Continued success and advancements are observed in treating hematologic malignancies as well as solid tumors with promising approaches such as chimeric antigen receptor (CAR) T cell and T cell receptor (TCR) therapies. Due to the high number of clinical trials emerging into the cell therapy and gene therapy markets every year, more automation is needed. When assessing automated technologies to support such complex manufacturing workflows, a critical first step is to investigate instruments that streamline the production process. The Sefia Select™ system is a functionally closed system developed for automated upstream and downstream cell processing steps, through dedicated application software. S-Wash and ReadySelect are two purpose-built applications for the Sefia Select™ system to perform cell washing and formulation. They can be used to automate multiple time-sensitive CAR T process steps.
Find out more about Sefia Select™ system here.
Introduction
In 2022 alone, an estimated 1.9 million new cases of cancer were diagnosed in the US, and about 600 000 Americans are expected to die of cancer (1). Although chemotherapy combined with drugs continues to be the standard of care, an increasing number of adults and children are refractory to conventional treatment modalities, motivating the “next generation” of approaches for cancer therapy. More and more autologous CAR T cell therapies are being developed to answer this need and have received approvals due to high patient response rates. Examples of approved CAR T products include Kymriah from Novartis, Yescarta and Tecartus from Kite Pharma, and Breyanzi from Juno Therapeutics (2).
To support the manufacturing of these new therapies at a commercial scale, solutions need to be implemented to secure the quality of these products and to guarantee patient safety. The use of a closed and automated system to process cells at different steps during cell therapy manufacturing is crucial to meet critical quality attributes (CQA). Several early phase clinical trial CAR T products (3,4) on the market are manufactured using manual processing, which is labor intensive, difficult to scale, and therefore prone to high risk of contamination and production failure due to open manipulations.
Reproducible manufacturing of high-quality, clinical-grade CAR T products is an essential prerequisite when designing your workflow. The implementation of a closed and automated system within every step of the workflow is required to standardize and secure your cell therapy manufacturing process. Automation also helps you to reduce manufacturing processing time and finally cost per dose of these therapies. A general workflow for manufacturing is shown in Figure 1, which also shows where the Sefia Select™ system can be used in in your cell processing workflow.
Fig 1. An example of an autologous (CAR)-T workflow.
A typical autologous CAR-T cell processing workflow begins with the collection of leukapheresis units (Fig 1). The units can be shipped to the manufacturing site fresh or frozen. Frozen leukapheresis units are thawed, diluted, and washed before moving to the next step. The cell isolation step can correspond to a peripheral blood mononuclear cells (PBMC) enrichment or to a T cell selection via magnetic beads. After cell isolation, selective activation, viral-based transduction, and large-scale bioreactor expansion are performed to achieve the target cell count. The cells are harvested to be concentrated, washed, and diluted with different solutions including cryoprotectant and finally split into several doses. CAR T doses are cryopreserved to be shipped and injected to the patient.
In this study, we present the data collected using healthy donors in a typical CAR T cell processing workflow without a transduction step from a cross-site validation study. We mainly focus on the capabilities and performances of the S-Wash and ReadySelect applications:
- S-Wash is a generic cell washing application mainly used at the cell harvest step to concentrate, wash, and resuspend cells. In the case of thawed cellular product, an initial dilution can be performed before the concentration step.
- ReadySelect is a versatile application used for product formulation. In the workflow described in Figure 2 the ReadySelect application can be used for the final formulation of the product including the cryopreparation and the splitting in up to four cryobags.
- Mononuclear cell collection protocol (MNC): dual-stage separation of cells based on specific gravity and cell size (volume range of 120 mL ± 40 mL).
- Continuous mononuclear cell collection protocol (CMNC): continuous separation of cells (single stage, no cycles), based on specific gravity alone (volume range of 150 to 220 mL).
- A 1 g = 1 mL conversion was applied when the solutions were aqueous.
- A 1.07 g = 1 mL conversion was applied when using 10% cryoprotectant solution or 20% human serum albumin (HSA) solution
Design of the experiment
Peripheral blood mononuclear cells (PBMC) were initially enriched from a fresh healthy donor leukapheresis using the PremierCell application (Fig 2). After their enrichment, PBMCs were split into two bags to cover two identical workflows and generate technical replicates. PBMCs were then cultured for 4 d in the presence of a T cell activator and then transferred into a Xuri™ cell expansion system W25 for an additional 4 d. On day 8, cells were harvested. Harvested cells were concentrated and washed using S-Wash application. On the same day, the final product from the S-Wash procedure was used as initial material for the ReadySelect application to prepare four cryogenic bags for the final formulation. There were two replicates for each of the four distinct leukapheresis donors giving a total of total of eight procedures performed.
