Feeder-free culture of human pluripotent stem cells drives MDM4-mediated gain of chromosome 1q

Summary Culture-acquired variants in human pluripotent stem cells (hPSCs) hinder their applications in research and clinic. However, the mechanisms that underpin selection of variants remain unclear. Here, through analysis of comprehensive karyotyping datasets from over 23,000 hPSC cultures of more than 1,500 lines, we explored how culture conditions shape variant selection. Strikingly, we identified an association of chromosome 1q gains with feeder-free cultures and noted a rise in its prevalence in recent years, coinciding with increased usage of feeder-free regimens. Competition experiments of multiple isogenic lines with and without a chromosome 1q gain confirmed that 1q variants have an advantage in feeder-free (E8/vitronectin), but not feeder-based, culture. Mechanistically, we show that overexpression of MDM4, located on chromosome 1q, drives variants’ advantage in E8/vitronectin by alleviating genome damage-induced apoptosis, which is lower in feeder-based conditions. Our study explains condition-dependent patterns of hPSC aberrations and offers insights into the mechanisms of variant selection.

).An ideogram depicting the frequency at which a particular cytoband is represented in abnormal hPSC karyotypes in the WiCell dataset.Green bars next to a chromosome represent a gain of the corresponding chromosomal region.Red bars represent losses.S2).An ideogram depicting the frequency at which a particular cytoband is represented in abnormal hPSC karyotypes in the CSCB dataset.Green bars next to a chromosome represent a gain of the corresponding chromosomal region.Red bars represent losses.

Figure S3 (related to Figures 1 and 3, and Table S1).
A. Effects of passaging numbers and cell line identity on the appearance of aberrations.Analysis of the appearance of aberrations in relation to passage numbers shows a trend in the increase of abnormal karyotypes with an increase in passage numbers.

B.
The relative frequency of the most used media in the WiCell dataset over time, shaded by normal and abnormal cultures identified.S1).Co-occurrences of a trisomy of a specific chromosome with another chromosome trisomy.S1).Co-occurrences of a trisomy of a specific chromosome with a partial gain or loss of a chromosome.

Figure S6
(related to Figure 4).Genetic and phenotypic analysis of the lines used in the study.
B. SNP array analysis of H7, H9, MIFF3 wild-type and v1q sublines confirmed the presence of chromosome 1q gain.MIFF3 has a relatively small amplification of chromosome 1q that is not readily detectable by G-banding.

