Reprogramming of Acute Myeloid Leukemia Patients Cells: Harboring Cancer Mutations Requires Targeting of AML Hierarchy

Abstract

Screening of primary patient acute myeloid leukemia (AML) cells is challenging based on intrinsic characteristics of human AML disease and patient-specific conditions required to sustain AML cells in culture. This is further complicated by inter- and intra-patient heterogeneity, and “contaminating” normal cells devoid of molecular AML mutations. Derivation of induced pluripotent stem cells (iPSCs) from human somatic cells has provided approaches for the development of patient-specific models of disease biology and has recently included AML. Although reprogramming patient-derived cancer cells to pluripotency allows for aspects of disease modeling, the major limitation preventing applications and deeper insights using AML-iPSCs is the rarity of success and limited subtypes of AML disease that can be captured by reprogramming to date. Here, we tested and refined methods including de novo, xenografting, naïve versus prime states and prospective isolation for reprogramming AML cells using a total of 22 AML patient samples representing the wide variety of cytogenetic abnormalities. These efforts allowed us to derive genetically matched healthy control (isogenic) lines and capture clones found originally in patients with AML. Using fluorescently activated cell sorting, we revealed that AML reprogramming is linked to the differentiation state of diseased tissue, where use of myeloid marker CD33 compared to the stem cell marker, CD34, reduces reprogramming capture of AML⁺ clones. Our efforts provide a platform for further optimization of AML-iPSC generation, and a unique library of iPSC derived from patients with AML for detailed cellular and molecular study.

Generating induced pluripotent stem cells (iPSCs) from patients with acute myeloid leukemia (AML) is challenging due to disease and patient heterogeneity, normal cell contamination, and reported refractory nature of AML to reprogram. Testing several strategies including de novo isolation of subsets, xenografting, and naïve versus prime states have allowed for the derivation of 129 iPSCs from AML patient samples for study.

acute myeloid leukemia, hematopoietic stem/progenitor cell, induced pluripotent stem cell, xenotransplantation, AML-induced pluripotent stem cell, reprogramming

Significance Statement

Establishment of a library of 79 acute myeloid leukemia (AML)-induced pluripotent stem cell lines harboring driver mutations from a broad range of patients with AML to increase capture of disease heterogeneity for modeling AML.

Introduction

Cancer is the leading cause of death in developed countries, firmly linked to genetics as well as environmental factors.¹ Cancer is considered a multistep disease involving several stages of development toward full malignancy, requiring a better understanding of the diverse genetic, and epigenetic steps in the alteration from healthy to diseased cells. Acute myeloid leukemia (AML) is one of the most heterogeneous cancers with diversity observed at the levels of genetics, epigenetics, and clonal organization across patients with AMLs.² This heterogeneity is believed to be the basis of difficulty in predicting patient responses to chemotherapy and related disease relapse.^3-6 Although some genomic DNA mutations carry prognostic value, these rarely provide a means for targeted therapeutic intervention, and their functional contributions to disease initiation, progression, and maintenance are largely unknown^7-9 creating further challenges in developing novel therapeutics.¹⁰

Patients with AML share a disease phenotype where dysfunctional cells accumulate in the myeloid compartment of the hematopoietic system and are blocked in their ability to differentiate and fully mature. AML results in the rapid accumulation of non-functional, immature hematopoietic cells in the bone marrow (BM) and peripheral blood (PB) of patients leading to hematopoietic system failure.¹¹ These aberrantly differentiated cells compete in the BM niche with healthy cells¹² where they also assemble in an hierarchy similar to normal hematopoiesis. In a non-diseased state, the healthy hematopoietic system has rare hematopoietic stem and progenitor cells (HSPCs) residing at the apex^13,14 to simultaneously maintain HSPC populations and undergo appropriate differentiation into to all mature lymphoid and myeloid blood cells.¹⁵ Unfortunately, primitive AML and healthy HSPCs cells share similar phenotypes.¹² Therefore, it is difficult to prospectively isolate diseased versus healthy primitive cells from each other for interventional studies or experimentally analysis. Accordingly, development for more diverse model systems able to capture and distinguish AML from healthy counterparts, as well as observe rare clonal and epigenetic diversity of AML disease evolution are needed.

Human-induced pluripotent stem cells (hiPSCs) represent a potential platform to achieve such goals. In the last decade, thousands of hiPSCs have been generated worldwide from healthy donors and from patients afflicted with various diseases.^16-19 Patient-derived hiPSCs have the potential to produce an endless number of specialized disease-associated cells and organoids, allowing researchers to replicate some pathological characteristics of human disease in vitro. Indeed, such models have already aided in the discovery of molecular processes of pathogenesis, paving the way for new treatments for some diseases.²⁰ Although monogenic inherited blood diseases were readily modeled with induced pluripotent stem cells (iPSC), malignant hematologic disorders such as AML have been more challenging to obtain. Unlike inherited genetic diseases, where disease-causing mutations are present in the germline to be passed to all somatic cells, most AML genetic lesions arise postnatally and accumulate sequentially in the somatic hematopoietic stem cell (HSC) compartment. Accordingly, while iPSC models of inherited monogenic diseases can be derived by reprogramming any accessible cell type, derivation of AML-iPSC requires reprogramming hematopoietic leukemic cells themselves and not skin fibroblasts or other cell types that would represent germline mutations of patients with AML.^21,22

Although some reports have successfully reprogrammed myeloid malignancies over the years, we and others have shown leukemic cells are relatively refractory to reprogramming and represent only a small minority of the diversity of genetic phenotypes observed in patients with AML and reprograming from patient with AML is a very rare event.^22-26 AML’s refractory behavior to reprogramming is similar to other highly proliferative malignant cells.^27-30 This results in an experimental predominance of normal iPSC from patient tissue defined by the absence of detectable clinically identified mutations.^22-25,30-34 Here, we deploy strategies to selectively reprogram AML and endogenous normal cells from patients with AML to increase successful reprogramming of bona fide aberration-containing AML cells. Using prospective purification, we show that AML and normal reprogramming correlates to the stage of hematopoietic differentiation. In total, we report the development of a library of 77 AML-induced pluripotent stem cell (AML-iPSC) lines from patients with AML. We also report 52 genetically normal iPSC lines developed from these patients referred to as aberration negative lines, for a total of 129 distinct, functionally and phenotypically characterized iPSC lines. Within this library of patient with AML iPSC, we demonstrate isogenic paired lines to be competent for lineage differentiation representing ectodermal, endodermal, and mesodermal germ layers illustrating the board utility of our lines.

Material and Methods

Primary Patient Samples

Healthy human hematopoietic cells were isolated from mobilized peripheral blood (MPB) of adult donors. Primary AML specimens were obtained from peripheral blood apheresis or BM aspirates of consenting patients with AML. AML samples and adult sources of healthy hematopoietic tissue were provided by Juravinski Hospital and Cancer Centre and London Health Sciences Centre (University of Western Ontario). All samples were obtained from informed consenting donors in accordance with approved protocols by the Research Ethics Board at McMaster University and the London Health Sciences Centre, University of Western Ontario. The only non-AML sample used was obtained from a healthy volunteer to generate MPB-iPSC, whereas all other samples reprogrammed were from AML patients.

Patient-Derived Xenografts

AML samples were thawed and CD3 depleted using EasySep Human CD3 Positive Selection Kit II (STEMCELL Technologies) and EasySep Magnet (STEMCELL Technologies). Immune-deficient NOD LtSz-scidIL2Rγnull (next generation sequencing, NGS) mice were bred in a barrier facility, and all experimental protocols were approved by the Animal Research Ethics Board of McMaster University. NGS mice 6-10 weeks of age were sublethally irradiated at 315 Rads using a ¹³⁷Cs γ-irradiator 24-hours prior to transplantation. 5-15 × 10⁶ cells were intravenous (IV) injected, and BM aspirates were performed to identify human chimerism prior to harvesting. BM was harvested from legs and spines 6-12 weeks post-engraftment and cells recovered by mechanical dissociation as previously described³⁵ and analyzed by flow cytometry.

Reprogramming of Primary AML Samples

Fluorescence-Activated Cell Sorting

Samples were thawed using 100% FBS and PBS supplemented with 3% FBS (HyClone FBS, Mississauga, ON, Canada), and 1 mM EDTA (Invitrogen, Waltham, Massachusetts, USA) referred to as PEF. Cells were counted, and a fraction of the cells were set aside to be sorted using fluorescence-activated cell sorting (FACS). Cells to be sorted were stained using the following antibodies at a 1:100 concentration: CD33-APC, CD34-FITC, CD45-v450 (BD Biosciences), and CD3- PE (Beckman Coulter). Cells were stained at 10 million cells per mL, for 45 minutes at 4°C. Subsequently, cells were then stained with 7-amino actinomycin D (7-AAD, Becton Dickinson) at 1:50 to exclude nonviable cells. Fluorescence Minus One controls and single stains of each antibody on compensation beads were used to ensure that gates were properly set, and sorted populations were pure. Cells were sorted using an Aria II flow cytometer (Beckman Coulter) into separate tubes for several target different populations, depending on what populations existed or had a substantial number of cells. This includes CD45⁺CD34⁺CD33⁺, CD45⁺CD34⁺CD33^-, and CD45⁺CD34^-CD33⁺. Collected cells were kept on ice during sorting, then centrifuged at 1500 g for 5 minutes, pooled and counted for viability and total cell count.

