Automating Human Induced Pluripotent Stem Cell Culture and Differentiation of iPSC-derived Retinal Pigment Epithelium for Personalized Drug Testing

Abstract

Derivation and differentiation of human induced pluripotent stem cells (hiPSCs) provides the opportunity to generate medically important cell types from individual patients and patient populations for research and development of potential cell therapies. This technology allows disease modeling and drug screening to be carried out using diverse population cohorts and with more relevant cell phenotypes than can be accommodated using traditional immortalized cell lines. However, technical complexities in the culture and differentiation of hiPSCs, including lack of scale and standardization and prolonged experimental timelines, limit the adoption of this technology for many large scale studies, including personalized drug screening. The entry of reproducible end to end automated workflows for hiPSC culture, and differentiation, demonstrated on commercially available platforms, provides enhanced accessibility of this technology for both research laboratories and commercial pharmaceutical testing. Here we have utilized TECAN Fluent automated cell culture workstations to perform hiPSC culture and differentiation in a reproducible and scalable process to generate patient-derived retinal pigment epithelial cells for down-stream use including drug testing. hiPSCs derived from multiple donors with age-related macular degeneration (AMD) were introduced into our automated workflow and cell lines were cultured and differentiated into retinal pigment epithelium (RPE). Donor hiPSC-RPE lines were subsequently entered in an automated drug testing workflow to measure mitochondrial function after exposure to “mitoactive” compounds. This work demonstrates scalable, reproducible culture and differentiation of hiPSC lines from individuals on the TECAN Fluent platform and illustrates the potential for end-to-end automation of hiPSC-based personalized drug testing.

Introduction

Human induced pluripotent stem cells (hiPSCs)^{1, 2} can now be efficiently generated from multiple sources of somatic cells and be subsequently differentiated to produce various cell types of the human body. This technology is now beginning to realize significant promise as a tool for generating cells for research and potential cell therapies^3–6. A key outcome of the ability to generate pluripotent stem cells from any individual is that it enables experiments to be conducted for communities and patient populations as well as targeting research, drug discovery, disease modeling and regenerative medicine therapies for individuals, an approach termed personalized medicine^{7, 8}. However, there is a need to develop and disseminate protocols demonstrating the utility of automated processes for reproducible and cost efficient hiPSC culture and differentiation to build the scale required for projects that will have many individual donors.

Standardized reagents and protocols for hiPSC derivation and maintenance culture under conditions compliant for both good laboratory practice (GLP) and current good manufacturing practice (cGMP) are being rapidly adopted world-wide in academia and industry. Conventional adherent human pluripotent stem cell culture requiring feeder cells and media containing undefined serum or xenobiotic products have been superseded by feeder and xenobiotic-free, fully defined culture media and culture substrates. These conditions have enabled the development of robust, reproducible and economic methods for manual hiPSC culture and provide the basis for improved standardization of hiPSC culture, as well as making the culture of hiPSCs amenable to automation.

Although there have been advances in batch culturing of human pluripotent stem cells for scaled production of cells, most current differentiation protocols for hPSCs still require an adherent culture format for all or part of the process. The basis of adherent hPSC culture is repetitive liquid handling procedures involving aspirating and dispensing various volumes of media, media supplements or cell suspensions. The benefits of automated cell culture workflows include improving precision and reproducibility, reduction of labor costs and increasing experimental scale. In addition, proof of principle demonstrations of processes performed on commercially available equipment allows immediate dissemination of both methods and laboratory practice beyond the initiating institution. Automated cell culture paired with automated imaging allows culture assessment and data acquisition over extended time periods including both real time measurements and endpoint assays that can be integrated into workflows.

In this report, we have utilized commercial TECAN Fluent workstations to demonstrate automated culture of multiple hiPSC lines and automated differentiation of multiple iPSC lines into retinal pigment epithelium (RPE) and their application in a personalized drug testing paradigm. RPE is a single layer of pigmented cells at the base of the retina that supports the health and function of the light sensing retinal neurons^{9, 10}. Diseases and genetic conditions that compromise the health or function of the RPE can result in loss of retinal neurons and vision impairment¹¹. RPE generated from human pluripotent stem cells including iPSCs has been shown to recapitulate the phenotype and function of primary RPE and is now being used extensively in research including disease modelling and drug screening^12–14. RPE derived from human embryonic stem cells (ESCs) and iPSCs has also entered clinical trials around the world to test safety and efficacy as a cell replacement therapy for macular diseases^15–18. We have been building on previous reports of automating retinal differentiation from hiPSCs¹⁹ to convert manual protocols for iPSC and RPE differentiation and develop automated procedures for hiPSC culture and iPSC-RPE differentiation that use fully defined media, substrates and differentiation molecules and can be conducted on the new generation of commercially available robotic cell culture workstations. This work is designed to increase the scale of generating donor- or patient specific lines of iPSC-RPE for clinically relevant populations. Increasing the cost effective scaling of iPSC RPE derivation and drug testing will support the implementation of personalized drug screening and population based disease modeling for patients with AMD.

