Durable engraftment of genetically modified FVIII‐secreting autologous bone marrow stromal cells in the intramedullary microenvironment

Abstract Genetically modified FVIII‐expressing autologous bone marrow‐derived mesenchymal stromal cells (BMSCs) could cure haemophilia A. However, culture‐expanded BMSCs engraft poorly in extramedullary sites. Here, we compared the intramedullary cavity, skeletal muscle, subcutaneous tissue and systemic circulation as tissue microenvironments that could support durable engraftment of FVIII‐secreting BMSC in vivo. A zinc finger nuclease integrated human FVIII transgene into PPP1R12C (intron 1) of culture‐expanded primary canine BMSCs. FVIII‐secretory capacity of implanted BMSCs in each dog was expressed as an individualized therapy index (number of viable BMSCs implanted × FVIII activity secreted/million BMSCs/24 hours). Plasma samples before and after implantation were assayed for transgenic FVIII protein using an anti‐human FVIII antibody having negligible cross‐reactivity with canine FVIII. Plasma transgenic FVIII persisted for at least 48 weeks after implantation in the intramedullary cavity. Transgenic FVIII protein levels were low after intramuscular implantation and undetectable after both intravenous infusion and subcutaneous implantation. All plasma samples were negative for anti‐human FVIII antibodies. Plasma concentrations and durability of transgenic FVIII secretion showed no correlation with the therapy index. Thus, the implantation site microenvironment is crucial. The intramedullary microenvironment, but not extramedullary tissues, supported durable engraftment of genetically modified autologous FVIII‐secreting BMSCs.


| INTRODUCTION
Haemophilia A treatment by protein factor replacement is invasive, expensive and only partially effective. 1  However, culture-expanded BMSCs engraft poorly in vivo.
BMSCs are intravenously infused in well over half of currently registered clinical trials. This is known to induce complement-mediated BMSC destruction which may partly explain equivocal results from hundreds of BMSC clinical trials. 6 The role of cell-specific tissue niches has not been adequately investigated as a factor that determines survival and engraftment of implanted BMSC. 7 We hypothesized that the microenvironment of the intramedullary bone marrow cavity would be more favourable than other tissue sites for FVIIIsecreting BMSC engraftment. Indeed, culture-expanded human BMSCs engrafted and differentiated after intramedullary transplantation in NOD/SCID mice. 8  All methods for primary BMSC culture, immunophenotyping and plasmid constructs for ZFN-mediated integration of a FVIII transgene ( Figure 1A-C) were as previously described. 4,9,10 ZFN-modified autologous BMSCs were implanted in vivo without genetic selection after taking an aliquot for secreted FVIII activity assay. Parallel electroporation of a GFP plasmid was performed to determine the efficiency of gene transfer in each case ( Figure 1D).

| In vivo implantation
Autologous BMSC implantation was performed in sterile surgical conditions. We suspended cells in an equal volume of sterile phosphate-buffered saline or Matrigel TM (final concentration 6 mg/mL; Corning Inc., USA). Intramedullary implantation was performed by injecting cells into the femoral medullary cavity. Intramuscular implantation was performed by injecting one-fifth of the cell suspension into 5 different sites in the middle gluteal muscle. Matrigel TMencapsulated cells were injected into fourteen subcutaneous sites (3 mL cell suspension/site) over the dorsal neck. Intravenous busulfan (6 mg/kg body weight; Otsuka Pharmaceutical Co., Japan) was administered 2 days before infusing cells into the cephalic vein.

| Plasma FVIII and anti-FVIII antibody assays
Citrated venous blood was drawn 3 weeks before cell implantation, immediately before implantation and thereafter at weekly intervals for 4 weeks, followed by monthly intervals. Plasma obtained by centrifugation (15 minutes at 1500 g; room temperature) was stored at À80°C until analysis.
Plasma samples were analysed for human FVIII protein using an antibody with negligible cross-reactivity against canine FVIII (VisuLize TM FVIII antigen ELISA kit; Affinity Biologicals, Canada). All plasma samples were also tested for anti-FVIII antibodies. 11 FVIII activity in conditioned media of ZFN-modified BMSCs was quantified as previously described. 10