Karyotyping database assembly and analyses
Karyotyping data analysed in this study was collected by WiCell (Madison, USA) and the Centre for Stem Cell Biology (CSCB) (Sheffield, UK) over a period of 2009-2021 and 2002-2019, respectively.While CSCB data was mainly generated from hPSC lines grown in-house (around 80 different lines), WiCell data contained hPSC samples that were grown not only in-house but also submitted to WiCell for cytogenetic analysis by different laboratories (estimated to contain over 1500 different cell lines).
The WiCell dataset builds on the dataset reported in (Taapken et al., 2011) (~1,700 karyotypes) and adds a further 10 years of data collection (~20,500 karyotypes), which equates to more than a 10-fold increase in cytogenetic data analysed.
The data was stored in a format of standard International System for Human Cytogenetic Nomenclature (ISCN), with some associated data, such as the data of submission, to the sample.WiCell data also contained media/matrix information according to customer submission.This information was screened, and blanket terms were assigned to both media and matrix for the purposes of tracking changes related to these variables but also to respect the anonymity of the WiCell customers.The CSCB dataset was curated to remove repeated sampling of the same cell line, especially given the CSCB's active work using genetically variant hPSC lines.This same curation was not possible for the WiCell datasets due to the anonymised data.Errors in the karyotype nomenclature were corrected where possible and speculative changes denoted with question marks (?) were treated as correct calls.The karyotypes for mosaic cultures were split up for classification of karyotypic changes but kept together when assessing the number of cultures with abnormal karyotypes.In assessing the matrices for growing the cells Cultrex, which had only a ~12% abnormality rate, was omitted from the matrix comparison as there was evident sampling bias with ~94% of karyotypes of cells grown on Cultrex coming from just one institute.
To facilitate further exploration of our collated dataset, we also compiled a browser-based database termed KaryoBrowser (https://karyo.group.shef.ac.uk/).The browser-based database was assembled using RMarkdown (https://bookdown.org/yihui/rmarkdown/) and FlexDashboard (https://pkgs.rstudio.com/flexdashboard/). CytoGPS data was used for plotting interactive ideograms using ChromoMap (Anand and Rodriguez Lopez, 2022) and bar charts using rplotly (https://plotlyr.com).A Cascading Style Sheets (CSS) structure was added to allow for browser-based selection of parameters.Using KaryoBrowser, the collated datasets in this study can be inspected and filtered based on different parameters, including culture media, matrix, whether the lines are embryonic or induced PSCs and their sex.Interactive ideograms can be generated as well as bar charts detailing the chromosomal alterations associated with each parameter.The filtered data containing the karyotypes can also be exported for further analysis.
Genetically variant hPSCs used in this study and their karyotypes were: H7 v1q (RRID:CVCL_A5KR) ([46,XX,dup(1)(q21q42)] (20 metaphases analysed), H9 v1q [46,XX,der(21)t(1;21)(q21;p11)] (20 metaphases analysed), XY] (20 metaphases analysed) and XY,der(21)t(1;21)(q12;p11.2) (20 metaphases analysed).Of note, although MIFF-3 v1q appeared diploid by G-banding analysis, a duplication of q32.1 on chromosome 1 was detected by qPCR and confirmed by SNParray analysis.The H7 v1q cell line and the fluorescently labelled H7 subline, H7-RFP, were established and described previously (Price et al., 2021).The WLS-1C v1q cells were kindly provided by STEMCELL Technologies.H9 v1q and MIFF-3 v1q were established in this study by cloning out spontaneously arising variants from mosaic cultures using single cell deposition by fluorescent activated cell sorting.Single cells from mosaic cultures were sorted directly into individual wells of a 96 well plate using a BD FACS Jazz (BD Biosciences) and cultured to form colonies over 2-3 weeks.The resulting colonies were expanded in culture and subsequently frozen to establish cell banks.At the time of freezing, sister flasks were sent for either karyotyping by G-banding and/or SNP array profiling and assessment of the relative copy number of commonly identified genetic changes by qPCR (Baker et al., 2016;Laing et al., 2019).To facilitate cell competition assays, fluorescently labelled sublines of wildtype H9 and MIFF-3 and v1q lines were generated, as previously described (Price et al., 2021).Briefly, cells were transfected with pCAG-H2B-RFP plasmid (a kind gift from Dr Jie Na, Tsinghua University, Beijing) or pCAG-H2B-GFP plasmid (Price and Barbaric, 2022) (Addgene, cat.no.184777) using Neon Transfection System (Cat.# MPK10025; Thermo Fisher Scientific).The stably transfected cells were selected by growing in medium supplemented with puromycin (Cat.# A11138; Thermo Fisher Scientific), bulk-sorted to enrich for RFP or GFP-expressing cells and subjected to karyotyping by Gbanding and/or SNParray and qPCR analysis for the presence of 1q gain.
Banks of wild-type cells were made at 17 passages from derivation from H7, 14 passages from H9, 29 passages from MIFF-3 and 38 passages for WLS-1C.Banks of H7 and H9 fluorescently labelled cells were made at 24 and 47 passages, respectively.For variant cells, banks were made at 32 passages for H7 v1q, 46 passages for H9 v1q and 56 passages for WLS-1C v1q.Upon defrosting from a bank, cells were used for experiments after 3 passages back in culture up to passage 10.After this a vial would be defrosted from the master banks.The majority of experiments place between passages 4 and 7 upon defrosting the bank.
For freezing, cultures were split in a 1:4 split ratio in STEM-cellbanker (Cat.# 11924; AMSBIO).Lines were defrosted in mTeSR1 (STEMCELL Technologies) with 10uM Y-27632.After one day Y-27632 was removed and after two days the medium was replaced with E8.

