Closed system bag spinoculation increases transduction efficiency for lentiviral vectors
Early experiments using a process designed for retroviral gene transfer (Additional file 1: Fig. S1A, B) showed that the effectiveness of transduction can be significantly enhanced with bag centrifuge spinoculation compared to static transduction and the results were comparable to plate spinoculation (Additional file 1: Fig. S1C). In addition, we showed that this increase in transduction due to the spinoculation did not come at the cost of lower cell expansion or viability (Additional file 1: Fig. S1D, E). Further, we investigated if there were any detrimental effects of the addition of Dynabeads during the spin process. In this experiment, 1.5 × 107 cells plus 4.5 × 107 Dynabeads (3:1 bead to T-cell ratio) were transferred to a PL30 bag, and the same number of cells without beads was transferred to a second bag as a control. A single 1000×g centrifugation was performed using cells with and without anti-CD3/CD28 beads. Cells were centrifuged for 2 h and cultured for 7 days. Cell number and viability were compared at post-centrifugation, and again at culture days 4 and 7. As shown in Additional file 1: Fig. S1F and G, cell number and viability of paired samples were nearly identical throughout the culture process, demonstrating that spinoculation with CD3/28 Dynabeads exhibited no detrimental effects on cell viability or expansion.
Automated closed system sepax is an alternative to bag spinoculation
Although bag spinoculation increases transduction efficiency, it suffers from the limitation of the maximum cell number for each bag, thus multiple bags and multiple rounds of centrifugations are required for high dose clinical products. This not only requires more total and hands-on time but increases the chance of microbial contamination. Therefore, we decided to test spinoculation of T-cells in the Sepax-C Pro to address these issues and streamline the process. To optimize the Sepax procedure, several durations of Sepax spinoculation (2-h, 1-h, 0.5-h) were compared to bag spinoculation using a 2-h bag centrifuge time. For these studies, we utilized a lentiviral vector encoding CD19/CD22 CAR at a MOI of 10. Transduction efficiencies from Sepax prepared cells were not significantly different from the 2-h bag spinoculation for CD19/22 CAR T-cells transduction efficiency, with the Sepax spinoculation 1-h results being significantly better than the static transduction (83.4% vs. 35.7%) (Fig. 2A, B). The fold expansion throughout the culture period was comparable for all transduction methods and was similar to the untransduced control cells (Fig. 2C). Donor 1 expanded more than the other two donors for all of the groups tested. There was no significant difference between the vector copy number between all three timepoints of the Sepax spinoculation compared to the 2-h bag spinoculation cells (data not shown). The vector copy number results were similar to the transduction efficiency results, with the 1-h Sepax spinoculation CAR T-cell transduction efficiency being significantly higher than that of static transduced cells (Fig. 2D).
The viability of the CAR T-cells was consistent among all groups throughout the culture period and by day 9 the viability was close to 100% (Fig. 2E). The 1-h Sepax spinoculation condition was chosen for the remaining the experiments due to having comparable results to the 2-h Sepax and bag spinoculation conditions and needing less total time for transduction compared to the 2-h Sepax spinoculation.
Sepax spinoculation is suitable for patient samples
The 1-h Sepax spinoculation condition was tested against the 2-h bag spinoculation and traditional static transduced conditions for a patient sample with the CD19/22 Bispecific CAR vector. The CD4/CD8 cells were prepared and transduced in the same manner as the previous optimization experiments. When testing the patient’s cells, the 1-h Sepax spinoculation condition was comparable to the 2-h bag spinoculation condition (Sepax: 50.4%, Centrifuge: 56.3%) (Fig. 3A, B) and both had higher transduction efficiencies than static transduction. Fold expansion was comparable between all groups throughout the culture period (Fig. 3C). The patient cells expanded more than the healthy donor cells in the previous experiments. The vector copy number results for the patient sample reflected the transduction efficiencies and the vector copy number results were within the FDA limit of < 5 copies/transduced cell (Fig. 3D). Cell viability was comparable between all groups throughout the culture period (Fig. 3E). For the patient cells the CAR T-cell CD4/8 ratio was consistent between the Sepax spinoculation, bag spinoculation, and static transduction cells and the untransduced cells (Fig. 3F). These results show that it is possible to use the Sepax spinoculation method for transduction of patient samples with acute lymphoblastic leukemia (ALL), which are often described as being more difficult to transduce and expand.
