Improving Gene Therapy Efficiency through the Enrichment of Human Hematopoietic Stem Cells

Abstract

Lentiviral vector (LV)-based hematopoietic stem cell (HSC) gene therapy is becoming a promising clinical strategy for the treatment of genetic blood diseases. However, the current approach of modifying 1 × 10⁸ to 1 × 10⁹ CD34⁺ cells per patient requires large amounts of LV, which is expensive and technically challenging to produce at clinical scale. Modification of bulk CD34⁺ cells uses LV inefficiently, because the majority of CD34⁺ cells are short-term progenitors with a limited post-transplant lifespan. Here, we utilized a clinically relevant, immunomagnetic bead (IB)-based method to purify CD34⁺CD38⁻ cells from human bone marrow (BM) and mobilized peripheral blood (mPB). IB purification of CD34⁺CD38⁻ cells enriched severe combined immune deficiency (SCID) repopulating cell (SRC) frequency an additional 12-fold beyond standard CD34⁺ purification and did not affect gene marking of long-term HSCs. Transplant of purified CD34⁺CD38⁻ cells led to delayed myeloid reconstitution, which could be rescued by the addition of non-transduced CD38⁺ cells. Importantly, LV modification and transplantation of IB-purified CD34⁺CD38⁻ cells/non-modified CD38⁺ cells into immune-deficient mice achieved long-term gene-marked engraftment comparable with modification of bulk CD34⁺ cells, while utilizing ∼7-fold less LV. Thus, we demonstrate a translatable method to improve the clinical and commercial viability of gene therapy for genetic blood cell diseases.

Graphical Abstract

Hematopoietic stem cell (HSC) gene therapy is a promising clinical strategy to treat a wide range of genetic blood diseases. Masiuk and colleagues demonstrate a clinically relevant method to improve HSC gene therapy efficiency through isolation and transplantation of highly purified HSCs prior to lentiviral (LV) transduction. Enhanced HSC enrichment allows for a 5- to 10-fold reduction in LV dose and culture scale compared with currently used methods.

Introduction

Lentiviral vector (LV)-based gene modification of autologous hematopoietic stem cells (HSCs) has demonstrated remarkable success for the treatment of genetic blood diseases.1, 2, 3, 4, 5, 6 In this approach, bone marrow (BM) or mobilized peripheral blood (mPB) is collected from a patient, enriched for HSCs, transduced with LV encoding the correct genetic sequence, and transplanted back into the patient. After transplant, HSCs engraft in the BM and produce mature, gene-corrected hematopoietic cells ideally throughout the patient’s lifetime. New clinical trials using anti-sickling β-globin vectors show promising early results in the treatment of sickle cell disease (SCD), the most common genetic blood disorder affecting millions of people.⁷ However, broad use of gene therapy will require an unprecedented amount of clinical-grade vector production, a process that is currently limited by high cost and commercial manufacturing capacity. Currently, production and clinical validation of β-globin vector for a single patient may require 20–40 L of viral supernatant and cost in the range of $250,000–$500,000. Therefore, new methods that reduce the amount of LV required per patient are needed to make gene therapy clinically and commercially feasible.

One strategy to reduce LV requirements is to improve enrichment of HSCs before transduction. Current clinical HSC enrichment protocols use immunomagnetic beads (IBs) to select for cells expressing the surface marker CD34 prior to LV transduction. This method is inefficient in a gene therapy setting, because the majority of CD34⁺ cells are short-term progenitor cells with a limited post-transplant lifespan. Further isolation of HSCs from short-term progenitors could improve LV economy by only transducing cells that can endure for the lifetime of the patient. We postulated that purifying CD34⁺CD38⁻ cells would further enrich for HSCs, reduce the absolute number of cells to be transduced with LV, and reduce the total LV dose required per patient. Importantly, these benefits could be achieved while still retaining the target HSCs required for long-term clinical benefit after transplantation.

To date, little is known about clinical transplantation of highly purified HSCs. A small number of clinical trials have shown successful transplantation of CD34⁺CD90⁺ cells,8, 9, 10 which represented ∼50% of total CD34⁺ cells. In contrast, CD34⁺CD38⁻ cells are highly enriched for HSCs and represent ∼5%–10% of CD34⁺ cells. While a large body of work has characterized the repopulation capacity of CD34⁺CD38⁻ cells, the majority of this work has been performed using fluorescence-activated cell sorting (FACS)-purified cord blood (CB). In contrast with FACS purification, IB-based cell sorting is currently available in Good Manufacturing Practice (GMP)-grade platforms with the capacity to purify large cell numbers quickly with high recovery. Here, we investigate an IB-based method to purify CD34⁺CD38⁻ cells from human BM and mPB samples with focus on practical implementation of our findings to current clinical gene therapy trials. We demonstrate the capacity of this method to reduce LV dose by ∼5- to 10-fold while maintaining early post-transplant recovery and long-term engraftment of gene-modified cells. These findings have the potential to reduce the total LV required per patient and improve the clinical and commercial viability of gene therapy for genetic blood cell diseases.

Results

More Than 90% of CD34⁺ Long-Term Repopulating Activity Is Contained within the Lowest 6% of CD38 Expression

Translating CD34⁺CD38⁻ purification to the clinic will require efficient HSC recovery to sustain engraftment of gene-corrected cells after transplant. In order to identify a target cell population to purify for gene therapy, we first sought to determine which subsets of BM CD34⁺ cells are capable of long-term engraftment after ex vivo culture required for LV transduction. We performed a competitive transplant assay in immune-deficient non-obese diabetic (NOD)/severe combined immune deficiency (SCID)/gamma (NSG) mice utilizing CD34⁺ cells with varying levels of CD38 expression (Figure 1A). IB-purified CD34⁺ BM cells were sorted by FACS to obtain three different cell populations with increasing, constant percentiles of CD38 expression (e.g., 6% intervals: 0%–6%, 6%–12%, and 12%–18% of CD38 expression). Each fraction of sorted cells was pre-stimulated with cytokines (interleukin-3 [IL-3], stem cell factor [SCF], fms-related tyrosine kinase 3 [Flt3]-ligand, and thrombopoietin) for 24 hr and then transduced for an additional 24 hr with one of three fluorescent LVs: CCL-ubiquitin C (UBC)-mCitrine, CCL-UBC-mStrawberry, or CCL-UBC-mCerulean. Following transduction, 15,000 cells from each of three differentially labeled CD38 intervals were competitively co-transplanted into sub-lethally irradiated NSG mice. Fluorescent reporter gene color choice for each sorted fraction was rotated to account for potential differences in transduction efficiency among vectors (Figure S1).