Materials and Methods

Tecan Fluent 780 Workstation

This work utilizes Tecan Fluent 780 robotic workstations, purchased by the Department of Ophthalmology and Visual Neurosciences and installed at the Laboratory for Stem Cell Automation located in the University of Minnesota Stem Cell Institute. The units comprise TECAN Fluent 780 base units and cabinets (TECAN, US, Morrisville, North Carolina) with HEPA and UV hoods (Bigneat Ltd, Waterlooville, Hampshire UK). The units are each customized with: 8 tip Air Flexible Channel Arm (FCA), Multiple Channel Arm (MCA) with the 96 well head adaptor; extended z-rail Robotic Gripper Arm (RGA) with finger exchange system (FES) eccentric fingers, FES centric fingers, FES tube fingers and the FES docking station; LiCONiC StoreX STX44 ICSA, Incubator linked by transporter; Integrated Hettich High Speed Robotic Centrifuge positioned below the deck level; two Echotherm^™ RIC20 Dry Baths (Torrey Pines Scientific) and a Cytation 1 high content imager (Biotek, Agilent Technologies, Santa Clara, California). Tecan software is Fluent Control Version 2.4.25.51907 (TECAN Trading Ltd., Switzerland).

Human induced pluripotent stem cell lines

The hiPSC lines used in this study are listed in Table 1.The derivation and characterization of hiPSCs lines from individuals with and without age related macular degeneration (AMD) utilized in this study has been previously described²⁰. Line UMN PCBC16iPS (lab designation 9–1) was previously derived from neonatal human dermal fibroblasts (ATCC PCS 201–010)²¹.

Table 1.

iPSC lines used in this study

iPSC Line	Donor			Cause of Death	Experiments
	Sex	Age	MGSa		Figures
UMN PCBC-16iPS (9–1)b	Male	-	-	NA	3
UMN MGS1-0237-3A3	Female	77	1	Anoxic brain injury	3,4
UMN MGS2-1747-1C1	Male	76	2	Sepsis	3,4
UMN MGS3-1775-6B4	Female	85	3	Breast cancer	3,4
UMN MGS3-0878-1B1	Male	79	3	Stroke	5
UMN MGS1-0698-1A2	Female	73	1	Pneumonia	5
UMN MGS1-0027-1A3+	Male	80	1	Brain hemorrhage	5
UMN MGS1-0027-1B3+	Male	80	1	Brain hemorrhage	3,4
UMN MGS3-1424-1A4#	Female	72	3	COPD	3,4
UMN MGS3-1424-2B1#	Female	72	3	COPD	3,4
UMN MGS3-1424-1A6	Female	84	3	Sepsis	3

^aMinnesota Grading System²⁶

^bDerived from primary neonatal dermal fibroblasts (ATCC PCS-201–010)²¹

^{+, #}Cell lines derived from the same donor

Manual hiPSC culture

Manual hiPSC culture was performed essentially as previously described²². Briefly, undifferentiated hiPSC cultures were maintained with Essential 8^™ Medium (ThermoFisher Scientific A1517001) on tissue culture dishes coated with recombinant human vitronectin (Peprotech AF-140–09). Cells were passaged using hypertonic citrate solution. Cells were released using gentle trituration with media and collected with centrifugation at 300 g for 5 minutes. Supernatant was discarded and cells re-suspended in fresh media. Media was changed daily.

Automated hiPSC culture on University of Minnesota Fluent 780 workstations

Custom TECAN Fluent Control scripts were optimized to automate the culture of hiPSC lines in a 6 well adherent format, adapting our manual culture system described in²⁰. For daily media exchange: The culture plate is retrieved from the Liconic incubator and transferred to the Cytation 1 imager where a preset Gen5 protocol is executed to image selected wells and reports the confluence to an excel file. The plate is then positioned on the tilt carrier. The plate lid and the media trough cover are relocated to the hotel and the tilt carrier moved to an angle of −5°. The FCA collects 1000μL filtered SBS pipette tips (Tecan 30057817) and media aspiration occurs from the lowest positon of each well. Used media is disposed in the liquid discard trough and used tips in the the trash and new tips are collected. Fresh media is collected from the media trough and dispensed into the wells, the plate lid and media trough cover are replaced and the tilt carrier rocks to distribute media. The tilt carrier then returns to 0° and the plate is moved to the transport position and returned to the incubator.

