Semi-automated closed system manufacturing of lentivirus gene-modified haematopoietic stem cells for gene therapy

Haematopoietic stem cell (HSC) gene therapy has demonstrated potential to treat many diseases. However, current state of the art requires sophisticated ex vivo gene transfer in a dedicated Good Manufacturing Practices facility, limiting availability. An automated process would improve the availability and standardized manufacture of HSC gene therapy. Here, we develop a novel program for semi-automated cell isolation and culture equipment to permit complete benchtop generation of gene-modified CD34+ blood cell products for transplantation. These cell products meet current manufacturing quality standards for both mobilized leukapheresis and bone marrow, and reconstitute human haematopoiesis in immunocompromised mice. Importantly, nonhuman primate autologous gene-modified CD34+ cell products are capable of stable, polyclonal multilineage reconstitution with follow-up of more than 1 year. These data demonstrate proof of concept for point-of-care delivery of HSC gene therapy. Given the many target diseases for gene therapy, there is enormous potential for this approach to treat patients on a global scale.

1) In figure 1b and c, a significant portion of the transplanted NSG mice have very little engraftment with human cells (less than 10%, particularly in APH donor 1). Would this be the same if the cells were processed manually and transduced in flasks ?
2) The transduction efficiencies are also relatively low as shown in figure 1d, with most mice showing less than 20% transduction of human cells. Is this due to the automated procedure, or rather intrinsic to the donor source and vector used. A "gold standard", concurrent control is needed to address this question.
3) The same criticisms apply to figure 3, very low VCNs in the bone marrow are present in 3C, ... process, cells, or vector ? 4) For the monkey experiments shown in figure 4, marking data is only shown for T and B lymphocytes, why is there no data from the myeloid cell populations ? Without myeloid marking data, it is very difficult to interpret the HSC transduction efficiency because lymphocytes are very long-lived in circulation. The data shown in supplementary figure S4 shows very low levels of marking in erythrocytes and platelets, further raising the concern that HSC transduction in these cases was relatively low.
Therefore, it seems critical to provide more experiments to determine if the semi-automated method is achieving similar transduction frequencies to what can be obtained with direct, manual manipulations.
Other points: 1) The Title: "Point of Care ..." is very vague and confusing and a scientific reader would have no idea what this means. I recommend changing the title with something that describes the new approach more clearly. Perhaps, "Semi-automated processing and transduction of human HSCs with lentiviral vectors in a close system device" or something along those lines.
2) Are the computer programs used in this study going to be made available to the readers ? This is an important question regarding reproducibility of these results in other centers and the ability to advance the field.
3) There is not enough description of the transduction process. How many vector applications were used, at what MOI, were the cells washed between transduction, what was the titer of the vector, etc. 4) While the clinical anti-HIV P140K vector used is appropriate as a clinical product, would the studies be easier to interpret if a high titer, simple, GFP LV were used ? This would allow better comparison with published transduction efficiencies using standard techniques.
Reviewer #2 (Remarks to the Author) 1. General comment: for RESULTS, it would help the reader if the Bone Marrow Processing section and the APH Processing section were marked as separate in the text. 2. Page 3: why the use of "APH" for a mobilized leukapheresis product? ISBT128 term is HPC-A. 3. Table S1: For samples 3 and 4, please explain the increased yield of CD34+ cells number after RBC depletion? Is it possible that the RBCs inhibited CD34 staining in the pre-depleted samples? 4. Page 5: line 19. Unless you mean this literally, the use of the word "release" is ambiguous and should be "completion". 5. Page 7: line 18-20. It does not seem possible that the cell viability decreased during transduction, despite an increase in viable cell number. Please correct or explain. 6. Page 8, line 15. How was the process modified when the NHP cell product was made? 7. Page 11, line 23. Please provide a reference that supports the lack of effect of fibronectin on VSV-G LV transduction. 8. DISCUSSION. Please comment further on the time of the procedure (~25 hr) and whether the ~4 hr direct involvement of the technician really means that there is 21 hr of 'walk away' time.

