Many multipotential gene-marked progenitor or stem cell clones contribute to hematopoiesis in nonhuman primates

An understanding of hematopoiesis, the process by which pluripotent stem cells give rise to the mature lineages of functional hematopoietic cells, has long been of great theoretical and biologic interest, and it has practical importance for clinical stem cell transplantation and gene therapy. Many important aspects of the proliferation and differentiation of hematopoietic cells have been elucidated using in vitro assays designed to study lineage-committed precursor cells, such as burst-forming units–erythroid (BFU-E) or granulocyte macrophage–colony-forming units (GM-CFU), or more primitive cells with multilineage potential, such as long-term culture-initiating cells. However, these cells clearly do not represent hematopoietic stem cells capable of reconstitution of the entire hematopoietic system in vivo.1-4 Despite progress toward the characterization of true repopulating stem cells by attributes such as cell surface phenotype, cycling characteristics, and drug or dye efflux, the only assay able to define and study these cells unequivocally has been the engraftment and repopulation of all hematopoietic lineages in vivo.1,5-8 

To study the in vivo behavior of repopulating progenitor and stem cells, 2 methods have been used. The most direct but technically challenging approach has been to “mark” these cells with replication-incompetent retroviral vectors that integrate identifiable exogenous DNA sequences randomly into target cell chromatin, thus allowing the tracking of cell progeny based on unique proviral insertion sites.9,10 Alternatively, characteristics and numbers of repopulating cells can be analyzed indirectly using competitive repopulation or X-linked gene expression assays.11 In the mouse, both approaches have been used to demonstrate that a single donor stem cell can repopulate all hematopoietic lineages for the life span of primary or secondary recipient animals.12-16 The life span and diversity of individual murine clonal contributing cells have also been analyzed. Initial retroviral tagging studies suggested that few stem cell clones are responsible for hematopoiesis after engraftment and that many clones are activated sequentially and finitely before they are replaced in a process termed clonal succession.9,14,15 However, subsequent marking studies found that after a 2- to 6-month period of instability, a very small number of clones1-6 accounted for all hematopoietic output, at least at the sensitivity level of Southern blot analysis.10,17 These studies might have been limited by the small numbers of viable repopulating cells surviving ex vivo culture and transduction, given the techniques available in the late 1980s. Data from murine competitive repopulation studies did not support clonal succession models involving a small number of contributing clones at a given time point, nor did they support the differential contribution of stem cells to myeloid as opposed to lymphoid populations.4,18 Both methods gave estimates of stem cell frequencies of approximately 1 in 10⁵ murine bone marrow cells.

There is less information regarding the characteristics of steady state and posttransplantation hematopoiesis in large animals and humans. Allogeneic transplantation recipients engrafted with marrow from donor females heterozygous for X-chromosome alleles had polyclonal hematopoiesis in all lineages studied at early and late time points after transplantation.19,20 Polyclonal stable hematopoiesis in all lineages also accounted for data on nontransplanted females using similar techniques.21Insertion site analysis of human hematopoietic cells engrafted in immune-deficient mice indicated that individual clones could contribute to T-cell and myeloid lineages in this model, but quantitative long-term analysis was not possible, and the relevance of these xenograft models to natural human in vivo hematopoiesis remains unclear.22 Abkowitz et al23,24 have carried out long-term analysis of X-linked heterozygous female cats undergoing autologous marrow transplantation with limiting marrow cell doses, and they used analysis of variance to estimate the numbers of clones contributing to hematopoiesis. In these studies the period of clonal instability extended as long as 1 to 5 years, perhaps reflecting the time required for reconstitution of the larger stem cell reserve necessary to support the much greater hematopoietic demands of large animals.25 Computer models suggested that at larger cell doses, the period of instability would be shorter.26 

For the first time in large animals, we have been able to follow directly the progeny of each primitive progenitor and stem cell in vivo using retroviral marking to track individual progenitor or stem cell clones. This was possible because of recent advances in the efficiency of gene transfer to primate repopulating cells.2,27,28 At all time points studied through 1 year, hematopoiesis was polyclonal. We have identified individual marked clones that contribute to myeloid and lymphoid lineages, and longitudinal analysis has identified clones that contribute to hematopoietic output early and late after engraftment.