For passaging, a vitronectin coated plate is placed in the first position of the tilt carrier. The passaging script can passage a single well or two wells simultaneously. The plate designated for passage is retrieved from the incubator and placed onto the tilt carrier, where the lid is removed and the tilt carrier assumes an angle of −5°. The FCA uses 1000μL filtered SBS tips to aspirate media and add 1mL of hypertonic citrate solution to each well. The tilt carrier tilts back and forth to coat the well, the solution is aspirated from the wells, and another 1mL of hypertonic citrate solution is dispensed to each well. The plate is rocked and the lid replaced before the plate is transported to the 37 °C incubator for six minutes. During the incubation time 1 mL of media is dispensed into each well that will receive cells. After incubation the plate is recalled from the incubator and 1000μL filtered wide bore SBS tips (Tecan 30115239) are used to detach and collect the cells. Cell aspiration uses a designed liquid handling microscript that uses aspiration and dispensing twelve times in a pattern across each well to detach cells before aspirating from the lowest point of the well. The cell suspension is dispensed into a 15mL conical tube and 1mL of media is dispensed into each well (to disperse any remaining adhered cells) using a modified microscript with six aspirations and dispensing steps across each well before the media is added to the cell suspension. An additional 3 mL of media are then also added into the 15mL tube. Tubes containing the cell suspension and balance tubes if needed are moved with the RGA arm into the below-deck centrifuge, set in opposing positions, and centrifuged at 800 g for three minutes before retrieval and repositioning on the tube rack. Using 1000μL filtered SBS tips, the FCA arm removes the supernatant from the 15mL tube aspirating 5.5 mL and discards the media in the liquid discard trough. After equipping new tips (six standard and one wide bore), a total of 6 mL of media is added incrementally onto the cell pellet. First, 1 mL is added, then the wide bore tip mixes three times to disrupt the pellet. This is followed by an additional 2 mL, where the mixing process is repeated and repeated again when the final 3 mL are added. At this point, the cell pellet is fully distributed throughout the media. The wide bore tip then aspirates a set volume of the cell mixture and distributes it into wells on the new plate. The split ratio is user defined for cell line maintenance, expansion or distribution for differentiation. The lids are placed back onto the plates and the media trough and the new plate is moved onto the transport position. Before being returned to the incubator, the RGA arm lifts and shakes the plate three times in the x-direction, twice in the y-direction, then twice more in the x-direction.

Automated RPE Differentiation of hiPSC lines

The 14 day RPE differentiation of hiPSC lines was conducted essentially as described previously^{20, 23} utilizing the TECAN workstation to exchange the media components automatically over the 2 week differentiation period for 5 hiPSC lines simultaneously. iPS cells were plated on vitronectin coated plates and starting on day 1 of differentiation the media components were exchanged daily. Stock solutions of the reagents for RPE differentiation were prepared and loaded onto the Tecan workstation distributed in V-Bottom 96 well plates (Sarstedt, 82.1583.001). The plates were kept frozen at −20 °C on the Echotherm^™ RIC20 Dry Baths. Basal neural induction media (DMEM/F12 with 1x N2 supplement (LifeTechnologies 17502–048), 1x B27 supplement (LifeTechnologies 17504–044) and 1x Non Essential Amino acids) was prepared by diluting the supplements in DMEM/F12 media. This media was used to thaw and re-suspend the appropriate reagents on each day of the differentiation The reagents were compounded daily in a 15 mL conical tube and DMEM/F12 was added to achieve the correct media concentration. The media was then changed on the appropriate wells. On days 1 to 4 the media was supplemented with 10 mM Nicotinamide (N0636 Sigma-Aldrich), 50 ng/ml Noggin (6057-NG R&D Systems), 10 ng/mL Dkk-1 (5439-DK R&D Systems) and 10 ng/mL IGF-1 (291-G1 R&D Systems). On days 3 and 4 the Noggin concentration was reduced to 10 ng/ml and bFGF was added at 5 ng/ml (R&D Systems). On days 5 and 6 the basal medium was supplemented with 10 ng/mL Dkk-1 (5439-DK R&D Systems), 10 ng/mL IGF-1(291-G1 R&D Systems) and 100 ng/mL Activin A (AF 120 14E PeproTech). Then from day7 to day 14 the basal medium was supplemented with 100 ng/mL Activin A and 10 μM SU5402 (SML0443 Sigma-Aldrich) with 3 μM CHIR 99021 (4423, Tocris, Bristol, UK.) added in addition from day 8 to day 14.

Expansion and maturation of iPSC-RPE

14 day RPE derived from hiPSCs was cultured with XVIVO-10 media (04–743Q, Lonza) on 6-well plates (CC7682–7506, CytoOne, USA Scientific) treated with vitronectin XF (07180, Stem Cell Technologies). Passaging was carried out by exposure to Accumax (A7089, Sigma-Aldrich) dissociation reagent for 25 minutes at 37°C and the cells detached with a cell scraper and centrifuged for 5 minutes at 200g. Cells were re-suspended in XVIVO-10 containing 10 μM ROCK inhibitor Y27632 (10005583, Cayman Chemical) with subsequent media changes of XVIVO-10 without inhibitor. Cells were passed through a 40 μm cell strainer (07-201-430, Corning) before plating at 1:3–1:4 split ratios (1 × 10⁵ cells/cm²). Cells were incubated at 37 °C with 5% CO₂ with twice weekly media changes for at least 30 days.

Workstation Consumables

Workstation consumable use is dependent on the workflow program. Pipetting steps using the 8 tip FCA utilized ANSI format, filtered, conductive tips in 1000μL (Tecan; 30057817) and 10μL (Tecan, 30104974) sizes. Autoclaved 1000μL wide bore (Tecan, 30115239) tips were primarily used in steps involving cell suspension pipetting. Pipetting with the MCA96 utilized nested, sterile, non-filtered, 200μL tips (Tecan; 30038619). Disposable tubes and troughs were utilized on the worktable including 15 mL conical tubes (Sarstedt, 62.554.002) and 300mL and 100mL media troughs (Tecan,; 10613048, 30077312), Autoclavable custom lids were manufactured in house for these. Cells for this work were grown in three plate formats, 6-well (CytoOne, CC7682–7506), 96-well V-bottom (Sarstedt, 82.1583.001) and Seahorse XF96 Cell Culture Microplates (Agilent, 101085–004).