Reprogramming

All AML or MPB samples were reprogrammed in media consisting of StemSpan SFEM II (STEMCELL Technologies) supplemented with 100 ng/mL stem cell factor, 100 ng/mL fms-related tyrosine kinase 3 ligand (FLT3-L), and 20 ng/mL thrombopoietin, all from R&D Systems, 8 μg/mL polybrene (Sigma-Aldrich) and 0.75 μM StemRegenin 1 (STEMCELL Technologies), referred to as “reprogramming media.” The factors used to reprogram the primary AML cells were delivered using a nontransmissible form of the Sendai virus, from the Cytotune iPS 2.0 Sendai Reprogramming Kit (ThermoFisher Scientific). Volume of virus used was variable depending on the lot number of the virus kit and its titer, but the general ratio used for these viruses is 5:5:3 (ThermoFisher Scientific). Any cells to be reprogrammed were counted, aliquoted, and centrifuged at 1500 g for 5 minutes. Cells were resuspended in reprogramming media with the appropriate volume of each virus added and plated at 200 000 cells per 24 well in 250 µL per well, in an ultra-low attachment plate (Corning). For each sample and population, at least 2 separate wells were reprogrammed as a technical replicate, or more if the number of cells and reprogramming resources allowed. Cells were transduced and incubated at 37°C for 48 hours, after which the cells in each 24 well were collected into individual tubes and centrifuged for 5 minutes at 1500 g. Cells were plated into Leukemia Inhibitor Factor (LIF), and 2 small molecule inhibitors (2i) of MEK and GSK (LIF2i media), which is SR media consisting of DMEM/F12 (1:1) (Gibco), 20% KOSR (Gibco), 1× NEAA (Gibco), 1 mM L-glutamine (Gibco), 0.1 mM β-mercaptoethanol, supplemented with 1 μM MEK inhibitor PD0325901 (Stemgent), 3 μM GSK3 inhibitor CHIR 99021 (Sigma-Aldrich) and 10 ng/mL LIF (Millipore). Each 24 well of cells was counted and plated into one tissue-culture treated 6 well, referred to as the mother plate. Six wells were coated with 0.5% gelatin (Millipore Sigma) and seeded with 180 000 irradiated mouse embryo fibroblast (iMEF) cells 24 hours prior to seeding the transduced AML cells. iMEFs were seeded in MEF media, which consists of KO-DMEM (Gibco), 10% FBS. 1% NEAA, 1% sodium pyruvate, 2 mM L-glutamine, and 0.1 mM β-mercaptoethanol. Reprogrammed cells were carefully fed every other day with LIF2i media, beginning on day 2 or 3 from plating them into the 6 wells, until colonies arose to be selected and expanded as individual clones (2-3.5 weeks posttransduction).

Clone Expansion

Each clone derived from individual colonies was expanded separately, after two to three and a half weeks posttransduction. Once colonies arose in the mother plate, colonies were stained using a live cell imaging kit (TRA-1-60 Alexa Fluor488 Conjugate Kit for Live Cell Imaging [Invitrogen]), according to manufacturer’s instructions. Colonies were picked individually and plated into a 12-well tissue culture-treated plate, referred to as passage zero (p0). Each 12 well was coated with 0.5% gelatin before being seeded with 100 000 iMEFs 24 hours prior to picking clones. iMEFs were seeded in MEF media previously described. After 24 hours, individual colonies were plated into 12 wells, and the media was carefully aspirated and fed with 1.5 mL of LIF2i media. Cells were then fed each day with LIF2i media and manually passed every 5-7 days to a new plate, increasing the number of wells or size of the well as needed to expand the clones. Between the second (p2) and third (p3) passage of the clones, they were transitioned to SR Media supplemented with 8 ng/mL basic fibroblast growth factor (bFGF). After 24 hours passing, the cells were transitioned by feeding with two-third LIF2i media and one-third SR media + bFGF, then two-third SR media + bFGF 48 hours from the initial pass date. SR media + bFGF was used to feed the cells on day 3 from the initial pass date of p2 or p3 and onwards. Cells reprogrammed using exclusively bFGF did not require this transition and were solely cultured in SR media + BFGF from the initial p0. Cells were cryopreserved in 1 mL of freeze media, consisting of DMEM/F12 (1:1) (Gibco), 30% KOSR (Gibco), and 10% DMSO (Sigma-Aldrich).

hPSC Subculture

All hPSC cell lines were maintained in an undifferentiated state and passed every 7 days onto a layer of iMEF in medium that contains DMEM-F12, KnockOut Serum Replacement (KOSR), 1 mM nonessential amino acids (NEAA), 1 mM L-glutamine, and β-mercaptoethanol (SR medium) supplemented with 8 ng/mL bFGF. In a subset of experiments, hPSCs were transitioned to mTeSR media (Stem Cell Technologies, Vancouver, BC, Canada) and maintained on matrigel with daily media changes. Experiments were performed using human ESC line H9, MPB-iPSC, and patients with AML-derived iPSCs, and daily morphological evaluation of cells was performed with light microscopy.

Aberration Detection of iPSC Lines/Clones

Droplet PCR

DNA was isolated from each iPSC clone by passing a subset of colonies to a tissue culture-treated plate coated with 1:15 Matrigel (Corning). The colonies were expanded for 2-3 days before DNA isolation. Upon collection, the colonies were treated with Collagenase IV (Thermo Fisher Scientific) for 10 minutes at 37°C then cell dissociation buffer (Gibco) for 10 minutes at 37°C. Colonies were then washed and spun down for 5 minutes at 1500 g. Cells were resuspended in PEF (3% FBS), filtered, and counted. Cells were then centrifuged for 5 minutes at 1500 g and resuspended in 200 µL of cold PBS. DNA was isolated using DNeasy Blood & Tissue Kit according to the manufacturer’s protocol (Qiagen). DNA was eluted in ultra-pure water and stored at −80°C. DNA samples were aliquoted at 50 ng/µL in 10 µL and diluted in ultra-pure water to be shipped to The Centre for Applied Genomics (TCAG) Genetic Analysis Facility (Toronto) for droplet digital PCR (ddPCR). For each probe used, a MPB sample was sent as a negative control, and the original primary AML sample associated with the iPSC clones was sent as a positive control. Primers were designed using RefSeq transcripts and https://mutalyzer.nl/positionconverter.

Karyotyping

iPSC were cultured and expanded into T25 tissue-culture flasks, preseeded with 350 000 iMEF each, 24 hours prior to being passaged. Twenty-four hours later, cells were washed and fed with SR media + bFGF. Flasks were filled with SR media + bFGF, sealed, and shipped to the karyotyping facility. Once received, the cells were harvested and screened for abnormalities and sent for a full karyotyping if abnormalities were present. Karyotyping of all samples was completed by TCAG Cytogenomics Facility (Toronto) following standard protocols.

Cytogenomics

Comparative genomic hybridization using Cytoscan HD Array (Thermo Fisher) was performed by TCAG, at TheHospital for Sick Children (SickKids), Toronto, Ontario, Canada. Cytoscan analysis was performed using Chromosome Analysis Suite (NetAffx 33.1, h19) using default settings.

Fluorescent In Situ Hybridization

Primary AML cells and iPSC were synchronized by adding 0.1 μg/mL KaryoMAX Colcemid (ThermoFisher) to cell media for 3-4 hours. Cells were collected as single cell suspension. iPSC were dissociated using cell dissociation buffer (Gibco) for 5-10 minutes at 37°C. Cells were incubated in a hypotonic solution of 0.075 M KCl for 15 minutes at 37°C. Cells were then fixed using 3:1 methanol: glacial acetic acid and pipetted onto a glass slide. Slides were dehydrated using a sequence of 70%, 80%, and 100% EtOH and prewarmed at 37°C. Probe was also prewarmed to 37°C. Sample and probe were simultaneously denatured on a hotplate at 75°C for 2 minutes and then placed at 37°C overnight. Slides were then washed, stained with DAPI, visualized, and scored manually. MYH11/CBFB probe (Empire Genomics) and PML/RARα translocation dual fusion probe (Cytocell) were used and scored as recommended by the manufacturer.

Flow Cytometry

All antibodies used for flow cytometry were titrated to generate signal-based populations consistent with those demonstrated by the antibody manufacturer. All extracellular staining was performed in PEF, where 10 million cells per mL were stained for 45 minutes at 4°C, washed with 10 volumes of PEF and then stained with 7-amino actinomycin D (7-AAD, Becton Dickinson) at 1:50 to exclude nonviable cells. The following antibodies were used at 1:100 unless otherwise specified as follows: CD45-v450, SSEA3-PE, TRA-1-60-AF647 at 1:1000, CD34-FITC or APC (1:200), mCD45-FITC, CD33-PE or APC, CD3-PE (Beckman Coulter), all from BD Biosciences unless otherwise specified. For pluripotent stem cells, cells were treated with Collagenase IV for 10 minutes, followed by a 10-minute treatment with cell dissociation buffer (Thermo Fisher Scientific) and then filtered through a 40-μm cell strainer. For intracellular staining, cells were fixed using Cytofix/Cytoperm (BD Biosciences) and Perm/Wash buffer (BD Biosciences) according to the manufacturer’s instructions. Prior to performing fixation and permeabilization, cells were stained with Live/Dead Violet discrimination dye at 1:7000 (Life Technologies) for 30 minutes at 4°C in the dark. Intracellular staining was performed in Perm/Wash buffer overnight at 4°C in the dark and washed with 10 volumes of PBS prior to analysis. The following antibodies were used at 1:000 for intracellular staining: Alexa Fluor 488 Mouse anti-OCT4, Alexa Fluor 488 Mouse anti-NANOG, and Alexa Fluor 488 Mouse anti-Sox2. Cells were analyzed with an LSRII Flow Cytometer (BD Biosciences) and resulting data analyzed using FlowJo software version 10.8.0 (FlowJo, LLC). https://fccf.sitehost.iu.edu/pdf/LSRIIBrochure.pdf

Teratoma Assay

To assess the developmental potential of patients with AML-derived iPSC, cells were collected by collagenase IV treatment and injected as clumps into NOD/SCID mice via intratesticular injection (IT). At 8 to 10 weeks, teratomas were harvested, dissected, and fixed with 4% paraformaldehyde. Samples were embedded in paraffin and processed for H&E staining. Images were acquired using Aperio ScanScope CS digital slide scanner (Leica).