| RESULTS AND DISCUSSION
Bone marrow-derived mesenchymal stromal cells (BMSC) immunophenotypes in this study resembled that of human MSCs and previously described canine BMSCs 12,13 ( Figure 1E). Table 1 summarizes characteristics of in vitro BMSC expansion. The theoretical number of BMSCs expandable from all 10 mL of bone marrow aspirate from each dog was 1.5-to 20.2-fold greater than the number of BMSCs used for FVIII transgene integration. These differences in canine BMSC yield and ex vivo expansion potential are similar to inter-donor differences for human BMSC.
FVIII activity was assayed in 24-hour conditioned media of an aliquot of ZFN-treated BMSCs taken on the day of implantation (Table 1). As the numbers of viable implanted BMSCs and secreted FVIII activities were variable among the dogs, a composite therapy index expressing the individualized "dose" of FVIII-secreting cells implanted was calculated as the number of viable cells implanted 9 FVIII secretion in mIU/10 6 cells/24 hours (Table 1). Figure 1F shows the time course of plasma human FVIII protein levels in all dogs. Each served as its own control in calculating the difference between pre-and post-implantation plasma human FVIII LEE ET AL.  4 Broken lines denote the corresponding isotype control. Bone marrow was obtained by aspiration through the trochanteric fossa of the femur. Summary of mesenchymal stromal cell markers of BMSCs of all dogs in our study compared with data of Takemitsu et al of primary bone marrow-derived mesenchymal stem cells of 4 beagle dogs. 13 The histogram subtraction method was used to calculate the mean percentage AE standard deviation of cells positive for each surface protein. 15 All antibodies used for immunophenotyping were reactive against the cognate canine protein except anti-CD105 antibody whose reactivity against canine CD105 was not specified by the manufacturer. F, Plasma levels of human FVIII protein after implantation of autologous ZFN-treated BMSCs by the intramedullary method (IMed), intravenous infusion (IV), intramuscular (IMus) and Matrigel TMencapsulated subcutaneous injection (SC). Summary of plasma human FVIII protein levels by implantation site and composite therapy index of each dog engraftment of BMSCs had occurred in the perivascular niche whose relative hypoxia may favour cellular quiescence and stability. Like clinical cell therapy with hematopoietic stem cells, implanted ZFNtreated BMSCs were not genetically pre-selected for FVIII transgene integration. 14 Nonetheless, successful BMSC engraftment in the intramedullary microenvironment enabled FVIII secretion from the unselected bulk population.
Our data showed no relationship between the composite therapy index and post-implantation plasma transgenic FVIII levels. The composite therapy index for intramuscular implantation (10742) and intravenous infusion (8884) which showed modest or undetectable plasma human FVIII, respectively, were more than twice as high as the composite therapy index for intramedullary implantation (3684) which resulted in plasma samples with consistently highest plasma human FVIII levels for at least 48 weeks. Intravenously infused BMSCs were completely ineffective despite a high composite therapy index, consistent with rapid destruction by complement activation. 5 Autologous cell therapy could be advantageous over factor replacement which is costly and requires frequent venous access.
Cryopreserved BMSCs remain functional and can be used for repeat treatments. However, several improvements are essential for genetically modified autologous BMSC therapy to be clinically feasible.

CONFLI CT OF INTEREST
The authors confirm that there are no conflicts of interest. BMSCs from fresh bone marrow were cultured in a 3:1 (v/v) mixture of low-glucose Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA, USA) as previously described. 4 BMSCs were expanded in culture for 3 weeks after bone marrow aspiration. The number of mononuclear cells in bone marrow aspirates was 3.54 9 10 8 AE 1.21 9 10 8 per mL (mean AE standard deviation; n = 4). The number of adherent cells which grew out initially after 9 days (P0) varied by 4.7-fold among the dogs (3.33 9 10 7 to 15.6 9 10 7 ). BMSC, bone marrow-derived mesenchymal stromal cells; IMed, intramedullary; IV, intravenous; IMus, intramuscular; SC, subcutaneous with Matrigel TM cell encapsulation. | 3701