Figure 1.

Long-Term Repopulating Activity of CD34⁺CD38⁻ Subsets

(A) Experimental setup. (1) FACS sort: BM cells were magnetically enriched for CD34 and sorted by FACS into distinct populations of increasing CD38 expression. Each gate contains an equivalent number of cells. (2) LV transduction: Cells from each interval were transduced with one of three LVs expressing the fluorescent proteins mCitrine, mStrawberry, or mCerulean. (3) Transplant: A combination of cells from each sorted interval (each differentially labeled) were combined and competitively transplanted into sub-lethally irradiated NSG mice. Each mouse received an equal number of cells from each sorted fraction. (4) BM analysis: NSG BM was analyzed at 16–18 weeks for the presence of fluorescently labeled hCD45⁺ cells. (5) Fluorescently labeled hCD45⁺ cells engrafted in the NSG BM were analyzed for their relative contribution to human hematopoiesis. (B) Relative abundance of labeled cells in NSG BM. Two separate transplants were performed using different intervals of CD38 expression (top and bottom). Initial sort for CD38 expression is shown in the left panels. At 16–18 weeks post-transplant, fluorescent, LV-labeled hCD45⁺, hCD34⁺, hCD19⁺, and hCD33⁺ cells were analyzed. y axis represents the relative frequency of each sorted and/or labeled CD38 fraction as a percent of all fluorescently labeled cells. Bars represent mean ± SEM. n = 7 mice for transplant 1; n = 3 mice for transplant 2.

At 16–18 weeks post-transplantation, engrafted hCD45⁺ cells in NSG BM were analyzed by flow cytometry for mCitrine, mStrawberry, and mCerulean expression to determine the long-term engraftment potential of each initially sorted population. Two separate transplants were performed using different intervals of CD38 expression (Figure 1B). In transplant 1, >90% of hCD45⁺ cells present in the BM were derived from CD34⁺CD38^0%–6% cells, whereas <10% of hCD45⁺ cells were derived from CD34⁺CD38^6%–12% and CD34⁺CD38^12%–18% cells. In transplant 2, 54% of hCD45⁺ cells were derived from CD34⁺CD38^0%–3% cells, with an additional 40% of hCD45⁺ cells derived from the CD34⁺CD38^3%–6% fraction. For each transplant, the observed patterns were consistent in myeloid (CD33⁺), lymphoid (CD19⁺), and hematopoietic stem and progenitor cells (HSPCs) (CD34⁺) lineages.

Collectively, these results suggest that after ex vivo culture and LV transduction, the majority of long-term repopulating activity in CD34⁺ adult BM cells is contained within the lowest 6% of CD38 expression. Furthermore, while the CD34⁺CD38^0%–3% fraction is highly enriched for long-term HSCs, the CD34⁺CD38^3%–6% also contains a substantial proportion of long-term HSCs; thus, purification beyond the lowest 6% of CD38 expression may lead to HSC loss. Based on these results, we defined “CD34⁺CD38⁻ cells” as the lowest 6% of CD38 expression for subsequent experiments tracking phenotypic HSC recovery.

Titration of CD38 Magnetic Labeling Optimizes Recovery and Purity of CD34⁺CD38⁻ Cells

We next sought to optimize a clinically relevant, IB-based method to purify CD34⁺CD38⁻ cells. We chose to first deplete CD38⁺ cells and subsequently select CD34⁺ cells. CD38 magnetic labeling was performed using a primary anti-CD38-phycoerythrin (PE)-conjugated antibody followed by incubation with anti-PE magnetic beads. Unlabeled CD38⁻ cells can be relabeled with CD34 beads, thereby preventing the necessity of a bead removal step. CD34⁺ cells exhibit a gradient pattern of CD38 expression rather than discrete positive and negative populations. We found that the proportion of cells separated into CD38⁻ and CD38⁺ fractions could be adjusted by varying the intensity of CD38 magnetic labeling with different concentrations of magnetic beads (Figure 2). Mononuclear cells (MNCs) from CB, BM, and mPB were separated into CD38⁻ and CD38⁺ fractions using three different concentrations of anti-PE beads. Strong magnetic labeling (anti-PE beads 1:5) greatly enriched for CD34⁺CD38⁻ cells (13-, 16-, and 15-fold for CB, BM, and mPB, respectively), but resulted in low recovery of CD34⁺CD38⁻ cells (50%, 19%, and 59% for CB, BM, and mPB). Weak magnetic labeling (anti-PE beads 1:25) resulted in high recovery of CD34⁺CD38⁻ cells (93%, 67%, and 96% for CB, BM, and mPB, respectively) but reduced enrichment of CD34⁺CD38⁻ cells (4.6-, 10.3-, and 4.7-fold for CB, BM, and mPB). Thus, there is a tradeoff between recovery and fold-enrichment of CD34⁺CD38⁻ cells that can be optimized by altering the strength of CD38 magnetic labeling. Here, we observe that anti-PE bead concentrations of 1:10 for mPB and >1:25 for BM (1:40 was used in all subsequent experiments) can optimize both enrichment and recovery of CD34⁺CD38⁻ cells.

Figure 2.