Sepax spinoculation is suitable for other lentiviral vectors
Two additional CAR lentiviral vectors, FGFR4 and CD22 were also tested for transduction using the Sepax spinoculation method to ensure results observed were not dependant on the CAR construct used. For all experiments using the CD19/22- and FGFR4-CAR vectors, a MOI of 10 was used. However, when using the CD22-CAR vector, a MOI of 2 was used. The same healthy donor cells were used for the studies with the CD19/22 Bispecific and FGFR4-CAR vectors and cells from a second healthy donor was used for the CD22 CAR vector studies. The transduction efficiency for the FGFR4-, CD22-, and CD19/22-CAR vectors were comparable with no significant differences among the bag and the Sepax spinoculation groups (Fig. 4A, B). Fold expansion was consistent among all conditions for each CAR (Fig. 4C). The vector copy number correlated with the transduction efficiency for each group, and there was no significant difference between the bag and Sepax spinoculation transduction methods (Fig. 4D). The viability was consistent throughout the culture and by day 9 the viability was close to 100% among all CAR T-cells produced using Sepax, bag, and static transduction methods and the untransduced controls (Fig. 4E). The CD4/8 ratio was compared for these experiments to see if there were any changes between conditions. The ratio was consistent among the transduction groups across each CAR vector (Fig. 4F). These results show that it is possible to use the Sepax spinoculation method for transduction of multiple CAR lentiviral vectors, and the transduction enhancement is not vector specific.
Functional characteristics of CAR T-cells are not affected by spinoculation
In order to confirm that the Sepax spinoculation did not affect the functional properties of the CAR T-cells, a killing assay was set up using NALM6 tumor cells, which expresses both CD19 and CD22 targets along with GFP that was used to quantitate the cells. CAR T-cells were cultured with the tumor cells in either a 1:1 or a 5:1 ratio for 24 h and 48 h. CD19/22 CAR T-cells derived from the patient and the healthy donor cells, along with the CD22 CAR T-cells from the additional healthy donor were used. In the CD19/22 experiments, the tumor cell counts per well decreased by approximately 75% for the static transduced CAR T-cells when tested at 5:1 while this cytotoxic effect was even greater for the CAR T-cells produced by a bag and Sepax spinoculation tested at 5:1. By 48 h there were almost no tumor cells remaining for CAR T-cells produced by bag and Sepax spinoculation tested at 5:1 (Fig. 5A, B Additional file 1: Fig. S2A, B). For the CD22 CAR T-cells, cytotoxicity between groups appeared to be more consistent. For CD22 CAR T-cells produced with each of the 3 transduced methods, target cell counts decreased by more than half by 24 h (Fig. 5C).
ELISA’s were used to detect IFNγ and TNFα from the supernatant from each of the above experiments. The CD19/22 CAR T-cells produced from healthy donor cells using each of the 3-transduction method released large quantities of IFNγ (~ 3000 pg/mL) and IFNγ released by the CD19/22 CAR T-cells produced using static transduction, bag spinoculation, and Sepax spinoculation was comparable (Fig. 5D). It is interesting to point out that the CAR T-cells produced using static transduction were observed to have less killing at the 24-h time point despite having approximately the same level of IFNγ production. The levels of TNFα released by each of these CAR T-cells were lower than the IFNγ levels. The quantities of TNFα released by CD19/22 CAR T-cells produced by bag and Sepax spinoculation were comparable and higher than the quantities released by CAR T-cells produced by static transduction at both 24 and 48 h (~ 400 pg/mL) (Additional file 1: Fig. S2C).
The cytokine release by CD19/22 CAR T-cells produced from patient cells was similar to that released by healthy donor CD19/22 CAR T-cells. For IFNγ production, the CAR T-cells manufactured by bag and Sepax spinoculation were comparable at both 24 and 48 h (~ 3500 pg/mL). However, the CAR T-cells produced by static transduction had a significant decrease in cytokine levels which correlated to the lack of killing of tumor cells (Fig. 5E). The absolute values of TNFα were again lower than IFNγ, as expected. The TNFα produced by CAR T-cells manufactured using bag spinoculation was higher than that produced Sepax spinoculation CAR T-cells at 24 h, but was comparable at 48 h. The TNFα levels dropped at 48 h for the bag and Sepax spinoculation CAR T-cells to ~ 200 pg/mL. We observed a statistically significant decrease in cytokine production by CD19/22 CAR T-cells manufactured by static transduction at both 24 h and 48 h compared to the Sepax and bag spinoculation CAR T-cells (Additional file 1: Fig. S2D).