Fig 2. Schematic representation of the workflow used.
Initial product collection
Leukapheresis was performed on four healthy voluntary donors. An agreement with the supplier mentioned that both parties agreed on the use of these products for “exclusive purposes of scientific research and/or laboratory evaluations”.
Leukapheresis collection was carried out in an authorized collection center in the United States (STEMCELL Technologies). The center performs the collection of autologous leukocytes on healthy donors using acid-citrate dextrose with solution A (ACD-A) as an anticoagulant. The volume of ACD-A anticoagulant in the final bag represented ~ 10% of the leukapheresis total volume.
The leukapheresis units were collected using Spectra Optia apheresis system (Terumo BCT) and Spectra Optia collection sets. The collection bags are made from gas-permeable PVC plasticized material with butyryl trihexyl citrate (BTHC). Collections were performed following one of the following protocols:
The collected leukapheresis units were maintained at 4°C after collection and during the shipment until delivery using refrigerant gel packs. The units were processed within 48 h after collection.
PBMC enrichment
The workflow described above required enriched PBMCs as initial material. For PBMC enrichment, the PremierCell application was used in combination with the CT-300.1 kit and Sefia™ S-2000 cell processing instrument. PremierCell is a multistep application that first depletes platelets via centrifugation steps and then removes red blood cells (RBC) using a density gradient medium (DGM)-based centrifugation method (Fig 3). The final bag contains a product enriched in PBMCs.
The list of parameters used for PremierCell application can be found in Appendix A-1
Fig 3. PremierCell application workflow.
Cell activation and expansion
After enrichment of PBMCs, a fraction of cells (~ 260 × 106 PBMCs) was removed from the PremierCell application final bag and seeded in T-flasks at a density of 1.2 × 106 cells/mL. Human CD3/CD28/CD2 T cell activator (ImmunoCult, STEMCELL Technologies) was added to the cells according to the manufacturer’s protocol. Cells were then maintained in an incubator (37°C, 5% CO2) for 4 d before being transferred into a 2 L Cellbag™ bioreactor container for expansion in Xuri™ cell expansion system W25.
On day 4, the 2 L Cellbag™ bioreactor container was placed on Xuri™ cell expansion system W25, gassed with a mixture of air and 5% CO2, and a reservoir containing cell culture medium was aseptically welded on. An initial 200 mL of cell culture medium was added to the bioreactor and allowed to equilibrate. The system parameters were set at 37°C, gas flow rate of 0.05 L/min, and a rocking speed of 10 rpm at a 6° angle. Activated T cells were inoculated into the bioreactor and cell culture medium was added to reach a total working volume of 500 mL in the bioreactor. Once initiated, the Cellbag™ bioreactor containers were left overnight while fresh medium was added into the bioreactor at 1 L/d to reach a total volume of 1 L. From day 5 onwards, 1 L of medium/d was continuously perfused to control lactate, ammonium, and glucose levels. Daily sampling was performed to monitor cell concentration, cell viability, pH, and levels of lactate, ammonium, and glucose. Cells were maintained in culture until day 8.
On day 8, the S-Wash application was used in combination with the CT-200.1 kit on the Sefia™ S-2000 cell processing instrument. The S-Wash application was performed downstream of T cell expansion to concentrate and wash the collected expanded T cells (Fig 4A). A schematic representation of the CT-200.1 kit is shown in Figure 4B.
Fig 4. (A) S-Wash application workflow. (B). Design of the single-use CT-200.1 kit. (C) List of parameters.
ReadySelect application
The ReadySelect application was used in combination with the CT-350.1 kit on the Sefia™ S-2000 instrument combined with the Sefia SelectTM module. On day 8, the ReadySelect application was used as the final step of the workflow, to split washed expanded T cells into several doses (up to four cryobags) and prepare them for cryopreservation (Fig 5A). A schematic representation of the CT-350.1 kit is shown in Figure 5B.