Neural Differentiation

For neural differentiation, hPSCs clones were harvested by collagenase IV treatment, and EBs were generated with EB medium without cytokines. Twenty-four hours after suspension culture, EBs were plated onto poly-L-lysine/laminin-coated plates (BD Biosciences) in neural differentiation medium composed of DMEM/F12 supplemented with B27 and N2 (Thermo Fisher Scientific), EGF (25 ng/mL, R&D Systems), and bFGF (8 ng/ml). Cell culture medium was changed every 3 days. After 10 days, neurospheres were collected, dissociated into single cells by Accutase treatment, and plated onto poly-ornithine- and laminin-coated plates. For mature neuron differentiation, cells were cultured in Neurobasal medium supplemented with N2 (1%), B27-RA (2%), non-essential amino acids (1%), brain-derived neurotrophic factor (20 ng/mL), glial cell-derived neurotrophic factor (20 ng/mL), dibutyryl cyclic adenosine monophosphate (1 µM), and ascorbic acid (200 µM). Cells were cultured for 15 days with half media change every other day.

Hepatocytic Differentiation

For hepatocyte differentiation, hPSC were transitioned to mTeSR media (Stem Cell Technologies,) and maintained on Matrigel with daily media changes. hPSC colonies cultured in mTeSR media were dissociate with 0.05% trypsin for 5 minutes at 37°C, pipetted thoroughly to form small aggregates, and subsequently washed twice with PBS + 2% FBS media. hPSC were subsequently seeded onto Matrigel-coated plates, and then cultured for 5 days with 8 ng/mL bFGF and then replaced with RPMI supplemented with 1 × B27, activin 1 (100ng/mL), hepatocyte growth factor (10 ng/mL) and 1 mM CHIR99021 (GSK3 inhibitor) for 3 days of endoderm formation as previously described.³⁶ Medium was subsequently replaced with knockout-DMEM with 20% knockout-SR, 1 mM L-glutamine, 1% nonessential amino acids, 0.1 mM β-mercaptoethanol and 1% dimethyl sulfoxide for 4 days. For hepatocyte maturation, the cells were cultured in hepatocyte maturation media consisting of Iscove’s modified Dulbecco’s medium (IMDM) containing 20 ng/mL Oncostatin M, 0.5 mM dexamethasone, and 50 mg/mL ITS premix as previously reported.^36,37

Cardiomyocyte Differentiation

Cardiomyocyte differentiation was performed based on a previous report.^38,39 Briefly, 10 000 cells were plated per U-bottom well in MEFCM containing 8 ng/mL bFGF and 10 μM Y-27632. At day 3 media was replaced with DMEM/F12, 20% FBS, 1 mM L-glutamine, 1 mM NEAA, 0.1 mM β-mercaptoethanol, and 50 µg/mL ascorbic acid. Aggregates were seeded on gelatin-coated wells at day 7 and differentiation medium was changed every 2 days.

Immunofluorescence

Briefly, cells were fixed in 4% paraformaldehyde and stained with appropriate antibodies. If permeation was needed, cells were fixed using the BD Cytofix/Cytoperm kit (BD Bioscience) and then stained in Perm/Wash buffer (BD Biosciences) according to the manufacturer’s instructions with the following Abs: Smooth Muscle Actin 1:100 (Millipore), FOXA2 1:100, GATA4 1:100, and β-Tubulin-III 1:100. Unconjugated antibodies were visualized with appropriated Alexa fluorochrome conjugated secondary antibodies (1:1000).

Statistical Analysis

Data are represented as means ± SEM. Prism (6.0c, GraphPad) software was used for all statistical analyses, and the criterion for statistical significance was P < .05. Statistics are described in each figure legend when applicable.

Results

Patient with AML Selection Strategy

To date, although primary AML reprogramming has been implicated in multiple studies,^{22-24,26,40,41} it is apparent that deriving AML-iPSC is challenging. Interestingly, AML-iPSC have been shown to reacquire leukemic properties, including AML reconstitution in xenograft models, as well as methylation and gene expression patterns providing indication that AML-iPSC can be used to successfully mimic the disease without the need for continuous use of primary patient samples.^24,40 The study by Kotini and colleagues used AML-iPSC to model myelodysplastic syndrome (MDS) to AML transition;²³ however, a limitation of this work was that the progression of disease stages iPSC were derived from different patients. Moreover, within the aforementioned study, only 4 patients with AML cases have been reprogrammed to iPSC, 3 of which harbor MLL gene family network alternations which represent only 2% of adult patients with AML⁷ thus failing to capture patients with AML heterogeneity. To overcome this, we sought to develop a library of AML-iPSC using a large number of heterogeneous AML patient samples with a range of different mutational categories (Table 1). For sample selection, samples were examined to see if they contained known, detectable genetic and/or molecular abnormalities, and if NGS data existed. The presence of various phenotypic markers based on preliminary flow cytometry data obtained from the clinic in which the sample was collected and was also considered to ensure that selected samples would be viable for reprogramming. This was to confirm that the populations of interest would yield an adequate number of cells based on the estimated percentage of the cells that contained that population (Table 1; CD34%). Similar to healthy hematopoietic cells, in AML, CD34 marks a more primitive cell compartment.⁴² These cells play a role in disease progression and relapse, hold a higher reprogramming potential, and are relevant in healthy hematopoiesis⁴³and, thus, were prioritized for reprogramming. In addition, if a sample had 2 clear populations such as CD34⁺CD33⁺ and CD34^-CD33⁺ (discussed later, see Fig. 1), this would be an opportunity to reprogram multiple populations from a sample and determine how these markers influence reprogramming, creating more interest for the sample. Samples were also prioritized if serial specimens were taken at different stages of a patient’s disease (such as sampling at both diagnosis and relapse) (Table 1; Patient’s 16308 and A472). Finally, samples were selected based on availability and cell quantity. Overall, 22 unique AML samples from 20 distinct patients were selected for reprogramming. By selecting a variety of patient samples with a broad range of genetic aberrations (Table 1), interpretations would not be limited and would strengthen the ability to be applied to AML disease wide conclusions.

Table 1.

Sample characteristics of reprogrammed AML samples.

Number of AML samples reprogrammed	AML patient ID number	Source	Clinical stage	AML patient age	AML number	Genetic/molecular abnormality	LSC	CFU	CD34+ (%)
1	A374.1	PB	Diagnosis	78	AML 1	46,XX,add(3),der(3),del(5),del(7),add(18),−20, −22,+mar1,+mar3	Y—myeloid	Y	80.3
2	A422	PB	Relapse	67	AML 2	45,XX, −7[19]/46,XX[1]	Y—myeloid	Y—myelo and erythro	88
3	13814.1	BM	Diagnosis	74	AML 3	43~46,XY,del(15)(q11.2q15)[cp5]/46,XY,+mar1[cp2]/46,XY[11]	Y—multilineage	N	60.1
4	16150	BM	Diagnosis	41	AML 4	CBFbeta/MYH11 inv(16)	Y—myeloid	Y—myelo and erythro	72.8
5	A151.1	PB	De novo AML	67	AML 5	46,XY,inv(16)(p13.1q22)[20]/47,sl,+8[5]	Y—myeloid	Y	30
6	15328	BM	Diagnosis	54	AML 6	45~46,XX,der(X)?t(X;11)(q28;q12)[2],-2[3],del(3)(q11.2)[3],-5[2],add(7)(p11.2)[3],-17[3],+3~5mar[3][cp3]	Y—multilineage	UNK	85.8
7	A295.1	PB	Diagnosis	52	AML 7	trisomy i21	Y—multilineage	Y	67
8	16158.1	BM	Diagnosis	69	AML 8	45, XX,del(5),-7	Y—multilineage	Y	92.8
9	16626	BM	Refractory AML	48	AML 9	inv (3)	Y—myeloid	Y	45
10	19447	PB	Diagnosis	56	AML 10	complex, isolated +8	Y—multilineage	N/A	64
11	A320-1	BM	De novo AML	38	AML 11	PML, RARA translocation (APL)	UNK	UNK	0.1
12	A477	Leuka	Diagnosis	62	AML 12	DNMT3A:c.2645G>A, p.(Arg882His) (48.8%)	N	Y—myelo and erythro	88
13	13051.1	Leuka	Diagnosis	56	AML 13	NPM+, FLT3-ITD+	N	Y—myelo and erythro	57.3
14	A485-1	Leuka	Diagnosis	24	AML 14	46,XY,del(7)(q22q34)[18]/46,XY[1] and FLT3:c.1794_1795insGGAGGAAACTACGTTGATTTCAGAGAATATGAA, p.(Glu598_Tyr599insGlyGlyAsnTyrValAspPheArgGluTyrGlu) (5.6%)	N	Y—myelo and erythro	92
15	16534	PB	Diagnosis	55	AML 15	NPM+, FLT3 ITD +, normal karyotype	N	N	22
16	A494-1	PB	De novo AML	72	AML 16	Karyotype: 47,XX,+11[20] and FLT3 −ve IDH2:c.515G>A, p.(Arg172Lys) (49.1%) PTPN11:c.179G>T, p.(Gly60Val) (46.3%)	N	Y—erythroid	90
17	A472-1	BM	Diagnosis	73	AML 17*	ASXL1:c.2725A>T, p.(Lys909*) (42.6%) IKZF1:c.476A>G, p.(Asn159Ser) (27.5%) RUNX1:c.656_657insAAGG, p.(Ser220Argfs*9) (27.8%)	UNK	Y—myelo and erythro	50
18	A063	PB	Diagnosis	77	AML 18	t(4;12)[25/25]	N	N	99
19	A492-2	PB	Relapse	70	AML 19	Karyotype: 46,XY,t(2:12)(p23;p13)[20] and FLT3 −ve NPM1:c.863_864insCCTG, p.(Trp288Cysfs*12) (28.6%)	N	N	30
20	16308A	PB	Diagnosis	44	AML 20#	NPM1+	Y—myeloid	Y—myelo and erythro	57.4
21	A472-3	PB	Refractory AML	73	AML 21*	ASXL1:c.2725A>T, p.(Lys909*) (42.6%) IKZF1:c.476A>G, p.(Asn159Ser) (27.5%) RUNX1:c.656_657insAAGG, p.(Ser220Argfs*9) (27.8%)	Y—myeloid	Y—myeloid	20
22	16308F	PB	Relapse	44	AML 22#	NPM1+	UNK	Y—myelo and erythro	33.5

