Process parameter development for the scaled generation of stem cell‐derived pancreatic endocrine cells

Abstract Diabetes is a debilitating disease characterized by high blood glucose levels. The global prevalence of this disease has been projected to reach 700 million adults by the year 2045. Type 1 diabetes represents about 10% of the reported cases of diabetes. Although islet transplantation can be a highly effective method to treat type 1 diabetes, its widespread application is limited by the paucity of cadaveric donor islets. The use of pluripotent stem cells as an unlimited cell source to generate insulin‐producing cells for implant is a promising alternative for treating diabetes. However, to be clinically relevant, it is necessary to manufacture these stem cell‐derived cells at sufficient scales. Significant advances have been made in differentiation protocols used to generate stem cell‐derived cells capable of reversing diabetes in animal models and for testing in clinical trials. We discuss the potential of both stem cell‐derived pancreatic progenitors and more matured insulin‐producing cells to treat diabetes. We discuss the need for rigorous bioprocess parameter optimization and identify some critical process parameters and strategies that may influence the critical quality attributes of the cells with the goal of facilitating scalable manufacturing of human pluripotent stem cell‐derived pancreatic endocrine cells.

1 diabetes, its widespread application is limited by the paucity of cadaveric donor islets. The use of pluripotent stem cells as an unlimited cell source to generate insulin-producing cells for implant is a promising alternative for treating diabetes.
However, to be clinically relevant, it is necessary to manufacture these stem cellderived cells at sufficient scales. Significant advances have been made in differentiation protocols used to generate stem cell-derived cells capable of reversing diabetes in animal models and for testing in clinical trials. We discuss the potential of both stem cell-derived pancreatic progenitors and more matured insulin-producing cells to treat diabetes. We discuss the need for rigorous bioprocess parameter optimization and identify some critical process parameters and strategies that may influence the critical quality attributes of the cells with the goal of facilitating scalable manufacturing of human pluripotent stem cell-derived pancreatic endocrine cells.

K E Y W O R D S
cell culture, cell transplantation, clinical translation, diabetes, differentiation, pancreatic differentiation, pluripotent stem cells, stem cell culture

Significance statement
Diabetes is a global pandemic that can potentially be treated using stem cell-derived pancreatic progenitors and hormone-secreting cells. The ability to generate stem cell-derived derivates at sufficient scales is a critical step toward using these cells to treat diabetes. This article reports on some critical process parameters and quality attributes that affect the final stem cell-derived product.

| INTRODUCTION
Diabetes is ranked within the top 10 most deadly diseases worldwide. 1,2 According to the World Health Organization, as of 2019, 463 million adults had diabetes, and that number is predicted to rise to 700 million by 2045. 3 It is estimated that by 2030, accumulated global diabetes management expenditures will rise to $2.48 trillion. 4 All forms of diabetes are characterized by chronically elevated blood glucose levels. Glucose homeostasis is regulated by the actions of alpha and beta cells contained within the islets of Langerhans in the pancreas. Alpha cells secrete more glucagon during periods of low blood glucose (hypoglycemia), while beta cells secrete higher levels of insulin during periods of elevated blood glucose (hyperglycemia). Type 1 diabetes (T1D) is an autoimmune disease caused by insulin insufficiency due to the destruction of insulin-producing beta cells. T1D requires lifelong insulin therapy either by exogenous or endogenous sources.
Although great strides have been made since the discovery and isolation of insulin by Drs. Banting, Best, Collip, and Macleod in 1922, 5 patients with T1D still experience compromised quality of life. The most common method for managing blood glucose levels is administering multiple daily insulin injections. However, improper dosing of insulin puts patients at risk of life-threatening hypoglycemia and a myriad of longterm complications resulting from prolonged hyperglycemia. Also, insulin injections cannot mimic the precise glycemic control of pancreatic islets.
