45+ cells.
cSF91m provirus was quantified by real-time PCR with MDR1 transgene-specific primers and related to human erythropoietin receptor gene copies as described .
dTransgene expression was measured by rhodamin dye exclusion .
eIntegration site sequences represented by # could be not mapped.
Chimeric BM DNA from seven mice was used to identify unique junctions between proviral and human genomic DNA sequence. Following an LM-PCR and cloning reaction (Fig. 2), we detected a median of 7 (range, 2–13) integration sites per chimeric BM (Table 1). Repeated LM-PCR analysis contributed to the total number of integration sites in sample E15M1 (Table 2). Six LM-PCR reactions did not seem to be enough to detect all integration sites, because only three reactions contained repeated junction fragments. Because of insufficient DNA from this mouse, we could not perform additional LM-PCR reactions. Identical integration site sequences were repeatedly found (Table 2, LM-PCR 1 and 2, E15M1EB9 and E15M1EB11). In some cases, PCR bands of different lengths contained the same integration site sequence (Fig. 2 and Table 2, E15M1EB9, PCR bands I and III, respectively), and one PCR band contained two integration site sequences (Fig. 2 and Table 2, LM-PCR 1, PCR band I). Because we cloned the PCR bands into plasmids, and several plasmids from the same cloning reaction were sequenced, it was possible to distinguish between the junction fragments with identical lengths but different sequences contained in the same PCR band. From the other six chimeric BM DNA samples, one LM-PCR analysis per sample was the basis for assembling the integration-site library. No integration sites were detected in mice transplanted with mocktransduced cells, confirming the specificity of our method (Fig. 2).
Table Table 2.. Reproducibility of the ligation-mediated polymerase chain reaction (LM-PCR) method
Nucleotide sequences of LM-PCR products found in sample E15M1. Six LM-PCR analyses were performed. Two integration sites were found two (E15M1EB11) or three times (E15M1EB9). In additional reactions, mostly new integration sites were detected.
The chromosomal localization of the integration sites is defined by the unique accession numbers. The obtained accession numbers are listed in the order of integration sites, as follows: E15M1EB–E15M1EB13: Z83847, Ac015823 Ac010746, Al445667, Al049569, Al355149, Ac005056, Al050317, Ac025471, Ac020904, Ac084854, Ac005632, Ac008533; E15M5K2–E15M5K8: Ac093759, Ap002907, Ap001437, Ac011979, Ac026703, Ac011501, Ac072052; E15M6K2–E15M6K6: AL157935.28, AL391239, Al137141, Al157877, Ac090117; E15M15K1–E15M15K7: Ac008760, Ac068134, Ac005208, Ac012100, Al138999, Al390961, Ac105245; E17M19K2–E17M19K5: AL162832.6, AC009116.8, AL136218.26, AL160169.12; E17M19K8–E17M19K10: AC130343.7, AC003030.1, AC026469.8; E18M22K2–E18M22K4: AL445903.3, BX005019.10, AC011456.2; E18M24K1–E18M24K2: AC036103.8 and AC005208.1. The integration sites not listed could not be mapped. The chromosomal localizations of the clones are given in Table 1.
Quantitation of Individual Clones
To quantify the clones detected by LM-PCR, we used a real-time quantitative PCR method. Unique primers specific for the flanking genomic DNA were designed for each integration site sequence, whereas the retroviral-specific primer and fluorogenic probe were identical for all integration sites. To test the specificity of the real-time quantitative PCR primer and probe set, we performed reactions with the corresponding plasmid and control plasmids. Reactions were only positive with the primer designed for a given clone (e.g., plasmid E15M1EB7 with primer for clone E15M1EB7) and were completely negative with primers designed for other clones (e.g., plasmid E15M1EB7 with primers for clones E15M1EB9 and E15M1EB11), thus demonstrating a very high specificity of the designed unique primers.
In a mixture of three plasmids each containing a different LM-PCR product, we were able to detect the proportion of an individual plasmid down to 10 copies in a total copy number of 105 (13 dilution steps; Fig. 3) using real-time quantitative PCR. To test the reproducibility of the results, three independent replicates were performed for each integration site sequence. The mean coefficient of variation was 0.16 (range, 0.0001–0.42).