IFNγ production by CD22 CAR T-cells was approximately equivalent for cells produced by static and the Sepax spinoculation transduction. Both cell types produced approximately 10,000 pg/mL at 24 h. All 3 types of CD22 CAR T-cells experienced a decline in IFNγ production at 48 h. At 24 h, the bag spinoculation group produced significantly less IFNγ than the two other groups, but the levels declined for all the groups at 48-h, making the differences non-significant (Fig. 5F). The TNFα levels were not different between the static transduction, bag spinoculation, and Sepax spinoculation CD22 CAR T-cells at 24 h. The TNFα levels appeared to rise in the static transduction CAR T-cells at the 48-h time point, whereas they dropped in the bag and Sepax spinoculation CAR T-cells at the same time point (avg, 1045 pg/mL, 558 pg/mL, and 343 pg/mL) (Additional file 1: Fig. S2E). We conclude that using the Sepax instrument for transduction did not affect the killing properties of the CAR T-cells, but may influence cytokine secretion at early time points after re-activation, and that the process of spinoculation in general may be beneficial for some patient samples, potentially by promoting activation and enhancing cytokine secretion.
Gene expression is not affected by spinoculation
CAR T-cell samples from two healthy donors cryopreserved at the end of the culture were thawed and rested overnight in complete medium in a 37 °C tissue culture incubator. Some cells were used to represent the clinical CAR T-cells that would be taken out of culture and prepared for patient injection (depicted as final). Another set of the cultured CAR T-cells were incubated for 24 h after an initial resting period with CD3/CD28 Dynabeads to simulate what would occur during T-cell re-activation following the infusion of CAR T-cells that had not be cryopreserved. When subjected to PCA analysis the final CD19/22 CAR T-cells clustered together nicely by donor and method, with the biggest difference being the static culture from healthy donor 2 (Fig. 6A). The re-stimulated cells showed even less variability than the final cells and formed very tight clusters (Fig. 6A). These results indicate that there are no large differences in gene expression that is a result of the transduction method.
PCA analysis of the FGFR4 CAR T-cells displayed a difference in final cultured cells, again with the bag and Sepax spinoculation group clustering separate from the static transduced CAR T-cells (Fig. 6B). Similar to the CD19/22 CAR, upon re-stimulation, there was much less variability between these groups (Fig. 6B).
The results of unsupervised hierarchical clustering analysis of the CD19/22 CAR T-cells and the FGFR4 CAR T-cells (Fig. 6C) were similar to the PCA analysis. Cells that were re-stimulated clustered separately from those that were not. When examining specific pathways important for T-cells, there was no observable difference between the CAR T-cells produced using the three different methods (Bag, Sepax, and Static) (Fig. 6D, E).
Discussion
The work described highlights the use of the automated closed system gene transfer system for clinical scale manufacturing of various CAR T-cells. Conventional methods typically use open 6-well plates or culture bags to perform spinoculation for the manufacture of clinical cell therapies. We tested an automated Sepax-II Pro instrument for spinoculation gene transfer of peripheral blood lymphocytes. The closed-system Sepax spinoculation described here reduces the risk of microbial contamination of cell preparations and reduces the risk of viral vector spread within the cell therapy facility clean room. Total time of the transduction process and technician hands-on time is also decreased due to having the ability to perform one large scale run compared to having to implement multiple centrifuge cycles for a patient. Sepax spinoculation more readily lends itself to clinical large-scale transduction and automation than the static or bag/plate centrifuge transduction, and is effective for both retroviral and lentiviral vectors. This new method was able to generate a large quantity of functional CAR T-cells that were capable of recognizing and killing tumor cells in culture. The closed system method is reproducible for various CAR lentiviral vectors and is shown here to be applicable to patient samples with little variation from healthy donor samples.