Immunocytochemistry

Cells were fixed in buffered formalin (10% w/v, Fisher Scientific 23-305-510) for 10 min at room temperature, permeabilized for 10 minutes at room temperature in DPBS with 0.2% v/v Triton X-100 (Millipore Sigma, T8787) and blocked in blocking buffer (DPBS with 1% w/v bovine serum albumin (BSA) (Millipore Sigma A3059) and 0.1% v/v Tween-20 (Millipore Sigma, P1379) for 1 hour. Cells were then incubated with primary antibodies diluted in blocking buffer overnight at 4 °C. Cultures were washed once with blocking solution and incubated 1 h with secondary antibodies diluted in blocking buffer. Nuclei were labelled with DAPI (4′,6-diamidino-2-phenylindole dilactate) (ThermoFisher D3571) for 10 minutes before washing two times in DPBS. Primary antibodies used: NANOG (1:100, BioTechne AF1997), OCT4 (1:500, Millipore Sigma MAB4401MI), PMEL17 (1:200, DAKO M063429–2), OTX2 (1:250, Abcam ab21990), ZO1 (1:500, Invitrogen 61–7300). Secondary antibodies used were Alexa Fluor 488 Donkey anti-Mouse (1:500, ThermoFisher Scientific A21202), Alexa Fluor 488 Donkey anti-Goat (1:500, ThermoFisher Scientific A-11055), Alexa Fluor 488 Donkey anti-Rabbit (1:500, ThermoFisher Scientific A-21206), Alexa Fluor 555 Donkey anti-Mouse (1:500, ThermoFisher Scientific A-31570). Negative controls included untreated cultures and cultures incubated only with secondary antibodies.

Automated imaging and image analysis

Cells were imaged using the Cytation1 and analyzed using the Gen5^™ data analysis software (Agilent Technologies, Biotek). The Cytation1 was equipped with 10x and 20x objective lenses and GFP and DAPI imaging filter cubes. For each well, 64 images were taken in an 8×8 square with each picture spaced 1500 μm apart. For each channel capture, exposure was set to “Auto” and kept consistent between samples. Unstained cultures or cultures with secondary antibodies only were used to establish negative expression. .Each image, was autofocused using the DAPI channel within a 10 micron offset for the fluorescent image from the appropriate secondary antibody images. For ZO1 imaging, using the Gen5 software, a primary mask was applied to the image to determine the center of each cell marked by DAPI and a secondary mask using the fluorescent image of the secondary antibody. For nuclear epitope imaging (OCT4. NANOG, OTX2) the DAPI primary mask was applied and a subpopulation was created for all cells where expression in the appropriate secondary antibody channel was also detected.

“Mitoactive” Drug testing

18 compounds were selected from literature reports of mechanisms of action that improved mitochondrial function (see Table 2). Aliquots of concentrated stock solutions were prepared and stored at −20 °C. iPSC-derived RPE cells were seeded into 96-well Seahorse plate using the Fluent workstation at a density of 4×10⁴ cells/well in a 96 well plate and cultured in MEM alpha medium (Gibco) containing 5% FBS (Hyclone), sodium pyruvate, 1X Glutamax, 1X non-essential amino acids, 1X N1 supplement, taurine (0.25 mg/mL), hydrocortisone (0.02 μg/mL) and 1X Pen Strep at 5% CO₂ and 37 °C. Medium was switched to 1% FBS 2 days prior to treatment. On the day of screening the Fluent workstation was utilized to prepare 500 μM solutions of each drug by adding media to the frozen stock plates (N-Acetyl-L-cysteine was prepared fresh) (See Figure S1C) and cells were treated with each drug at a final concentration of 10 μM for 2 days in MEM alpha medium with 1% FBS prior to Seahorse assay.

Table 2.

Compounds used in this study

Compound		Manufacturer	Cat #	Solvent	Stock	Ref
AICAR	AIC	Sigma-Aldrich	A9978	H2O	10 mM	30
Amiodarone	AMI	Sigma-Aldrich	A8423	H2O	500 mM	32
Carbamazepine	CAR	Sigma-Aldrich	C4024–1G	DMSO	20 mM	32
7 hydroxy 4H chromene	CHR	sigma-Aldrich	683949	DMSO	500 mM	30
Clonidine	CLO	Sigma-Aldrich	C7897	H2O	500 mM	32
DETA-NO	DETA	Sigma-Aldrich	D185	H2O	500 mM	30
Diazoxide	DIA	Sigma-Aldrich	D9035	DMSO	250 mM	34
Fluspirilene	FLU	Sigma-Aldrich	F100	DMSO	42 mM	32
Gemfibrozil	GEM	Sigma-Aldrich	91823	DMSO	1 M	30
GW501516	GW50	Sigma-Aldrich	SML1491	DMSO	10 mM	30
Lithium	LIT	Sigma-Aldrich	L4408	H2O	1 M	32
Metformin	MET	Sigma-Aldrich	PHR1084	H2O	100 mM	39
N-Acetyl-L-cysteine	NAC	Sigma-Aldrich	A9165	H2O	Fresh	33
Papaverine	PAP	Sigma-Aldrich	P3510	H2O	100 mM	31, 37
Pyrrolo-quinoline-quinone	PQQ	Sigma-Aldrich	80198	H2O	5 mM	36
Quercetin	QUE	Sigma-Aldrich	Q4951	DMSO	500 mM	35
Rapamycin	RAP	Sigma-Aldrich	553211	DMSO	50 mM	38
Resveratrol	RES	Sigma-Aldrich	R5010	DMSO	500 mM	30, 35

Measuring RPE mitochondrial function using the Seahorse Cell Mito Stress Test (CMST).

