Biomanufacturing and Scale-Up of Clinical-Grade Induced Pluripotent Stem Cell Lines

Ahlberg G¹, Conti C², Szabo R³

Volume 8 Issue 1 - 1004

Introduction

Induced pluripotent stem cells arrived in the clinic as a manufacturing problem before they were solved as a biological one. Deriving a line is routine; producing one that a regulator will accept as the starting material for a therapeutic product, and then expanding it reproducibly to the numbers a dosing schedule demands, is not. The first fully cGMP-compliant hiPSC banks were described a decade ago, together with the donor consent, traceability and testing framework needed to release them [1]. Consensus quality-control guidance for clinical-grade lines followed [2], building on earlier points-to-consider for pluripotent seed stocks [3].

What has changed since is where the difficulty sits. Volumetric productivity is no longer the rate-limiting step: perfused stirred-tank processes now deliver 5.25 x 10^9 cells from a 150 mL working volume in 7 days [4], and two-step spinner-to-bioreactor trains take 16 x 10^6 cells to 2 x 10^9 in 14 days [5]. Those numbers already exceed what most allogeneic indications require per lot. The binding constraints are elsewhere: the genomic drift that accumulates over the passages such expansion consumes, the analytical sensitivity of residual-undifferentiated-cell testing, and the strategic choice between autologous manufacture, an HLA-homozygous haplobank and a gene-edited universal line — a choice that determines cost structure more decisively than any bioreactor decision [6].

This review takes those three constraints in turn, after a short account of how a clinical line is derived and qualified. Throughout, process figures are quoted with the vessel, working volume and feeding regime at which they were obtained, because peak densities from different formats are not interchangeable claims about the same quantity.

Reprogramming and the qualified starting line

Clinical manufacture uses non-integrating reprogramming. Sendai virus vectors deliver the reprogramming factors as an RNA genome that does not integrate [7]; episomal plasmid systems combining p53 suppression with L-MYC substitution provided the other main route and were used to generate lines from HLA-homozygous donors matching approximately 20% of the Japanese population [8]; synthetic modified mRNA offers a third, footprint-free option [9]. Head-to-head comparison established that these methods differ materially in aneuploidy rate, efficiency, reliability and workload [10], and the choice is made on those grounds rather than on efficiency alone.

The practical constraint on the front end is throughput per operator, not yield per experiment. A chemically defined, small-format reprogramming platform reduced reagent cost per patient sample by more than 80% and medium usage by 70%, compressed experimental scale up to 50-fold, and raised patient-sample throughput 10-fold, so that one technician runs 20 to 40 lines simultaneously in 24- to 96-well format and a facility supports more than 200 well-characterised lines annually; more than 800 lines from over 160 samples were established this way [11].

Vector clearance is a release attribute of the resulting line rather than a property of the method. In cGMP banks derived from cord blood CD34+ cells, residual episomal sequences were tracked by qPCR against EBNA and OriP elements of the pEB-C5 and pEB-Tg plasmids at multiple passages before final expansion and banking [12]. No fetched primary record states a numeric acceptance limit, and none should be inferred; what is documented is the practice of demonstrating clearance across passage, not a threshold. The same caution applies to reprogramming efficiency: the comparative study establishes that the methods differ, but the per-method figures belong to its primary data and are not reproduced here [10].

Expansion at the litre scale

Planar culture on defined matrices remains the derivation and qualification format, but scale-up moves to stirred systems. Microcarrier culture in a 50 mL spinner with Essential 8 and vitronectin reached 1.4 x 10^6 cells mL^-1 and a maximum 3.5-fold yield over 10 days, with the optimum set jointly by inoculation density of 55,000 cells cm^-2 and agitation at 44 rpm, identified by three-level factorial design [13]. Matrix-free aggregate suspension has since displaced it. A 60 mL Vertical-Wheel vessel reached 2.3 +/- 0.2 x 10^6 cells mL^-1 and 9.3 +/- 0.6-fold expansion in 5 days, a volumetric productivity of 4.6 +/- 0.3 x 10^5 cells mL^-1 d^-1, with mean aggregate diameter 346 +/- 11 um; dextran sulfate to suppress agglomeration raised maximum cell number by 97% in mTeSR1 and 106% in mTeSR3D [14]. Computational-fluid-dynamics-guided operation of 100 mL and 500 mL Vertical-Wheel vessels gave over 30-fold expansion in 6 days from a single-cell inoculum of 2.0 x 10^4 cells mL^-1, with day-5 aggregate diameters from