The most effective therapy for reversing hyperglycemia in T1D is the transplantation of islets isolated after brain or cardiac death. [6][7][8] Islet transplantation demonstrates the efficacy of beta cell replacement for the treatment of T1D. However, the scarcity of donors necessitates the need for other sources of insulin-producing cells in order to support widespread adoption of this approach (Figure 1). Several differentiation protocols use a stepwise multistage procedure to generate pancreatic progenitors and endocrine cells ( Figure 2). D'Amour et al were the first to report human PSC-derived pancreatic progenitors and immature hormone-producing cells following the efficient induction of definitive endoderm using a combination of small molecules and growth factors to direct the differentiation. 10,11 Implant of these stem cell-derived pancreatic progenitors and immature hormone-producing F I G U R E 1 Schematic of potential cell sources of insulin-producing cells that could be used for replacement therapy in diabetes. ESC, embryonic stem cell; iPSC, induced pluripotent stem cell cells into immunodeficient mice produced grafts with mature insulinproducing beta cells that protected the mice against chemically induced hyperglycemia. 12 Three phase 1/2 clinical trials led by ViaCyte Inc are currently active (ClinicalTrials.gov IDs: NCT04678557, NCT03163511, and NCT02239354) using product candidates PEC-Encap and PEC-Direct, that consist of a mix of human PSC-derived pancreatic progenitor cells and immature hormone-producing cells that are contained in retrievable macroencapsulation devices and implanted subcutaneously. The first device, PEC-Encap, was designed to isolate and protect the graft from the immune cells while allowing insulin, glucose, oxygen, and waste products to diffuse through the device membranes. The initial clinical trial (ClinicalTrials.gov ID: NCT02239354) revealed that these cell-containing devices were safe, free of off-target cell growth, and protected against allo-and autoimmune rejection. 13 However, unlike results observed in mouse studies, 12 there was generally poor cell engraftment and survival, mostly likely due to hypoxia caused by a foreign body response which hampered vascularization of the cell-containing device. 13  with those microencapsulated in alginate, which matured mostly into glucagon-producing cells. 19 Pepper et al showed that pancreatic progenitors implanted in a prevascularized subcutaneous site more effectively reversed hyperglycemia in mice in comparison to when the cells were implanted in the fat pad or non-vascularized subcutaneous F I G U R E 2 A representative schematic of the different stages during pluripotent stem cell (PSC) differentiation toward insulin-producing cells. PSCs go through a seven-stage protocol (adapted from Rezania et al 9 ) using a combination of growth factors and small molecules. Each stage is identified by key proteins and transcription factors (depicted in gray text) space. 20 It has also been shown that the host sex, species (rats vs mice), and thyroid hormone levels can affect the rate of maturation and the acquisition of glucose competence in implanted pancreatic progenitor cells. [21][22][23] The degree of microenvironment variability would likely be greater in the human population than it would in the lab setting using inbred strains of mice. As such, the host microenvironment, including levels of various circulating factors, could differ significantly in humans and might have an impact on the outcome of the pancreatic progenitor cell implants.

Pluripotent stem cells (PSCs) include embryonic stem cells (ESCs
Despite potential caveats with implanting cells that are not terminally differentiated, pancreatic progenitors may offer significant advantages as a product compared with fully differentiated endocrine cells. For instance, pancreatic progenitors can be generated within a shorter culture time than mature islet-like cells, potentially lowering cell production costs. A recent study showed that human PSC-derived insulin-producing cells and isolated human islets had a similar oxygen consumption rate (OCR) when challenged with high glucose. 24 In contrast, PSC-derived pancreatic progenitors have a lower OCR compared with human islets. 20 These findings suggest that pancreatic progenitors may be relatively metabolically quiescent, a characteristic that could better support graft survival in a hypoxic environment.
Cells are proliferative during the first four stages of differentiation to pancreatic progenitors 17 ; however, cell losses typically occur during the later stages 9 and will thereby contribute to the cost of cell manufacturing. Collectively, the time and cost to manufacture, cell yield, metabolic state, and the proven ability to effectively reverse hyperglycemia in rodent models make the use of PSC-derived pancreatic progenitors appealing for the treatment of diabetes.