The system described in this study is based on a semiautomated process that utilizes the CliniMACS Pr4odigy device, which is commercially available. The authors developed a set of custom programming for this device that allows for grafts to undergo 1) RBC depletion, 2) CD34 selection, 3) transduction of selected cells with a lentiviral vector, and 4) harvest and formulation of cells for patient infusion. This can be accomplished with a modest amount of operator input and in limited space. The system was tested with human bone marrow and G-CSF mobilized peripheral blood samples as well as non-human primate bone marrow samples and high quality data is provided regarding the CD34+ cell yields, purity, RBC depletion efficiency, and lentiviral gene marking in transplanted NSG mice (human cells) and transplanted pigtail macaques (NHP samples).
While there is great enthusiasm for the significance and novelty of this study, the effectiveness of this new approach is not clearly established based on the data presented. In general, engraftment and transduction data are variable and at times raise questions regarding the engraftment capacity of the transduced grafts as well as the transduction efficiency. What is lacking in these experiments are controls showing grafts transduced in the conventional manner. Therefore, one cannot be sure whether this automated approach is as efficient as the "gold standard" approach of operator-dependent CD34+ purification and transduction in retronectin coated flasks. A direct comparison of the new technology with the old methodology, done in parallel using the same lentiviral vector and CD34+ cell source, would greatly improve the interpretability of these data as regards the efficiency of the new system.

RESPONSE:
The intention of our study was not to demonstrate the inferiority of manual processing of autologous grafts for gene therapy approaches, but to establish the proof of concept that this can be done in an automated setting, which, as this reviewer contends, will be necessary for the field of gene therapy to advance. We believe that the effectiveness of this approach is clearly established based on the data we present demonstrating manufacture transduced CD34 + cell populations which not only engraft in autologous nonhuman primate recipients in the myeloablative setting, but at levels equivalent to previous reports with VSV-g pseudotyped lentivirus vectors transduced using traditional methods including manual processing and transduction on retronectin at the same multiplicity of infection. To emphasize this in in the revised manuscript, we have added experimental data comparing standard transduction of CD34 + cells to our automated approach in the nonhuman primate. We compared the two animals presented in the original manuscript to two separate animals that received CD34 + cells transduced with the exact same LV vector under standard, manual conditions in our laboratory. We observed equivalent levels of transduction, hematopoietic reconstitution and gene modified cell engraftment across these four animals. We have summarized this comparison into the table shown below, which is now included in the main body of the revised manuscript as Table 2. Per this reviewer's request, this replaces the previous Supplemental Table S7 since it represents a critical control for these studies.

RESPONSE:
To address this reviewer's concern that a significant proportion of the NSG mice show very little engraftment with cells selected and processed in the closed system described here, we have compared our data to NSG mice transplanted by another laboratory in the published literature. We have cited Greene, M., et al. Human Gene Therapy Methods, 2012 Part B, Methods. 23: 297-308 (PMID:23075105) wherein the authors transduced human CD34 + hematopoietic cells from mobilized apheresis products with a VSV-G pseudotyped LV vector expressing the human gamma globin gene at a multiplicity of infection of 135 infectious units/cell and transplanted these into NSG mice at 1 × 10 6 cells per mouse following conditioning with 35 mg/kg busulfan. The authors observed median engraftment of 22.5% human CD45 + cells in the bone marrow of these mice at 12-16 weeks after transplantation, with a mean vector copy number ranging between 0.15 and 0.25 copies per human cell. In comparison, we observed 21% and 46% human CD45 + cells in the bone marrow of mice transplanted in our study (original Figure 1C) with VCN in human cells ranging from 0.1 to 0.17 per cell. Importantly, our levels of gene modified cells are only modestly lower to that reported by Greene and colleagues despite a dramatically lower MOI of 20 IU/cell during transduction, nearly 7-fold lower than that used in the published report. For reference, we have excerpted the relevant figures from Greene et al. and our current data below for direct comparison. PB: peripheral blood; ND: Not determined; CFC: colony-forming cells; ANC: absolute neutrophil count † Number of cells in the infusion product with potential to generate colonies in a standard CFC assay. Value is extrapolated from the percentage of colonies generated from 3000 cells plated in a single assay. ε Number of gene modified cells in the infusion product with potential to generate colonies in a standard CFC assay. Value is extrapolated from the percentage of gene modified colonies generated from 3000 cells seeded in a single assay.

2)
The transduction efficiencies are also relatively low as shown in figure 1d, with most mice showing less than 20% transduction of human cells. Is this due to the automated procedure, or rather intrinsic to the donor source and vector used. A "gold standard", concurrent control is needed to address this question.