Karyotyping by G-banding
Karyotyping by G-banding for lines used in experiments in this study was performed by the Sheffield Diagnostic Genetics Service (https://www.sheffieldchildrens.nhs.uk/sdgs/), as previously described (Baker and Barbaric, 2022;Baker et al., 2016).Briefly, hPSCs were treated with 0.1 µg/ml KaryoMAX Colcemid (Cat.# 15212012; Thermo Fisher Scientific) for up to 4h.The cells were then harvested with trypsin, re-suspended in pre-warmed 0.0375M KCl hypotonic solution.After incubating for 10 min at room temperature, cells were pelleted and resuspended in fixative (3:1 methanol:acetic acid).Metaphase spreads were prepared on glass microscope slides and G-banded by brief exposure to trypsin and stained with 4:1 Gurr's/Leishmann's stain (Sigma-Aldrich).Slides were scanned, metaphase images captured and analysed using the Leica Biosystems Cytovision Image Analysis system (version 7.3.2build 35).

SNP arrays
SNP array profiling was performed on Infinium Global Screening Array-24 v3.0 BeadChip GSA.Array data was obtained from the HuGe-F as a Genome Studio vs. 2.0.4 (Illumina, Eindhoven, The Netherlands) project using the hg38 reference genome, as described previously (Timmerman et al., 2021).

Competition assays
To assess selective advantage of variants and MDM4 overexpressing cells, mixing experiments were performed using matched fluorescently labelled and unlabelled lines, as follows: H7 v1q with H7-RFP cells, H9 v1q with H9-RFP cells, MIFF3 v1q-H2B-GFP with MIFF3 counterparts, and MIFF3-GFP cells with MIFF3 v1q or MIFF3-MDM4 cells.Cells were harvested to single cells by treating with Accutase (A6964; Sigma-Aldrich) and counted.Cells were then mixed to contain 10% v1q cells, or MDM4 overexpressing cells, with the respective wild-type population.Cell mixes were plated into either E8/VTN or KOSR/MEF conditions with the addition of 10µM Y-27632 for 24h.Cultures were grown and passaged as described above.At each passage, one flask was kept for further passaging while a parallel flask was harvested for assessment of the ratio of individual sublines using flow cytometry.

Clonogenic assays
Cells were harvested by treating with Accutase (A6964; Sigma-Aldrich) at 37°C for 10 min to create a single cell suspension.Cells were washed with DMEM/F12, pelleted at 200g and resuspended in fresh DMEM/F12.Cells were then seeded at a density of ~500 cells per cm 2 in 24 well plates, which had been pre-coated with either vitronectin or plated with mouse embryonic fibroblasts (MEF), and incubated with 0.5ml of pre-warmed media containing 10µM Y-27632 (Cat.# A11001-10; Generon).After two days, a fresh 0.5ml of media containing 10µM Y-27632 was added on top of the existing media.After four days, the media was fully replaced with 0.5ml of media without Y-27632.On day 5, cells were fixed using 4% PFA and stained for NANOG (Cat.# 4903; Cell Signalling Technology) and nuclei were counterstained with Hoechst 33342.Images were taken on the InCell 2200 and tiled across the well.The tiled images were stitched together for analysis and colonies were counted from images using a CellProfiler (Carpenter et al., 2006) pipeline.