Young rhesus macaques (Macaca mulatta) were housed and handled in accordance with the guidelines set forth by the Committee on the Care and Use of Laboratory Animals of the Institute of Laboratory Animal Resources (DHHS publication no. NIH 85-23), on a protocol approved by the Animal Care and Use Committee of the National Heart, Lung and Blood Institute. Mobilization, harvesting, and CD34⁺ cell enrichment were performed as previously described.29 The animals received recombinant pegylated human stem cell factor (SCF) 200 μg/kg (Amgen, Thousand Oaks, CA) and recombinant human granulocyte macrophage–colony-stimulating factor (GM-CSF)10 μg/kg (Amgen) subcutaneously for 5 consecutive days and then underwent apheresis of 2.5 times the blood volume on day 6. The mononuclear fraction was purified by density-gradient centrifugation over lymphocyte separation media (Organon Teknika, Durham, NC). Enrichment for primitive progenitor and stem cells was performed using the Ceprate LC CD34 immunoabsorption column as directed (Cellpro, Bothell, WA).

Producer cell lines G1Na and LNL6 were grown to confluence in Dulbecco's modified Eagle's medium (Mediatech, Herndon, VA) supplemented with 10% fetal calf serum (Gibco/BRL, Gaithersburg, MD).30,31 Repeated tests using polymerase chain reaction (PCR) and mus dunnii amplifications as described32,33 have shown both producer cell lines to be negative for replication-competent helper virus. For each animal, CD34⁻-enriched peripheral blood cells were divided into 2 equal fractions, and each was transduced under identical conditions with 1 of the 2 vectors. Transductions were carried out as previously described, for a total of 96 hours at a starting concentration of 1  × 10⁵cells/mL with daily replacement of vector supernatant and cytokines.2,33 All transduction cultures were supplemented with 20 ng/mL recombinant human interleukin-3 (IL-3) (Sandoz, East Hanover, NJ), 50 ng/mL recombinant human interleukin 6 (IL-6) (Sandoz), 100 ng/mL SCF (Amgen), and 100 ng/mL recombinant human flt-3 ligand (FLT; Immunex, Seattle, WA) and either a preformed irradiated autologous marrow stromal layer (STR) or the CH-296 fibronectin fragment (FBN) (Retronectin; TaKaRa, Otsu, Japan).34,35This stromal layer was generated as described2; 4 μg/mL protamine sulfate (Sigma, St. Louis, MO) was added to the stromal transduction cultures. Animals were given 500 cGy total body γ-irradiation daily for 2 days. On the third day, all cells were collected from both STR and FBN flasks by trypsinization, washed, and reinfused through a central venous catheter. Standard supportive care was given after transplantation.36 

Peripheral blood stem cell samples obtained before and after transduction or bone marrow samples obtained at the time of recovery of the neutrophil count to more than 1000/μL and at 1, 2, 3, 4, 6, 8, 10, and 12 months after transplantation were analyzed. Bone marrow mononuclear cells were purified by density centrifugation over lymphocyte separation media. CFU assays were carried out using methylcellulose media (StemCell Technologies, Vancouver, BC, Canada) supplemented with 5 U/mL recombinant human erythropoietin (Epo) (Amgen), 10 ng/mL IL-3 (Sandoz), 10 ng/mL GM-CSF (Sandoz), and 100 ng/mL SCF (Amgen) with and without G418 (Gibco BRL; 1 mg/mL active). At day 14, colonies containing at least 50 cells were enumerated, and the percentages that were G418-resistant were calculated. Two to 5 days later, well-separated individual GM-CFU or BFU-E colonies grown in the presence of G418 were plucked for subsequent PCR assay. This concentration of G418 was chosen because it was high enough to enrich significantly for transduced colonies but low enough to allow colonies containing vector to reach optimal cell numbers for insertion site analysis (more than 500-1000 cells).

Peripheral blood mononuclear cells were stained with fluorescein-conjugated anti-CD2 or phycoerythrin-conjugated anti-CD20 (Immunotech, Marseille, France) or with isotype controls, and positive cells were sorted using a Coulter Epics Elite instrument (Coulter, Hialeah, FL). Sorted populations had purities of more than 99%. Limiting number dilutions of collected cell fractions were made and frozen as cell pellets for subsequent DNA extraction and inverse PCR analysis.