Contribution of Individual Human Marrow-Repopulating Cells to Hematopoiesis
The amount of all transduced cells was determined as the copy number of the MDR1 transgene by real-time quantitative PCR and ranged between 263 and 6,807 copies per μg DNA (median, 1,380.5 copies per μg DNA; Table 3). A total of 17 integration site sequences from three different donors detected by LM-PCR were analyzed. Individual clone copy numbers ranged from 0.02–124 per μg DNA (median, 2.2 per μg DNA; Table 3). The ratio of the mean copy number of an individual clone and the mean copy number of the MDR1 transgene gives an estimate of the contribution of this clone to human transduced hematopoiesis. The analyzed individual marrow-repopulating cell clones contributed to 0.01%–1.82% of human retrovirally transduced hematopoiesis (median, 0.15%; Table 3). Considering the level of engraftment and the proportion of MDR1 transgene-marked cells, the transduced clones analyzed here amounted to 0.001%–0.4% of all human cells and to 0.0001%–0.14% (median, 0.03%) of all human and mouse cells in chimeric BM (Table 3). Interestingly, the clones, which were repeatedly found in separate LM-PCR analyses (Fig. 2 and Table 2), showed the highest engraftment (1.82% for E15M1EB11 and 1.23% for E15M1EB9).
Table Table 3.. Proportion of individual clones among the transduced human severe combined immunodeficient mouse repopulating cells
Seventeen clones were quantified by real-time quantitative polymerase chain reaction in seven chimeric BM DNA samples. The SD was calculated from three repeats, except of mouse E15M15, where two repeats were performed.
aProviral multidrug resistance 1 gene (MDR1)-containing human cells (%) from Table 1.
bCalculated from percent of human CD45+ cells and percent of provirus-containing human cells from Table 1.
Abbreviations: BM, bone marrow; SD, standard deviation.
In this study, we show that individual human retrovirally transduced marrow-repopulating cell clones can be quantified with high sensitivity and specificity [8–10].
We optimized the engraftment of human transduced PBPCs in a NOD/SCID mouse model , which allowed us to demonstrate high-level transgene expression . In contrast to the initial hypothesis regarding establishment of immunodeficient mouse models for human hematopoiesis [17–19], there is growing evidence that not only pluripotent long-term repopulating cells but also primitive, lineage-restricted progenitors with short- and long-term repopulating capacity contribute to the SRC population [20–22] that was analyzed here. In a recent study, hematopoietic repopulation in NOD/SCID mice was directly compared with autologous reconstitution in baboons . It was concluded that the NOD/SCID assay preferentially measures progenitor cells with short-term engraftment capability. However, long-term clinical gene marking studies with normal human donors are not available, and the NOD/SCID data as presented are the closest approximation to normal adult human engrafting hematopoietic cells that is currently possible in an experimental or clinical setting.
The basis for the identification of individual hematopoietic clones is the detection of retroviral-genomic DNA junctions by LM-PCR. This technique was originally described by Rosenthal and Jones  and has first been applied to hematopoietic gene therapy in a broader way by Schmidt et al. . We modified the LM-PCR method at several steps, including cloning of PCR products . Other groups use LAM-PCR , which includes an additional DNA amplification step at the beginning. These two different methods have not yet been directly compared. However, the average insert size in our LM-PCR assay was 510 bp; therefore we could analyze genomic flanking regions between approximately 127 and 1,600 bp. Genomic flanking regions are the critical parameter for gene mapping. In this way, we were able to identify a PCR band that harbored two junction fragments with different sequences and to detect less-abundant junction fragments (Fig. 2), pointing to the necessity of sequencing all of the LM-PCR bands. The integration site library we generated has now allowed us to set up a real-time quantitative PCR using standard plasmid dilutions as a reference to determine the contribution of individual repopulating cells to hematopoietic engraftment.
We found the same integration in fragments of different lengths. This may be attributable to an incomplete extension reaction in the first step of the LM-PCR.