The Mitochondrial function of iPSC-RPE was measured using an XFe96 Extracellular Flux Analyzer (Agilent Technologies) and the Cell Mito Stress Test (CMST) assay conditions described previously²⁴. After incubating iPSC-RPE cells with drugs or solvent only (control) for 48 hrs, cells were washed (2X) with CMST assay medium (XF base medium DMEM supplemented with 2 mM glutamine, 5.5 mM glucose, and 1 mM sodium pyruvate, pH 7.4), and then incubated in CMST medium for 1 h at 37°C in a non-CO₂ incubator. The CMST assay protocol was performed according to the manufacturer’s instructions (Agilent Technologies). Oxygen consumption rate (OCR) was detected in cells pre-treated with either drugs or solvent (control) followed by the sequential addition of oligomycin (2 μM), FCCP (1 μm), and finally rotenone (1 μM) and antimycin A (1 μM). The resultant measurement of OCR allowed for calculation of: basal respiration, ATP production and maximal respiration. Hoechst dye was added in the third and final injection to enable post assay cell count at 10X magnification using a Cytation 1 imager. Data processing used Wave software version 2.6.1.56 (Agilent Technologies) normalizing OCR to cell count.

Results

Description of the TECAN Fluent Hardware Platform

The goal of this report is to demonstrate that commercially available automated workstations can be used to perform complex cell culture workflows and functional assays with hiPSCs. The University of Minnesota (UMN) Laboratory for Stem Cell Automation at the University of Minnesota Stem Cell Institute (Minneapolis, MN) is equipped with two TECAN Fluent 780 units. The Fluent units can be customized to meet customer specifications with multiple optional components and a wide range of peripheral instruments that can be integrated with the integral workstation components and managed with the TECAN Fluent control software. Figure 1 shows the layout of Fluent 780 workstations installed at the UMN Laboratory for Stem Cell Automation. The workstation layout and range of optional peripheral instruments was planned and integrated in consultation with TECAN engineers to provide a flexible workstation designed to be capable of a wide range of mammalian cell culture procedures with the inclusion of imaging, centrifugation and long-term cell incubation (Figure 1A). The UMN Fluent 780 worktable (78 × 165 cm²) (Figure 1A) has a tilt carrier with locations for up to four plates for controlled angled tilting of plates and shaking. Two EchoTherm dry bath devices provide temperature control between −20 °C and 110 °C. A storage area is available for short term plate and lid storage. Optional trough runners are available for 100 mL troughs, 50 and 15mL conical tubes and micro-centrifuge tubes. About 25% of the deck space is available to pre-position racked pipette tips. Each UMN workstation is equipped with a Liconic STX44 automated incubator programmed to maintain 37°C, 85% humidity, and 5% CO₂ where two interchangeable incubator racks each hold either 11 tissue culture plates of 6,12,24 or 48 well formats or 22 96 well plates for a total cell culture surface area of approximately 1250 cm². A Cytation1 imager utilizing Gen5 software is integrated for bright field and fluorescent image acquisition and analysis. The UMN Fluent 780 machines are equipped with three robotic arms (Figure 1B). The 8-channel air displacement liquid handling arm (air LiHa) allows pipetting of liquids at a volume range from 0.5 to 1000 μL per channel. The MultiChannel Arm (MCA) is equipped with a 96 channel pipetting head adaptor that allows simultaneous pipetting of liquids into microplate formats. The extended Robotic Manipulator Arm (RoMa) is able to transfer plates and other labware on the worktable and to peripheral devices, including transferring tubes to and from the high speed Hettich robotic centrifuge situated below the worktable. An integrated laminar flow HEPA hood and workspace enclosure maintains air quality to reduce contamination of deck components.

Figure 1. TECAN Fluent 780 workstations at the UMN Laboratory for Stem Cell Automation.

A) Diagram of worktable layout: 1,2. Staging area for racked pipette tips. 3,4. EchoTherm dry baths. 5. MultiChannel Arm (MCA Nest for 96 and 384 well MCA adaptor heads. 6. Liquid Waste Reservoir. 7. Elevated stage for tips, reservoirs and plates. 8. Tilt carrier. 9. Tube runners. 10. on-deck storage area. 11. Storage area for adaptors for extended robotic manipulator arm. 12. Transport for labware to shuttle between incubator and workspace. 13. Cytation 1 imager. 14. Centrifuge (below deck). B) (i) External view of TECAN fluent workstations showing the enclosed work surface and external computer controller and Cytation 1 imager. (ii) Close up of the worksurface front view and (iii) side view showing 3 arms. 1. 8 channel, adjustable width, air displacement liquid handling arm (air LiHa). 2. MultiChannel arm with 96 channel pipetting head adaptor (MCA) 3. Extended Robotic Gripper Arm (RGA) (iv) Close view of RGA finger exchange area. 4. RGA eccentric fingers adaptor. 5. RGA centric finger adaptor. 6. RGA tube gripper adaptor

Building an automated process to derive iPSC-RPE for personalized drug screening.