Over the last decade, significant progress has been made toward developing differentiation protocols that yield functionally mature insulin-producing cells. 9,12,[24][25][26][27][28][29] In some cases, PSC-derived insulinproducing cells generated entirely in vitro were capable of glucosestimulated insulin secretion 9,24-26,30 as well as increased calcium signaling and mitochondrial respiration, similar to primary human islets. 24 An implant of insulin-producing cells can reverse diabetes in mice faster, and at a lower dose, than PSC-derived pancreatic progenitors. 9 More differentiated cell types may have a lower risk of outgrowth 9,17 following implant. Compared to pancreatic progenitors, cells further along in their differentiation may be less susceptible to becoming offtarget cell types resulting from uncontrolled environment cues in vivo. during the last 11 days of a 24-day differentiation protocol to make iPSC-derived glucose responsive insulin-producing cells. 33 Using a 20-day differentiation protocol, Zhu et al demonstrated that overexpressing PDX1, NGN3, and MAFA during stage 1 (definitive endoderm), stage 4 (pancreatic progenitors), and stage 6 (immature beta cells), respectively, resulted in glucose-and GLP-1 responsive beta cells. 34 Although presently there is no clinical data for the treatment of diabetes using implanted PSC-derived insulin-producing islet-like clusters, results from preclinical studies have demonstrated their ability to reverse or protect against streptozotocin (STZ)-induced hyperglycemia, 9,[24][25][26]28,29 suggesting that they may be appropriate for cell replacement therapy. Vertex has recently announced a phase 1/2 clinical trial for VX-880, a stem cell-derived islet product, for the treatment of T1D in patients with hypoglycemia unawareness and severe hypoglycemia (ClinicalTrials.gov ID: NCT04786262). 35 It remains to be established whether progenitor-based or islet-based stem cellderived cell products exhibit any relative enhancements in durability and efficacy toward the treatment of T1D.

| A NEED FOR GREATER PROCESS UNDERSTANDING TO FACILITATE ROBUST, SCALED CELL PRODUCTION
The dose of donor pancreas cells typically infused during islet transplantation using the Edmonton protocol is 7000 to 12 000 islet equivalents (IEQ)/kg body weight. 6,36 Based on the estimated number of beta cells within an islet, this would translate into approximately a billion stem cell-derived cells per recipient as a therapeutic dose, assuming equivalent survival and potency. 37,38 To facilitate manufacturing of cells at sufficient scale, adopting a scientific risk-based analysis using a well-structured quality-by-design strategy is beneficial. 39 The maintenance of a cell product's critical quality attributes (CQAs), or physical, chemical, biological, or microbiological characteristics that the production process should control to be within appropriate limits, ensures the desired product quality required for therapeutic benefit.
CQAs are controlled by the performance of the manufacturing process which drives their establishment. A control strategy during cell production is designed to monitor critical process parameters (CPPs), defined here as process criteria whose variability impacts CQAs. 39,40 Control of the various CPPs that impact the cells' CQAs should facilitate robust and consistent large-scale production of a therapeutic stem cell-derived cell product. Such CPPs could include a wide range of process inputs and outputs: parameters that influence process performance, material attributes that feed into the process, and defined user requirements for equipment utilized to facilitate the process ( Table 1).