Besides the Sepax, there are other commercially available devices that are emerging that may be able to enhance transduction efficiency using a spinoculation technique. The Miltenyi Prodigy instrument, for example, is another closed system transduction system that is currently beta testing a process for the same purpose. The spin duration, speed (up to 400×g), and temperature (+ 4 to + 38 °C) can be programed by the technician, which may be useful for development purposes. The Sepax is similar in that each step in the transduction protocol is modifiable, and therefore, each step can be fine-tuned to ensure optimal transduction occurs. Sepax spinoculation also allows for larger volumes, spin times, and spin speeds from 500 to 1200×g. One caveat to the Sepax instrument is that it currently only operates at room temperature. Although, this appears to be sufficient to enhance transduction efficiency, literature in the field would suggest increasing the temperature to 32 °C may result in even more dramatic effects [24, 25].
The study described here used a single 1-h Sepax program, transducing 60 × 106 T-cells, but this number can be doubled to 120 × 106 T-cells for larger patient transductions. The transduction process was designed to be incorporated at clinical cell processing facilities and utilized standard equipment that are commonly available at these sites. First, CD4 + and CD8 + T-cells were selected by the automated CliniMACS (Miltenyi). Once the T-cells were activated with CD3 + /CD28 + Dynabeads (Invitrogen, Camarillo, Ca), they were transduced using either the Sepax spinoculation, bag spinoculation, or remained static during the transduction. Cells were fed and expanded over a period of 7 days after transduction in culture bags. Once the cells are harvested they are washed with plasmaLyte-A containing HSA and resuspended at a final concentration on 10 × 106 cells/mL. These cells were then assayed for vector copy number, flow cytometry, and frozen down for future functional assays.
We confirmed that this process results in very consistent products, capable of recognizing and killing tumor cells. We did however experience some differences in cell growth and transduction efficiency based on different donors and different CAR lentiviral vectors. The patient samples expanded better than the healthy donors, but some healthy donors had better transduction efficiency than the patient. Variability between patient samples, and even between different healthy donor samples is commonly observed. Therefore, we tried to keep our focus on differences between the three groups (static, bag centrifuge, Sepax) within a particular healthy donor or patient sample.
When comparing multiple Sepax spin times against the 2-h bag centrifuge and static transduction, we found that we were able to use a Sepax spin time of 1-h that was comparable to the two-hour centrifuge approach, eliminating at least an entire hour from the total processing time (Fig. 2). If only one centrifuge is available within a tissue culture room and > 4 spins are required, this could potentially result in saving several hours rather than only one. The 1-h Sepax spinoculation was then tested against the bag centrifuge and static transduction using a sample from a patient with B-cell-ALL. Transduction efficiency was similar between the Sepax and bag centrifuge spinoculation groups (Fig. 3A, B), and both of these were significantly higher than static transduction (~ 10% TE static vs. ~ 55% for centrifuge and Sepax). This highlights the power of using the spinoculation approach for patient samples that tend to be more difficult to transduce. The 1-h Sepax spinoculation was also tested for the FGFR4 CAR (Lentigen Technology, Gaithersburg, MD) and CD22 CAR vector (Lentigen Technology, Gaithersburg, MD). The transduction efficiencies were comparable among the different CAR vectors for the Sepax spinoculation groups (Fig. 4A, B). It is important to note that the MOI used for the CD22 CAR vector was reduced to MOI = 2 for the studies presented here. Initial experiments using a MOI = 10 resulted in very little difference between static and spin approaches (data not shown). We speculate that the vector was saturating under these circumstances and required dilution to observe effects on transduction. The obvious positive attribute is that we are able to obtain high transduction with the Sepax spinoculation approach when using significantly less vector, resulting in a substantial cost-savings since GMP vector production tends to be very expensive.
Functional tumor cell killing assays and cytokine release assays showed that the Sepax spinoculation transduction method did not affect the tumor recognition and killing functions of the CAR T-cells (Fig. 5). The main differences between the CAR T-cells produced with the different CAR lentiviral vectors were due to differences in vector constructs, rather than differences in the cell manufacturing process. Nanostring gene expression comparison also helped conclude that the Sepax spinoculation had very little effect on the functional properties of CAR T-cells, and major pathways utilized in the T-cells ability to kill tumor cells appeared unaffected (Fig. 6).
Concurrent with this study, optimization of the Sepax spinoculation method for gamma-retroviruses is in the process of being developed. Testing to determine if the Sepax spin chamber can be coated with material commonly used to enhance transduction is underway.