Automated Cell Culture Systems and Their Applications to Human Pluripotent Stem Cell Studies

Various Automated Platforms for Cell Culture

Several companies manufacture automated platforms. These incorporate robotic arms for liquid and labware handling, Class II biosafety cabinets to ensure sterile conditions for cell handling, and incubators providing precise control over temperature and CO₂ and O₂ levels. Many of these systems are highly modular and may be easily upgraded with additional devices, such as automated cell counters, imaging systems, centrifuges, cell sorters, PCR thermocyclers, plate readers, and thermostats. Currently available robotic platforms may be distinguished based on the type of labware used, that is, microplates or cell culture flasks. For example, the Cellerity and subsequent Tecan Integration Group (TIG) customized Freedom EVO (Tecan, Zürich, Switzerland), Cellhost (Hamilton Robotics, Reno, NV), and BioCel (Agilent, Santa Clara, CA) systems are designed for microplate applications. Cell Culture Automatic Laboratory (CCAL) employs Petaka bioreactors (Celartia, Columbus, OH), Biomek Cell Workstation (Beckman Coulter, Brea, CA), and CELLSTAR Autoflasks (Greiner, Monroe, NC) for handling flasks, while SelecT and CompacT SelecT platforms (TAP Biosystems, Hertfordshire, UK) may utilize either microplates or T-flasks ( Table 1 ). Another liquid handler was recently reported for the maintenance and passaging of human pluripotent stem cells (PSCs). This smaller system is based on a self-contained Gilson’s pipette Max liquid handler, and can accommodate up to 96-well plates. ⁷ Some limitations include the absence of a connected incubator and the requirement of offline steps and substantial human contribution. Each system has a different capacity for labware handling, offering a high degree of flexibility for various applications, with the emphasis on large-scale studies that would otherwise be difficult to undertake.

OPEN IN VIEWER

Table 1.

Labware Capacities.

	Type of Labware
	Plates (Wells)			Flasks
	96	384	1536	T-75	T-175	HYPER Flask
Cellerity/Freedom EVO a	860	860	1360	—	—	—
SelecT	420	420	420	182	182	182
CompacT SelecT	210	210	—	130	130	130
CompacT SelecT S.C.	210	210	—	90	90	90
BioCel	500	500	500	—	—	—
Cellhost b	42/189	42/189	56/252	—	—	—

a

When combined with LiCONIC STX1000 incubator.

b

When combined with Kendro Cytomat 6000 or Kendro Cytomat 2 C450 incubator.

All systems can accommodate 6- and 24-well plates, although their capacity is not specified.

Current Usage

Initially, automated cell culture systems were developed to validate their capabilities to replicate tasks that were principally performed by human operators. The successful outcome of such studies and rapid progress in the iPSC field have prompted the establishment of large-scale repositories of patient-specific iPSC lines derived from both disease-affected individuals and healthy controls. The overarching goal of these banks is to provide researchers with a reliable source of cells in order to facilitate studies of numerous conditions, including diabetes and cardiovascular and neurodegenerative diseases. McKernan and Watt recently described both existing banks and those under development. ⁸ Several of these initiatives plan to utilize robotic systems to achieve their goal. The New York Stem Cell Foundation (NYSCF) established in 2005 aims to generate 2500 human iPSC lines using a self-designed robotic platform composed of three connected platforms using STAR liquid handling systems (Hamilton Robotics).^8,9 This system remarkably demonstrated its potential in the automation of the entire process of iPSC production, including the automated culture of fibroblasts, reprogramming, the selection of iPSCs, maintenance, passaging, differentiation, thawing, and the assessment of pluripotency, differentiation, and karyotype. ⁹ The California Institute for Regenerative Medicine (CIRM), in collaboration with Cellular Dynamics International, Inc. (CDI), plans to generate 9000 iPSC lines from 3000 individuals from the United States with nonintegrating episomal vectors. ⁸ Several consortiums in Europe also aim to generate large biobanks of iPSCs. The European Bank for induced pluripotent Stem Cells (EBiSC) intends to bank more than 10,000 lines covering a wide range of diseases. ⁸ The StemCellFactory initiative’s objective is to automate production of iPSC lines, and their subsequent characterization and differentiation into neural and cardiac derivatives with a self-built system, ¹⁰ while I-Stem is an initiative that aims to automate the screening of compounds with therapeutic potential and subsequent disease modeling using the CompacT SelecT platform. Moreover, the AUTOSTEM Consortium is undertaking a project to expand the automated platform developed for the StemCellFactory project and utilize it for the production and banking of therapeutic stromal cells. ¹¹ Finally, our group aims to establish a large repository of patient-specific iPSCs and differentiated cells for ophthalmic research utilizing a Tecan Freedom EVO platform.^12–14