Feeding regime, not vessel geometry, sets the ceiling. Perfusion in single-use stirred tanks gave 2.85 x 10^6 cells mL^-1 and 47% higher yield than repeated batch [16]. Adding metabolic control and in silico process modelling raised peak density an order of magnitude further, to 35 x 10^6 cells mL^-1 and 5.25 x 10^9 cells per 150 mL run at 70-fold expansion in 7 days, while cutting medium demand by 75%; the process was reported as transferable to 500 mL [4]. A two-step train — 100 mL spinner then 1000 mL bioreactor — achieved 16-fold then a further 10-fold expansion, taking 16 x 10^6 cells to 2 x 10^9 in 14 days with free scalability stated between 50 and 2400 mL, at the cost of dissociating day-7 aggregates of 324 +/- 71 um back to single cells for re-inoculation, after which aggregates settled at 198 +/- 58 um [5].

Genomic stability across passage

Every fold of expansion is passages, and passages select. Screening of 125 human ES and 11 human iPS cell lines from 38 laboratories (N = 136) found recurrent changes on chromosomes 1, 12, 17 and 20 accumulating with prolonged culture, and a minimal amplicon at 20q11.21 containing ID1, BCL2L1 and HM13 in more than 20% of lines [17]. The mechanism is a genuine growth advantage: BCL-XL mediates the selective benefit conferred by that amplicon [18]. Point mutations behave the same way. Sequencing the protein-coding genes of 140 human ES cell lines identified six TP53 mutations across five unrelated lines, all dominant negative and all selectively advantaged in culture [19]. The consequences of these recurrent genetic and epigenetic variants for product quality are now the organising question for the field [20].

Format and scale Peak Fold Stated limitation density expansion (x10^6 per (duration) mL) Aggregate, stirred 2.85 47% above Repeated-batch tank, perfusion repeated feeding less batch homogeneous Aggregate, stirred 35 70-fold (7 Requires metabolic tank, 150 mL, days) control; stated metabolic control transferable to 500 mL Aggregate, 100 mL Not stated 16-fold then Day-7 aggregates spinner to 1000 mL as density 10-fold (14 must be dissociated tank days) for re-inoculation This makes the expansion train and the stability specification a single design problem rather than two. The detection methods in routine use are not matched to the task. Chromosome banding, qPCR, FISH and digital PCR differ substantially in their ability to see low-level mosaicism, and the limits of the standard cytogenetic approach are significant [21]. A targeted droplet digital PCR panel built from a meta-analysis of more than 100 publications covers 738 catalogued recurrent abnormalities and detects 94.3% of them from DNA extracted from culture supernatant, which makes in-process monitoring between passages practical rather than terminal [22]. Because that panel runs on spent medium, it can be scheduled against the passage plan of an expansion train rather than against banking milestones, which is the practical shift: variant surveillance becomes a process parameter instead of a release test. Seed-stock guidance has long framed genetic characterisation as a continuing obligation across the life of a bank rather than a one-time qualification [3], and consensus quality-control criteria for clinical lines assume the same [2].

