Adaptation to ex vivo culture reduces human hematopoietic stem cell activity independently of the cell cycle

Key Points • Substantial attrition of HSC function occurs within 24 hours of ex vivo culture independently of cell cycle progression.• Inhibition of JAK/STAT signaling during culture adaptation via ruxolitinib improves HSC function ex vivo.

(Polyplus) added to enhance transfection.Plasmids were obtained from the GSK research facility (Stevenage, UK).HEK293T cells were cultured (37°C & 5% CO2) and 5mM of Sodium Butyrate (Sigma) was added to enhance transfection at 24 h.Cells were harvested at 72 h post transfection and LV containing medium was separated from cell debris by centrifugation (1,000g for 20 minutes at 4°C) with the supernatant clarified using 0.8µm and 0.45µm vacuum filters (Corning).LV particles were concentrated by ultracentrifugation (5,000g for 20 h at 4°C), then the supernatant was discarded and the pellet air dried.Pellets were resuspended in SCGM and stored at -80°C.
LV transduction.Flow-sorted cells were cultured in GT media in a 96 well flat bottom plate coated with 33.3 µg/ml of Retronectin (Takara).The 62h protocol consisted of pre-stimulation in GT media (24h), transduction with a LV containing GFP (14h), an interim incubation in GT media without the vector (10h) and a second hit of transduction (14h).

Cell cycle assays.
For time to first division assays, LT-HSCs were single cell sorted into 96 well u-bottom plates containing 100µl of indicated media per well either untreated (UNTR) or PD treated (200nM), centrifuged at 500g for 5 minutes and manually counted every 12h for 96h using a light inverted microscope.For phoshoRb stainings, cells were harvested from culture at indicated time-points and fixed with neat Cytofix/Cytoperm (BD Biosciences) for 10 minutes at RT. Cells were washed with 1x Permwash (BD Biosciences) and stained overnight with anti-Rb (phospho S807/S811, conjugated to Alexa647 fluorochrome) (Cell Signalling Technologies).Cells were again washed with 1x Permwash and stained with 0.5µg/ml of DAPI (ThermoFisher Scientific) for 15 minutes.Cells were resuspended in PBS + 3% FCS for flow cytometry analysis.DAPI was recorded on a linear scale and samples analysed at £ 30 events/second.For Ki-67 staining, cells were harvested from culture at indicated time-points and fixed with neat Cytofix/Cytoperm (BD Biosciences) for 10 minutes at RT. Cells were washed with 1x Permwash (BD Biosciences) and stained with anti-Ki-67 (conjugated to FITC fluorochome) (BD Biosciences) for 20 minutes at room temperature.Cells were again washed with 1x Permwash and stained with 0.5µg/ml of DAPI (ThermoFisher Scientific) for 15 minutes.Cells were resuspended in PBS + 3% FCS for flow cytometry analysis.DAPI was recorded on a linear scale and samples analysed at £ 30 events/second.Cell size measurements.100 CB LT-HSCs per well were sorted into a 384 well plate containing MEM media (see Methods) either UNTR or PD treated (200nM).Cell images were taken every 24 h in bright field (20x magnification) on a Leica DMI300 B microscope using the MetaMorph Microscopy Automation and Image Analysis Software.Cell size was analyzed using ImageJ measuring 25 cells/well and expressed as cell diameter (µm).
Mitochondrial activity assay.CB LT-HSCs were cultured in EXPER media, harvested at indicated time-points and stained with the cationic dye Tetramethylrhodamine (TMRM) (100nM) (Life Technologies) which accumulates in active mitochondria for 40 minutes at 37°C + 5% CO2.Cells were washed and resuspended in an appropriate volume of PBS + 3% FCS for flow cytometry analysis.
Apoptosis assay.mPB LT-HSCs were cultured in GT media and CB LT-HSCs were cultured in EXPER media.Cells were harvested at indicated time-points, washed and resuspended in fridge cold PBS containing Annexin-V/PE (1/20) (BD Biosciences) and 7-AAD (1/20) (BD Biosciences) for 15 minutes at RT. 1X Annexin-V binding buffer (BD Biosciences) was added before immediate flow cytometry analysis.
Bulk LT-HSC differentiation assay with RUX.250 mPB LT-HSCs were sorted into 96 well flat bottom plates containing MEM media.After 2.5h in culture, RUX (5nM, 10nM, 50nM, 100nM or 500nM, Selleckchem) or DMSO vehicle control was applied to cultures.Cells were harvested at day 14 of culture into a 96 well u-bottom plate, stained with an antibody mix (Table S15, Panel F) for 20 minutes at RT and washed (100µl per well of PBS + 3% FCS) before high-throughput flow cytometry analysis of colony size and lineage output.Representative gates for analysis are shown in Fig. S8.
Serial colony replating: mPB HSC pool cells (CD34 + CD38 -CD45RA -) were isolated by flow cytometry and seeded at 400 cells per well in 96 well plates and cultured in GT, GT media with low TPO (20ng/ml) or EXPER media. 2 h after cells were placed into culture, RUX (5nM, 10nM, 50nM or 500nM, Selleckchem), pan-Caspase inhibitor (Z-VAD(OH)-FMK, Cayman Chemical; 100nM), UM171 (StemCell Technologies; 35nM) or DMSO (Sigma) was added to the media.After 72h of culture, cells were harvested and washed with FACS buffer (PBS + 3% FCS).Cells from each condition were split into two dilutions: either containing 20% or 80% of cells.These two dilutions were placed into H4034 Methocult medium (Stemcell Technologies) supplemented with 10ng/ml Flt3-L (Peprotech), 10ng/ml IL-6 (Peprotech) and 10U/ml Penicillin-Streptomycin (ThermoFisher Scientific) and split into duplicate wells (2 x 10% cell suspension and 2 x 40% cell suspension) of a 6 well plate.After 14 days, the number of colonies in each condition was counted using the Stemvision analyser (Stemcell Technologies) and then manually counted to adjust StemVision automated analysis (primary plating analysis).After counting, the duplicate well contents were harvested and washed with FACS buffer (3%FBS in PBS) before replating.The harvested cells were again split into two dilutions (either containing 20% or 80% of cells) and placed into H4034 Methocult (Stemcell Technologies) with added cytokines and Penicillin-Streptomycin and were then split into duplicate wells.After a further 14 days, the colonies were again counted by Stemvision followed by manual adjustment (secondary plating analysis) and replated once again.After a further 14 days, the colonies were again counted (tertiary plating analysis) as before.Colony counts shown are from the replating with 80% of cells and assigned by manual image analysis.