Maintenance

The manual maintenance of iPSCs introduces several limitations for transition into large-scale experiments. First, the maintenance of the iPSC culture to retain pluripotency and for directed differentiation protocols requires highly trained and experienced staff. Moreover, technician variability and human error pose major limitations when high numbers of samples are being processed in parallel. This variability also contributes to significant differences between cell lines generated and maintained in various laboratories,^15,16 which are exacerbated by factors affecting the reprogramming efficiency itself. These include the tissue of origin, ¹⁷ the transcription factors employed in the reprogramming cocktail, their stoichiometry, ¹⁸ the method of reprogramming, the genetic background, and the stress that cells are exposed to during their reprogramming and subsequent culture. ¹⁹ Overcoming these limitations by moving toward an automated process will allow the handling of a greater numbers of cells, which will in turn facilitate the optimization of existing methods and generation of new protocols for cell maintenance and directed differentiation. Furthermore, it will enable standardization of liquid handling and incubation times, which are two factors underlying the volatility of the manual approach. Therefore, the use of fully automated platforms for iPSC derivation, expansion, and differentiation may be key in transitioning to large-scale cell cultures. Cells derived from such cultures would provide a means for thorough disease modeling in vitro, as well as a broad range in toxicity screening of new chemical compounds, and pave the way for therapeutic cell replacement.^3,4

Several groups have already adopted an automated approach and thoroughly tested its suitability for everyday use in various applications. Initially, research in the field of automation was focused on assessing whether employing the automated platforms for culture maintenance would have any effect on cellular characteristics, for example, cell morphology or gene expression, comparing them with cultures that relied solely on manual handling. Joannides et al. explored the capability of automated mechanical passaging (AMP) to expand human ESCs. ²⁰ Their analysis revealed that AMP-expanded cells were more uniform in colony shape and colony numbers than the manual control. The stem cells also retained both morphological features and pluripotency marker expression, characteristic of pluripotent cells.

Successive groups examined both ESCs and iPSCs of human and rodent origin maintained on automated platforms in feeder-free conditions^9,14,21–23 or on a mouse embryonic fibroblast (MEF) feeder layer.^20,24–26 Paull and colleagues are the only study that report on the entire process of human iPSC generation, from fibroblast culture to iPSC maintenance. ⁹ They carefully examined gene expression profiles, revealing that automatically processed cells retained their round-shaped morphology and pluripotent characteristics, such as markers of pluripotency; for example, OCT4, TRA-1-60, TRA-1-81, NANOG, and SSEA-4 were expressed at the same or higher level compared with manual controls.^{9,20,21,24,26} Stem cells were also tested for any karyotypic anomalies that could be introduced during the cell culture itself—most groups reported that no anomalies occurred during the time of their experiments,^20,21,24,26 while one ⁹ identified aneuploidy in 11% of lines tested, that is, the acquisition of chromosome 22. This is not surprising given that chromosomal abnormalities are to be expected with reprogramming and culturing of PSCs. ²⁷ The variation observed in the number of chromosomal anomalies is most likely a reflection of the size of sample analyzed, or time of culture prior to analysis or the sensitivity of the analysis itself. For instance, Konagaya and colleagues analyzed one iPSC line by G-banding, ²⁶ Thomas et al. reported on two human ESC lines by G-banding, ²¹ and Paull et al. analyzed 38 iPSC lines using a NanoString karyotype assay, with improved sensitivity compared with traditional G-banding. ⁹ Of importance, three of the four abnormal lines tested by Paull and colleagues originated from the same fibroblasts and shared the same anomalies, suggesting that those originated within the fibroblasts and passed on, but did not arise from the automated reprogramming or maintenance. ⁹ The high-resolution analysis of copy number variant in iPSCs maintained manually or on the automated platform revealed a similar amount of variation, hence suggesting that continued assessment of cell genetic stability is essential. ⁹