Fig 5. (A). ReadySelect application workflow. (B). Design of the single-use CT-350.1 kit. (C) List of parameters.
Methods used to assess performance
Cell samples were taken to assess performance obtained at each step of the workflow. Table 1 describes all criteria evaluated during the process. Each method used has been described in the Appendix section.
Table 1. List of criteria evaluated, and description of test executed to evaluate these criteria at each step of the workflow
| Criteria to evaluate |
Test execution description |
| Volume accuracy error in final bags |
All volume measurements were performed using a calibrated and tared scale. |
| Cell viability |
Samples were withdrawn from initial bag and all final bags. Cell viability was assessed in triplicates using a NucleoCounter® NC-200 automated cell counter (Chemometec). The triplicate average and standard deviation were reported for each sample (Appendix 2). |
| Cell recovery |
Samples were withdrawn from the initial bag and all final bags. Cell counts were performed in triplicates using the cell counter and cell recovery was calculated by dividing the total viable cells (TVC) obtained in the final bag by the TVC measured from the initial bag. The average and standard deviation were reported for each sample (Appendix 2). |
| Washout efficiency |
Samples were withdrawn from the initial bag and all final bags. We performed ELISA on both samples to assess the washout efficiency of a specific solute (applicable only for S-Wash procedures) (Appendix 4). |
| Cell repartition in cryobags |
Appropriately sized samples were withdrawn from all cryobags. Cell counts were performed in triplicates using the cell counter. As all cryobags volumes targeted are identical for the scenario tested in this document; the deviations in cell count between cryobags were calculated by dividing the cell count obtained in the two cryobags to be compared (applicable only for ReadySelect procedures). |
| Cell subpopulation repartition |
Samples were withdrawn from the initial bag and all final bags. We performed flow cytometry to identify cell population distribution at different time points of the experiment (Appendix 3). |
Results and discussion
Donor variability
A total of four donors were used to generate the performances using ReadySelect and S-Wash applications. A sample was withdrawn from the initial product and cell analysis was performed internally using methods described in Table 1 above.
Table 2 below shows the variability between donors in terms of white blood cell (WBC) content as well as T cell proportion in each leukapheresis.
Table 2. Table summarizing WBC content range and T cell repartition within leukapheresis from a pool of donors
|
Fresh healthy donor leukapheresis (US donor center) n = 4 |
|
| Min. WBC content | 5.00 × 109 |
| Max. WBC content | 6.86 × 109 |
| Average volume | 112 ± 34 mL |
| Min. T cell | 20.1% |
| Max. T cell | 65.6% |
| Average cell viability | 99.50 ± 0.1% |
Concentration and washing of cellular products
S-Wash is an application designed for concentrating and washing cellular products while maintaining a high cell recovery.
To evaluate the washout efficiency of the S-Wash application, we performed ELISA on n = 12 procedures to quantitate the concentration of a specific solute in the initial vs final bags. We used bovine serum albumin (BSA) as solute, which was manually added into the initial bag containing only water before launching the S-Wash application.
Over 12 procedures, we obtained an average 4.3 ± 0.3 log reduction of BSA between initial and final products.
- A 3-log reduction means 99.9% reduction of the solute between the initial and the final bag.
- A 4-log reduction means 99.99% reduction of the solute between the initial and the final bag.
Details about the theoretical washout efficiency are summarized in Appendix 5.
The average total cell recovery obtained when using the S-Wash application for downstream processes including a volume reduction of the initial product and a washout of the resuspension media for n = 8 procedures (two replicates for each of the four leukapheresis units) was equal to 97.8 ± 6.2%.
An assessment of the viability was done by comparing TVC viability in the final bag vs the initial bag for all procedures (n = 8). The average viability drop was equal to -0.2 ± 1.1% when starting with collected expanded T cells as the initial product.
The last key feature of the S-Wash application is to provide an accurate final bag volume. Over eight procedures (Table 3), we found that the average volume accuracy error between the measured and the expected final volumes was equal to 0.6 ± 2.3 mL.
Table 3. Volume measured in all final bags post S-Wash procedures
| Exp. no. ID | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
| Target volume (mL) | 35 | 35 | 35 | 35 | 35 | 35 | 35 | 35 |
| Obtained volume (mL) | 33.4 | 39.4 | 34 | 38.2 | 32.6 | 36.2 | 34.8 | 35.8 |
Evaluating ReadySelect dosing of cellular products
ReadySelect is a modular application providing solutions to prepare accurate dose(s) of products that have been optionally mixed with cryoprotectant upstream.