The decision to use either ESCs or iPSCs may be the first consideration when developing a large-scale manufacturing plan. Although iPSCs were initially thought to be similar to ESCs based on morphology, growth rate, pluripotency markers, epigenetic status, and trilineage differentiation capability, 41,42 several studies observed differences in gene expression, 43 DNA methylation, 44 and persistent epigenetic memory, 45 which may predispose the iPSCs to differentiate more efficiently to particular lineages. Some of the reported difference between ESCs and iPSCs may be due to the reprogramming procedure, incomplete reprogramming, as well as the method and duration of cell culture. 46 Furthermore, Yamanaka noted that studies that reported differences between ESCs and iPSCs analyzed a relatively small number of clones (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12) compared to studies where no differences were observed . 46 The use of iPSCs may have less potential ethical and religious concerns associated with ESCs. 47  CellCube ® , and HYPERFlask ® vessels have been developed to provide a higher surface area for cell attachment and higher yield than conventional T-flasks and multilayered CellSTACK ® chambers. A major challenge with scalable 2D cell culture is the need to harvest the cells from the vessel surface. Large PSCs banks typically need to be generated, characterized for pluripotency and genetic stability, tested for potential contamination (eg, mycoplasma), and cryopreserved prior to the start of any downstream applications. It is imperative to develop strategies for a seed train that will allow for efficient culture and potential cryopreservation of both PSCs and their derivatives at the required scale. As the field of regenerative medicine expands, there is growing interest in developing bioprocesses that are performed in 3D dynamic suspension in order to address some of the challenges experienced with adherent monolayer culture, including high-density culture via perfusion. 59 Many differentiation protocols used to make insulin-producing cells are initiated with 3D PSC aggregates ( Figure 3). 18,24,25,60,61 After T A B L E 1 An example of the quality-by-design process for the generation of pluripotent stem cell (PSC)-derived insulin-producing cells. The quality product profile outlines the properties of the desired clinical product based on the critical quality attributes (CQA). The critical process parameters are bioprocess parameters that influence the CQA generating enough cells, PSCs are typically harvested as single cells and seeded at a significantly higher density to initiate aggregate formation prior to the start of differentiation. There is also a short period between the initial high-density seed of PSCs and the induction of the definitive endoderm stage, typically between 1 and 3 days. The generation of PSC aggregates is a common bottleneck to larger scale cell production because of the low efficiency of aggregate formation from single cells due to anoikis. 62 56 PSC aggregate growth rate can be variable and slower compared to adherent cells, 53 thus impacting the potential cell yield. The size and compaction of PSC aggregates affects the proliferative capacity of the clusters; smaller aggregates may have a slower unstable proliferative capacity, whereas larger aggregates may experience oxygen and nutrient deprivation resulting in poor viability. [68][69][70][71] Other challenges associated with 3D suspension cultures may include aggregate agglomeration which could result in necrotic cores due to hypoxia, and apoptosis due to the shear stress as a result of mixing.
Several groups are actively working on improving the yield of PSC aggregates by improving bioreactor design, optimizing seeding density, feeding frequency, and the media formulation. 56,72,73 Given the short time between seeding and initiating the differentiation, coupled with the low aggregate formation efficiencies and the slower growth of PSC aggregates, it is imperative that optimal process parameters be identified to minimize excessive material, labor, and financial costs associated with manufacturing large batches of cell clusters.
The formation of 3D aggregates appears to facilitate stem cell differentiation to endocrine lineages and promote regulated insulin secretion. For instance, generation of PSC-derived pancreatic progenitor clusters increased gene expression of endocrine makers. 17,74 F I G U R E 3 Schematic of sample bioprocessing strategies used for generating insulin-producing cells from cryopreserved pluripotent stem cells (PSCs). Following the expansion of PSCs on a monolayer, the appropriate time point to generate cell aggregates still needs to be determined. Cell clusters may be generated using different available cell culture platforms such as low attachment plates, AggreWell plates, spinner flasks, roller bottles, or PBS bioreactors Furthermore, while pancreatic progenitors implanted in rodents as aggregates differentiated to insulin-producing cells, similar non-aggregated pancreatic progenitors failed to effectively differentiate into insulin-producing cells within 16 weeks post-transplant. 74,75 3D aggregate architecture may be important for insulin secretion in primary islets and insulin-secreting cell lines. Compared to purified single rat beta cells, both intact and reaggregated islets secreted four to five times more insulin in response to elevated glucose and leucine. 76 These results were attributed to higher levels of cellular adenosine Purified rat beta cell aggregates implanted in diabetic mice could effectively reverse hyperglycemia, similar to intact islets, within 1 week, suggesting that the normal beta cell to non-beta cell relationship may not be necessary for adequate glycemic control post-transplant. 81 However, the incorporation of non-beta cells with reaggregated rat beta cells promoted long-term survival of the implanted cells. 82 It is noteworthy that Hogrebe et al achieved glucose-responsiveness from PSC-derived insulin-producing cells in 2D planar culture through manipulation of the actin cytoskeleton, similar to that of cells made using a 3D suspension protocol. 26 During the endocrine induction stage, depolymerization of filamentous actin using latrunculin A led to increased NGN3 expression, and its downstream targets NEUROD1 and NKX2.2, relative to that of untreated 2D controls at an equivalent time point of the differentiation. Furthermore, latrunculin A treated cells displayed glucose-stimulated insulin secretion while the untreated 2D controls did not. Nevertheless, the cells differentiated in 2D were aggregated prior to implant into diabetic mice and resulted in the reversal of hyperglycemia more rapidly than cells generated using the 3D suspension protocol. Collectively, these studies highlight the impact of aggregate formation on nutrient and hormone-mediated insulin secretion, and endocrine specification, based on the presence of key markers.