Several automated platforms are also able to be adapted for processing other stem cell types. Some examples are detailed below. Thomas et al. successfully shifted the manufacturing of current good manufacturing practice (cGMP)–compliant human neural stem cell line CTX0E03, from manual to automated processes, maintaining the cell number and transcription profile according to strict quality control specifications, that is, level of expression of neural markers, together with cell survival and expansion rate. ²⁸ The CTX0E03 cell line is currently in Phase II clinical trials for the treatment of patients with ischemic stroke (NCT02117635). On the other hand, Kato et al. evaluated the possibility of expanding primary cell lines, that is, human fibroblasts and bone marrow stromal cells (BMSCs), using a benchtop platform with a footprint not exceeding 0.5 m². ²⁹ Cells obtained using the automated system displayed characteristics similar to those of manually cultured BMSCs and the ability of bone formation when transplanted into nude mice. Furthermore, Liu et al. tested the human osteosarcoma (HOS) cell line to identify sources of variation during cell expansion and the influence of residual trypsin on cell viability after passaging. ³⁰ Recently, a Biomek Cell Workstation was successfully used to cultivate and passage suspension cultures from four different cell lines, demonstrating the feasibility of this platform for automation of floating cell culture. ³¹

Passaging

Differentiation

Having established that the maintenance and passage of PSCs by automated platforms does not lead to changes in cellular morphology, gene expression, or karyotype anomalies, it is important to assess their retention of pluripotency and their capacity to differentiate. Several teams have analyzed the ability of rodent and human PSCs maintained on automated platforms to differentiate into the three germ layers. Each group observed that EBs generated by automation express differentiation markers from the three germ layers,^{9,21,22,24–26,38} while Soares et al. confirmed this ability in an adherent setting. ²³ Moreover, Kowalski et al. optimized an EB formation method in a 384-well microplate format for high-throughput cardiomyocyte differentiation, achieving an efficiency exceeding 40%. ⁶¹ The positive outcome of these experiments prompted further experiments differentiating cells into various lineages in monolayers. Automated platforms have been utilized to successfully differentiate stem cells into neural lineages, ²² including midbrain dopaminergic neurons;^9,26 neural stem cells and oligodendrocytes; ⁹ endodermal derivatives, such as hepatocytes, definitive endoderm, ⁹ and pancreatic islet cells; ²⁶ and cell types of mesodermal origin, such as cardiomyocytes and metanephric mesenchymal cells. ⁹ It is noteworthy that the efficiency of differentiation experiments with automated platforms was comparable to those performed manually,^9,26 while the intersample variation was noticeably lower.^9,22 Broader use of automation will facilitate optimization of differentiation protocols to increase the homogeneity of produced cells and also increase statistical power, allowing researchers to analyze low-penetrance phenotypes.

Microfluidic Platforms

Numerous groups have highlighted that successful implementation of cell-based therapies in the clinic requires thorough and high-throughput quality control testing to ensure efficacy and safety (e.g., Yamanaka, ⁶² Gonzalez et al., ⁶³ Andrews et al., ⁶⁴ Seifinejad et al. ⁶⁵ , and Miura et al. ⁶⁶ ). Microscale technology or microtechnology offers multiple improvements over the standard-sized tissue culture dishes. With microtechnology, it is possible to miniaturize assays and use microfluidic platforms to conduct multiple experiments simultaneously. Parallelization of the analysis decreases variability between samples, improving the consistency of acquired data. It also significantly reduces the amount of reagents required to conduct an experiment, and shortens analysis time, leading to obtaining results faster, for example, allowing the detection of HIV and syphilis infections within 20 min. ⁶⁷ Moreover, it provides an unprecedented precision in the control of the in vitro microenvironment. ⁶⁸ Indeed, as the in vitro environment is artificial, it is essential to optimize the settings to achieve near-native conditions. Microfluidics gives the ability to meticulously define cell seeding density to achieve the cell number required and to control tissue-specific spatiotemporal changes in concentrations of small molecules, which can be delivered to cells in femtoliter volumes. ⁶⁹ It is also feasible to organize cells into a characteristic three-dimensional structure to reproduce tissue complexity, for instance, using synthetic scaffolds and coating with extracellular membrane proteins to facilitate adhesion. ⁷⁰ Microfluidic platforms can be easily automated, and thus could eventually be used for the long-term cell culture without human intervention. This would be an advantageous feature over a conventional approach, as it would standardize the cell culture process. Furthermore, microfluidic platforms are characterized by low reagent consumption, which would be beneficial for experiments involving continued medium changing, often required with long-term cell cultures and differentiation experiments. ⁷¹ Microfluidic platforms could also enable the production of high-resolution live images of cell cultures, a useful feature not only for tracking cells throughout the experiment, but also for observing the effects of external stimulation with drugs or toxins on their survival.^70,72 Although these are key advantages of microfluidics, and some aspects of stem cell culture can be performed using them (reviewed in Wu et al. ⁷³ ), most of the current platforms remain experimental, and none have yet been successfully used for the long-term maintenance of stem cells.