Characteristics of 22 reprogrammed AML samples, including AML patient ID number, primary AML specimen source, clinical stage of disease, genetic/molecular abnormality, leukemic stem cell (LSC) content and engraftment type if applicable, colony forming unit (CFU) capacity, and CD34⁺ percentage. Samples with engraftment are described as myeloid or multilineage engraftment. Samples in which CFU capacity was known are described as either containing myeloid and erythroid CFU’s (Y—myelo and erythro) or only myeloid (Y—myelo

* and # represent the same AML patient but different samples, respectively.

Abbreviations: AML: acute myeloid leukemia; LSC: leukemic stem cell; CFU: colony forming unit; Y: yes; N: no; UNK: unknown; PB: peripheral blood; BM: bone marrow.

Figure 1.

Novel Reprogramming Strategy for LSC Selection (A) Schematic diagram depicting the way primary AML samples were reprogrammed using fluorescence-activated cell sorting (FACS) directly by de novo reprogramming, or post-xenograft sorting (xeno reprogramming) on CD45, CD34, and CD33 populations. (B) Workflow of select AML samples (AML 1, 2, 3, 4, and 9) that underwent both methodologies of reprogramming to generate iPSC colonies, de novo and xenograft reprogramming, respectively. AML samples were engrafted into NSG mice and assessed for human chimerism by pan-hematopoietic marker, CD45. Exclusive myeloid grafts were identified by exclusive expression of CD33 and considered leukemic and LSC⁺. Multilineage grafts displayed expression of CD33 and CD19 and considered to be healthy and LSC^-. Samples that were LSC⁺ (Table 1) were candidates for reprogramming for both de novo and xenografted methods. (C) Workflow of select AML samples (AML 5-8, 10-22) that underwent de novo reprogramming solely to generate iPSC colonies. Samples in which clear and distinct populations existed underwent reprogramming from multiple populations (CD34⁺CD33⁺, CD34^-CD33⁺, CD34⁺CD33^-, CD34^-CD33^-). (D) AML iPSC derived and MPB-iPSC imaged using phase-contrast microscopy. Scale bars are 500 μm. Morphologically, clones do not look distinct from each other and are consistent with a typical morphology of iPSC on iMEF.^22,23,38,49(E) Percent frequency expression of pluripotency markers SSEA3, TRA-1-60, OCT3/4, NANOG, and SOX2 by flow cytometry in representative iPSC colonies derived from AML patient sample in comparison to healthy control MPB iPSC and H9’s a hESC cell line. Statistical analysis done using two-way ANOVA and Šídák’s multiple comparisons test where no significance difference was detected. N = 2-5, iPSC samples; n = 7-12. Comparative expression levels have been previously reported.^50-52(F) Teratoma analysis of AML 1-iPSC lines derived via de novo reprogramming or xenograft reprogramming. Hematoxylin and eosin staining of teratoma sections at 10 weeks. Arrows denote pigmented epithelium (ectoderm), goblet cells (endoderm), and cartilage (mesoderm). Scale bar 100 μm. Frequency of tissue per independent mouse teratomas after one scored tissue section is recorded in bottom right corner. N = 3 mice transplanted per iPSC line.

Reprogramming to Naïve Versus Primed States

Patient samples selected for reprogramming were subsequently categorized based on the percent distribution of cytogenetically and molecularly defined subsets of AML described by Medinger and Passweg.^44,45 Eight out of the fourteen samples yielded classification (Fig. 2A) demonstrating representative breadth of molecular heterogeneity of AML disease from the proposed AML-iPSC library of which samples were detectable using ddPCR, Karyotyping, Cytoscan HD Array, and FISH (Fig. 2A). ddPCR uses digital PCR by separating DNA molecules into individual droplets based on water-oil emulsion droplet technology and amplifies the DNA within each droplet and analyzed based on Variant Allele Frequency (VAF) representing the number of droplets positive for the mutation compared to control,⁴⁶ whereas clinical cytogeneticists allowed analysis of human karyotypes involving several megabases of DNA and Karyotyping revealing chromosome number associated with aneuploid conditions, or chromosomal deletions, duplications, translocations, or inversions.

Figure 2.

Reprogramming strategy and AML sample section (A) Size distribution of cytogenetically and molecularly defined subsets of AML described by Medinger and Passweg^44,45 of the 22 AML patient samples analyzed in this study. Further details concerning patients and samples are included in Table 1. (B) Illustrative workflow depiction of reprogramming strategy using primed (SR+bFGF) versus naïve (LIF2i) reprogramming. Reprogramming is carried out on sorted populations using Sendai virus (SeV) containing polycistronic Klf4-Oct3/4-Sox2, cMyc, and Klf4 vectors in naïve (LIF2i) media for 21-28 days until iPSC arose and then transitioned to primed (SR+bFGF) media. For samples undergoing primed reprogramming, no transitional phase was required, as cells were always maintained in SR+bFGF medium. (C) Morphology and live cell TRA-1-60 staining of derived iPSC colonies prior to being picked and propagated. Scale bar 500um. (D) Average reprogramming efficiency of 6 AML patient samples reprogrammed by both naïve and primed methodologies (N = 6, AML samples; n = 3-5, the number of wells reprogrammed as a technical replicate), (E) Average reprogramming efficiency of 6 AML patient samples stratified by reprogramming a CD34⁺ population or CD34^- population independent of methodology used. Reprogramming efficiency is calculated as the number of colonies that arose divided by 200 000 cells reprogrammed, per well. (N = 6, AML samples; n = 8-12, the number of wells reprogrammed as a technical replicate). Data shows average with error bars representing ±SEM. All comparisons were made using unpaired student t-test *P<.05.

We first subjected AML patient cells to 2 distinct methods of reprogramming referred to as naïve or primed reprogramming, LIF2i or SR + bFGF, respectively (Fig. 2B). Uniquely, iPSCs are like embryonic stem cells in terms of their ability to exist in 2 different states, referred to as the naïve and primed states. Naïve cells are derived from the preimplantation blastocyst inner cell mass, whereas primed cells are derived from postimplantation epiblast cells.⁴⁷ They differ in terms of their cell morphology, gene expression, growth factor dependency, and presence of X chromosome activity in female cells. Although iPSCs derived from somatic cells maintain epigenetic signatures from their origin cell,⁴⁸ iPSCs derived into naïve conditions erase this epigenetic memory of their origin cell type, as seen by a reactivation of the X chromosome.⁴⁷ Cells in the naïve state can re-establish epigenetic memory by switching to the primed state, and epigenetic memory of iPSCs in the primed state impacts their differentiation by favoring differentiation into the cell type of their origin.^37,48 We have previously shown the refractoriness of reprogramming primary AML samples. Specifically, we demonstrated that a single AML patient sample (AML 15331) was able to reprogram and yield iPSCs with a leukemic aberration matching the primary patient sample only when reprogramming was carried out in naïve conditioned media (LIF2i).²²

To compare the reprogramming efficiency between naïve reprogramming (LIF2i) and primed (bFGF), we hypothesized that reprogramming in SR + bFGF media would enhance our overall reprogramming efficiency as well as our efficiency at deriving isogenic healthy iPSC clones devoid of leukemic aberrations. Reprogrammed cells were fed every other day with LIF2i media, beginning on day 2 or 3 from plating them into 6 wells, until colonies arose to be selected and expanded as individual clones (2-3 week’s posttransduction) (Fig. 2B; see method section for additional details). Once colonies arose, cultures were stained using a live cell imaging kit, for detection of the PSC marker TRA-1-60. iPSC colony emergence demonstrated positive expression denoting colonies as truly pluripotent (Fig. 2C and Supplementary Fig. S1A). Colonies were then plucked and expanded for approximately 2 weeks. For naïve reprogramming, iPSC colonies were subsequently transitioned to SR media supplemented with bFGF (SR + bFGF) following the 2 weeks. Samples undergoing primed reprogramming did not undergo this transition and were always maintained in SR + bFGF medium (Fig. 2B). A 2-tailed unpaired t-test was used to compare the two distinct methods of reprogramming, and no significant difference was observed among the 6 AML samples tested (Fig. 1D). Reprogramming efficiency is calculated as the number of colonies that arose divided by the number of cells reprogrammed (ie, 200 000 cells) per well. Thus, reprogramming efficiency is determined as a frequency on a patient-specific manner. Reprogramming efficiency ranged from ~0 to 5.9 (% ×10⁻²), compared to a previously reported efficiency of 10% (% ×10⁻²) for healthy blood cells using Sendai Virus.⁵³ This highlights the extreme rarity of reprogramming, and a frequency has not been reported in this manner for iPSC from non-cancerous cells, as the rate is so low. However, when samples were collapsed on either methodology used, then stratified based on reprogramming a CD34⁺ population versus a CD34⁻ population, a significantly greater reprogramming efficiency was achieved by reprogramming the CD34⁺ population (Fig. 2E) similar to previous observations.⁵⁴ Despite yielding no significant difference among LIF2i reprogramming and bFGF statistically speaking, a higher number of reprogrammed iPSC were observed among the AML patient samples via LIF2i reprogramming (Fig. 2D). Given our previous report only yielded one bona fide AML-iPSC clone using LIF2i,²² and others showed low-iPSC skewed often in favor of residual normal cells over cells of the premalignant or malignant clone, we used LIF2i for future reprogramming in this study to ideally achieve a higher yield of AML-iPSC aberration positive AML-iPSC lines.