Release testing and residual undifferentiated cells

Regulatory expectations differ by jurisdiction and by vintage. In the European Union, the CAT reflection paper adopted and effective 14 January 2011 states that pluripotent cells carry a relatively high potential risk of tumour formation that should be carefully addressed, that hESC- and iPSC-derived cells should be shown to be lineage-committed before administration, and that the presence of proliferative and pluripotent cells tolerated in the final product should be limited and justified [23]. In the United States, CBER's April 2008 CMC guidance for somatic cell therapy INDs specifies sterility per 21 CFR 610.12 or USP <71>, mycoplasma testing at the manufacturing stage most likely to detect contamination, master and working cell bank characterisation, and in vitro viral testing on both banks with a one-time test on end-of-production cells [24]; that document predates iPSC products and contains no pluripotent-cell-specific text. A multi-regulator consensus effort has since set out points to consider for tumorigenicity assessment of cell therapy products [25].

The analytical answer to the EU purity expectation is a sensitivity question. Flow cytometry for TRA-1-60 detected 0.1% undifferentiated hiPSC spiked into primary retinal pigment epithelium, while qRT-PCR for LIN28 in the same system reached signal equivalent to one hiPSC in 5.0 x 10^4 RPE cells, or 0.002% [26]. Droplet digital PCR for LIN28 pushed this to 0.001% — one hiPSC in 1 x 10^5 cardiomyocytes — with LIN28 confirmed extremely low across seven human tissues [27]. A highly efficient colony-formation assay on laminin-521 in Essential 8 detected spiked hiPSC at 0.001% to 0.01% depending on line and host cell type, across two lines and two host types [28]. For malignant potential rather than impurity, systematic comparison of embryoid body assays, teratoma histology, TeratoScore and PluriTest found only the teratoma assay assesses pluripotency and malignant potential together [29]. Identity, sterility, endotoxin, karyology and pluripotency panels round out the release package: STR profiling tracked from donor material to the cryopreserved final product, high-resolution typing at HLA-A, -B, -C, -DQA1, -DQB1 and -DRB1, blood group typing, G-banding to FDA GLP and ACMG standards with 20 metaphases per line, and flow cytometry for SSEA-4, TRA-1-60, TRA-1-81 and OCT3/4 with embryoid body trilineage differentiation and PluriTest as potency evidence; two cord-blood-derived clones were banked under this package, one reported with a normal karyotype after 17 passages [12].

Cryopreservation, banking and cost of goods

Banking is where the process meets the balance sheet. Cryopreservation of aggregates without DMSO is now feasible: an optimised DMSO-free solution gave post-thaw reattachment of 104 +/- 5.73% of fresh control at a 1:6 split ratio, against 48.7 +/- 9.85% for the non-optimised DMSO-free formulation and 58.4 +/- 6.58% for the 7.5% DMSO control, while DMSO-frozen samples ice-nucleated at -12 C recovered only 9.53% to 10.4% [30]. Since the aggregate is the unit both of expansion and of freezing in suspension processes, removing DMSO without a recovery penalty simplifies the handoff from bioreactor to vial rather than adding a step to it.

Traceable cost figures are scarce, and the widely circulated bioprocess cost models are not among them. What is documented comes from an ISCBI workshop at which named banking centres contributed their own 2019-vintage numbers in mixed currencies: cell bank production at approximately US$300,000 per batch of 300 vials produced over one month; vials priced at US$1,000 to 1,500 across biobanks; GMP facility maintenance at approximately US$2.2 million per year excluding raw materials and product QC; GMP process development from US$3 million minimum. Per-item testing costs were itemised as US$500 for patient viral testing, US$2,000 per plasmid per reprogramming round, US$25,000 to 30,000 for media and reagents, US$450 per clone for karyology, US$1,600 per clone for oncogene screening (under US$500 if non-CLIA) and US$220 per clone for STR analysis [31]. No primary source reports a batch failure rate for clinical-grade iPSC manufacture, and none should be assumed.

Sourcing strategy: autologous, haplobank or universal line

The dominant cost lever is upstream of all of this. The first autologous iPSC clinical case cost nearly 100 million yen, approximately US$1 million, and required the patient to wait over 10 months for surgery; using an allogeneic iPSC stock brought the overall cost under 20 million yen per patient [6]. That single comparison, not bioreactor productivity, explains the field's migration to banked allogeneic material.