A workflow is being developed using automation where possible and advantageous, to generate a scalable process to derive iPSC lines, differentiate and utilize iPSC-RPE from multiple individual donors. Our research has utilized somatic conjunctival cells obtained by manual biopsy. The conjunctival cells grown from the donor biopsies can be manually reprogrammed into iPSCs and differentiated into iPSC-RPE using fully defined manual methods described previously²⁰. We are now generating a suite of Fluent control scripts (Figure 2) that could be linked to form an end to end process automating iPSC derivation, culture and RPE differentiation. When combined with downstream drug screening this work is intended to increase the scale at which we are able generate personalized drug screening data for multiple individuals. Processes we have now successfully demonstrated using automation are shown in green in Figure 2 and we are now working to link the automated processes where this advantageous and appropriate to improve the workflow. In this current study we automated iPSC line expansion for banking (not shown) and consecutive seeding of undifferentiated iPSCs for automated RPE differentiation. Separately iPSC-RPE lines were plated, cultured and exposed to drug compounds using the automated workstations prior to determining oxygen consumption rate (OCR) using the Agilent Seahorse XFe96 Flux Analyzer (see figure 5). Further details of the automated workflows are shown in Supplemental Figure 1

Figure 2. Building Automation for the processes involved in the production of iPSC-RPE derived from epithelial cultures of conjunctival biopsies from human donor eyebank tissue for the purpose of drug testing.

Processes shown in the Green blocks are fully automated. Automation of processes in blue are currently under development. Full details of each step in the process is shown in Supplementary Figure 1.

Figure 5. Donor specific drug testing with hiPSC-RPE.

hiPSC-RPE lines from three different donors with (MGS3) or without (MGS1) AMD were cultured in the presence of active compounds or solute vehicle for 48 hours before oxygen consumption rates were determined using a Seahorse XF Flux Analyzer. Results calculated for basal respiration, maximal respiration and ATP production for each cell line after exposure to each drug are presented as the ratio of the calculated values determined for treated cells/vehicle. Results for two separate experiments are presented for each cell line (Squares). Abbreviation for each drug is shown at the bottom. Responses to Clonidine (Green), Metformin (Red) and Amiodarone (Blue) are highlighted for each line.

Human Induced Pluripotent Stem Cell Maintenance Culture

A component of all processes using human iPSCs is the maintenance of undifferentiated iPSC lines. This step is critical for successful hiPSC line derivation, undifferentiated iPSC line maintenance during expansion and characterization pre-banking, recovery post-cryopreservation and in preparation for differentiation protocols. Using the TECAN Fluent workstations described above we simultaneously cultured eight different hiPSC lines over expansion and passage cycles using conditions designed to maintain iPSCs with undifferentiated phenotypes (Figure 3). We adapted previously published c-GMP compliant defined conditions for manual adherent hiPSC culture^{22, 25} for the TECAN system. For the demonstration described in Figure 3, cells from 8 different previously derived iPSC lines were thawed and cultured in defined Essential 8^™ media on plates coated with recombinant vitronectin with complete media changes every day. Using automation each line was simultaneously passaged every 3 days with hypertonic citrate reagent using workflows controlled by custom TECAN Fluent control software scripts (see Supplementary Figure 1A). Cells were initially seeded to achieve a 24 hour post-plating colony confluency between 5–20% of available culture area to maintain discrete colony isolation and expanded in culture with daily media changes for 2 further days before passage. Five consecutive passage cycles were performed with brightfield imaging of each iPSC line culture conducted each day at the time of media exchange to monitor culture appearance, colony growth and confluence. Representative bright field images of expanding iPSC colonies for one line during one passage cycle are shown in Figure 3A. Brightfield images and confluence data are stored digitally and available to the workstation user. Culture confluence for the different iPSC lines cultured simultaneously for 15 days over 5 passage cycles is shown in (Figure 3B). This monitoring of growth kinetics can be used to modify workflow script parameters for individual cell lines as necessary during culture runs. To determine if each cell line had maintained an undifferentiated phenotype following automated culture, immunohistochemistry was used to detect expression of the pluripotency associated transcription factors NANOG and OCT4 in cells at Passage 1 and Passage 5 (Figure 3C). Quantification of NANOG and OCT4 expression indicated that each cell line retained nuclear expression of these pluripotency associated proteins, consistent with maintaining an undifferentiated phenotype.

Figure 3. Simultaneous automated culture of undifferentiated hiPSC lines.