Primary xenograft transplantation.
All experimental cohorts were >11 weeks old.Only female cohorts were used for experiments involving NSG animals.For primary transplantation experiments, NSG mice were sub-lethally irradiated (2.4 Gy) 24 h prior to transplantation.For intrafemoral (IF) injections, mice were anesthetized with isoflurane and transplanted with the indicated cell dose in PBS + 0.1% Pen/Strep (ThermoFisher) (25µl).Following transplantation, mice were injected subcutaneously with the analgesic buprenorphine (Animalcare) at 0.1mg/kg.For intravenous injection, mice were transplanted with a cell suspension (max 150µl volume) in PBS + 0.1% Pen/Strep by tail vein injection.For all xenograft experiments involving cell culture, injected cell doses at time-points are representative of the cell count at the time of sort (0 h). Mice were bled at 8 weeks post-transplant from the tail vein.5 drops of blood were collected and Pancoll (PAN-Biotech) added to each sample.Density gradient centrifugation was performed at 500g for 25 minutes with the brake off.The MNC layer was collected and taken for antibody staining (Table S15; Panel C) for 20 minutes before washing and resuspension in PBS + 3%FCS.Mice were culled and bone marrow harvested 18-20 weeks for primary transplantation experiments.The femur and tibia bones from the two hind legs were taken and for IF injected mice, the injected femur was analysed separately.Bone marrow was stained in an antibody panel (Table S15; Panel C) for 20 minutes before washing and resuspension in PBS + 3% FCS for flow cytometry analysis.