To provide accurate doses of cellular product, the application must generate very precise volumes and guarantee efficient mixing of the cellular product to obtain an accurate cell repartition between all final bags.
The average volume accuracy error between the measured and the expected volumes in all cryobags (n = 32, from four cryobags across eight procedures) containing cells resuspended in cryoprotectant corresponded to 1.4 ± 1.8%.
The volume accuracy error in cryobag number 4 (last cryobag prepared) was slightly below the others due to the accumulation of errors obtained from the previous final bag preparations (Fig 6).
Fig 6. Final cryobag volume accuracy error using ReadySelect application.
To obtain a precise cell splitting between final cryobags, the efficiency of the mixing, as well as the split of the product in the final bags, must be performed with high accuracy. Figure 7 represents the average cell repartition error obtained in each of the final cryobags. The average cell repartition accuracy in all final cryobags (n = 32) was 93.5 ± 4.6%.
Fig 7. Cell repartition accuracy in final cryobags (n = 32).
To understand the degree of cell loss during the ReadySelect step, we assessed cell recovery across n = 8 runs. We found that the ReadySelect application allowed recovery on average for more than 97.4 ± 5.4% of the initial cellular product.
Finally, the cell viability drop in final bags was on average 1.8 ± 1.3% (n = 32). We hypothesized that the viability drop in cryobags could be increased because of the presence of cryoprotectant in the cryobag.
The summary of the performance we obtained for S-Wash and ReadySelect applications is shown in Table 4.
Table 4. Table summarizing the performance obtained from S-Wash and ReadySelect applications
| S-Wash application | ReadySelect application | |
| Volume accuracy error in final bags (between expected and measured final bag volumes) | (n = 8) = 0.6 ± 2.3 mL | (n= 32) = 1.4 ± 1.8% |
| Viability drops in final bags (%) (between initial and final products) | (n = 8) = -0.2 ± 1.1% | (n = 32) = 1.8 ± 1.3 |
| Total cell recoveries (%) | (n = 8) = 97.8 ± 6.2% | (n = 32) = 97.4 ± 5.4% |
|
Cell repartition accuracy (%) (between expected and obtained total cell counts in final bags) |
NA | (n = 32) = 93.5 ± 4.6% |
| Washout efficiency (log) | 4.3 ± 0.3 (n = 12) | NA |
Conclusion
The study described in this document shows that ReadySelect and S-Wash applications deliver consistent and repeatable results in terms of cell recovery, cell viability, desired volume accuracy of final products, and efficiency of product dosing in the case of ReadySelect.
The added value of both applications is their flexibility/versatility and adaptability to the user needs. Both applications can handle multiple initial products and support multiple steps of the CAR T workflow.
In the current study (Fig 3), we used both applications for downstream steps only:
- S-Wash was used for the wash and concentration of expanded T cells.
- ReadySelect was used for the final formulation (dosing and cryopreparation) of expanded T cells.
These two applications could also be used in some upstream steps of the CAR T workflow as described in Figure 8.
Additional uses include:
- For ReadySelect: the dosing and/or cryopreparation of extra enriched cells (i.e., PBMC – T cells). In case the enriched cells from the patient leukapheresis is higher than the minimum required to generate CAR T cell doses; there is the possibility to dose the quantity required and cryoprepare the extra enriched cells for a backup therapy in case the first one does not work the first time.
- For S-Wash: the washout of cryoprotectant from thawed product. This can be applicable to wash out cryoprotectant from thawed patient leukapheresis or thawed enriched extra cells (i.e., PBMC – T cells).
The summary of the performance we obtained for S-Wash and ReadySelect applications in these upstream steps is shown in Table 5.