Growth curve analysis
Initial assessment of population growth of wild-type and v1q cells was done using MIFF3 H2B-GFP and MIFF3 v1q H2B-GFP cells plated at 30,000 cells per cm 2 in 96 well plates.Cells were plated directly into test conditions, i.e., E8/VTN or KOSR/MEF with 10µM Y-27632 (Cat.# A11001-10; Generon).After 24 hours, the media was replaced and Y-27632 was removed.Cells were imaged for GFP signal at day 0, 1, 2 and 3 using the InCell Analyzer with 16 set positions in each well.The GFP signal at each position was calculated as area of image covered and tracked over the days.The resulting average of the fields was used to calculate the fold change in growth.
To assess the effect of KOSR addition to E8, MIFF3 cells and MIFF3 v1q cells growth rate analysis was performed, as previously described 7 .In brief, cells were plated at 30,000 cells per cm 2 in 96 well plates.Cells were plated directly into test conditions, i.e., E8/VTN, KOSR/MEF or E8+20% KOSR/VTN with 10µM Y-27632 (Cat.# A11001-10; Generon).After 24 hours, Y-27632 was removed and the media replenished daily.Plates were fixed daily with 4% PFA and nuclei counterstained with Hoechst 33342 prior to imaging on InCell Analyzer 2200.The resulting images were analysed using a CellProfiler pipeline (Carpenter et al., 2006) to calculate cell numbers.

Time-lapse analysis
Time-lapse microscopy was performed at 37°C and 5% CO2 using a Nikon Biostation CT.To perform lineage analysis, cells were imaged every 10 min for 96 hours using 10x air objective.Image stacks were compiled in CL Quant (Nikon) and exported to FIJI (Image J) (Schindelin et al., 2012) for analysis.Lineage trees were constructed manually from FIJI movies.Individual cells were identified in the first frame and then tracked in each subsequent frame until their death, division or the end of the 72 hours.The timing of cell death or division for each cell was noted and then used to reconstruct lineage trees of founder cells using Interactive Tree Of Life (iTOL) (Letunic and Bork, 2007) software.Tracking cells beyond 72 hours was not feasible as variant cultures were becoming too dense.Lineage trees were used to calculate the cell cycle time and the proportion of daughter cells surviving/dying following the division, as previously described (Barbaric et al., 2014).

Flow cytometry
To assess the ratio of individual sublines in mixing experiments, cells were harvested with TrypLE (Cat.# 11528856; Thermo Fisher Scientific) or Accutase (A6964; Sigma-Aldrich) and resuspended in DMEM/F12.The sample was then pelleted by centrifugation at 200 g for 5 min and subsequently fixed with 4% PFA for 10 min.The sample was washed with PBS and stored at 4°C until analysis.Samples were run on the BD FACS Jazz (BD Biosciences) flow cytometer to assess ratios of unlabelled to labelled.Gates were set based on 100% labelled and 100% unlabelled samples.
Flow cytometry for cleaved caspase-3 was performed to assess levels of apoptotic cells in cultures, as previously described (Price and Barbaric, 2022;Price et al., 2021).In brief, the old media, containing apoptotic cells which had detached from the flask, was collected into a 15ml Falcon tube.The remaining cells within the flask were harvested with TrypLE (Cat.# 11528856; Thermo Fisher Scientific) and added to the 15ml tube containing the collected culture medium and apoptotic cells.The collated sample was pelleted by centrifugation at 270 g for 5 min and subsequently fixed with 4% PFA.Cells were permeabilised with 0.5% Triton X-100 in PBS and then incubated with anti-cleaved caspase-3 primary antibody (Cat.# 9661; Cell Signalling Technology) in blocking buffer (1% BSA and 0.3% Triton X-100 in PBS).Samples were gently agitated for 1 hour at room temperature or overnight at 4°C, prior to washing and staining with secondary antibody (Goat anti-Rabbit AffiniPure IgG+IgM (H+L), Cat.# 111-605-003-JIR; Stratech) for 1 hour at room temperature in the dark.Cells were then washed and analysed on BD FACS Jazz (BD Biosciences).Baseline fluorescence was set using secondary antibody-only stained samples.