Haplobanking exploits HLA homozygosity. Modelling on UK donor data showed that 150 selected HLA-homozygous volunteers would match 93% of the UK population with minimal immunosuppression [32]. In Japan, 10, 75 and 140 homozygous lines are estimated to match approximately 50%, 80% and 90% of the population; the released CiRA stock comprised 21 clones from 5 donors covering approximately 32% [6]. The curve is steeply concave — the first few donors buy most of the coverage, and the tail is expensive — which is the quantitative case for the gene-edited alternative. Targeted CRISPR-Cas9 disruption of HLA genes yields iPSC with enhanced immune compatibility from a single line [33], and hypoimmunogenic derivatives have been shown to evade rejection in fully immunocompetent allogeneic recipients [34]. Each edit, however, adds genomic characterisation burden to a product already constrained by the stability and purity testing above.

homozygous volunteers); the two series use different populations and matching assumptions [6,32].

Conclusion

Expansion of clinical-grade hiPSC is largely a solved engineering problem at the litre scale: perfused, metabolically controlled aggregate suspension reaches 35 x 10^6 cells mL^-1 and billions of cells per run, and multi-step trains scale predictably across two orders of magnitude of working volume. The residual difficulties are analytical and strategic. Recurrent variants at 20q11.21 and in TP53 are selected by the very passages that scale-up requires, and standard cytogenetics sees mosaicism poorly, so stability specifications must be built around in-process molecular monitoring rather than terminal karyotype alone. Residual undifferentiated cell testing now reaches 0.001% by digital PCR and colony formation, but the EU expectation that such cells be limited and justified is a framework, not a number, and each product must argue its own case. Sourcing decides economics: the gap between roughly US$1 million for one autologous case and under 20 million yen per patient from a stock dwarfs any process improvement, and the choice between haplobank breadth and a universal edited line is now the most consequential decision in an hiPSC manufacturing programme.