Disease Modeling

Direct examination of human cells affected in disease is not always feasible due to their anatomical location; thus, there is often reliance upon isolation from postmortem tissues. A significant drawback of this is the fact that it generally allows observing the phenotype at the terminal stages of the disease, after aging and environmental impacts, which are difficult to control. Thus, postmortem tissues cannot be readily employed to investigate the development of disease and progression of disease pathophysiology. ⁷⁴ Human iPSCs offer a solution to overcome this problem, as these cells represent an unlimited source of cellular material for analysis. Furthermore, iPSCs have a genetic background identical to that of the person they were derived from; thus, generating iPSCs from those with genetic diseases or susceptibilities, and subsequently differentiating them to the pathological cell types, offers the possibility of directly examining dysfunction in cells with disease-specific genotypes and phenotypes. Utilizing such disease-affected cells to track molecular changes may give insight into genetic interactions and/or mechanisms that bear a disease risk or contribute to disease development. Indeed, numerous groups have successfully reported derivation of patient-specific iPSCs from monogenic disease or multifactorial disorders. ⁷⁵ However, modeling of the multigene disorders remains a great challenge because of the interplay between genetics and the environment. Additionally, the tissue-specific growth conditions and differentiation efficiency must be optimized first to ensure that cellular phenotypes displayed in vitro are in concordance with those observed in vivo. ⁷⁶ The automated approach will greatly facilitate optimization of both processes, allowing researchers to streamline experiments and thus test multiple combinations of conditions at the same time.

It is worth noting that modeling late-onset diseases using iPSCs may require additional modifications of cell culture conditions in order to more closely recapitulate the age-related disease phenotype. Cell reprogramming erases epigenetic marks of somatic cells, and at the same time, pluripotency-associated marks are created.^77,78 However, several groups have shown that reprogramming is a continuous process, and iPSCs at an early stage retain an epigenetic memory of the somatic cell of origin, which is gradually lost upon passaging.^79–81 Nevertheless, iPSC reprogramming results in the generation of rejuvenated cells that lost age-associated characteristics, such as shortened telomeres, decreased mitochondrial activity, and senescence markers,^82–84 and that also require a prolonged maturation to a functional state. ⁸⁵ To address this issue, fully differentiated iPSCs could be exposed to senescence-associated compounds, such as progerin ⁸⁵ or ceramide, ⁸⁶ to simulate aging in vitro. Interestingly, the use of progerin reactivates the expression of age-related markers that were lost during reprogramming. ⁸⁵ This strategy might not be sufficient to recapitulate all features of aging, but could allow capturing specific cell characteristics that would not be observed with fetal-like iPSC-derived progeny, hence improving our understanding of the mechanisms involved in the progression of late-onset diseases.