Reprogramming Strategy and Pluripotency Validation

Despite the “barriers” that exist with successful reprogramming including the mixture of co-isolating normal and clonal leukemic cells in the BM and PB of patients with AML, their clonal heterogeneity holds the potential to be positively leveraged to derive both disease and normal iPSC lines in the same reprogramming process to derive paired isogenic and AML-iPSC from individual patients with AML (termed “de novo” samples). AML reprogramming to iPSC has been shown to be a rare event and limited to extremely rare subtypes of AML.^22-24 To improve the frequency of generating AML-iPSC with aberrations of the patient’s somatic leukemic cells, we proposed that reprogramming competency in the AML hierarchy may be unique to specific compartments. Using FACS, we isolated AML patient samples into cells expressing cell surface markers for CD34, CD33, and CD45 which are also used to classify patient disease in the clinic.^55,56 Samples were, thus, sorted for CD45 expression (CD45⁺) hematopoietic cells and then divided into 4 populations based on expression of CD33 and CD34 (Fig. 1A).

In addition to purification of de novo samples, a subset of AML de novo samples were transplanted into immunodeficient mice to determine whether leukemic cell progeny of cancer stem cells had adopted changes to reprograming capacity due to a human-mouse xenograft environment. Successful engraftment enabled reprogramming FACS-purified populations to enrich for functionally defined stemness in vivo (Fig. 1A). Furthermore, we hypothesized that progeny of leukemic initiating cells (LIC)⁴² that engraft these recipient mice, would potentially bestow an enhanced reprogramming efficiency of AML-iPSC and would also allow selective enrichment of leukemic versus normal hematopoietic cells as LIC dominantly reconstitute in mice. In addition, some reports have also speculated that AML cells do not reprogram due to the lack of cell division,³⁴ whereas our approach to reprogram active LSCs that engraft mice could allow an increase AML-iPSC generation. We compared reprogramming of AML samples by de novo reprogramming, xenograft reprogramming, or a combination in which select samples underwent both methods of reprogramming (Fig. 1B and 1C). Samples that underwent xenografting were assessed for human engraftment using pan-hematopoietic marker CD45, myeloid marker CD33, B-cell marker CD19, and stem and progenitor cell marker CD34 after 8 weeks postinjection (Fig. 1B). Human chimerism was assessed by human pan-hematopoietic marker, CD45. Myeloid grafts were identified by exclusive expression of CD33 and considered LSC⁺. Multilineage grafts displayed expression of CD33 and CD19 and were healthy HSCs-based engraftment and devoid of LSCs (LSC^-, Fig. 1B). To best maximize the rare event of reprogramming known to be enhanced in the CD34⁺ hematopoietic compartment, we reprogrammed FACS purified primitive (CD45⁺CD33⁺CD34⁺) and more mature myeloid populations (CD45⁺CD33⁺CD34^-) from each sample.⁴³ In rare instances, select samples had a CD45⁺CD33^-CD34⁺ purified and subjected to reprogramming in vitro (Fig. 1C). Since healthy hematopoietic cells are known to have reduced reprogramming potential with increased differentiation, CD34 was used to separate cells into more primitive (CD34⁺) and terminally differentiated (CD34^-) cell fractions.⁵⁴ To further promote CD34⁺ cell culture, in vitro reprogramming was attempted together with supplementation of StemRegenin 1 (SR1). SR1 is a small molecule which is well established to promote in vitro expansion of CD34⁺ cells⁵⁷ and carried out in naïve conditions media,²² which has been shown to have better success at generating AML-iPSC (Fig. 2). A total of 5 AML samples underwent both de novo and xenograft reprogramming (Fig. 1B), and an additional 17 samples underwent de novo reprogramming (Fig. 1C). All iPSC colonies that arose on iMEFs had indistinguishable morphology in comparison to a healthy MPB-iPSC control (Fig. 1D) and expanded similarly when picked and subsequently passaged every 6 days. Once cultured for approximately 3 to 4 weeks post-derivation, iPSC clones were dissociated on day 6 post-passage and tested for their expression of external pluripotency markers TRA-1-60 and SSEA3, and internal pluripotency markers OCT3/4, NANOG, and SOX2. No significant differences were observed among pluripotency markers from AML-derived iPSC and healthy hiPSC controls (Fig. 1E). When functionally interrogated for pluripotency using the teratoma assay⁵⁸ both methods of deriving iPSC from AML, de novo reprogramming or xenografting, demonstrated that AML-derived iPSC can give rise to all 3 germ lineages—endoderm, mesoderm, and ectoderm in vivo (Fig. 1F). A comprehensive iPSC characterization of a representative clone is provided from patient AML A374.1 (AML 1; Supplementary Fig. S1B-S1D).

Overall, despite the historically refractoriness to reprogram, 15 out of 22 samples obtained from 14 patients with AML were successfully reprogrammed the highest number reported to date, yielding a total of 129 patient with AML-derived iPSC lines (Supplementary Table S1).

Classification of Derived iPSC

To determine if leukemic or isogenic clones were derived, the mutation status of each clone was tested for the presence of established patient with AML mutations (DNA aberration) and subsequently classified as either aberration negative or aberration positive (Fig. 3A). Of the 129 iPSC lines generated, see Supplementary Table S1, all isolated iPSC colonies were treated as independent clones, expanded, and cryopreserved at multiple early passages into iPSC lines (Fig. 2B). Independent of method, of the 22 patient samples reprogrammed, 7 patient samples (AML 8-10, 12, 18-20) were unsuccessful in generating any iPSC and were thus not assessed (Fig. 3B and Table 2). All clones derived were assessed by methodologies previously reported and representative results of each methodology are shown in Fig. 3. Individual clonal information is described later in detail and provided in supplemental figures. Classification of each clone derived via aberration detection was imperative to determine whether reprogramming efficiency had been altered using our distinct novel methods. Moreover, classification would allow for clonal and subclonal representation to be captured.

Figure 3.

Detection of primary AML mutations in derived iPSC lines. (A) Illustrative depiction of aberration detection in AML-derived iPSC lines using selective methods previously described. (B) AML patient samples that did not yield any iPSC. (C) Karyotype results of primary AML A374.1 (AML 1) and 2 associated clones derived via de novo reprogramming and xenograft reprogramming showing an abnormal karyotype mirroring the primary sample, as well as karyotype of iPSC clone derived from patient with AML 15328 (AML 6) that did not contain any mutations corresponding to the expected karyotype based on the primary AML sample. (D) Gene copy number was assessed in AML A422 (AML 2) and derived iPSC by Affymetrix HD CytoScan microarray and analyzed using ChAS software. All 4 clones contain del(7) present in primary AML 2. (E) FISH was performed on AML A151.1 (AML 5) using CBFβ-MYH11 hybridization probe. When the gene fusion is present the foci pattern is 1 green, 1 red, 1 red-green. All clones derived from AML 5 show no gene fusion with a 2 green, 2 red pattern. (F) AML A295.1 (AML 7) and AML7-iPSC clone assessed by ddPCR for long arm (q) of chromosome 21. Copy number determined relative to control probe. Error bars represent Poisson 95% confidence intervals. (G) Droplet digital PCR results of AML A494 (AML 16) iPSC clones 1-8 by comparison to primary AML and healthy controls. Variant Allele Frequency (VAF) is displayed for droplet PCR results, probing for IDH2:c.515G>A. The primary AML sample contained the mutation, and the mobilized peripheral blood (MPB) used as a healthy control did not, as expected. All 8 clones tested contained the mutation, with VAFs similar to that of the primary sample.

All AML samples reprogrammed were divided into 4 categories based off the iPSC lines derived. Table includes a summary of the total number of clones derived per AML sample, patient ID number, source of sample, clinical stage, patient age, AML number provided in the manuscript, and the genetic/molecular abnormality associated with the primary sample. Category 0, made up of 7 samples, represents samples in which no iPSC were derived. Category 1 represents 8 samples in which aberration negative iPSC were solely derived after reprogramming. Category 2 represents 5 samples in which aberration positive iPSC were solely derived after reprogramming. Category 3 represents 2 samples in which both aberration positive and aberration negative lines were derived there by producing paired isogenic and AML-iPSC lines. Specifically, AML patient 13814.1 (AML 3) gave rise to 6 isogenic lines and 2 AML iPSC lines. Whereas, AML patient A485-3 (AML 14) gave rise to 4 isogenic lines and 4 AML iPSC lines.

Abbreviations: iPSC: induced pluripotent stem cells; AML: acute myeloid leukemia; PB: peripheral blood; BM: bone marrow.

