We report here an integrated manufacturing process for a hESC-based therapeutic candidate for type 1 diabetes using the CyT49 cell line (). High-density, single cell banks of CyT49 were thawed and expanded with efficient population doublings. Expanded cultures were aggregated in suspension, generating uniform clusters of undifferentiated cells, which were then differentiated in suspension en masse. An optimized 4-stage protocol directed the stepwise formation of highly enriched pancreatic populations that functioned robustly in vivo. The process integrates a standardized cell source and scaled differentiation with the ability to cryopreserve the end-stage pancreatic aggregates so that function is retained in vivo (not shown). This provides the critical ability to test function and safety of scaled manufactured lots, prior to pre-clinical studies or clinical application. Our approach represents the first demonstration of a practical system for manufacturing a hESC-based treatment for type 1 diabetes, mitigating many of the perceived hurdles to clinical development.
Schematic representation of the manufacturing process for pancreatic progenitors.
In contrast to previous reports of cGMP banking of clinically-relevant hESC lines, which by necessity relied on earlier clump-passaging methodologies 
, our single-cell suspension cGMP MCB/WCB are a starting point for acute scaled expansion for producing large batches of pancreatic progenitors. WCB4B vials could theoretically be thawed on a monthly basis, generating batches of >1010
cells each month, and at that rate would provide starting material for more than 20 years of manufacturing. If exhausted, additional WCBs derived from MCB4 could be generated as required. Our strategy also minimizes the potential for karyotypic drift in euploid hESC cultures 
, as banked cells are only expanded for 3–6 passages over a 15–20 day period before differentiation. It is notable that while theoretical considerations implicate population size as a risk factor in accumulating aneuploidies in hESC cultures 
, our conditions could support the expansion of euploid cells on a clinical-manufacturing scale. The simplicity of this expansion system for CyT49 is also likely to be amenable to automation.
Aggregation of undifferentiated hESC in dynamic rotational suspension culture is a unique methodology. Current examples of suspension culture of hESC use aggregates generated from collagenase passaging 
, shear of suspension clusters 
, in static culture conditions 
, require microcarrier support 
, or heat shock in combination with the ROCK inhibitor Y27632 to support aggregation and survival 
. Such limitations add hurdles to maintaining cells in an undifferentiated state, the ability to scale and automate, or support effective population expansion in the face of passaging inefficiencies. Conversely, we have demonstrated that uniform clusters of undifferentiated hESC could be aggregated with high incorporation efficiencies in rotational culture, by taking advantage of the inherent self-associative properties of hESC. The particular conditions we optimized were critical for effective aggregation, as the circular movement imposed a shallow central vortex drawing cells into a high local density in the middle of each well. The epithelial characteristics of undifferentiated hESC, presumably E-cadherin expression in particular 
, may mediate effective self-association of cells as they collide 
. Aggregates did not form efficiently in static suspension culture, rocked, or stirred cultures, or even in centrifuged cell pellets (not shown), indicating that fluid movement and limited shear forces played an important role in transitioning cell-cell contact into stable adhesion, a well characterized phenomenon 
. The ability to serially passage aggregates of euploid hESC in dynamic rotational culture suggests that future manufacturing processes may also utilize scaling of hESC cultures in such a format.
The pairing of scaled expansion of CyT49 in adherent culture with aggregation and differentiation en masse is also a novel strategy. Effective step-wise lineage specification en route to pancreatic cell types was achieved by adapting our previous adherent-based system to dynamic suspension culture. This approach is essentially unrelated to traditional embryoid body differentiation, which is stochastic at best and not capable of directing uniform specification. Our strategy is also conceptually different from efforts to expand endoderm intermediates such as DE 
. In the present method, differentiation was directed in a highly controlled manner and resulted in consistency and uniformity of cellular composition superior to that we reported previously. In contrast to differentiation of hESC in adherent culture, which in our hands appears to be sensitive to variations in local densities, hESC aggregates by their very nature have high and uniform local cellular density, which may exert a positive influence on the uniformity of pancreatic differentiation. We demonstrated with our previous methods that multiple hESC lines could be differentiated to pancreatic lineages 
, including CyT203-derived glucose responsive, insulin secreting cells in vivo
. Furthermore, others have recently generated functional grafts from the WA1 hESC line 
. Given the likely requirement for cell line-specific optimization of culture conditions and timing 
, it is reasonable to expect that these suspension methodologies could be applied to other pluripotent cell lines. The mechano-physical properties of the cellular microenvironment are also likely to be quite different in suspension as compared to adherent conditions, which may also contribute to consistency of cell fate determination. We have achieved defined cell compositions without the requirement for cell sorting and the associated poor yields that accompany it 
, although sorting of the dissociated Stage-4 aggregates to enrich PE or endocrine cells for profiling analyses was achieved using CD142 and CD200, respectively (data not shown). Furthermore, for a cellular therapy based on implanting pancreatic lineages, the use of cellular aggregates offers a significant advantage over microcarrier-based suspension technologies. Pancreatic aggregates can be implanted without disrupting the maturing cellular architecture, avoiding substantial losses that would occur when harvesting from microcarriers.