The real-time quantitative PCR protocol described here was specific and highly sensitive. In a plasmid mixture, no cross-reactivity was detected when quantifying an individual plasmid. We were able to quantify clones down to 0.01% of transduced cells in chimeric BM DNA. Considering the engraftment and transduction data (Table 1), the detection limit of our real-time quantitative PCR was 0.0001%. Data were highly reproducible, and the coefficient of variation was in a range reported in clinical BM transplantation studies monitoring donor chimerism . Even at this level of sensitivity, some clones might be not detectable by real-time quantitative PCR. The LM-PCR method including a nested-PCR step followed by a cloning step is more sensitive than real-time quantitative PCR. Six LM-PCR reactions on one chimeric BM DNA did not detect all integration site sequences, because in additional reactions, mostly new integration site sequences were detected (Fig. 2 and Table 2). The detection of additional junction fragments in repeated LM-PCR reactions can be explained by enrichment of fragments using paramagnetic beads. It was expected to work more efficiently for short fragments than for long fragments as provided by the manufacturer . Therefore, some fragments may have been lost, and the number of bands visible in the LM-PCR underestimates the real number of integration sites.
One must be cautious, because these data may underestimate the frequency of marrow-repopulating cells contained in the graft. Two opposing scenarios can be envisaged. Assuming that all clones in a sample amount to 1.82% of transduced cells (Table 3), 56 clones would make up the transduced cell population. If the least-frequent clone amounts to 0.01% of the transduced cells (Table 3), 10,000 clones could be found. The real number of simultaneously active clones ranges between these estimates. Because the transduced cells represent only a fraction of the human cells in the chimeric mouse BM, the number of simultaneously active clones will still be considerably higher. In earlier studies , up to 12 repopulating human clones were found in NOD/SCID mice. These studies were qualitative and not quantitative, so the actual number of active clones may have been higher. To get an estimation of the clonal distribution of the marked human hematopoiesis by LM-PCR, several reactions need to be performed. For example, with the clone with a frequency of approximately 2% (E15M1EB11, Tables 2 and 3, Fig. 2) the integration site was detected in only three of the six LM-PCR reactions, and for the clone with a frequency of approximately 1% (E15M1EB9, Tables 2 and 3, Fig. 2), the integration site was detected in only two of six LM-PCR reactions, suggesting that two to three reactions are sufficient to detect more abundant clones. Considering the frequency of provirus-containing human cells and the frequency of transgene-expressing cells, more than one provirus integration per cell is possible. This would decrease the possible number of transduced cell clones. The frequency of SRC that can be estimated from our analysis of individual marrow-repopulating cells is consistent with the data from limiting dilution experiments . Bhatia et al.  reported a frequency of one SRC in 617 CD34+CD38− cells. SRCs were exclusively found in the CD34+CD38− fraction. In leukapheresis products of normal donors, approximately 4% of all CD34+ cells express the CD34+CD38− phenotype . It then follows that one SRC is contained in 15,425 CD34+ PBPCs. Therefore, in mouse E15M1 that received 8.8 × 105 MDR-transduced CD34+ PBPCs, 56 SRCs or 1.8% of transduced hematopoiesis/SRCs are expected. This is in the range of the results obtained in our study. The methods presented here allow the study of human hematopoiesis at the level of single repopulating cells and the quantification of their progeny in an experimental or clinical setting. Our assay would help in predicting an adverse event, because more abundant clones comprising more than 1% of transduced cells can be detected reliably. Once these integration sites are characterized, quantitative PCR can be performed and clonal size can be monitored. Because an overt leukemia is defined as more than 20% blast cells in the BM , the monitoring protocol described here opens a diagnostic window that has not been available before. The assay could be additionally refined by sampling T cells, B cells, granulocytes, and monocytes separately over time. In the light of the current leukemia cases following retroviral gene transfer, the protocol described here would be of particular relevance for patient monitoring .
The technical assistance of Bernhard Berkus, Hans-Jürgen Engel, Sigrid Heil (German Cancer Research Center), and Carmen Hoppstock (Department of Internal Medicine V, University of Heidelberg) and the support of the animal facility team of the German Cancer Research Center are gratefully acknowledged. We thank Birgit Stähle (EUFETS AG) for her help at the start of this project. This work was supported in part by grant 10-1294-Ze3 from the Deutsche Krebshilfe/Dr. Mildred-Scheel-Stiftung, grant M 20.4 from the H.W. & J. Hector-Stiftung, and by a grant from the AiP plus Forschung program of the University of Heidelberg.
K.Z.N. and S.L. contributed equally to these results.