Twelve clones derived from AML patient sample A374.1 (AML 1) were tested by G-band karyotyping and determined to all have inv(3) and del(7), along with other mutations detected in the primary sample (Fig. 3C and Supplementary Fig. S2A and S2B). Some clones had identical karyotyping patterns and thus likely originated from the same leukemic clone, for example, clones AML1-2 and AML1-4 (Supplementary Fig. S2A and S2B). Thus, 7 unique clones were captured from AML-1. Four clones were derived from AML patient sample A422 (AML 2) and tested using HD CytoScan Array and determined to capture del(7) coinciding to the primary AML sample in all clones (Fig. 3C). Of the 4 clones derived via de novo reprogramming from AML patient sample 13814.1 (AML-3) G-band karyotyping was used for assessment. Of these clones, 2 clones were found to have unique leukemic aberrations, and 2 clones were found to have a normal karyotype (Supplementary Fig. S2C and S2D). Four additional clones via xenograft reprogramming were derived from AML-3. However, karyotypic analysis revealed that these were all aberration negative and, therefore, presumed to be healthy isogenic clones (AML 1-5 to AML 1-12; Supplementary Fig. S2D). Interestingly, our karyotyping data demonstrated that select abnormalities were detected in the iPSC clones derived but not detected in the primary patient sample suggestive that these mutations are unique to the reprogramming process. However, this is unlikely as these are shared among AML-iPSC lines (Supplementary Fig. S2B) and, thus, is not generalizable. Alternatively, these may be patient-derived mutations that are below the limit of detection, and thus, the development of other high-resolution methods would be beneficial to interrogate the primary AML patient samples for these mutations, as these could represent rare clones only derived by reprogramming. Another AML patient sample assessed by karyotype 15328 (AML 6) was also found to be devoid of any leukemic aberrations in comparison to the primary sample (Fig. 3B and Supplementary Fig. S2E). Finally, AML A485 (AML 14) was also characterized using G band karyotyping (Supplementary Fig. S2F). The expected karyotype included common AML-associated mutations, such as del(7), which are typically associated with a poorer prognosis.⁵⁹ However, the resultant karyotype was abnormal for 2 out of 6 clones derived, but these clones all had a translocation between chromosomes 3 and 12, with breaks in 3q26 and 12p13 (Supplementary Fig. S2D). This mutation was not detected in the primary patient sample. The remaining clones derived from AML 14 all had normal karyotypes (Supplemental Fig. S1D) and, thus, could not be defined as AML in origin.

FISH was used to test AML patient sample 16150 (AML 4), A151.1 (AML 5), and A320 (AML 11) (Fig. 3E and Supplementary Fig. S3A and S3B). Ten clones derived from AML 4, and 3 clones derived from AML 5 did not contain inv(16) tested using the CBFβ-MYH11 fusion probe (Fig. 3F and Supplementary Fig. S3A). Five clones derived from AML 11 did not contain PML-RARα gene fusion found in the primary patient sample (Supplementary Fig. S3B).

AML patient samples A295.1 (AML 7), 16534 (AML 15), A494 (AML 16), 16308F (AML 20-2), A472-1 (AML 17-1), and A472-3 (AML 17-2) were probed for mutations using ddPCR. As previously mentioned, ddPCR is a method for performing digital PCR by separating DNA molecules into individual droplets based on water-oil emulsion droplet technology and amplifies the DNA through PCR within each droplet. The data are analyzed based on VAF, which shows the number of droplets that were positive for the mutation being probed for in each sample or control.⁴⁶ Only a single clone was derived from A295.1 (AML 7) and when tested using ddPCR for the presence of an isodecentric chromosome 21, it was found to be normal (Fig. 3F). AML 16 (A494) contained the point mutation IDH2:c.515G>A, a mutation in the isocitrate dehydrogenase gene involved in intermediary metabolism (IDH2 isocitrate dehydrogenase (NADP(+)) 2 [Homo sapiens (human)]—Gene—NCBI). All 8 clones derived and tested contained this mutation, as seen by the fractional abundance of the clones being similar to that of the primary AML sample (Fig. 3G). Of the 15 clones tested of 13051.1 (AML 13) and the 3 clones tested of 16534 (AML 15), none of them contained the NPM1c.863_864 ins_TCTG mutation present in their respective primary AML sample. This is a mutation in the gene encoding nucleophosmin, which is involved in centrosome duplication, cell proliferation, and protein chaperoning (NPM1 nucleophosmin 1 [Homo sapiens (human)] —Gene—NCBI) (Supplemental Fig. S4A and S4B). A472-3 (AML 17-2) was reprogrammed twice and was probed for 2 different mutations: ASXL1:c2725 A>T and IKZF1:c.476A>G. ASXL1 are the transcriptional regulators involved in chromatin remodeling (ASXL1 gene—Genetics Home Reference—NIH), while IKZF1 encodes a zinc-finger transcription factor also associated with chromatin remodeling (IKZF1 gene—Genetics Home Reference—NIH; Supplementary Figure S4C and S4D). Of the 10 clones tested in the first round of reprogramming this sample, all clones contained the ASXL1 mutation where, 1 of 10 clones contained just the ASXL1 mutation which may allow to capture disease progression (Supplementary Fig. S4B; Clone 1.4 absent for IKZF1). This patient (AML 17-2) also contains a RUNX1 mutation (Table 1; RUNX1:c.656_657insAAGG). RUNX1 is a transcription factor which commonly forms a complex with the cofactor, core binding factor beta (CBFβ), to activate genes that regulate the differentiation of HSCs into myeloid and lymphoid lines (RUNX1 gene—Genetics Home Reference—NIH). Unfortunately, we were unable to validate a successful probe that was able to detect this mutation in the primary patient sample; therefore, any clones that were derived were not probed for the RUNX1 mutation. The second time AML 17-2 was reprogrammed, and two different reprogramming methods were done LIF2i vs bFGF, respectively (Supplementary Fig. S4D). Of the 11 clones derived using primed reprogramming (SR + bFGF), all clones contained the ASXL1 mutation as well as the IKZF1 mutation. Note clones are denoted with the letter F in front to distinguish clones derived using primed reprogramming vs naïve reprogramming denoted by the letter L. Of the 6 clones derived using our standard reprogramming media LIF2i, all clones contained both mutations as well (Supplementary Fig. S4D). AML 17-1 (A472-1) is the diagnosis AML patient sample of A472-3 and harbors the same 3 mutations as previously described for A472-3, ASXL1:c2725 A>T and IKZF1:c.476A>G, and RUNX1:c.656_657insAAGG. Of the 10 clones derived using SR media + bFGF to reprogram this sample, 3 did not contain both mutation (Supplementary Fig. S4E; Clone F1.1, F1.7, F1.10 absent for IKZF1). Of the 8 clones derived using LIF2i reprogramming media, 4 clones did not contain both mutations (Supplementary Fig. S4E; Clone L1.4, L1.6, L1.11, L1.1, and L1.3 absent for IKZF1). Finally, one clone was derived from AML 20-2 (16308F), and it did not contain the point mutation NPM1 c.863_864 ins_TCTG present in the primary AML sample (Supplementary Fig. S4F).

Overall, although not all 22 samples donated from 14 patients with AML reprogrammed yielded iPSC, the detection of AML-associated mutations in iPSC may uniquely represent the heterogeneity of AML and capture the progression of the disease in various lines derived. In terms of detecting aberrations within the 15 AML samples successfully reprogrammed, it is imperative to attempt to probe for as many mutations in a patient sample as possible to better understand the heterogeneity of the disease. Based on these results, we successfully generated both aberration positive (AML-iPSC), aberration negative, and paired AML-iPSC and Isogenic iPSC for the first time.

Reprogramming Efficiency Correlates to AML Hierarchy

In establishing a novel method of reprogramming via partitioning AML samples based on immuno-phenotyping and FACS, we sought to determine if de novo reprogramming versus xenograft reprogramming would yield a greater efficiency in producing patient with AML-derived iPSC. Reprograming efficiency is calculated by dividing the number of resulting colonies by the number of input cells⁶⁰ which in our cases was 200 000 cells, per well. Five AML samples subjected to both methods of derivation were thus assessed independent of the type of iPSC generated, ie, aberration positive (AML-iPSC) or aberration negative. Of the 5 samples, only one sample AML 16626 (AML 9) was unsuccessful in generating patients with AML-derived iPSC (Fig. 4A). Intriguingly, the reprogramming efficiency via de novo reprogramming was on average higher in comparison to the same sample that underwent xenograft reprogramming, though not statically significant (Fig. 4B). Because there was no significant difference among methods of reprogramming, we separated the 5 individual samples and assessed their individual reprogramming efficiency based on the specific immuno-phenotype population that was subjected to reprogramming. This included the following populations for AML 13814.1 (AML 3) and 16150 (AML 4): CD45⁺CD34⁺CD33⁺, CD45⁺CD34^-CD33⁺, CD45⁺CD34⁺CD33⁻ and CD45⁺CD34⁻CD33⁻. Whereas A374.1 (AML 1) had 3 populations reprogrammed, CD45⁺CD34⁺CD33⁺, CD45⁺CD34⁻CD33⁺, and CD45⁺CD34-CD33, and AML A422 (AML 2), only had 2 populations reprogrammed, CD45⁺CD34⁺CD33⁺and CD45⁺CD34⁻CD33⁺ (Fig. 4C). Across all 3 samples in which a CD45⁺CD34⁻CD33⁻ population was reprogrammed, no colonies were derived (Fig. 4C). AML 13814 (AML 3) and 16150 (AML 4) both displayed significantly higher reprogramming efficiency in the more primitive population (CD45⁺CD33⁻CD34⁺) in comparison to any other phenotypic population reprogrammed. In contrast, both AML A374.1 (AML 1) and AML A422 (AML 2) had significantly higher reprogramming efficiency in the more mature double positive population (CD45⁺CD33⁺CD34⁺) in comparison to any other phenotypic population successfully reprogrammed (Fig. 4C). Interestingly, AML 1 and AML 4 were samples that gave rise to both aberration positive and negative iPSC lines, whereas AML 1 and AML 2 solely gave rise to aberration positive lines (Fig. 5A). We next sought to assess the reprogramming efficiency of the samples that solely underwent de novo reprogramming of which the 17 samples reprogrammed, 7 did not produce any iPSC colonies (Fig. 4D). Of the 17 samples, one sample AML A320 (AML 11) was excluded from the analysis, as its reprogramming efficiency was extremely high (0.078%; Supplementary Fig. S5A) and would thus skew the data, since its reprogramming efficiency was even superior to healthy MPB which demonstrated an average of 0.00825% (Supplementary Fig. S5B). Using these samples, we sought to discern the relationship between leukemic reprogramming and CD34, and thus, the efficiency was again stratified on either CD34⁺CD33⁺ or CD34⁺CD33⁻ populations. Although no significant difference was observed, there was on average a higher reprogramming efficiency in the double positive population presumed to contain an increase probability of capturing a leukemic stem or progenitor cell for reprogramming (Fig. 4E). Given that there are age-related phenotypic alterations in AML, we assessed whether age would correlate to reprogramming efficiency within the different subsets of populations reprogrammed and did not discern a relationship (Supplementary Fig. S5C-S5D). Overall, the efficiency of reprogramming of the AML patient samples in this study had a wide and unpredictable range from 0% to 0.078% (Supplementary Fig. S5C). Notably, reprogramming blood cells using Sendai virus has been reported to have an efficiency of 0.1%⁵³ and more recently, generation of iPSC from a variety of human primary fibroblast lines using an RNA-based approach has shown an even higher efficiency of approximately 7% (3132 colonies per well, per 500 input cells per well).⁶⁰

Figure 4.