Secondary xenograft transplantation.
Both male and female animals were used for experiments involving NSG-SGM3 animals.For secondary transplantation experiments in the EXPER culture system, NSG-SGM3 mice were irradiated using 2.25 Gy 24 h prior to transplantation and primary BM samples were thawed in X-VIVO 10 media (Lonza) + 50% FBS (Wisent) supplemented with Dnase (100 μg/ml, Roche).Viable (SytoxBlue -) (ThermoFisher Scientific) human CD45 ++ cells were sorted (Table S15; Panel D) on the Aria Fusion (BD Biosciences).Cells were pooled based on condition and intrafemorally injected in three doses.Mice were culled at 8 weeks, bone marrow harvested, stained in an antibody panel (Table S15; Panel E) and analysed by flow cytometry with the same methodology as in primary transplantation experiments.Secondary transplantation experiments for EXPER conditions were kindly performed at UHN, Toronto.
For secondary transplantations of experiments in the GT culture system, primary mouse BM was thawed by dropwise addition of pre-warmed Iscove's Modified Dulbecco's Medium (IMDM) (Thermo Fisher Scientific) supplemented with 0.1mg/ml Dnase (Lorne Laboratories) + 50% FCS.Cells were counted and injected in three doses by IV injection (max 150µl volume) in PBS + 0.1% Pen/Strep.Mice were culled at 8 weeks, bone marrow harvested, stained in an antibody panel (Table S15; Panel C) and analysed by flow cytometry with the same methodology as in primary transplantation experiments.Mice were considered engrafted with the same criteria as for primary xenograft transplantation experiments.
Smart-Seq2 adapted protocol.Single cell RNA-Sequencing (scRNA-Seq) libraries were prepared using an adapted Smart-Seq2 protocol 2 .A cell lysis buffer was prepared of 20U/µl SUPER-In RNase inhibitor (Thermo Fisher Scientific) and 0.4% Triton-X100 (ratio of 1:19).The lysis buffer was added to a mix of DTT (Thermo Fisher Scientific) (final concentration of 0.57mM), dNTPs (Invitrogen) (final concentration 0.11mM) and nuclease free water (Thermo Fisher Scientific) and 4µl aliquoted per well in a 96 well PCR plate and stored at -80°C.Upon thawing and single cell sorting, an annealing mix was added (containing External RNA Consortium Controls (ERCC) RNA Spike-In mix diluted 1:3x10 6 (Thermo Fisher Scientific)), 10µM Oligo-dT30 VN (Thermo Fisher Scientific) and nuclease free water (ThermoFisher Scientific) and annealing performed (72°C for 3 minutes).The plate was briefly centrifuged (1,000g for 30 seconds at 8°C) to collect liquid at the bottom of the well, the reverse transcription master mix added (containing 10U/µl SMARTScribe reverse transcriptase (Thermo Fisher Scientific), 1U/µl SUPER-In RNase inhibitor (Thermo Fisher Scientific), 2µM Template Switching Oligo (TSO), 5X First Strand Buffer (Thermo Fisher Scientific) and nuclease free water (Thermo Fisher Scientific)) and reverse transcription performed (~2.5 h).The plate was again briefly centrifuged (1,000g for 30 seconds at 8°C), the PCR master-mix added (containing final concentration of 166nM IS-PCR primer, 2X KAPA HiFi HotStart ReadyMix (Roche) and nuclease free water) with 23 PCR cycles performed to account for the low RNA content of quiescent HSCs (~3 h).Plates were stored at -20°C until PCR purification.RT Ampure XP Beads (Beckmann Coulter) were mixed and incubated with PCR product (1:0.6/0.7 ratio) and PCR purification performed as in 2 , to remove fragments of a small, suboptimal length.The Biomek FXP Automated Workstation was used for experiments involving >3 plates.The quality of material was checked using the Agilent High Sensitivity DNA kit (Agilent) following manufacturer's instructions.The Quant-iT PicoGreen dsDNA kit and reagents (Thermo Fisher Scientific) were then used to determine the concentration of DNA (triplicate values per plate per cell condition) and dilution plates created (using elution buffer as diluent) for all samples to achieve 0.1-0.15ng/µl.Tagmentation was performed (55°C for 10 minutes) using the Nextera XT DNA library preparation kit (Illumina) mixing 2.5µl Tagment DNA Buffer, 1.25µl Amplicon Tagment Mix and 1.25µl of DNA per sample.The Tn5 transposase was stripped using the NT buffer (Illumina) (1.24µl per sample) and a PCR mix added (containing 3.74µl of NPM mix (Illumina) and 1.24µl of appropriate i7 and i5 primers (Illumina) per sample) with PCR performed (~30 mins) to amplify adapter-ligated sequences.PCR product purification was performed in two successive steps using RT Ampure XP beads (1:0.5, 1:0.3) to yield final average amplicon lengths of ~400-700 bp.
scRNA-Seq experimental design.Two scRNA-Seq datasets were generated in this study: a timecourse of LT-HSCs cultured in EXPER conditions (Dataset 1), and a time-course of LT-HSC cultured in GT conditions (Dataset 2).PD treated (200nM) LT-HSCs were included in both datasets at indicated time-points.Dataset 1 was acquired over 2 independent experiments (Batch 1 and 2) using LT-HSCs isolated from CB samples from 2 independent pools of male donors at the time-points and conditions indicated in Table S16.Dataset 2 was acquired over 4 independent experiments (Batches 1 to 4) using LT-HSCs isolated from mPB samples from 4 independent healthy male donors at the timepoints and conditions indicated in Table S17.UNTR and PD treated conditions were always paired at matched culture durations in the same batches.
Two integrations were performed in this study: i) all CB LT-HSCs from 0 h and UNTR timepoints of Dataset 1 (Integration 1); ii) all single cells of Dataset 1 and Dataset 2 (UNTR and PD treated; Integration 2).Integration 1 was performed with 2 independent methods: i) Seurat 4 method; ii) Scanpy method.Integration 2 was performed with the Seurat 4 method.All bioinformatic methods described in extended description of bioinformatic methods below.