Titration of CD38 Immunomagnetic Labeling Influences Recovery and Enrichment of CD34⁺CD38⁻ Cells

(A) CD38 depletion of BM mononuclear cells (MNCs) using three different concentrations of anti-PE beads (one representative experiment is shown). Plots are gated to show CD34⁺ cells present in starting MNCs (far left), and after column separation into CD38⁺ (top) and CD38⁻ (bottom) fractions. Gates show CD34⁺CD38⁻ cells (defined as the lowest 6% of CD38 expression) and list absolute counts of CD34⁺CD38⁻ cells in each population. (B) Efficiency of CD38 depletion step at different PE bead concentrations. Bar graphs show “percent recovery” and “fold-enrichment” of CD34⁺CD38⁻ cells before and after CD38 depletion using three different anti-PE bead concentrations. Bars represent mean ± SEM. Data represent three independent experiments each for CB, BM, and mPB.

Co-depletion of CD15⁺ and CD38⁺ Cells Increases the Purity of Isolated CD34⁺CD38⁻ Cells

In BM and CB, we observed that CD38 depletion followed by CD34 selection led to a final cell product with <50% CD34⁺ cells because of contamination of CD15⁺ granulocytes (Figure 3A, top). Similar absolute numbers of granulocytes were isolated by CD34⁺ or CD34⁺CD38⁻ purification (Figure 3B), suggesting that the CD38 depletion step does not increase non-specific granulocyte selection. To obtain more highly purified CD34⁺CD38⁻ cells, we included a CD15 magnetic bead during the CD38 depletion step, allowing us to simultaneously co-deplete CD38⁺ cells and CD15⁺ granulocytes prior to CD34 selection (Figure 3C). In BM, addition of a CD15 bead increased purity of CD34⁺CD38⁻ cells from 24.5% of total cells to 70.5% of total cells, thus further reducing total number of cells isolated prior to transduction. Based on these findings, all subsequent CD34⁺CD38⁻ purifications were performed using CD38/CD15 co-depletion.

Figure 3.

Co-depletion of CD15⁺ and CD38⁺ Cells Increases CD34⁺CD38⁻ Purity

(A) CD34⁺CD38⁻ cells were purified from CB MNCs using one of two methods: CD38 depletion followed by CD34 selection (top) or CD38/CD15 co-depletion followed by CD34 selection (bottom). Leftmost FACS plot shows CD34 and CD38 staining of all viable leukocytes (CD45⁺DAPI⁻) isolated by either method. Histograms indicate the percentage of CD15⁺ and CD34⁺ cells within total viable leukocytes isolated by each method. (B) IB purification of BM 1 (no CD15 co-depletion). Left panel shows CD34 purification, and right panel shows parallel CD34⁺CD38⁻ purification (using CD38 depletion only). FACS plots show viable leukocytes (CD45⁺DAPI⁻) isolated by each method. Gates show CD34⁺CD38⁻ cells (top) and granulocytes (defined by high side scatter [SSC], bottom) purified by each method. Absolute cell counts are listed in parentheses. (C) IB purification of BM 2 (with CD15 co-depletion). Left panel shows CD34 purification, and right panel shows parallel CD34⁺CD38⁻ purification (using CD38/CD15 co-depletion). FACS plots show viable leukocytes (CD45⁺DAPI⁻) isolated by each method. Gates show CD34⁺CD38⁻ cells (top) and granulocytes (defined by high SSC, bottom) purified by each method. Absolute cell counts are listed in parentheses.

Optimized CD34⁺CD38⁻ Dual IB Purification Enriches CD34⁺CD38⁻ Cells with High Recovery and Results in Efficient HSC Gene Marking

We next compared the optimized CD34⁺CD38⁻ IB purification strategy with standard CD34⁺ selection in multiple BM and mPB samples from healthy donors. For each sample analyzed, two identical fractions of MNCs were purified by either CD34⁺ single-step IB purification or CD34⁺CD38⁻ dual IB purification. The total cell product obtained after purification by each method was analyzed by flow cytometry and counting beads to determine absolute CD34⁺CD38⁻ cell counts (Figures 4A and 4B; Figure S2). In BM, CD34⁺CD38⁻ dual IB purification of starting MNCs recovered a median 62.5% (range 57.2%–65.6%) of phenotypically defined CD34⁺CD38⁻ cells, whereas standard CD34 selection recovered a median of 93.1% (range 80%–95.6%) of CD34⁺CD38⁻ cells. CD34⁺CD38⁻ dual IB purification enriched for CD34⁺CD38⁻ cells an additional 8.7-fold (range 4.8–12) beyond standard CD34⁺ purification. Similar results were achieved in mPB, with a median CD34⁺CD38⁻ recovery of 75.1% (range 58.4%–91.0%) by dual CD34⁺CD38⁻ purification and 87.9% (range 83.7%–93.2%) by standard CD34 selection. CD34⁺CD38⁻ dual IB purification enriched for CD34⁺CD38⁻ cells an additional 12-fold (range 8.7–13.2) beyond standard CD34⁺ purification.

Figure 4.

Comparison of CD34⁺ Single-Step IB Purification versus CD34⁺CD38⁻ Dual IB Purification

(A) Comparison of a single representative BM sample purified in parallel by either CD34⁺ single-step IB purification (top) or CD34⁺CD38⁻ dual IB purification (bottom). Gates show CD34⁺CD38⁻ cells (defined as the lowest 6% of CD38 expression) with absolute cell counts listed in parentheses. (B) Recovery and enrichment of CD34⁺CD38⁻ cells obtained from IB purification of BM and mPB MNCs. 5 BM samples and 3 mPB sample from independent donors were purified in parallel by either CD34⁺ single-step IB purification or CD34⁺CD38⁻ dual IB purification. For each sample, CD34⁺CD38⁻ cells were defined as those with the lowest 6% of CD38 expression within CD34⁺ cells. Recovery indicates the percentage of CD34⁺CD38⁻ cells recovered after purification from starting MNCs, and it is calculated for both CD34⁺ single-step (row 1) and CD34⁺CD38⁻ dual IB purification (row 2). Fold-reduction (row 3) indicates the reduction in the total number of cells purified by CD34⁺CD38⁻ dual IB selection as compared with standard CD34 selection. This represents the clinically relevant reduction in the number of cells to be transduced and concordant vector dose reduction. Fold-enrichment (row 4) represents the enrichment of CD34⁺CD38⁻ cells as a percentage of total cells obtained by CD34⁺CD38⁻ dual IB purification compared with standard CD34⁺ IB purification. (Figure S2 shows absolute cell counts for each IB-purified fraction.) (C) Poisson statistical analysis of limiting dilution NSG xenotransplant assay of IB-purified BM. Plots demonstrate SRC frequency of CCL-UBC-mCitrine transduced CD34⁺ or CD34⁺CD38⁻ IB-purified cells (n = 5–7 mice transplanted at each cell dose per condition; see Figure S3 for individual values). y axis shows the percentage of recipient NSG mice containing <0.015% hCD45⁺ cells (non-engrafted) in the BM at 16 weeks post-transplantation versus the number of cells injected per mouse (x axis). (D) In vivo VCN analysis. NSG mice were transplanted with IB-purified CD34⁺ or CD34⁺CD38⁻ cells transduced with CCL-UBC-mCitrine LV. 16 weeks post-transplant, engrafted human cells in NSG BM were analyzed for VCN by digital droplet polymerase chain reaction (ddPCR). Data represent three independent experiments performed using BM from independent donors; n = 21–23 mice/group. n.s., not significant. Bars represent mean ± SEM.