Aggregate size can have an impact on the survival and function of implanted cells. 24,25 Despite species differences in pancreas size, islet size is well conserved 83 with an average of $150 μm in diameter and a range of between 50 and 500 μm in mammals. 84,85 Lehmann et al demonstrated that islet size might play an important role in determining human islet transplantation outcome. 86 In comparison to large islets (150-300 μm diameter), smaller islets (50-150 μm diameter) had a higher percentage of insulin-positive cells per islet, higher insulin production, almost double the glucose-stimulated insulin secretion based on perifusion assay, and less cell death following culture in hypoxic and normoxic conditions. 86 Interestingly, following islet transplantation, two patients had equivalent stimulated C-peptide levels despite the fact that one recipient received smaller islets and significantly less IEQ than the other that received larger islets (3352 IEQ/kg vs 11 625 IEQ/kg respectively). 86 Previous reports have demonstrated that larger human and mouse islets secrete less insulin per IEQ compared to smaller islets. [87][88][89][90] The reaggregation of human islets has been employed to generate smaller and more homogeneously sized pseudoislets with improved posttransplant survival and function. 91,92 These results indicate that the size of islets is an important variable affecting graft survival and function.
Similar reaggregation strategies used to generate pseudoislets have been adopted with stem cell-derived insulin-producing cells. 24,25 Velazco-Cruz et al demonstrated that the reaggregation of clusters at the beginning of stage 6 (PSC-derived beta cells), combined with the removal of ALK5i, a TFGβ inhibitor, resulted in the acquisition of glucose-stimulated insulin secretion. 25 Nair et al observed that reaggregation of PSC-derived insulin-producing cell clusters alone was not sufficient to induce glucose competence. 24 In order for the cells to develop glucose-stimulated insulin secretion, insulin positive cells were sorted prior to reaggregation and then cultured for an additional week. 24 Several groups have used various sorting strategies to purify the cells based on different stage-specific surface markers. 16,60,93,94 Pronounced cell losses associated with the dispersal, purification, and reaggregation of PSC-derived clusters, although scarcely reported, 16  cells is yet to be determined and undoubtedly efficiencies will be improved upon.

| CONCLUSION
Although islet transplantation demonstrates proof-of-concept for cell replacement therapy to treat T1D, the widespread implementation of this procedure is limited by paucity of donor islets. The potential of stem cellderived insulin-producing cells for the treatment of diabetes has been demonstrated using rodent models. Furthermore, the safety and efficacy of PSC-derived pancreatic progenitors is currently being evaluated in ongoing clinical trials. Differentiation protocols are being developed using multiple cell lines. It is desirable to identify critical parameters that will affect the CQAs of the target product profile. Adopting quality-by-design methodology can help unify the work being done across the field and accelerate the efforts toward successful clinical translation, large-scale cell manufacturing and commercialization.