The manufacturing process we have developed serves as a foundation for additional scaling, development of conditions for cGMP manufacturing and production of qualified material for preclinical and clinical studies. The Edmonton protocol calls for a patient dose of 10,000 islet equivalents (IEQ)/kg body weight to achieve the primary endpoint of insulin independence 
. A projected dose suggests that a large number of hESC-derived pancreatic progenitors will be required for clinical application, estimated to be a minimum of 108
. A sensitivity analysis of the scale required to enter a phase 1 clinical trial needs to account for the number of patients, absolute doses to be tested, amount of product utilized in quality control testing, and efficiencies at each step of the manufacturing process (). Given the assumptions that can be made for each variable and allowing for the range within these may fall, we can speculate that a batch of somewhere between 2.5×109
CyT49 cells will be necessary to generate sufficient cell product for a phase 1 clinical trial of ten patients at a dose of 108
cells/patient. In this report we have demonstrated the ability to reach the lower end of this predictive window using our current technology. A single vial of 107
cells was thawed, expanded over two weeks, and differentiated to produce pancreatic aggregates of 3.3×109
cells, which functioned appropriately in vivo
(data not shown). The upper end of this prediction is also well within reach, as additional passages in multilayer chambers are neither technically difficult nor approaching the limit of the technology. A single 40-stack cell factory has a surface area of 25,000 cm2
, which conservatively would yield >6.2×109
CyT49 cells under our present conditions. Given the progress in automating hESC expansion with robotics, the application of hESC expansion in suspension culture, or the logical adaptation of differentiation to controlled, expandable, and closed bioreactor manufacturing systems, we anticipate that we will be able to increase the scale of our process by several additional orders of magnitude.
Assessing the consistency that could be achieved with a scalable manufacturing process for hESC-derived pancreatic progenitors was the primary objective of this study. CyT49 could be differentiated with high reproducibility using an optimized process, achieving robust function in vivo
. In concordance with our previous reports, we generated pancreatic grafts that could sense blood glucose and respond by releasing human insulin 
. The grafts differentiated and matured over time, as shown by the statistically significant increase in baseline and amplitude of response observed after week 10, and significant 5 to 10 min GSIS response at weeks 16–50. Critically, grafts could maintain blood glucose homeostasis in an endogenous β-cell ablation model. The combination of scaled differentiation and functional outcome in hundreds of animals represents an experimental magnitude far greater than previously reported, with reproducibility that enables progression to formal preclinical development. We envision developing an allo-compatible neo-pancreatic product, by engrafting pancreatic progenitors within a vascularizing and durable immunoisolation, or macroencapsulation, device.
Suspension-based pancreatic differentiation runs consistently yielded only minimal amounts of non-pancreatic tissue upon implantation, in contrast with the variability in teratoma rates displayed in our previous reports 
. Nonetheless, some grafts also contained dilated ducts and/or cysts derived from these ducts. While not a particular safety concern pathologically, an enlarged cyst could potentially impinge upon surrounding tissue. Cell implantation within a durable macroencapsulation device could potentially constrain such structures, and offer an additional level of safety by enabling retrievability of implanted cells. In any format, formal demonstration of product safety requires both regulated preclinical studies, and eventual batch release qualification of cryopreserved material produced under cGMP.
In summary, we have assembled and demonstrated the utility of a system for the manufacturing of a functional hESC-based therapeutic product for type 1 diabetes. Our approach coordinates many discrete steps into a highly regulated process, linking a scaled and standardized cell source with expansion and scalable differentiation, through to qualification in vivo. The process generates implantable material with reproducibility and the compatibility required for industrialization.