References

1. Baghbaderani, B.A., Tian, X., Neo, B.H., Burkall, A., Dimezzo, T., Sierra, G., et al. (2015) cGMP-manufactured human induced pluripotent stem cells are available for pre-clinical and clinical applications. Stem Cell Reports, 5(4): 647-659.
2. Sullivan, S., Stacey, G.N., Akazawa, C., Aoyama, N., Baptista, R., Bedford, P., et al. (2018) Quality control guidelines for clinical-grade human induced pluripotent stem cell lines. Regenerative Medicine, 13(7): 859-866.
3. Andrews, P., Baker, D., Benvinisty, N., Miranda, B., Bruce, K., Brustle, O., et al. (2015) Points to consider in the development of seed stocks of pluripotent stem cells for clinical applications: International Stem Cell Banking Initiative (ISCBI). Regenerative Medicine, 10(sup2): 1-44.
4. Manstein, F., Ullmann, K., Kropp, C., Halloin, C., Triebert, W., Franke, A., et al. (2021) High density bioprocessing of human pluripotent stem cells by metabolic control and in silico modeling. Stem Cells Translational Medicine, 10(7): 1063-1080.
5. Kwok, C.K., Ueda, Y., Kadari, A., Gunther, K., Ergun, S., Heron, A., et al. (2018) Scalable stirred suspension culture for the generation of billions of human induced pluripotent stem cells using single-use bioreactors. Journal of Tissue Engineering and Regenerative Medicine, 12(2): e1076-e1087.
6. Umekage, M., Sato, Y., Takasu, N. (2019) Overview: an iPS cell stock at CiRA. Inflammation and Regeneration, 39(1): 17.
7. Fusaki, N., Ban, H., Nishiyama, A., Saeki, K., Hasegawa, M. (2009) Efficient induction of transgene-free human pluripotent stem cells using a vector based on Sendai virus, an RNA virus that does not integrate into the host genome. Proceedings of the Japan Academy, Series B, 85(8): 348-362.
8. Okita, K., Matsumura, Y., Sato, Y., Okada, A., Morizane, A., Okamoto, S., et al. (2011) A more efficient method to generate integration-free human iPS cells. Nature Methods, 8(5): 409-412.
9. Warren, L., Manos, P.D., Ahfeldt, T., Loh, Y.-H., Li, H., Lau, F., et al. (2010) Highly efficient reprogramming to pluripotency and directed differentiation of human cells with synthetic modified mRNA. Cell Stem Cell, 7(5): 618-630.
10. Schlaeger, T.M., Daheron, L., Brickler, T.R., Entwisle, S., Chan, K., Cianci, A., et al. (2015) A comparison of non-integrating reprogramming methods. Nature Biotechnology, 33(1): 58-63.
11. Beers, J., Linask, K.L., Chen, J.A., Siniscalchi, L.I., Lin, Y., Zheng, W., et al. (2015) A cost-effective and efficient reprogramming platform for large-scale production of integration-free human induced pluripotent stem cells in chemically defined culture. Scientific Reports, 5: 11319.
12. Baghbaderani, B.A., Syama, A., Sivapatham, R., Pei, Y., Mukherjee, O., Fellner, T., et al. (2016) Detailed characterization of human induced pluripotent stem cells manufactured for therapeutic applications. Stem Cell Reviews and Reports, 12(4): 394-420.
13. Badenes, S.M., Fernandes, T.G., Cordeiro, C.S.M., Boucher, S., Kuninger, D., Vemuri, M.C., et al. (2016) Defined Essential 8 medium and vitronectin efficiently support scalable xeno-free expansion of human induced pluripotent stem cells in stirred microcarrier culture systems. PLOS ONE, 11(3): e0151264.
14. Nogueira, D.E.S., Rodrigues, C.A.V., Carvalho, M.S., Miranda, C.C., Hashimura, Y., Jung, S., et al. (2019) Strategies for the expansion of human induced pluripotent stem cells as aggregates in single-use Vertical-Wheel bioreactors. Journal of Biological Engineering, 13(1): 74.
15. Borys, B.S., Dang, T., So, T., Rohani, L., Revay, T., Walsh, T., et al. (2021) Overcoming bioprocess bottlenecks in the large-scale expansion of high-quality hiPSC aggregates in vertical-wheel stirred suspension bioreactors. Stem Cell Research & Therapy, 12(1): 55.
16. Kropp, C., Kempf, H., Halloin, C., Robles-Diaz, D., Franke, A., Scheper, T., et al. (2016) Impact of feeding strategies on the scalable expansion of human pluripotent stem cells in single-use stirred tank bioreactors. Stem Cells Translational Medicine, 5(10): 1289-1301.
17. Amps, K., Andrews, P.W., Anyfantis, G., Armstrong, L., Avery, S., Baharvand, H., et al. (2011) Screening ethnically diverse human embryonic stem cells identifies a chromosome 20 minimal amplicon conferring growth advantage. Nature Biotechnology, 29(12): 1132-1144.