RNA extraction, sequencing and analysis
Three independent replicates of wild-type and v1q sublines of the H7 and H9 hPSC lines from E8/VTN cultures were used for RNA extraction and RNAseq analysis.Prior to RNA isolation, cells were lysed using Buffer RLT (Cat.# 79216; Qiagen) and stored at -80°C.RNA was isolated, libraries constructed and sequenced by GENEWIZ (Azenta Life Sciences, UK).
Briefly, libraries were prepared for Illumina (New England Biolabs, Ipswich, USA) and the library preparations were sequenced on an Illumina Novaseq platform (Illumina, San Diego, USA) to generate 150 bp paired-end reads.The sequencing reads were aligned to the GRCh38 human reference genome using STAR aligner v.2.5.2b.Unique gene hit counts were calculated using featureCounts from the Subread package v.1.5.2 and were normalized into transcript per million mapped reads (TPM), based on the length of the gene and reads count mapped to it.Differential gene expression analysis was performed for each pair of cell lines independently (H7-WT vs H7-v1q, H9-WT vs H9-v1q).Highlyexpressed genes (expressed in the top quartile of at least half of the samples) were considered for further analyses.Gene set enrichment analysis (Subramanian et al., 2005) was performed using GSEA software 4.0.3 using the following parameters: gene set, 1000 permutations, and 'Collapse' analysis.

Quantitative PCR (qPCR)
Expression of MDM4 in siRNA knockdown experiments was assessed using qPCR.RNA was isolated using a Qiagen RNAeasy Plus Mini Kit (Cat.# 74134; Qiagen), and the RNA concentration and purity determined using a NanoPhotometer (Implen, Munich, Germany).cDNA was synthesised using a high-capacity reverse transcription kit (Cat.# 4368814; Thermo Fisher Scientific).qPCR reactions were set up in triplicate, with each 10µl PCR reaction containing 1X PowerTrack SYBR Green Master Mix (Cat.# A46110; Thermo Fisher Scientific), 10µM Forward and Reverse Primers (Integrated DNA Technologies) and 10ng of cDNA.PCR reactions were run on a QuantStudio 12K Flex Thermocycler (Cat.# 4471087; Life Technologies).Following the first two steps of heating the samples to 50°C for 2 min and denaturing them at 95°C for 10 min, reactions were subjected to 40 cycles of 95°C for 15 s and 60°C for 1 min.The Ct values were obtained from the QuantStudio 12K Flex Software with auto baseline settings.Data was normalised to control GAPDH and 2 -ddCt calculated for relative expression in comparison to non-transfected v1q cells.

MDM4 overexpression
The pCAG-MDM4 expression vector was established by cloning of MDM4 sequence into a pCAG vector (Liew et al., 2007) containing a multiple cloning site (pCAG-MCS) using In-Fusion® Snap Assembly Starter Bundle kit (Cat.#638945; Takara Bio).A single restriction digest was performed on the pCAG-MCS vector using XhoI (Cat.#0146, New England Biolabs) to linearize plasmid.The MDM4 sequence in pLVX-TetOne-Puro-MDM4, kindly gifted to us by Jason Sheltzer (Cat.# 195140; Addgene), was PCR amplified and the resulting fragment cloned into the linearized pCAG-MCS vector as per manufacturer's instructions.
To generate the MIFF3 wild-type MDM4 overexpressing line, cells were transfected using the Neon Transfection System as described previously (Price and Barbaric, 2022;Price et al., 2021).In brief, MIFF3 cells were dissociated to single cells using TrypLE and resuspended at 2,0x10 4 cells/ml in "R buffer".Transfection was performed with 5µg of plasmid DNA using 1 pulse of 1600V, 20msec width.After electroporation, the cells were immediately transferred to a vitronectin-coated 60mm diameter culture dish (Cat.# 150288; Thermo Fisher Scientific) containing E8 media supplemented with 10µM Y-27632 (Cat.# A11001-10; Generon).To select for stably transfected cells, 48h post transfection cells were subjected to puromycin (Cat.# A11138; Thermo Fisher Scientific) drug selection.The cells were then expanded in the presence of puromycin selection and subsequently frozen to establish cell banks.At the time of freezing, cells from sister flasks were karyotyped by G-banding and assessed for the relative copy numbers of commonly identified genetic changes by qPCR (Baker et al., 2016;Laing et al., 2019).Upon defrosting and subsequent culture, cells were also regularly genotyped by karyotyping and screened for common genetic changes by quantitative PCR (Baker et al., 2016;Laing et al., 2019).