Single, well-isolated colonies of at least 200, but optimally more than 500, cells were plucked from methylcellulose in a total volume of 40 μL, expelled into 1 mL phosphate-buffered saline, and incubated for 1 hour at room temperature in microcentrifuge tubes. This was followed by lysis in 200 μL proteinase K buffer (0.01 mol/L Tris HCl, pH 7.4, 0.15 mol/L NaCl, 0.01 mol/L EDTA, pH 8.0, and 0.01% sodium dodecyl sulfate) with 30 μg proteinase K (Gibco BRL) for 2 hours at 56°C. One extraction with 200 μL buffered phenol–chloroform–isoamyl alcohol (25:24:1; vol/vol) was then performed. The aqueous phase was precipitated by the addition of 2 μg glycogen (Boehringer Mannheim GmbH, Mannheim, Germany), 18 μL of 10 mol/L NH4Ac, and 500 μL absolute ethanol, and this was incubated at −20°C overnight. DNA was centrifuged at 10000 rpm for 20 minutes, rinsed in 70% ethanol, and dried on the bench-top for 1 hour. DNA pellets were resuspended in 25 to 50 μL H₂O, and 2 μL was used to assess for vector neo sequences by standard PCR as described below. Simultaneous PCR for β-actin sequences was performed on each plucked colony, and the percentage transduction was calculated by dividing the number of CFU positive for the neogene by the number of CFU positive for β-actin as previously described.30,33 DNA from each colony confirmed to contain vector sequences was subjected to integration analysis.

The clonal identity of vector proviral integrants was determined using inverse PCR by a modification of the method previously described.22 To 20 μL DNA from each sample, 10 μL React 2 Buffer (Gibco/BRL), and 2 μL (20 U) TaqI restriction enzyme (Gibco/BRL) was added with H₂O to a total volume of 100 μL. Samples were digested for 2 hours at 65°C with the readdition of 2 μL TaqI after the first hour of incubation. A 16-μL sample was then ligated by the addition of 4 μL 5X T4 ligase buffer (Gibco/BRL) and 1 μL T4 ligase (Gibco/BRL) at 15°C for a minimum of 6 hours. The first round of amplification of 10 μL circularized DNA used the primers 5′-AGGAACTGCTTACCACA and 5′-CTGTTCCTTGGGAGGGT in Perkin–Elmer (Foster City, CA) PCR buffer. The first cycle was at 95°C denaturation for 5 minutes, 50°C annealing for 2 minutes, and 72°C extension for 4 minutes. The subsequent 29 cycles were identical, except that the denaturation time was reduced to 1 minute. Nested PCR was then performed on 2 μL of the amplified product with primers 5′-TCCTGACCTTGATCTGA and 5′-CTGAGTGATTGACTACC using the same reaction conditions and cycles. Resultant PCR products were separated on a 1% agarose gel (Gibco/BRL) with ethidium bromide. Bands were purified from the agarose gel and subjected to DNA sequence analysis. Amplified PCR products were ligated directly into the TA vector (Invitrogen, Carlsbad, CA) and sequenced using an automated sequencer and Taq polymerase (Perkin–Elmer). Sequences were aligned using DNAStar (DNAStar, Madison, WI).

Peripheral blood mononuclear cell DNA from the 6- to 9-month time points were assayed for helper virus sequences by PCR as described.32,33 All were negative, with PCR sensitivities as low as 1 in 100,000 cells.

Figure 1 summarizes our rhesus macaque hematopoietic cell retroviral transduction and autologous transplantation models. The in vivo marking levels achieved in these 2 animals have been reported more fully as part of a larger study28 analyzing several transduction conditions. The animals had stable in vivo marking levels of 0.01 to 0.20 vector copies per cell in peripheral blood granulocytes and lymphocytes in the first year after transplantation.