RESPONSE:
As demonstrated in the above responses, we believe the levels of transduced cells observed in xenotransplanted mice are in line with previously published studies performing similar experiments (Greene et al., 2012) at nearly seven times the multiplicity of infection used in our study. Importantly, we would also like to emphasize the autologous nonhuman primate model data presented in our manuscript as noted in the first response above and in Table  2 of the revised manuscript. In this clinically relevant setting, we did not observe reduced engraftment of transduced autologous cells in the two animals who received cells manufactured in the automated system in this study compared to two separate animals that received CD34 + cells transduced with the exact same LV vector under standard, manual conditions in our laboratory. Again, this data is now included in Table 2. Per this reviewer's request, this replaces the previous Supplemental Table S7 since it represents a critical control for these studies.
3) The same criticisms apply to figure 3, very low VCNs in the bone marrow are present in 3C, .. process, cells, or vector?

RESPONSE:
We acknowledge in the Results and Discussion sections of the original manuscript that processing of CD34 + cells from steady state bone marrow products for LV vector transduction was not as efficient as observed for mobilized apheresis products. However, as noted in the above response to points 1 and 2, the human CD45 + cell engraftment and VCN in human cells observed in these studies are not dramatically different than what was previously reported by Greene and colleagues after LV transduction of mobilized peripheral blood CD34 + cells at seven times the MOI used in the current studies. There is no equivalent data on LV transduced steady state bone marrow derived CD34 + cells that we are aware of in the published literature. Therefore, we do not have a cited reference for comparison here. To address the potential difference between CD34 + cell sources, we have revised the text in the discussion section of the manuscript as follows: While extrapolated cell doses from human HPC-A products manufactured in this study would easily achieve this threshold, CD34 + cell numbers from BM products would be more clinically useful for children, requiring further optimization for adult patient populations for whom BM is the desired CD34 + cell source (e.g. sickle cell disease). Additionally, we observed lower levels of transduction and engraftment of gene modified cells in the xenotransplantation model when unmanipulated BM was the source of CD34 + cells, suggesting that culture and transduction conditions will need to be modified to improve efficiency.  figure S4 shows very low levels of marking in erythrocytes and platelets, further raising the concern that HSC transduction in these cases was relatively low.

4) For the monkey experiments shown in figure 4, marking data is only shown for T and B lymphocytes
Therefore, it seems critical to provide more experiments to determine if the semi-automated method is achieving similar transduction frequencies to what can be obtained with direct, manual manipulations.

RESPONSE:
We would like to clarify that gene marking in both granulocytes (myeloid) and lymphocyte compartments of nonhuman primates is shown in the graphs included in Figures 4a and 4b of the original manuscript (green lines). We have clarified this in the results section of the revised manuscript. Figures 4c and 4d were meant to further classify gene modification levels in lymphocyte sub-populations since T cell engraftment is critical for the target disease referenced in this study, (i.e. HIV/AIDS). To further support the successful transduction and engraftment of HSC, we have now extended these analyses in the revised manuscript to more than 450 days after transplant for each of these animals.
Here, HSC transduction and engraftment is supported by long-term, stable levels of gene modification in both granulocytes and lymphocytes at > 1 year since transplant. We have also included engraftment levels of autologous LV gene modified cells transplanted into two control nonhuman primates after standard culture and transduction for comparison as shown in the response to point 2 above ( Table 2 in the revised manuscript). This comparison demonstrates equivalent engraftment of LV gene modified cells using the point-of-care approach compared to standard manufacturing in a highly clinically relevant large animal model of autologous HSC transplantation in the myeloablative setting.

8.
DISCUSSION. Please comment further on the time of the procedure (~25 hr) and whether the ~4 hr direct involvement of the technician really means that there is 21 hr of 'walk away' time.

RESPONSE:
We have revised the Discussion section of the manuscript to clarify that this could be considered "walk away" time but that it is not a continuous 21 hours. Furthermore, to reliably continue manufacturing during the hands-off time without direct oversight, an alarm and status notification system should be in place to remotely track progress of the semi-automated system.

Reviewer #1 (Remarks to the Author)
The comparison shown in the newly included Table 2 addresses my main concern that the automated process had not been compared to the standard manual transduction procedure. This shows that in transplanted non-human primates, the transduction and engraftment with the automated process is giving equivalent results. The other changes that have been made also result in an improved manuscript including a more direct title, description of the key transduction parameters used in the automated process, and inclusion of extended reconstitution data in the primate system.