A) Representative brightfield images of adherent hiPSC colonies from UMN MGS1-0237-3A3, one of the lines expanded using automation over one passage cycle. B) Colony confluence for 8 different hiPSC lines cultured simultaneously and measured over five consecutive automated passage cycles. C) Expression of pluripotency associated proteins OCT4 and NANOG was determined by immunohistochemistry at Passage 1 and Passage 5. Representative images for line UMN MGS1-0237-3A3 are shown together with epitope expression quantified for each line. The percentage of cells with epitope expression was determined by comparing the total number of nuclei detected with the number of nuclei where co-expression of fluorescent secondary antibody was detected. Scale bar 100μm

Automated differentiation of hiPSC into retinal pigment epithelium

After successfully automating the undifferentiated culture of multiple hiPSC lines, we extended the workflow to demonstrate automated differentiation of 6 of the hiPSC lines into RPE (Figure 4). We utilized a published 14 day defined differentiation protocol²³ that we had previously successfully implemented in manual culture²⁰. The protocol utilizes timed addition of growth factors and small molecules to adherent cells and is suitable for transition to automated simultaneous differentiation of multiple hiPSC lines (Figure 4A). Aliquots of growth factor small molecules appropriate for each timed addition to the culture wells were prepared, aliquoted and frozen in a light protected 96 well plate that was positioned on one of the temperature-controlled deck positions. For this demonstration hiPSCs from six selected lines were passaged onto vitronectin coated wells and cultured in Essential 8 media in order to maintain an undifferentiated phenotype until differentiation was initiated. On each day of the differentiation the appropriate aliquots of differentiation factors were re-suspended by addition of media and required volumes introduced to cells during the daily media change as illustrated in Figure 4A. After day 14, differentiated cells from each line were fixed for analysis by immunohistochemistry (Figure 4B) or continued in manual culture in X-Vivo10 media as described previously²⁰. Consistent with published reports of manual differentiation using this protocol, hiPSCs differentiated using the automated differentiation workflow were seen by day 14 to have formed a cobblestone monolayer morphology with cells expressing the tight junction protein ZO1, homeobox transcription factor OTX2 and pre-melanosome protein PMEL17 (Figure 4C). These characteristics are indicative of the transition to RPE phenotype. As published previously^{20, 23}, functional RPE can be maintained following extended culture in defined X-VIVO10 media.

Figure 4. Automated differentiation of hiPSC lines into retinal pigment epithelium.

For this experiment six different hiPSC lines were entered into the automated iPSC-RPE differentiation workflow on the TECAN workstation. A) Diagram showing the sequence of differentiation media and growth factor and small molecule additions that were automatically implemented over the 14 day protocol before expansion. B) Brightfield imaging and fluorescent Immunohistochemistry at d14 post differentiation indicated that the cells had adopted morphology and protein expression appropriate for early RPE differentiation. Representative images for line UMN MGS2-1747-1C1 are shown. Scale bars 100μm C) Quantification of protein expression for ZO1, OTX2 and PMEL17 for each differentiated hiPSC line.

Drug testing with donor-specific iPSC-RPE

The TECAN Fluent workstation is especially capable of executing in vitro drug testing and screening protocols involving exposing adherent cells to different compounds. These protocols require the plating of cells in appropriate formats, controlled exposure to soluble compounds and an appropriate functional test. We have previously utilized metabolic analysis determining oxygen consumption rate (OCR) using the Agilent Seahorse XFe96 Flux Analyzer to examine changes in mitochondrial function in primary RPE from donors with and without AMD²⁴. To extend this work to iPSC-RPE derived from multiple individual donors graded for the stage of AMD²⁶, we employed the TECAN Fluent workstation to conduct iPSC-RPE cell culture, plating and compound addition for subsequent functional analysis after exposure to different drugs. iPSC-RPE differentiated from three different individual donors was plated in 96 well format and cultured for 48 hours before exposure to a targeted library of 18 compounds shown previously to affect mitochondrial metabolism (Table 2). The cells were exposed to the compounds at a standard 10μM final concentration for 2 days before analysis (Figure 5). Three metabolic measurements (basal respiration, maximal respiration and ATP production) were calculated from direct measurement of OCR in the Seahorse Flux Analyzer. For each compound the mean OCR from duplicate experiments performed on different days is shown as the ratio between cells exposed to the compound and cells exposed to the solute vehicle for three different cell lines. Drugs with a response over 1 were deemed beneficial.

Results from each of the three donor lines are shown arranged from lowest to highest response to drug exposure for basal respiration, maximal respiration, and ATP production (Figure 5) The raw Seahorse OCR data for these experiments is available from the authors on request. Overall, donor lines 1B1 and 1A3 were more responsive to drug exposure than donor line 1A2, where the response to all drugs was minimal. Each donor line exhibited an individualized response to the different drugs. For example, Clonidine (Green boxes in Figure 5), an α2-adrenoreceptor agonist previously used to treat glaucoma, had a positive effect on basal respiration and ATP production in donor lines 1B1 and 1A3, but had an unfavorable response in donor line 1A2. Exposure to Amiodarone (blue boxes in Fig 5) demonstrated improvement of basal respiration and ATP production in line 1B1 and basal respiration in line 1A3 but had the most detrimental reduction on all three measured parameters for line 1A3. Of note, some of the drugs also seemed to have a differential effect on the mitochondrial parameters in individual lines. For example, in considering the mitochondrial response to Metformin (red boxes in Fig 5) ATP production was slightly impacted in line 1B1 but improvements in both basal and maximal respiration were observed for this line.