To track the descendants of retrovirally marked hematopoietic progenitor or stem cells, individual G418-resistant colonies were isolated from bone marrow regularly sampled after transplantation (Table 1). At the concentration of G418 used, approximately 80% of the G418-resistant colonies contained the vectors as documented by PCR (data not shown). This concentration was chosen to enrich for vector-containing colonies but was low enough to permit relatively normal colony size, critical for obtaining enough DNA to perform the insertion site analysis. It is likely that some CFU containing the vector were unable grow under this degree of G418 selection because of inadequate neomycin phosphotransferase expression from some vector insertion sites. The percentage of CFU from nonselected plates found to contain vector were generally higher than the percentage of phenotypically G418-resistant CFU. Given the high-level marking achieved in these animals, sensitive analysis for helper genome sequences within peripheral blood mononuclear cell populations was performed and were repeatedly negative (data not shown)

The vector insertion site in individual myeloid and erythroid colonies was identified using the inverse PCR technique adapted for small numbers of cells, and a representative analysis of 7 individual GM-CFU from animal 1 is shown (Figure2).22 The PCR products were separated on agarose gels, and the bands were isolated, cloned, and sequenced to identify unequivocally the genomic DNA flanking the provirus. Of CFU known to contain vector by standard PCR, only approximately one third were informative by inverse PCR (Table 1). This reflects the dependence of the technique on the efficiency of ligation of the TaqI-digested colony DNA; presumably it was unsuccessful in some CFU because of extremes of length or other characteristics of the digested DNA fragments.

The use of individual, well-separated CFU ensured that insertions at the single cell level were analyzed, and it allowed the differentiation of clonal patterns resulting from multiple insertions in 1 transduced parental cell compared with 1 insertion in multiple parental cells. Analysis of individual colonies also overcame the large variability in efficiency of ligation and amplification of various insertions using the inverse PCR technique. Mixing studies demonstrated that performance of the technique on bulk populations of cells derived from multiple clones underrepresented the clonal diversity of the population. In most informative CFU, a single band was obtained, derived from either the 5′ or the 3′ long terminal repeat (LTR) insertion site. This has been true in other applications of this technique to retrovirally transduced cell populations. Presumably it reflects the differential efficiency of ligation and amplification of the 2 fragments based on size discrepancies. Repeat inverse PCR reactions on DNA from the same CFU always resulted in the same band or bands. Thus it was unlikely that CFU containing the same insertion site would be scored as derived from different clones because of isolation of the 5′ site from 1 CFU and the 3′ site from another. Inverse PCR products from 5 CFU of animal 1 and from 8 CFU of animal 2 had 2 bands. In all such CFU, an identical 4-base direct repeat was found at the actual insertion site flanking both the 5′ and the 3′ LTR sequences, suggesting that these double bands represent a single insertion site rather than multiple proviral insertions in the same clone.37 In the murine model, multiple proviral insertions into 1 primitive cell appear to be common, as shown by secondary CFU-S analysis, but thus far we have not documented this phenomenon in the large animal model.17,38 

Results of the analysis of informative myeloid and erythroid colonies from each animal at different time points after transplantation, from 4 weeks to 1 year, are summarized in Table 1 and Figure3. DNA from 238 well-separated, G418-resistant colonies from animal 1 and 292 similar colonies from animal 2 were subjected to inverse PCR. Seventy-five colonies from animal 1 and 111 from animal 2 yielded distinct bands on inverse PCR, and the amplified DNA segments (containing the LTR along with the flanking genomic DNA) were sequenced (Tables 1, 2). Thirty-five unique flanking genomic sequences were detected in the first 12 months after transplantation in animal 1, and 12 distinct integration sequences were found at 2 or more time points (Figure 3A). In 1 instance, common integration sequences were found in myeloid CFU and in BFU-E (Table2). In animal 2, followed up until 12 months after transplantation, 45 unique flanking genomic sequences were detected. Twenty distinct integration sequences were detected at 2 or more time points, and, in 6 instances, common integration sites were found among myeloid and erythroid cells (Figure 3B, Table 2). In both animals, some clones that contributed early after engraftment (weeks 8-12) continued to be detectable up to 10 months later (Figure 3). Overall, approximately one third of the individual insertions were detected at 2 or more time points. Standard PCR analysis of colonies and bulk circulating cell populations showed that approximately equal levels of marked cells were derived from cells transduced on stroma compared with fibronectin support.28 Insertion site analysis could not distinguish the 2 vectors used at the 5′LTR end, but, for informative 3′LTR insertions, approximately half were derived from each vector and thus from stromal versus fibronectin transductions.(data not shown)