Cell attachment and survival assays in different matrix/media combinations
Cells were seeded at a density of 35,000 cells per cm 2 in 96 well plates, which had been pre-coated with either vitronectin or plated with mouse embryonic fibroblasts (MEF).Each matrix was tested in combination with E8 and KOSR-based medium.After 2 hours, some plates were fixed and stained with Phalloidin (Alexa Fluor® 647 Phalloidin, Cat.#8940, Cell Signalling Technology) and a phospho-FAK (Tyr397) antibody at 1:200 (phospho-FAK (Tyr397), Cat.#44-624G, ThermoFisher Scientific).The secondary antibody (Alexa Fluor 488 donkey anti-rabbit, Cat.#A21206, Invitrogen) was used at 1:200.Nuclei were counterstained with Hoechst 33342.The remaining plates were kept until 48 hours postplating, then fixed and stained with Hoechst 33342.Images were taken on the InCell 2200 and analysed using Cell Profiler or ImageJ pipelines.Cells fixed at 2 hours post-plating were analysed for cell number, cell area and number of cells, whereas the cells fixed at 48 hours post-plating were analysed for cell numbers.

Genome damage induction
To induce genome damage, hPSCs plated on VTN were treated with either 10µM CPT or 500nM KBrO3 in E8 for 1 hour.Cells were then washed with media, fixed and processed for immunocytochemistry, as described above.

QUANTIFICATION AND STATISTICAL ANALYSIS
Statistical analysis of the data presented was performed using either GraphPad Prism version 9.0.2,GraphPad Software, La Jolla California USA, www.graphpad.comor Real Statistics Resource Pack for Excel, Charles Zaiontz, www.real-statistics.com.Differences were tested by statistical tests as indicated in figure legends.

Figure
Figure S2 (related to Figure 1 and TableS2).An ideogram depicting the frequency at which a particular cytoband is represented in abnormal hPSC karyotypes in the CSCB dataset.Green bars next to a chromosome represent a gain of the corresponding chromosomal region.Red bars represent losses.

Figure S4 (
Figure S4 (related to Figure 1 and TableS1).Co-occurrences of a trisomy of a specific chromosome with another chromosome trisomy.

Figure
Figure S5 (related to Figure 1 and TableS1).Co-occurrences of a trisomy of a specific chromosome with a partial gain or loss of a chromosome.

Figure
Figure S7 (related to Figures6 and 7).KOSR-based medium, rather than MEFs, causes diminished advantage of v1q cells in KOSR/MEF compared to E8/VTN. A. Similar numbers of H9 wild-type and v1q hPSCs attach regardless of the medium (E8 and KOSR) or the matrix (VTN and MEFs) at 2h post-plating.Data shown are the mean ± SD of three independent experiments.ns, non-significant; Unpaired t test.
A. Similar numbers of H9 wild-type and v1q hPSCs attach regardless of the medium (E8 and KOSR) or the matrix (VTN and MEFs) at 2h post-plating.Data shown are the mean ± SD of three independent experiments.ns, non-significant; Unpaired t test.or KOSR/MEF (lower panels) stained with antibodies against MDM4 and OCT4 (POU5F1).The nuclei are counterstained with Hoechst 33342.Scale bar: 50µm.D. Quantification of the MDM4 nuclear expression in E8/VTN versus KOSR/MEF condition in H9 line.Data shown are the mean ± SD of three independent experiments.**p<0.01;Unpaired t test.E.