Reprogramming efficiency of AML samples via immunophenotyping on distinct populations. (A) Illustrative depiction of 5 AML samples that underwent both de novo and xenograft reprogramming. Note AML 9 did not give rise to iPSC using either methodology. (B) Bar graph in which de novo reprogrammed samples versus xenografted samples are compared. Reprogramming efficiency was first calculated by determining the number of colonies that arose divided by the number of cells reprogrammed, per well, per sample. Technical replicates of individual AML samples were then averaged. N = 5 AML samples, n = 7-22, the number of wells reprogrammed as a technical replicate. (C) Reprogramming efficiency of AML samples that were FACS separated by CD34 and CD33 independent of methodology used for reprogramming. Populations that did not exist in the patient sample and, thus, were not reprogrammed are labeled as N/A. N = 1 AML sample, n = 2-30, the number of wells reprogrammed as a technical replicate per immunophenotype. (D) Illustrative depiction of 17 AML samples that de novo reprogramming. Note AML 8-10, 12, and 18-20 did not give rise to iPSC. (E) Average reprogramming efficiency based of the immunophenotype CD34⁺CD33⁺ or CD34⁺CD33^- in samples that underwent de novo reprogramming. Plotted dots represent reprogramming efficiency calculated by taking the average of each individual AML sample. N = 10 AML samples, n = 3-14, the number of wells reprogrammed as a technical replicate. (F) Average reprogramming efficiency based of the immunophenotype CD34⁺CD33⁺ or CD34⁺CD33^- in samples that produced aberration negative, or aberration positive iPSCs. Plotted dots represent reprogramming efficiency calculated by taking the average of each individual AML sample. N = 5-7 AML samples, n = 3-20, the number of wells reprogrammed as a technical replicate. (G) All reprogramming efficiency based of the immunophenotype CD34⁺CD33⁺ or CD34⁺CD33^- in samples in which paired isogenic and AML-iPSC lines were produced. N = 2 AML samples, n = 5-6, the number of wells reprogrammed as a technical replicate. Data points represent individual values with error bars representing ±SEM. All comparisons were made using either a Two-way ANOVA with Tukey’s multiple comparison or an unpaired student t-test *P<.05, **P<.01.

Figure 5.

Differentiation potential and cytogenetic summary of iPSC derived. (A) Graphical representation of the total number of iPSC colonies generated from 22 patient samples that were reprogrammed. All clones (iPSC lines) derived were tested for patient aberrations using FISH, karyotyping HD cytoscan array, and ddPCR. (B) Percent distribution of the type of derived iPSC lines within the 4 categories—no iPSC, aberration negative, aberration positive, or mixed (paired isogenic and AML-iPSC). 7/22 samples produced no iPSC colonies, 15/22 produced colonies, of which 5 produced aberration positive lines solely (AML-iPSC), and 7 produced only normal/healthy clones (aberration negative). (C) Summary of AML samples reprogrammed and the category of mutations they fall under. 22 primary AML samples were reprogrammed, representing 7 different categories of mutations. Representative immunocytochemical images of paired isogenic and AML-iPSC lines of (D) βIII-Tubulin counterstained with Hoechst 33342, (E) early hepatocyte marker Alpha-fetoprotein (AFP) counterstained with Hoechst 33342, and (F) smooth muscle actin (SMA) counterstained with Hoechst 33342. MPB-iPSC and hESC line H9’s were used as control hPSC. All scale bars represent 100 μm.

Based on our results obtained and goal of characterizing AML-iPSC on the presence or absence of driver mutations, we stratified patients with AML into 4 categories (Table 2). Category 0 encompasses 7 samples in which no iPSC were derived. Category 1 represents 8 samples in which aberration negative iPSC were solely derived, whereas category 2 represents 5 samples in which aberration positive iPSC were solely derived after reprogramming. Finally, category 3 represents 2 samples in which both aberration positive and aberration negative lines were derived thereby producing paired isogenic and AML-iPSC lines. Specifically, patient with AML 13814.1 (AML 3) gave rise to 6 isogenic lines and 2 AML iPSC lines, whereas patient with AML A485-3 (AML 14) gave rise to 4 isogenic lines and 4 AML iPSC lines (Table 2). Using this stratification, we next determined whether samples that gave rise to aberration negative iPSC lines solely (category 1), aberration positive iPSC lines solely (category 2), or a mixture of both types of lines (category 3; paired isogenic and AML-iPSC lines) would yield distinct reprogramming efficiencies (Fig. 4F and 4G). Interestingly, samples in which aberration negative colonies were only derived (category 1) had significantly higher reprogramming efficiency in the CD34⁺CD33⁻ population. In contrast, samples in which aberration positive colonies were only derived (category 2) yielded significantly higher reprogramming efficiency in the CD34⁺CD33⁺ population (Fig. 4F). In rare instances in which a sample gave rise to both aberration positive containing clones and aberration negative containing clones (see category 3; paired Isogenic and AML-iPSC clones of Table 2), the CD34⁺CD33⁻ subpopulation demonstrated a significantly higher reprogramming efficiency (Fig. 4G). This suggests that AML clones reside in a variety of immunophenotypic compartments (compare Fig. 4F and 4G). However, there is a greater probability of obtaining AML-iPSC (aberration positive) clones when reprogramming the double positive CD34⁺CD33⁺ population. Moreover, there was no difference in TRA-1-60 expression across these categories (Supplementary Fig. 5E and 5F). Thus, in future studies, it may be advantageous to use the myeloid marker CD33 in combination with the hallmark stem cell marker (CD34) in deriving AML-iPSC or disadvantageous if the goal is to produce aberration negative lines (devoid of the primary patient mutation) from an patients with AML.

Stratification of Derived iPSC Lines and Germ Layer Differentiation Capacity

In previous reports from our group and others, AML patient cells have been notoriously difficult to reprogram compared to healthy counterparts. AML either does not produce any reprogrammed colonies or it produces colonies that do not contain leukemic aberrations and thus likely arise from healthy progenitors within the patient sample. Despite these barriers, we successfully reprogrammed 15 of 22 AML samples, of which 7 AML samples were found to contain aberrations related to the primary patient sample and thus yielded bona fide AML-iPSC lines (Fig. 5A). Having successfully achieved AML-iPSC from 7 diverse AML samples gives us the unique opportunity to use this library for further investigations and proves the hypothesis that reprogramming primary AML samples to iPSC is possible, but achieving AML-iPSC is a rare event, as seen by the frequency (23%) of only 6 out of 22 samples yielding aberration positive clones, 2 of which yielded a combination of both (9%) versus (32%) failing to reprogram or solely generating aberration negative iPSC lines (36%; Fig. 5B). Excitingly, this library of reprogrammed AML patient samples can model 8 different categories of AML-associated mutations (Fig. 5C). Previously, the collective efforts from other groups who have published results related to AML iPSC excluding our previous work, have only represented 2 categories: TP53 mutations and MLL fusion genes, using a total of 5 patient samples across 4 publications^23,24,26,40 whereby the same lines have been studied across publications. Intriguingly, samples either gave rise to aberration negative iPSC (presumed to be isogenic healthy) or aberration positive (AML-iPSC) solely, except for AML 3 and AML 14 in which both iPSC were derived (Fig. 5C), allowing for the library to be stratified into 4 distinct categories providing unique opportunities for future studies using a broad and representative library of AML-iPSC.

Further validation of the pluripotent potential of our library was next assessed by performing guided differentiation of paired isogenic and AML-iPSC clones into neural (ectoderm), hepatocytes (endoderm), and cardiomyocytes (mesoderm) (Fig. 5D-5F). Neural lineage differentiation was evaluated in 2 stages; the initial analysis encompassed formation of neural tube-like rosette structures within 5-7 days of differentiation followed by differentiation into neurons, which represent 1 of 3 major cell types found in the central nervous system.^61,62 All clones tested successfully gave rise to neural precursors (rosette cells; data not shown) as well as expressed the pan-neuronal marker βIII Tubulin to similar levels to that of control healthy MPB iPSCs (MPB-iPSC) and the hESC line H9’s when qualitatively assessed (Fig. 5D). We also differentiated additional AML-iPSC lines independent of the paired sample and found no observable differences in neural differentiation capacity in comparison to control hPSC lines (Supplementary Fig. S6A). Next, paired iPSC were differentiated into hepatocytes through culture in endoderm-supportive media followed by hepatocyte maturation media containing hepatocyte growth factor as previously described.^36,37 No clear expression differences were observed for the early hepatocyte marker Alpha-fetoprotein (AFP) (Fig. 5E and Supplementary Fig. S6B). Finally, patient with AML-derived iPSC were differentiated in cardiomyocytes based on a previous report whereby embryoid body aggregates are seeded on gelatin-coated wells at day 7 of differentiation and subsequently cultured in cardiomyocyte-specific supportive media.³⁹ Both paired isogenic and AML-iPSC (Fig. 5F), as well as multiple iPSC lines per patient per patients with AML-derived iPSC lines (Supplementary Fig. S6C), yielded qualitatively similar expression levels of smooth muscle actin (SMA) in comparison to healthy control hPSC. A summary of all 129 cell lines and their respective characterization stratified by the categories of patient with AML lines previously described in Table 2 (Supplementary Tables S2-S4).