To further assess recovery of functional HSCs, we transplanted cells from each purified population into NSG mice at limiting doses. Limiting dilution analysis of CCL-UBC-mCitrine LV-transduced cells demonstrated a 12-fold enrichment of SCID repopulating cell (SRC) frequency within CD34⁺CD38⁻ cells (1 in 2,314 cells; 95% confidence interval: 1/1,241–1/4,314) compared with CD34⁺ cells (1 in 28,248 cells; 95% confidence interval: 1/14,701–1/54,278) (Figure 4C; Figure S3). Analysis of long-term (>16 week) engrafted hCD45⁺ cells in NSG BM revealed no significant difference in average vector copy number per human cell between groups, suggesting that the dual CD34⁺CD38⁻ IB purification method allows for efficient transduction of HSCs while using less cells and LV (Figure 4D).

CD34⁺CD38⁻ Cells Show Delayed Myeloid Recovery that Can Be Rescued by Increasing Cell Dose or Adding Back Non-transduced CD38⁺ Cells

Purifying CD34⁺CD38⁻ cells may provide the advantage of utilizing less LV to transduce HSCs. However, transplanting a smaller population of quiescent, stem-enriched cells may cause delayed recovery of neutrophils and leave a patient susceptible to infection early after transplant. Therefore, we compared the early myeloid potential of IB-purified CD34⁺ and CD34⁺CD38⁻ cells. Here, we transplanted purified cells into the NSG-SGM3 mouse model,¹¹ an immune-deficient mouse model that supports human myelopoiesis through constitutive expression of human cytokines IL-3, granulocyte-macrophage colony-stimulating factor (GM-CSF), and SCF. In order to reflect a clinically relevant scenario, transplanted cell doses for each group represented the total cells obtained from IB purification of a standard volume of marrow (marrow equivalent [ME]). 1 marrow equivalent of CD34⁺CD38⁻ cells showed decreased early production of circulating human myeloid cells as compared with 1 marrow equivalent of CD34⁺ cells (Figure 5A). 2.5 marrow equivalent of CD34⁺CD38⁻ cells achieved myeloid reconstitution comparable with 1 marrow equivalent of CD34⁺ cells, whereas 6 marrow equivalent of CD34⁺CD38⁻ cells resulted in myeloid reconstitution ∼3-fold higher than 1 marrow equivalent of CD34⁺ cells. These results demonstrate that dual IB-purified CD34⁺CD38⁻ cells are capable of early myeloid reconstitution, but contain less early myeloid potential than CD34⁺ cells purified from an equivalent volume of BM.

Figure 5.

Early Myeloid Potential of IB-Purified CD34⁺ and CD34⁺CD38⁻ Cells

(A) Early myeloid engraftment of CD34⁺ and CD34⁺CD38⁻ IB-purified cells. Peripheral blood hCD45⁺hCD33⁺ cells per milliliter was quantified at 4 weeks from NSG-SGM3 mice transplanted with CD34⁺ cells or increasing doses of CD34⁺CD38⁻ cells IB-purified from BM. Cell doses for each group reflect the total cells obtained from IB purification from a standard volume of marrow (marrow equivalent). n = 3 mice/group. Bars represent mean ± SEM. (B) CD38 add-back experimental setup. BM MNCs were IB-depleted of CD3⁺ T cells to avoid xenogeneic graft-versus-host disease. CD3⁻ cells were further purified to obtain CD34⁺, CD34⁺CD38⁻, or CD38⁺ populations. CD34⁺ and CD34⁺CD38⁻ cells were transduced with a CCL-UBC-mCitrine LV, while CD38⁺ cells were cultured in parallel without the addition of LV. 4 experimental arms were transplanted, with each mouse receiving 1 marrow equivalent of the pictured graft. Gray boxes represent UBC-mCitrine transduced populations; white boxes represent non-transduced populations. (C) Early myeloid engraftment of IB-purified cells in NSG-SGM3 mice. Two transplants were performed using the four arms described in (B). Flow plots show CD34⁺ or CD34⁺CD38⁻ IB-purified cells prior to transplant (left: highly purified CD34⁺CD38⁻ cells; right: moderately purified CD34⁺CD38⁻ cells). Bar graphs represent mean ± SEM of myeloid engraftment in peripheral blood. Gray bars represent total myeloid engraftment, and overlay of black bars represents gene-marked (mCitrine⁺) myeloid engraftment. Myeloid engraftment is expressed as the percent of hCD45⁺CD33⁺/CD66b⁺/CD15⁺ cells out of total CD45⁺ (human and murine) cells. Each graph represents one experiment performed using BM from independent donors with n = 4–5 mice/arm. *p < 0.05, **p < 0.01. n.s., not significant.