To confirm that integrants detected in GM-CFU corresponded to progenitor or stem cells that contributed to circulating myeloid populations, inverse PCR was performed on DNA from circulating granulocytes collected 6 months after transplantation. Two bands were isolated and sequenced, and 1 corresponded to an insertion identified in GM-CFU at multiple time points after transplantation (insertion 15, animal 1). When DNA from individual CFU with well-characterized insertion sites was mixed and the entire inverse PCR procedure was performed again, bands frequently disappeared; thus, simple analysis of DNA from bulk populations of cells may grossly underestimate the number of marked clones contributing to hematopoiesis (data not shown). However, granulocyte analysis did demonstrate that clones contributing to the marrow GM-CFU compartment also may be found to make up the circulating granulocyte population, as expected from models of hematopoiesis and our prior results in the primate gene transfer model.39 

For analysis of the lymphoid lineages, CD2⁺ or CD20⁺ T or B cells were isolated by FACS with very stringent gating. Sorted populations had purities greater than 99% (Figure 4). Inverse PCR was then performed on limiting dilutions of sorted cells because attempts to grow individual T-cell clones from these animals were unsuccessful. The overall level of marking in T and B cells in these 2 animals was high, 5% to 20%, during the first year.28 Nine proviral integration sequences from the first animal and 7 from the second animal were identified (Figure 3A, B) between 24 and 40 weeks after transplantation. In animal 1, common integration sequences were found in myeloid CFU, T-cells, and B-cells (insertion 20) in 1 case and in myeloid CFU and T cells in a second (insertion 23) (Table 2). Four common integration sequences in myeloid CFU and T cell populations were identified in the second animal (Table 2). Myeloid and lymphoid cells harboring the same proviral integrant demonstrated that at least some of the marked cells were pluripotent.

Important insights into stem cell biology have been gained using genetic tagging or competitive repopulation in murine models, but extrapolation of these findings to larger animals and humans may not be advisable given the orders of magnitude difference in hematopoietic demands between mice and larger species.25 The overall low (generally 0.1% or less) levels of vector-containing hematopoietic cells in large animals and humans after transplantation of retrovirally transduced marrow or peripheral blood stem cells has precluded the tracking of progeny of individually marked engrafting cells.33,40 However, others and we,2,27 working in nonhuman primate models, have recently achieved much higher levels of marked cells in all lineages by optimizing transduction conditions, which has allowed the identification and tracking of individual progenitor clone progeny for up to a year after transplantation.

The inverse PCR methodology used in our studies has some limitations, and the full spectrum of clonal diversity contributing to hematopoiesis must be calculated based on several extrapolations. Only 75 of 238 analyzed G418-resistant CFU in the first animal and 111 of 292 in the second animal yielded bands by inverse PCR, despite evidence that at the G418 concentrations used, more than 80% of colonies did contain vector as determined by conventional PCR for internal vector sequences (Table 1). We assume that the clones detectable in the individual colonies by inverse PCR represent a valid random subset, determined by the distance between the TaqI site and the actual LTR insertion. Additionally, in most informative clones, a single band was generated by inverse PCR despite the theoretical minimum of 2 bands derived from the 5′ and 3′ LTR ends of each inserted provirus. The 3′ end appeared to result in informative products more frequently, with 53 of 75 proviral integration sites from CFU analyzed in animal 1 and 75 of 111 similar sites in animal 2 limited to the 3′ LTR insertion. Multiple independent amplifications of the DNA from an individual CFU in no case yielded a different band or sequenced product containing LTR and flanking genomic DNA, strengthening the assumption that each LTR-containing product could track a unique clone. Furthermore, in most colonies from which both 5′ and 3′ LTR insertions were identified, a 4-bp direct repeat adjacent to the LTR indicated that these LTRs were the 2 ends of the same proviral integrant.37 

Using this method, we demonstrated that multiple clones contributed to hematopoiesis during the first 12 months in each animal. Most clones were detected at only a single time point; in CFUs, 26 of 37 were detected in animal 1, and 25 of 45 were detected in animal 2. Given the observed polyclonality, this lack of longitudinal repeat detection in many clones may have resulted from sampling, not from true clonal disappearance. Only analysis of hundreds of colonies at single time points would allow the statistical exclusion of persistent contributions, even at levels greater than 5%. This type of analysis should be more practical in these animals if we can apply a newly described technique that allows more complete detection of clonal diversity in DNA from bulk cell populations; such collaborative studies are ongoing (Von Kalle C, unpublished data).41 