Table 5. Table summarizing the performance obtained from S-Wash and ReadySelect applications in upstream steps
| S-Wash application | ReadySelect application | |
| Volume accuracy error in final bags (between expected and measured final bag volumes) | (n = 10) = 1.3 ± 2.3 mL |
(n= 16)
|
| Viability drops in final bags (%) (between initial and final products) | (n = 10) = 1.7 ± 1.0% |
(n= 16)
|
| Total cell recoveries (%) | (n = 10) = 86.6 ± 7.2% | (n = 16) = 92.4 ± 3.7% |
|
Cell repartition accuracy (%) (between expected and obtained total cell counts in final bags) |
NA |
(n= 16)
|
Moreover, the large set of parameters available gives a lot of flexibility to the user to use these two applications for other needs:
- S-Wash:
- The washout of any solute at any step of the process (i.e., IgG from magnetic beads; HSA from isolation buffer).
- The concentration of any large volume up to 1.2 L.
- The dilution of any thawed material (i.e., leukapheresis, T cells, PBMC, CAR T cells, and cord blood).
- ReadySelect
- A precise cell dosing at any step of the autologous CAR T process.
- The cryopreparation of any cell content in up to 4 cryobags for a maximum total volume equal to 280 mL.
- The cryopreparation of a small volume of cells (20 mL) dedicated for future quality control sampling.
Fig 8. How ReadySelect and S-Wash applications can be used in a CAR T process workflow.
These two applications have been developed to accommodate a maximum of user needs and give a lot of flexibility to cover many different steps of the CAR T cell workflow. Except for scenarios tested and shared in this application note, any other use of these two applications must be tested and characterized by you as they have not been validated by Cytiva.
CY36272
Table 6. List of parameters used for PremierCell process
| Parameters | Value |
| Dilution volume |
0 |
| Dilution temperature | N/A |
| Enable dilution temperature | Disable |
| Dilution mix time | N/A |
| Dilution rate | N/A |
| Process temperature | N/A |
| Enable process temperature | Disable |
| Sampling screens | Disable |
| Washing cycles—concentration |
2 |
| g-Force—concentration | 85 |
| Sedimentation time - concentration | 420 |
| Intermediate volume - concentration | 20 |
| Enable density gradient separation | Enable |
| High purity | Enable |
| Separation medium retention volume |
45 |
| Hematocrit | Determined by the cell analyzer |
| Washing cycles—separation | 2 |
| g-force—separation | 200 |
| Sedimentation time—separation | 210 |
| Intermediate volume—separation | 20 |
| Final volume 1 | 50 |
| Final volume 2 | 50 |
| Final product temperature | Disable |
| Enable final product temperature | N/A |
| Final product conditioning time | N/A |
Cell concentration and viability were measured using the NucleoCounter® NC-200® automated cell counter (Chemometec). For each sample, the exact volume of the initial product was measured using a calibrated scale (1g = 1 mL conversion was applied when the solutions were aqueous; 1.07g = 1 mL conversion was applied when using 10% cryoprotectant solution or 20% HAS solution) and at least 3 dilutions were prepared using a diluent that was matched to the media of the initial product. The dilution factor was established to obtain a cell concentration value within the linear range of the NC-200 (5 × 104 to 5 × 106 cells/mL). If the initial product contained a cryoprotectant agent, the dilution factor was not lower than 5.
Each diluted sample was carefully mixed by pipetting and then loaded into a Via-Cassette (Chemometec). The cassette was then inserted into the NC-200 for analysis. For each sample, the dilution factor was set up in the parameters in the NC-200 window to obtain the correct cell concentration post-sample processing.
In case the number of aggregates was higher than 2%, the cell concentration measurement was discarded and repeated using a new sample properly resuspended by pipetting to avoid any aggregates in the analysis of the sample.
Once all triplicates were analyzed, the average cell concentration and cell viability were calculated. The total cell number was then calculated by multiplying the average cell concentration by the volume of the product.
Briefly, 5 × 105 cells were washed with 1 mL of BD stain buffer containing fetal bovine serum (FBS) and centrifuged at 400 × g for 5 min. Cells were resuspended with 50 µL blocking buffer per test, containing 10 µL human FcR blocking reagent (Miltenyi Biotec) in BD stain buffer, and incubated for 10 min at 4°C. Cells were then stained for 20 min at 4°C with the following antibody cocktail: CD45-APC-Cy7 (2D1, BD), CD3-PE (UCHT1, BD), CD56-BV421 (NCAM16.2, BD), CD14-FITC (M5E2, BD), and CD19-APC (555415). Cells were washed with 1 mL of BD stain buffer, centrifuged at 400 × g for 5 min, and resuspended in 500 µL BD stain buffer. The stained cells were analyzed on the CytoFLEX flow cytometer (Beckman Coulter). Beckman Coulter QC beads were used to assess instrument performance prior to use. A minimum of 30 000 single-cell events were acquired. Within the PBMC (CD45+) population, monocytes (CD14+ cells), NK cells (CD3-CD56+ cells), T cells (CD3+CD56-), NKT cells (CD3+CD56+), and B cells (CD19+) were identified.