18. Avery, S., Hirst, A.J., Baker, D., Lim, C.Y., Alagaratnam, S., Skotheim, R.I., et al. (2013) BCL-XL mediates the strong selective advantage of a 20q11.21 amplification commonly found in human embryonic stem cell cultures. Stem Cell Reports, 1(5): 379-386.
19. Merkle, F.T., Ghosh, S., Kamitaki, N., Mitchell, J., Avior, Y., Mello, C., et al. (2017) Human pluripotent stem cells recurrently acquire and expand dominant negative P53 mutations. Nature, 545(7653): 229-233.
20. Andrews, P.W., Barbaric, I., Benvenisty, N., Draper, J.S., Ludwig, T., Merkle, F.T., et al. (2022) The consequences of recurrent genetic and epigenetic variants in human pluripotent stem cells. Cell Stem Cell, 29(12): 1624-1636.
21. Baker, D., Hirst, A.J., Gokhale, P.J., Juarez, M.A., Williams, S., Wheeler, M., et al. (2016) Detecting genetic mosaicism in cultures of human pluripotent stem cells. Stem Cell Reports, 7(5): 998-1012.
22. Assou, S., Girault, N., Plinet, M., Bouckenheimer, J., Sansac, C., Combe, M., et al. (2020) Recurrent genetic abnormalities in human pluripotent stem cells: definition and routine detection in culture supernatant by targeted droplet digital PCR. Stem Cell Reports, 14(1): 1-8.
23. European Medicines Agency, Committee for Advanced Therapies (2011) Reflection paper on stem cell-based medicinal products. EMA/CAT/571134/2009, adopted 14 January 2011, effective 14 January 2011. Jurisdiction: European Union.
24. US Food and Drug Administration, Center for Biologics Evaluation and Research (2008) Guidance for FDA reviewers and sponsors: content and review of chemistry, manufacturing, and control (CMC) information for human somatic cell therapy investigational new drug applications (INDs). CBER, April 2008. Jurisdiction: United States.
25. Sato, Y., Bando, H., Di Piazza, M., Gowing, G., Herberts, C., Jackman, S., et al. (2019) Tumorigenicity assessment of cell therapy products: the need for global consensus and points to consider. Cytotherapy, 21(11): 1095-1111.
26. Kuroda, T., Yasuda, S., Kusakawa, S., Hirata, N., Kanda, Y., Suzuki, K., et al. (2012) Highly sensitive in vitro methods for detection of residual undifferentiated cells in retinal pigment epithelial cells derived from human iPS cells. PLoS ONE, 7(5): e37342.
27. Kuroda, T., Yasuda, S., Matsuyama, S., Tano, K., Kusakawa, S., Sawa, Y., et al. (2015) Highly sensitive droplet digital PCR method for detection of residual undifferentiated cells in cardiomyocytes derived from human pluripotent stem cells. Regenerative Therapy, 2: 17-23.
28. Tano, K., Yasuda, S., Kuroda, T., Saito, H., Umezawa, A., Sato, Y. (2014) A novel in vitro method for detecting undifferentiated human pluripotent stem cells as impurities in cell therapy products using a highly efficient culture system. PLoS ONE, 9(10): e110496.
29. Allison, T.F., Andrews, P.W., Avior, Y., Barbaric, I., Benvenisty, N., Bock, C., et al. (2018) Assessment of established techniques to determine developmental and malignant potential of human pluripotent stem cells. Nature Communications, 9(1): 1925.
30. Li, R., Hornberger, K., Dutton, J.R., Hubel, A. (2020) Cryopreservation of human iPS cell aggregates in a DMSO-free solution - an optimization and comparative study. Frontiers in Bioengineering and Biotechnology, 8: 1.
31. Kim, J.-H., Kawase, E., Bharti, K., Karnieli, O., Arakawa, Y., Stacey, G. (2022) Perspectives on the cost of goods for hPSC banks for manufacture of cell therapies. npj Regenerative Medicine, 7(1): 54.
32. Taylor, C.J., Peacock, S., Chaudhry, A.N., Bradley, J.A., Bolton, E.M. (2012) Generating an iPSC bank for HLA-matched tissue transplantation based on known donor and recipient HLA types. Cell Stem Cell, 11(2): 147-152.
33. Xu, H., Wang, B., Ono, M., Kagita, A., Fujii, K., Sasakawa, N., et al. (2019) Targeted disruption of HLA genes via CRISPR-Cas9 generates iPSCs with enhanced immune compatibility. Cell Stem Cell, 24(4): 566-578.e7.
34. Deuse, T., Hu, X., Gravina, A., Wang, D., Tediashvili, G., De, C., et al. (2019) Hypoimmunogenic derivatives of induced pluripotent stem cells evade immune rejection in fully immunocompetent allogeneic recipients. Nature Biotechnology, 37(3): 252-258.