Discussion

Our study has tested Sendai and Lentiviral delivery of reprogramming factors, naïve versus primed conditions, de novo versus xenograft conditions, and prospective cell purification and fractionation using 20 independent patients with AML. Despite the technical challenges of reprogramming human cancer cells, our group has established an unprecedented library of 129 patient with AML-derived iPSC of which 77 are leukemia-specific genetically aberrant AML-iPSC. Through directed differentiation assays, we demonstrated the pluripotent potential of select iPSC lines and thus the board utility of our library. We also derived aberration-negative iPSC from a subset of these same patients with AML, providing an invaluable isogenic control for future direct molecular comparisons. This library aims to capture both intra- and inter-patient heterogeneity and provides a novel model system for the study of AML biology.

Using a combination of immunophenotyping and FACS, we successfully reprogrammed 15 patients with AML samples harboring a wide variety of genetic aberrations using nonintegrating plasmids from a diverse genetic background. Utilizing our novel approach of phenotyping CD34(+)CD45(+)CD33(+/−) cells prior to reprogramming de novo and xenografted AML patients cells, successful derivation of multiple subclones and aberration negative iPSC lines were derived, the largest library to date. Given that no additional genetic analysis was done to further characterize our patient with AML-derived iPSC library beyond what has been presented, we cannot presume that aberration negative iPSC lines are truly “healthy.” Nevertheless, by generating both iPSC devoid of leukemic mutations (aberration negative), containing leukemic mutations (aberration positive), or paired isogenic and AML iPSC lines, we were able to stratify our library into 3 categories that will enable us to evaluate hematopoietic differentiation and intracellular signaling differences between different cytogenic patients with AMLs-derived iPSC in future studies. Moreover, our iPSC lines within category 1 (devoid of leukemic mutations) may serve as high importance, as these lines are excellent candidates for future CRISPR-Cas9 gene editing experimentation. Recently, iPSC have been utilized to investigate leukemogenesis and to identify compounds targeting AML.^{24,26,40,63-65} Previous studies using CRISPR-Cas9 gene editing have demonstrated the possibility to introduce distinct mutations in iPSC and study stepwise, stage-specific leukemia progression.²³ By using this approach, leukemia models can be created to compare various mutations (such as missense vs. nonsense mutations) in endogenously produced proteins (eg, RUNX1).⁶⁶ Interestingly, within this category, we observed almost completely successful reprogramming in the CD34⁺/CD33⁻ population. Thus, future studies interested in deriving patient with AML-specific iPSC devoid of clinically defined leukemic mutations would benefit from incorporating such methodologies. In terms of the phenomenon associated with cancer stem cells (CSC) referred to as “phenotype switching,” a term first introduced by Hoek in 2008,⁶⁷ now commonly used to describe transitions between phenotypic states without implying anything about the nature of the alterations in the cells’ biological characteristics, we assessed the relationship of age and reprogramming efficiency. Given that our study incorporated selecting samples expressing some level of CD34 positivity, and age definitively changes the phenotype of patients with AML, we assessed and did not find a correlation between age and reprogramming efficiency within any stratified subpopulations isolated in this study, for example, CD34+CD33−, CD34−CD33− or CD34+CD33+. Notably, AML samples that are CD34 positive initially can become CD34 negative and simultaneously express cKit. Thus, we caution CD34 enrichment for pluripotent reprogramming, as this could still be dependent on phenotype, for example, CD34-AML samples that enrich for CSC features in c-Kit+ cells, as an example.⁶⁸ This is something that will have to be further investigated in future reprogramming of AML samples that present with CD34 negativity.

Consistent with previous results, our data indicate that clonal representation of the original cells in the iPSC is often in favor of normal cells instead of the malignant clones. Moreover, it is reprogramming, and not the in vitro stimulation, that accounts for this bias, which appears to be conferred by AML-associated genetic lesions, but not others, while some genetic abnormalities seem to be incompatible with reprogramming. In contrast, elevated reprogramming efficiency was observed in category 2 iPSC in the CD34⁺/CD33⁺ population suggesting a greater probability of reprogramming a leukemic stem and/or progenitor cell harboring a leukemic mutation. We suggest this finding to be independent of varying levels of pluripotency factors given the demonstration that TRA-1-60 frequency was consistent across AML-iPSC lines from distinct stratification categories. Thus, if bona fide AML-iPSC lines are of a priority to derive, future studies would benefit by utilizing our immunophenotyping approach as a means of overcoming the refractory nature of reprogramming AML patient cells. This refractoriness to reprogramming is paradoxical, and the refractory nature of leukemic cells remains unclear.

Despite utilizing the limited techniques of ddPCR and karyotyping, we successfully demonstrating that our AML-iPSC lines harbor cytogenetic mutations originating from donor patient disease cells. In addition to cytogenetic abnormalities, it is well established that the epigenome, which regulates gene expression at the level of chromatin conformation, is aberrant in AML.^69,70 NGS and array-based approaches capable of interrogating the epigenetic landscape (ie, DNA methylation and histone modifications via Methyl-seq or ATAC-seq) in future studies would provide an in-depth characterization of our library and in turn could lead to the identification of rare epigenetic elements that describe AML etiology. Given that our efforts also provided aberration-negative iPSC from a subset of the same patient with AML, our study also offers an invaluable genomic isogenic control for such proposed molecular comparisons independent of background DNA.

The AML-iPSCs model has the potential to study broad disease concepts while eliminating the need to use quantity-limited patient samples, for which the ideal ex vivo culture conditions have recently improved but have not been completely elucidated.⁵⁷ Although we have yet to assess this, previous reports have successfully demonstrated that AML-iPSC can re-establish a leukemic phenotype upon hematopoietic differentiation with engraftment capacity.^23,24 As this is the only report of engraftment of leukemogenic engraftment from cells derived from hPSC, our study can add to these efforts in assessing the leukemia-initiating potential of a larger cohort of AML-iPSC. In addition, the ability of AML-iPSC to break down the disease into clones can provide an avenue to pursue targets of differentiation therapy, the discovery of which has proven to progress to previously unheard of cure and remission rates of 80% and 90%, respectively for patient with acute promyelocytic leukemia .⁷¹ Our group invested in a new campaign to develop techniques to reprogram primary AML leukemic blasts into iPSC to serve as a future library for investigating leukemogenesis, whereby implementation of this cellular models may lead to the characterization of rare clonal contributions of genetic and epigenetic abnormalities and lead to a better understanding of patients with AML etiology.

Acknowledgments

We want to thank Ronan Foley, Brian Leber, and Anargyros Xenocostas for providing primary human MPB and AML samples. We would also like to thank B.T., for their assistance with initial lentiviral transduction; J.R for in vivo experimental assistance, and A.L.B for valuable discussions and input. We would also like to acknowledge D.G, M.D, and J.C.R in editing and reviewing the manuscript. Finally, we thank the Centre for Applied Genomics at the Hospital for Sick Children for performing all karyotypic screening, droplet digital PCR, and Affymetrix HD CytoScan microarray analyzed using ChAS software.

Funding

D.P.P. was supported by the Jans Graduate Scholarship in Stem Cell Research, the Michael G. DeGroote Doctoral Scholarship, and the Ontario Graduate Scholarship. D.G. was supported by the Jans Graduate Scholarship in Stem Cell Research. J.C.R. was supported by the Canadian Institutes of Health Research Doctoral Award, the Michael G. DeGroote Doctoral Scholarship, and an Ontario Graduate Fellowship. This work was supported by Canadian Institutes of Health Research foundation grant FRN 159925 to M.B. and Braley Foundation as was approved by the Stem Cell Oversight Committee in Canada. M.B. holds the Canada Research Chair Program CRC Tier 1 in Human Stem Cell Biology and the Michael G. DeGroote Chair in Stem Cell Biology.

Conflict of Interest

The authors indicated no potential conflicts of interest.

Author Contributions

D.P.P., D.G., M.D. performed the experiments. J.C.R. and A.L.B. aided with the transplant assay design. D.P.P., D.G., M.D., and M.B. designed the experiments and interpreted data. D.P.P. and M.P. wrote the manuscript. M.B. directed the study.

Conception and design, collection and/or assembly of data, data analysis and interpretation: D.G. Conception and design, collection and/or assembly of data, data analysis and interpretation, manuscript writing, final approval of manuscript: D.P.P. Collection and/or assembly of data, data analysis and interpretation: M.D. Collection and/or assembly of data, data analysis, and interpretation: J.C.R. Collection and/or assembly of data: B.T. Data analysis and interpretation, administrative support, provision of study material or patients: A.L.B. Collection and assembly of data: K.V. Collection of data: A.E. Collection of data: A.Q. Conception and design, financial support, administrative support, provision of study material or patients, manuscript writing, final approval of manuscript: M.B.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Institutional Review Board Statement

This study and all the experimental protocols were conducted in accordance with the Animal Research Ethics Board of McMaster University (AUP: 19-11-29).

Author notes

Contributed equally.