We next investigated whether co-transplanting non-transduced CD38⁺ cells (obtained during IB depletion of MNCs) alongside transduced CD34⁺CD38⁻ cells could produce early hematopoietic recovery comparable with CD34⁺ cells. In order to avoid xenogeneic graft-versus-host disease (GVHD) in recipient mice receiving CD38⁺ cells, we purified CD34⁺, CD34⁺CD38⁻, and CD38⁺ populations from CD3-depleted BM MNCs (Figure 5B). IB-purified CD34⁺ or CD34⁺CD38⁻ cells were transduced with CCL-UBC-mCitrine LV, whereas CD38⁺ cells were cultured in parallel transduction conditions without LV. NSG-SGM3 mice received 1 marrow equivalent of the grafts depicted in Figure 5B: (1) CD34⁺ cells (+mCitrine LV), (2) CD34⁺CD38⁻ cells (+mCitrine LV), (3) CD34⁺CD38⁻ cells (+mCitrine LV) + CD38⁺ cells (non-transduced), or (4) CD38⁺ cells (non-transduced). Irradiated CD34⁻ cells were added as needed so that each mouse received the same number of total cells (Table S1).

Two transplants were performed using BM from independent donors. Transplant 1 used highly purified CD34⁺CD38⁻ cells (∼1% of total CD34⁺ cells), and transplant 2 used more moderately purified CD34⁺CD38⁻ cells (∼12% of total CD34⁺ cells) (Figure 5C). In transplant 1, mice receiving CD34⁺CD38⁻ cells showed significantly reduced levels of circulating human neutrophils at 3 weeks compared with mice transplanted with CD34⁺ cells. The addition of non-transduced CD38⁺ cells to the graft significantly increased circulating neutrophil levels. The same pattern was observed in transplant 2, but it did not reach significance when using more moderately purified CD34⁺CD38⁻ cells. Thus, when compared with CD34⁺ cells, moderately purified CD34⁺CD38⁻ cells (∼12% of total CD34⁺ cells) do not show as great of an early myeloid deficit when compared with stringently purified CD34⁺CD38⁻ cells (∼1% of total CD34⁺ cells). These results are consistent with the retention of CD38^low myeloid progenitors within the CD34⁺CD38⁻ fraction when using a more moderate purification strategy (∼12% of total CD34⁺ cells). In both transplants, the majority of human neutrophils present at 3 weeks in arm 3 were mCitrine⁻, confirming that these cells derive from non-transduced CD38⁺ cells. In the NSG model, we observed a similar defect in early myelopoiesis from transplanted CD34⁺CD38⁻ cells and rescue by the add-back of CD38⁺ cells (Figure S4), suggesting that these results are not unique to the NSG-SGM3 model. Collectively, these data show that addition of cultured, non-transduced CD38⁺ cells can restore early myeloid engraftment when added to transduced CD34⁺CD38⁻ cells.

Addition of CD38⁺ Cells Enhances Long-Term Engraftment of Gene-Modified HSCs

We next assessed the effect of adding back non-transduced CD38⁺ cells on long-term gene-marked engraftment. Here, we transplanted IB-purified CD34⁺, CD34⁺CD38⁻, and CD38⁺ cells into NSG mice using the same four experimental arms depicted in Figure 5B. To reflect a clinically relevant scenario, each mouse received the total cell product purified from a set volume of BM. At 16 weeks post-transplant, recipients of CD34⁺CD38⁻ cells showed decreased engraftment of total (hCD45⁺) and gene-marked (mCitrine⁺hCD45⁺) cells as compared with recipients of CD34⁺ cells (Figure 6A). Addition of non-transduced CD38⁺ cells to the graft of transduced CD34⁺CD38⁻ cells significantly increased engraftment of gene-marked (hCD45⁺mCitrine⁺) cells to levels achieved by CD34⁺ cell transplants. Importantly, this was accomplished using 6.2- and 7.8-fold less LV (in two independent experiments) compared with the volumes required to transduce bulk CD34⁺ cells. All transplanted arms showed similar lineage distribution, with the exception of mice transplanted with CD38⁺ cells alone, which showed a significant myeloid bias in engrafted cells (Figure 6B). PB analysis showed comparable engraftment of total hCD45⁺ cells at all time points after transplant of transduced CD34⁺ cells or transduced CD34⁺CD38⁻ cells co-transplanted with non-transduced CD38⁺ cells (Figure 6C, left). At early time points (3–12 weeks), transduced CD34⁺ cells showed higher levels of gene-marked (mCitrine⁺) engraftment than transduced CD34⁺CD38⁻ cells co-transplanted with non-transduced CD38⁺ cells (Figure 6C, right). However, after 12 weeks, we observed equivalent levels of gene-marked engraftment in these two groups. These data suggest that non-transduced CD38⁺ cells drive early hematopoiesis (3–12 weeks), while gene-marked CD34⁺CD38⁻ cells take over hematopoiesis 12 weeks after transplant.

Figure 6.

Long-Term Engraftment of IB-Purified Cells

(A) Long-term engraftment of IB-purified cells in NSG BM (top) or spleen (bottom) 16 weeks post-transplant. Gray bars represent mean ± SEM of total human engraftment (%hCD45⁺); black bars represent gene-marked engraftment (%hCD45⁺mCitrine⁺). Engraftment is expressed as the percent of hCD45⁺ or hCD45⁺mCitrine⁺ cells out of total CD45⁺ (human and murine) cells. Data represent two experiments using BM from independent donors with a total of eight to nine mice per arm. (B) Lineage analysis of total hCD45⁺ in NSG BM 16 weeks post-transplant. Bars show mean ± SEM of each lineage (CD19⁺ B cells, CD33⁺ myeloid cells, CD3⁺ T cells, and CD34⁺ HSPCs) expressed as percent of total hCD45⁺ cells. Data represent two experiments using BM from independent donors with a total of eight to nine mice per arm. (C) PB engraftment over time after transplant of CCL-UBC-mCitrine transduced CD34⁺ cells (filled circles) or CCL-UBC-mCitrine transduced CD34⁺CD38⁻ cells combined with non-transduced CD38⁺ cells (open squares). Left panel shows total PB engraftment (%hCD45⁺); right panel shows gene-marked engraftment (%hCD45⁺mCitrine⁺). *p < 0.05, ***p < 0.001. n.s., not significant.