It may be more appropriate to address issues such as clonal succession in these animals at a later time because equilibration after transplantation may take longer than 1 year.25 However, we did detect many clones at 2 or more time points, in some cases spanning 10 months after transplantation (Figure 3). In our earlier primate studies, at least the overall level of marking did not change significantly in the first 3 to 6 months after transplantation.39 In the myeloid lineage, the prolonged detection of individual clones suggests the self-renewal of early primitive cell populations and the ability of at least some primitive precursor cells to contribute to hematopoiesis for prolonged periods. We believe it argues against rapid clonal succession as the primary pattern of hematopoiesis in vivo.

We also detected individual marked clones able to contribute to lymphoid and myeloid lineages, including 1 clone contributing to B cells, T cells, and myeloid CFU. Our analysis of the lymphoid lineages was less extensive than that of myeloid cells, and it was limited by the use of bulk cell populations, with the resultant identification of only the most efficiently ligated and amplified inverse PCR products. Thus, many multipotent clones might have been missed in their contribution to the lymphoid lineages. Further study will be required to better assess whether lymphoid- or myeloid-restricted engrafted progenitors exist. Most murine data do not support the concept of lineage-restricted progenitor or stem cell engraftment.4,10In humans analyzed in steady state rather than after transplantation, X-linked gene expression analysis supported the concept of similar myeloid and lymphoid contributions by 1 population of stem cells.21 Taken together, our findings of long-lived, multipotential clones contributing to hematopoiesis in vivo suggest the successful transduction of at least a practically defined primitive hematopoietic stem cell. Formal single-cell engraftment studies or serial transplantation cannot be carried out in any animal larger than a mouse.

Although our data did not permit a sophisticated estimate of stem cell number, the minimum number of clones contributing to hematopoiesis over the first year could be estimated as approximately 1000 in animal 1. Forty-four unique insertion sites were detected and represented only 5% of hematopoiesis—the level of overall gene marking quantitated by Southern blot analysis and CFU-PCR. To address the theoretical number of transduced clones contributing to hematopoiesis at the time points analyzed, mathematical models for capture and release can be applied.42 Using this method at time points for which there were common (recaptured) clones, a range (±1 SD) of 5 to 44 clones in animal 1 and 8 to 60 clones in animal 2 contributed to hematopoiesis in the first year. Given that the starting number of CD34-enriched cells in animal 1 was 20 million and that CD34⁺ cells represented approximately 1% of the bone marrow mononuclear cell compartment, we estimate that 5 cells per 10⁷ bone marrow mononuclear cells contributed to hematopoiesis in the first year. This is in agreement with calculations presented by Abkowitz et al26 using a completely different approach in the cat, a large animal of similar size and presumably of similar hematopoietic demand.

These data are in contrast to marking data from murine studies in which monoclonal or oligoclonal patterns were found at all time points, though the use of potentially limiting numbers of marked engrafting cells confound interpretation of these studies.12-14,16Recent murine studies using in vivo bromodeoxyuridine, or BrdU, incorporation to distinguish cycling from noncycling cells found that 99% of phenotypically defined stem cells were recruited in the cell cycle within 58 days.43 This argues against a large number of continuously quiescent stem cells that contribute sequentially and then disappear.

In summary, we have shown that multiple transduced clones contributed to hematopoiesis for at least 1 year after transplantation in a primate animal model that has many similarities to humans.44Successful transduction of a pluripotent hematopoietic stem cell is inferred from the prolonged contribution to hematopoiesis by some clones and the finding of common insertion sites among various hematopoietic lineages. Although our results do not support the clonal succession model as the only operating framework for in vivo stem cell behavior, the follow-up of these animals was relatively short and more definitive assessment will require years of careful monitoring. Finally, marking derived from multiple transduced clones in the first year after transplantation is encouraging for the eventual successful application of hematopoietic stem cell-based gene therapy in humans.

We thank Jan Nolta for her much appreciated guidance regarding the inverse PCR technique and Christof von Kalle and Manfred Schmidt for their helpful discussions. We thank Immunex for supplying flt-3 ligand and Amgen for supplying rhG-CSF and rhSCF.