In addition, disease modeling with patient-specific iPSCs faces other obstacles. First, analysis between cell lines and between early and late passages may introduce variability and hinder the comparison of results. Second, many current studies employ samples from healthy individuals as controls for iPSC disease modeling. ⁷⁶ This approach may introduce unwanted variability between cells derived from different individuals. Therefore, recently there has been a strong push in the stem cell field to incorporate isogenic controls for iPSC disease modeling. This involves the use of genome editing technologies to correct mutations in the patient-derived iPSCs ⁸⁷ with the use of programmable endonucleases, including zinc-finger endonucleases (ZFNs), ⁸⁸ TALENs, ⁸⁹ or CRISPR/Cas9. ⁹⁰

Drug Screening

Human PSCs and their progeny could be used for drug screening, but also for toxicity assessment, using cells differentiated into those affected in disease, as well as from other tissues, in order to assess secondary side effects and toxicity. The large-scale use of patient cells for this type of approach would greatly benefit from automation of cell culture. For instance, McLaren et al. performed a medium-throughput screening of 1,000 compounds using human iPSC–derived neuroepithelial-like stem cells and identified 24 compounds promoting proliferation/survival. ⁹¹ Microfluidic platforms would be particularly useful, as those can capture single-cell responses to external stimuli and allow visualization behavior of rare cell populations that otherwise would be missed from analysis. ⁹² With the advantages offered by combining an automated approach with microfluidics, iPSCs could be used to conduct large-scale adsorption, distribution, metabolism, elimination, and toxicity (ADMET) studies of newly developed drugs. ⁹³

Despite many advantages, the two-dimensional system commonly used in the in vitro culture cannot fully recapitulate the process of development and cell differentiation occurring during organ development in vivo. The main reason behind this phenomenon may be a lack of proper cell–cell interactions that are normally required for tissue patterning during organogenesis. However, this issue may be addressed thanks to recently developed three-dimensional systems and use of organoids, ⁹⁴ which are composed of multiple cell types that self-organize into an organ-like structure. ⁹⁵ The first PSC-derived organoids were established by Sasai’s team, who generated three-dimensional cortical tissues from both mouse and human ESCs. ⁹⁶ Since then, organoids have been developed for multiple organs, including the brain, liver, heart, lung, pancreas, and kidney (reviewed in Huch and Koo ⁹⁴ and Lancaster and Knoblich ⁹⁵ ). Organoids are likely to mimic organ responses to external stimuli more accurately than monolayer cell culture, and therefore could prove beneficial for toxicity screening. ⁹⁵ Moreover, they have great potential to help unravel the mechanisms governing the development of human organs, such as the brain ⁹⁷ and the retina, ⁹⁸ and also malignant transformation. ⁹⁹ The use of automation would offer greater consistency of generated organoids, resulting in the increased reproducibility of this method, and help to elucidate the mechanisms involved in tissue development that may also play a role in disease progression.

Therapeutic Cell Replacement

The most anticipated application of human PSCs is their widespread use for cell replacement therapies. Both ESCs and iPSCs are considered potential sources of cells for transplantation;^100–102 however, patient-specific iPSCs have the advantage to match the genetic background of the potential patient. This holds great promise for regenerative medicine, as it would theoretically significantly reduce the risk of immune rejection. ¹⁰³ iPSC-based therapies have already been examined in rodent models of Parkinson’s disease ¹⁰⁴ and sickle cell anemia, ¹⁰⁵ where in vitro–derived cells engrafted into host tissues offered significant improvement of observed disease symptoms. Yet, several challenges need to be addressed before iPSC-based therapies will be successfully implemented in humans. First, the generation of iPSCs for clinical applications needs to be thoroughly assessed for safety, as the viral delivery of reprogramming factors poses a risk of malignant transformation of such cells. ⁶⁵ However, Sendai virus (SeV)–based vectors are nonintegrative and, at the same time, can transfect cells with high efficiency. ¹⁰⁶ The recent advances in technology have enabled virus-free generation of iPSCs with the use of proteins,^107,108 chemical compounds, ¹⁰⁹ nonintegrating episomal vectors, ¹¹⁰ or miRNAs. ¹¹¹ Moreover, much progress has been made to increase the efficiency of in vitro differentiation into the cell type of interest, but it is still far from obtaining a pure, homogenous cell population. Optimizing the efficacy of iPSC generation, their differentiation, and integrating methods for the removal of undifferentiated cells from the cell culture to cGMP standards will ultimately allow us to obtain high-quality biological material for safe and durable transplantation.