T cell populations were analyzed by flow cytometry on days 0, 4, and 8 of culture, as well as after S-Wash procedure on day 8. Briefly, 5 × 105 cells were washed with commercially available FACS Buffer (BD) and centrifuged at 400 × g for 5 min. Cells were stained for 20 min at 4°C with the following antibody cocktail: CD3-V450 (UCHT1, BD), CD4-PerCP-Cy5.5 (SK3, BD), CD8-APC-Cy7 (RPA-T8, BD), CD45RO-PE-Cy7 (UCHL1, BD), CD62L-PE (DREG-56, BD), CD279-APC (MIH4, BD) and CD25-AF700 (M-A251, Biolegend). Cells were washed with 1 mL of FACS buffer, centrifuged at 400 × g for 5 min, and resuspended in 500 µL of FACS buffer. The stained cells were analyzed on the CytoFLEX flow cytometer (Beckman Coulter). Beckman Coulter QC beads were used to assess instrument performance prior to use. A minimum of 10 000 single cell events were acquired. Subsets of the T cell population (CD3+) were identified by CD4 and CD8 expression. Activation and exhaustion were assessed by CD25+ and CD279+ expression, respectively. Memory phenotype was assessed based on CD62L and CD45RO expression.
All data were analyzed using FlowJo v10.7.1 software.
To evaluate the washout efficiency of S-Wash application, 12 additional procedures were performed. In this case, we used a 0.9% NaCl solution containing a fixed amount of bovine serum albumin (BSA) as initial product and as washing solution. The same parameters used for S-Wash application and summarized in Figure 4C were used for these procedures.
Samples were taken from the initial and final products of S-Wash procedures and BSA content was measured by ELISA using a commercially available kit and essentially following the recommended protocols from the supplier (Cygnus Technologies, LLC, BSA ELISA product insert). Briefly, samples containing BSA were reacted in microstrips coated with an affinity-purified capture antibody (anti-BSA). A second anti-BSA antibody labeled with horseradish peroxidase (HRP) was reacted simultaneously, forming a sandwich complex. After a wash step, tetramethylbenzidine (TMB) was added. The amount of hydrolyzed substrate was directly proportional to the concentration of BSA and was determined on a microplate reader.
The working range of the assay was between 0.50 ng/mL and 32.0 ng/mL in sample diluent. Quality control samples (QCs) at concentrations of 1.60, 8.00, and 24.0 ng/mL in sample diluent were used as standards.
ELISA analysis was performed by an external company (Swiss BioQuant AG, Reinach).
The following calculation assumes that the to-be-washed-out part is homogenously distributed in the washing solution during sedimentation.
The washout efficiency is determined by the two phases of cell concentration (steps 1–3), and wash cycles (step 4) as shown in the process graph below.
Vin = initial volume
d = dilution ratio
Vint = intermediate volume
nwash = number of wash cycle
ncycles = number of concentration cycle
1. Siegel RL, Miller KD, Fuchs HE, Jemal A. Cancer statistics, 2022. CA Cancer J Clin. 2022 Jan;72(1):7-33. doi: 10.3322/caac.21708.
2. De Marco RC, Monzo HJ, Ojala PM. CAR T Cell Therapy: A versatile living drug. Int J Mol Sci. 2023 Mar 27;24(7):6300. doi: 10.3390/ijms24076300.
3. Wang X and Rivière I. Clinical manufacturing of CAR T cells: foundation of a promising therapy. Mol Ther Oncolytics. 2016 Jun 15;3:16015. doi: 10.1038/mto.2016.15.
4. Kaiser AD, Assenmacher M, Schröder B, Meyer M, Orentas R, Bethke U, et al. Towards a commercial process for the manufacture of genetically modified T cells for therapy. Cancer Gene Ther. 2015 Mar;22(2):72-8. doi: 10.1038/cgt.2014.78.