Discussion

We have demonstrated a clinically relevant, IB-based method for purifying CD34⁺CD38⁻ cells from CB, mPB, and BM (Figure 7). We further show that modification of CD34⁺CD38⁻ cells with a reduced LV dose coupled with co-transplantation with non-modified CD38⁺ cells produced early myeloid reconstitution and long-term engraftment of gene-marked cells comparable with traditional methods of modifying bulk CD34⁺ cells.

Figure 7.

Proposed Clinical Cell Processing Pathway

BM or mPB will be RBC/platelet depleted, followed by labeling with CD38-PE, anti-PE beads, and anti-CD15 beads. The CD38⁺/CD15⁺ fraction (bound to the column) will undergo short-term culture followed by cryopreservation and release testing. The CD38⁻/CD15⁻ fraction (column flow-through) will be labeled with anti-CD34 beads, followed by selection of CD34⁺CD38⁻ cells. CD34⁺CD38⁻ cells will undergo transduction with LV followed by cryopreservation and release testing. After successful release testing, transduced CD34⁺CD38⁻ cells and non-transduced CD38⁺/CD15⁺ cells will be thawed and co-transplanted into a conditioned patient.

The strategy explored here is broadly applicable to a number of HSC-based gene therapy applications and may be especially useful in gene therapy for hemoglobinopathies, where clinical-scale LV production has proven difficult because of the large size and complexity of the human β-globin gene expression cassette. We estimate that the additional cost to purify 1 × 10¹⁰ total BM or mPB cells beyond standard CD34⁺ IB purification is ∼$10,000 per patient (based on the cost of commercially available CD38 antibody, CD15 antibody, and anti-biotin beads). Therefore, the per-patient cost of improving stem cell enrichment is highly cost-efficient compared with current vector production costs (∼$500,000 per patient). Furthermore, this strategy may improve efficiency in clinical-scale HSC gene-editing applications such as CRISPR/Cas9.

While a number of in vivo repopulating studies have shown that HSCs are enriched within the CD34⁺CD38⁻ fraction,12, 13, 14, 15, 16 we sought to more thoroughly define a target population to isolate for gene modification. The majority of prior work has evaluated the engraftment capacity of FACS-sorted/uncultured cells, whereas current gene therapy protocols require a period of ex vivo culture that likely alters repopulating capacity. Additionally, prior work has used variable definitions for CD34⁺CD38⁻ cells ranging from the lowest 1% to 30% of CD38 expression. Some work has also suggested that a CD34⁺CD38^low fraction contains long-term repopulating activity.14, 16 Here, we find that within freshly isolated BM CD34⁺ cells, only the lowest 6% of CD38 expression is capable of long-term NSG repopulation after LV transduction/ex vivo culture. Thus, purifying CD34⁺ cells based on CD38 expression alone could lead to a theoretical ∼16-fold enrichment of long-term repopulating activity.

High individual donor variability in CD38 expression could be a potential hurdle to clinical translation. Because the lowest 6% of CD38 expression may not encompass >90% of HSCs for every individual, a conservative cutoff for CD38 expression may increase the chance of retaining HSCs. Our results show that the level of CD38 expression retained within the CD38⁺ and CD38⁻ populations can be adjusted by titration of magnetic beads, thereby achieving optimal recovery of CD34⁺CD38⁻ cells while also maximizing HSC enrichment. Clinical scale-up to GMP-grade IB sorting systems will likely require additional titration of CD38 magnetic labeling specific to the system parameters (magnetic field strength, column flow rate, etc.).

In the dual CD34⁺CD38⁻ IB protocol, we found that CD38⁺/CD15⁺ co-depletion increased purity of CD34⁺CD38⁻ cells by preventing the non-specific enrichment of CD15⁺ granulocytes. Due to the relative rarity of CD34⁺CD38⁻ cells (∼0.1%–0.3% of MNCs), non-specific enrichment of granulocytes can represent >50% of total cells after CD34⁺CD38⁻ purification. In contrast, CD34⁺ cells are ∼10 times more abundant than CD34⁺CD38⁻ cells (∼1%–3% of MNCs); thus, the relative contribution of granulocytes is minor in CD34⁺ purified cells (<10%). Because clinical vector doses are calculated on a per-cell basis, contaminating granulocytes present in CD34⁺CD38⁻ purified populations could reduce vector MOI and limit the vector dose reduction achieved by CD34⁺CD38⁻ IB purification. Furthermore, cryopreservation of a granulocyte-rich cell population could potentially lead to poor cell recovery upon thawing because of granulocyte death and/or clumping.

Application of the optimized CD34⁺CD38⁻ dual IB method to samples of healthy BM and mPB achieved ∼75% recovery and ∼10-fold enrichment of phenotypically defined CD34⁺CD38⁻ cells in mPB, and ∼60% recovery and ∼5-10-fold enrichment in BM. In BM, CD34⁺CD38⁻ purification results in a 12-fold enrichment of SRC beyond standard CD34⁺ selection. We observed no difference in transduction efficiency of IB-purified CD34⁺CD38⁻ or CD34⁺ cells when evaluated by vector copy number (VCN) analysis of long-term NSG-engrafted human cells. This is in contrast with our lab’s previous demonstration that highly purified (FACS-sorted) CB CD34⁺CD38⁻ cells demonstrate enhanced transduction in vitro compared with CD34⁺ cells.¹⁷ This discrepancy could be because of differences in the tissue analyzed (BM versus CB), CD34⁺CD38⁻ purity (moderately purified IB-sorted cells versus highly purified FACS-sorted cells), or post-transduction assay (in vitro culture versus in vivo xenografts). Importantly, our results here suggest that the dual IB CD34⁺CD38⁻ purification does not impede LV transduction of long-term HSCs.