Although results from only three donors are shown here, similar variability of response to drugs by individual iPSC-RPE lines has been noted in other studies^{14, 27}. We have also seen a variable drug response in primary RPE cultures from donors with AMD²⁸.This highlights the importance of testing the drug response of RPE derived from individual donors and to accomplish this task at a clinically relevant scale will require the adoption of automated processes such as demonstrated here.

Discussion

The successful deployment of automated robotic cell culture systems will be critical to realizing the clinical and commercial potential of human pluripotent stem cell technology. Cell reprogramming to generate donor specific iPSC lines, that can then be differentiated into specific cell types, provides the ability to pursue disease modelling, drug testing or therapeutic applications at the level of both the individual or large populations in a manner that has not previously been possible. Technological improvements in the culture and differentiation of human pluripotent stem cells, with the introduction of defined culture media, substrates and small molecule differentiation reagents, has allowed for the design of reproducible protocols that can be converted to workflows for the new generations of cell culture workstations that have begun to enter the commercial sector. The successful use of these machines can reduce the cost and number of skilled bench technicians, improve reproducibility and greatly increase sample capacity. This current work demonstrates the use of the Tecan Fluent automated cell culture platform in conducting maintenance culture, differentiation and a drug screening protocol using human induced pluripotent stem cells. The combination of a fully defined iPSC maintenance culture system with a defined RPE differentiation protocol allowed us to demonstrate the generation of RPE from multiple individual donor iPSC lines for use in a downstream drug screening process. This successful validation of a standardized, hiPSC culture and RPE differentiation process demonstrates a practical approach for using commercial automation equipment for applications requiring the cost-effective generation of RPE from multiple individuals.

Use of automated systems for human pluripotent stem cell culture and differentiation is beginning to gain traction as the potential to increase scale and reproducibility whilst reducing personnel costs becomes apparent. A number of both custom built and commercial machines have now been demonstrated (eg, NYSF Global Stem Cell Array, TAP Biosystems, Beckmann Coulter I series and others) for various applications including pioneering work deriving hiPSC-derived retinal cells on the TECAN Freedom EVO platform¹⁹ or differentiating and expanding RPE from hESCs in the flask based CompacT SelecT automation platform from Sartorius²⁹.

For our current application we are developing automated protocols to assist research that includes the derivation of hiPSC lines from multiple individuals both with and without AMD and the subsequent differentiation of these cells into RPE for disease modelling and drug screening. We have adapted manual culture and differentiation processes that use fully defined reagents and substrates to build automated protocols that can be executed in plate based adherent workflows on the TECAN Fluent platform. The hiPSC maintenance and expansion protocol employing Essential 8 media and recombinant vitronectin coated plates, combined with passaging with hypertonic citrate results in a straightforward protocol that is highly reproducible between cell lines. We utilized a 3 day passage cycle for simplicity where the daily confluence measurement allows the user to adjust spit ratios on passage to maintain optimal confluence. Multiple cell lines can be cultured simultaneously and this allows coordinated initiation of RPE differentiation for batches of donor lines. The rapid RPE differentiation process was initially described by Buchholz et al in 2013 and efficiently differentiates iPSC-RPE from most hiPSC lines tested in 14 days. These cells are expanded and matured over 30 day passage cycles in defined X-Vivo10 media. Using the 6 well plate format adopted here and seeding undifferentiated iPSCs at 15–30% confluency prior to RPE differentiation generated approximately 15 × 10⁶ iPSC-RPE cells at P0 for expansion. Each 96 well Seahorse plate requires ~ 4 × 10⁵ iPSC RPE so this scale is appropriate to provide sufficient cells for repeated drug screening and any subsequent downstream analysis for each donor.

Combining the ability to maintain undifferentiated hiPSC cultures with efficient iPSC differentiation that can be carried out in a convenient, cost effective adherent plate format, without the need for extensive adjustments between different donor lines, provides the basis for a scalable process to generate hiPSC-RPE from large patient populations. In contrast to the bulk expansion of cell numbers that are necessary to build large banks of allogeneic cell products for potential therapeutic applications, our population based drug screening protocol uses donor specific RPE and requires expanding the scale of the number of individual lines that can be processed efficiently and cost effectively. The plate format we have adopted reduces the cost of culture and differentiation reagents while allowing the expansion of sufficient cells for the drug screening process. We are now employing our workflows to generate RPE from a large bank of iPSC lines we have generated from patients and donors with and without dry AMD.

The ability to develop and demonstrate the utility of protocols for hiPSC culture and differentiation on commercially available automated workstations without having to design or modify lab-specific equipment has enabled the introduction of automation to a complex experimental process in our academic laboratory. We are able to thaw multiple iPSC lines simultaneously and maintain them in undifferentiated culture, allowing seeding for simultaneous RPE differentiation, without manual handling. Currently RPE expansion and maintenance, is conducted manually due to space limitations but the recurring tasks of plating the iPSC-RPE lines in multiwell plates and culturing them with timed exposure to drugs prior to Seahorse analysis, requiring repetitive accuracy and reproducibility are conducted with automation. We have worked to design and test these automated protocols to replace manual methods for discrete parts of the iPSC maintenance, RPE differentiation and testing workflow. Linking these parts together to generate an end to end automated process still requires further development of scheduling software currently underway. Further information of our experiences with the workstations together with details of the control scripts are available on request form the authors.