We next explored the early myeloid potential of IB-purified CD34⁺CD38⁻ cells. Prior work has shown that short-term repopulating activity is enriched in the CD34⁺CD38⁺ fraction,¹⁸ suggesting that CD34⁺CD38⁺ progenitors may be necessary for early myeloid recovery. However, murine studies have suggested that high doses of purified stem cells can lead to rapid hematopoietic recovery.19, 20 Clinical trials using FACS-sorted CD34⁺CD90⁺ cells have demonstrated prompt neutrophil and platelet engraftment in patients transplanted with >0.8 × 10⁶ CD34⁺CD90⁺ cells/kg.8, 9, 10 However, the purified CD34⁺CD90⁺ populations in these studies represented ∼50% of total CD34⁺ cells, whereas the IB-purified CD34⁺CD38⁻ cells evaluated here represent ∼10% of total CD34⁺ cells. Therefore, it is largely unknown whether high doses of purified CD34⁺CD38⁻ cells could provide rapid myeloid reconstitution in a clinical transplant setting, or if co-transplant of CD34⁺CD38⁺ cells will be required.

Here, we observed that IB-purified CD34⁺CD38⁻ cells were capable of early myeloid reconstitution, but contained less early myeloid potential than an equivalent marrow volume of CD34⁺ cells. Our xenograft data suggest that approximately two to three times as many IB-purified CD34⁺CD38⁻ cells (when transplanted alone) may be required to achieve early myeloid reconstitution comparable with CD34⁺ cells. In a clinical scenario, it is possible that the total product of CD34⁺CD38⁻ cells obtained will be above the threshold required for sufficient myeloid reconstitution. However, we favor the safer approach of adding non-transduced CD38⁺ cells (containing CD34⁺CD38⁺ progenitors) to the graft, because transplant of bulk CD34⁺ cells in gene therapy clinical trials has historically provided prompt myeloid reconstitution.

Furthermore, add-back of non-transduced CD38⁺ cells may also play a critical role in achieving optimal long-term engraftment of gene-modified HSCs. In long-term NSG transplant assays, we observed decreased engraftment of both total and gene-marked cells in recipients receiving CD34⁺CD38⁻ cells alone as compared with recipients receiving CD34⁺ cells. However, when non-transduced CD38⁺ cells were added to transduced CD34⁺CD38⁻ cells, levels of gene-modified engraftment were equivalent to those achieved by standard CD34⁺ transduction/transplant. These data suggest that the CD34⁺CD38⁻ dual IB purification method efficiently recovers HSCs but requires co-transplantation of CD34⁺CD38⁺ cells for optimal long-term engraftment. In agreement with our findings, prior work has demonstrated that cycling CD34⁺CD38⁺ cells facilitate HSC engraftment21, 22, 23 through enhancement of stromal cell-derviced factor 1 (SDF-1)-mediated homing and secretion of metalloproteinase (MMP)-9.²³

An important advantage of using CD38 as an enrichment marker is that it can cleanly separate HSC-containing and HSC-devoid fractions; virtually no repopulating activity is contained within the CD38⁺ fraction, and almost all repopulating activity is contained within the CD34⁺CD38^0%–6% fraction. Although additional HSC markers have been identified24, 25 (e.g., CD90, CD45RA, and CD49f), in vivo xenograft analysis of human HSC populations suggests that these markers enrich for HSCs but do not exclusively define HSCs. For example, CD34⁺CD38⁻CD45RA⁻CD90⁺ cells contain more HSCs per cell than CD34⁺CD38⁻CD45RA⁻90⁻ cells; however, CD34⁺CD38⁻CD45RA⁻90⁻ cells do contain long-term HSCs capable of multi-lineage reconstitution of NSG mice.²⁴ Thus, discarding the CD34⁺CD38⁻CD45RA⁻90⁻ population prior to gene modification or transplant would likely result in a lower total HSC dose and could be disadvantageous in a clinical scenario.

In our CD38 add-back transplant studies, we used a conservative CD38 depletion strategy (6.2- to 7.8-fold reduction in cell number/LV) in order to maximize the number of HSCs retained in the CD38⁻ fraction. It is possible that further reductions in total cell number and LV dose could be achieved using a more stringent CD38 depletion or using additional cell surface markers (e.g., CD90, CD49f). These strategies may be advantageous if highly purified HSCs can be modified and expanded with compounds such as StemRegenin-1 (SR-1),26, 27 prostaglandin E2 (PGE2),28, 29 or UM171.³⁰

In summary, we demonstrate a method to improve the efficiency of gene therapy for genetic blood cell diseases through improved HSC enrichment and reduced LV dose.

Materials and Methods

MNC Isolation

Healthy adult BM and mPB were obtained from commercial sources (All Cells and Hemacare). Umbilical CB was obtained after vaginal and cesarean deliveries at UCLA Medical Center. All specimens obtained have been deemed as anonymous medical waste exempt from institutional review board review. MNCs were isolated using Ficoll-Paque PLUS (GE Healthcare) density centrifugation within 48 hr of collection. The total number of MNCs used for IB purification varied with each cell product and ranged from 4 × 10⁶ to 4 × 10⁷ (absolute cell numbers are summarized in Figure S2B).

IB Purification of CD34⁺ and CD34⁺CD38⁻ Cells

All microbeads and magnetic columns used for cell separation were purchased from Miltenyi Biotec. All incubation steps were performed in magnetic-activated cell sorting (MACS) buffer (PBS/0.5% BSA/2 mM EDTA) at 4°C with 10⁷ cells/100 μL total volume. Total reagent/buffer volumes were scaled accordingly based on MNC number in order to keep cell density and reagent concentration constant for each processed sample. Concentrations described for each reagent represent the reagent volume:total volume of reagent and MACS buffer.

MNCs were first stained with CD38-PE (1:15, Clone IB6; Miltenyi Biotec), CD34-fluorescein isothiocyanate (FITC) (1:10, Clone 581; BD Biosciences), and CD45-allophycocyanin (APC) (1:10, Clone HI30; BD Biosciences) for 30 min. Stained MNCs were washed once with MACS buffer and divided into equal fractions for further IB processing. Each experiment comparing CD34⁺ and CD34⁺CD38⁻ IB purification (cell counts and xenograft studies) utilized the total purified cell product isolated from an equal starting volume of MNCs (defined in the text as marrow equivalent).

To purify CD34⁺ cells, we incubated MNCs with anti-CD34 microbeads (1:5) for 30 min, washed them, and added them to an LS column. The column was washed with 3 × 3 mL of MACS buffer. CD34⁺CD38⁻ cells were purified by first incubating MNCs with anti-PE microbeads (1:5–1:25 for bead titration studies, 1:10 for mPB purification, 1:40 for BM purification and xenograft studies) and anti-CD15 microbeads (1:5) for 15 min. Cells were washed and applied to an LD column. The column was washed with 2 × 1 mL fractions of MACS buffer. After washing, the CD38⁺/CD15⁺ fraction was flushed from the column. The CD38⁻/CD15⁻ flow-through fraction was collected, washed, and subsequently selected with anti-CD34 microbeads as described above. To obtain absolute CD34⁺CD38⁻ cell counts in each fraction, we added 3 × 100 μL aliquots to 300 μL of MACS buffer and 50 μL of counting beads (eBioscience) and DAPI (1:1,000; Life Technologies), and analyzed them on an LSRII or LSR Fortessa flow cytometer (BD Biosciences) (Figure S2). In xenograft experiments where CD38⁺ cells were added back to the graft, CD3-depleted MNCs were used as a starting material to deplete mature human T cells and avoid xenogeneic GVHD. MNCs were incubated with anti-CD3 microbeads (1:5) at 4°C for 15 min, washed, and separated on an LD column. CD3⁻ cells were immediately processed by further CD34⁺ or CD34⁺CD38⁻ IB purification.

Cell Sorting

BM MNCs were enriched for CD34⁺ cells as described and were stained with CD38-PE (1:15, Clone IB6; Miltenyi) and CD34-APC (1:10, Clone 581; BD Biosciences) in MACS buffer for 30 min at 4°C, followed by washing to remove unbound antibody. DAPI (1:1,000; Life Technologies) was added just before analysis. Cells were gated on viable, single CD34⁺ cells (Figure S1A) prior to defining CD38 intervals. Cells were sorted according to defined intervals of CD38 expression using a FACSAria II (BD Biosciences).

LV Transduction

Construction, packaging, and titering of CCL-UBC-mCitrine-PRE-FB-2XUSE, CCL-UBC-mStrawberry-PRE-FB-2XUSE, and CCL-UBC-mCerulean-PRE-FB-2XUSE have been described.¹⁷ Cells were plated on retronectin (Takara Shuzo)-coated plates (20 μg/mL) at a density of 0.5–1 × 10⁶ cells/mL in X-VIVO15 medium (Lonza) containing 1× glutamine/penicillin/streptomycin (Gemini BioProducts), 50 ng/mL SCF, 50 ng/mL fms-related tyrosine kinase 3 ligand (Flt3-L), 50 ng/mL thrombopoietin (TPO), and 20 ng/mL IL-3 (PeproTech). LV was added to a final concentration of 2 × 10⁷ TU/mL and incubated with cells for 24 hr prior to transplant. All LV concentrations (transducing units [TU]/mL), cell concentrations (cells/mL), and cell plating densities (cells/cm²) were kept constant during transduction of CD34⁺ and CD34⁺CD38⁻ cells by adjusting the total number of wells for plating/transduction proportional to cell counts (CD34⁺ cells required ∼5–10 times more cell culture wells than CD34⁺CD38⁻ cells).

Xenografts

All work with mice was done under protocols approved by the UCLA Animal Care Committee. 6- to 10-week-old male and female NOD.Cg-Prkdsci-dIl2rgtm1Wjil/SzJ (NSG) and NOD.Cg-PrkdcscidIl2rgtm1WjlTg(CMV-IL-3, CSF2, KITLG)1Eav/MloySzJ (NSG-SGM3) mice (Jackson Laboratory) were used as transplant recipients. Mice were irradiated 4–6 hr prior to transplantation using a ¹³⁷Cesium source at a total dose of 250 rad (∼101 rad/min). Cells were harvested from retronectin-coated dishes by gentle pipetting and PBS washes. Collection of all cells from each well was verified by visual inspection under the microscope. Collected cells were washed, resuspended in PBS, and administered via intravenous injection into the retro-orbital sinus. CD34⁻ cells were irradiated (10 Gy) and added to grafts as filler cells (total cell doses per mouse are listed in Table S1).

Engraftment Analysis

Engraftment of human cells in PB, BM, and spleen of xenograft recipient mice was evaluated by flow cytometry using anti-human CD45-APC (HI30), anti-murine CD45-PE (Clone 30-FII), anti-human CD33-V450 (Clone P67-6), anti-human CD66b-V450 (Clone G10F5), anti-human CD15-V450 (Clone HI98), anti-human CD3-PerCP-Cy5.5 (Clone SK7), anti-human CD19-APC-Cy7 (SJ25CI), and anti-human CD34-PE-Cy7 (Clone 581) (all antibodies are from BD Biosciences).

Determination of Vector Copies per Human Cell

The average vector copies (VCs) per human cell was measured in engrafted NSG BM samples as previously described.¹⁷ In brief, LV DNA content was quantified using a digital droplet PCR probe specific to the HIV-1 Psi region and normalized to the autosomal human gene SDC4.

Statistics

Pairwise comparison was performed by unpaired t test within the framework of one-way ANOVA. Two group comparisons by Wilcoxon rank-sum test were performed when the assumption of normality was not met. Hypothesis testing was two-sided, and a significance threshold of p = 0.05 was used. Limiting dilution analysis was performed using online software provided by Walter and Eliza Hall Institute (WEHI) bioinformatics.³¹

Author Contributions

K.E.M. designed, performed, and analyzed experiments and wrote the manuscript. D.B. and J.L. performed portions of the research studies. R.P.H. provided study materials and advised experiments. F.U. advised experiments. D.B.K. participated in conception and design of the studies, provided financial support through grant funding, and co-wrote the manuscript.

Conflicts of Interest

The authors declare no conflict of interest.