Statistical analysis.
Analysis of the HSC frequency/Long term repopulating cell frequency (%LTRC) from transplanted populations was performed by using Extreme Limiting Dilution Analysis (ELDA) software (https://bioinf.wehi.edu.au/software/elda)taking into account the number of engrafted mice, the total mice used and the cell dose transplanted.For analysis of a statistical difference in the HSC frequency within two groups, a Chi-Squared test was performed within the ELDA software.For the analysis of in vitro serial replating data, raw colony counts per technical duplicates were average and rounded up.These colony counts (Table S12) were fitted to a generalized linear mixed-effects model (lme4 library, glmer function in R, Poisson distribution) using donor as a random effect.EM means pairwise comparison was then performed with emmeans.
Flow cytometry data was analysed using FlowJo software (v10).For analysis of colony data derived from single sorted LT-HSCs, FlowJo v.10 gating statistics were exported and data further analysed in R Studio (v.1.2).Graphpad Prism (v9.3), python (v3.8.6) and R Studio (v1.2) were used for the creation of all plots.Statistical analysis between multiple groups was performed in Graphpad Prism or R Studio.Normality of data was deduced from the Shapiro-Wilks normality test.For statistical analysis between two groups a parametric test (Students t-test) or a nonparametric test (Mann-Whitney U-test) was performed.For analysis between multiple groups, an analysis of variance (one or two-way ANOVA) test was performed.All statistical tests were performed with a confidence interval of 95%.

Extended description of bioinformatic methods
scRNA-Seq quality control.Read alignment was performed using GSNAP 3 against Ensembl genes and initial quality control (QC) was performed by FastQC 4 .The count matrix was generated by HTSeq 5 .Additional QC was then performed in the bglab package 6 using the determined thresholds shown in Table S18, yielding the number of cells passing QC reported in Table S16 for Dataset 1 and Table S17 for Dataset 2.
Seurat 4 pipeline for batch correction and batch/dataset integration.Seurat objects were created using the Seurat package (v4) 7 .Genes from raw counts were filtered if not detected in >3 cells.The function SCTransform was used to perform normalisation and variance stabilisation, the mitochondrial genes percentage was regressed out.All other parameters were left as default if not mentioned specifically.For cell cycle regression, S and G2-M cell cycle scores were calculated using the CellCycleScoring function on the object following SCTransform.The values of G2-M scores were subtracted from the values of S scores resulting in the difference of the cell cycle scores.SCTransform was applied on the Seurat object again regressing the mitochondrial genes percentage and the difference of the cell cycle scores.For batch/dataset integration, Seurat objects to be integrated or batch corrected were curated into a list.3,000 features were selected with the function SelectIntegrationFeatures. The list of objects was prepared to integrate using the function PrepSCTIntegration.FindIntegrationAnchors function was used to find a set of anchors for integration with the k.filter parameter set to 100.Integration was performed with the function IntegrateData.PCA was computed on the batch corrected/aligned Seurat objects with the RunPCA function.UMAPs were computed using the RunUAMP function with the dims parameter at 1:30.Pseudotime analyses.The Monocle3 package (version 1.2.9) 8 was used to determine the pseudotime ordering of samples.Seurat integrated objects were converted to Monocle3 objects with the function as.cell_data_set from the package seurat-wrappers 9 .The cluster_cells function was run with the parameter reduction_method set to 'UMAP'.The principal graph was generated with the function learn_graph (with use_partitition parameter set to TRUE for Integration 1 and set to FALSE for Integration 2).Cells were ordered on the principal graph using the order_cells function, with a manually chosen root cell.
Scanpy normalisation and processing of counts.Raw counts of both batches from Dataset 1 were processed using the Scanpy package (version 1.4.5.1) 10 .Scanpy anndata objects of both batches were created and concatenated.The function filter_genes was run with parameter min_cells=3.After filtering, 33,774 genes were left.Cells were normalised with the function normalize_total with parameter target_sum=1e4.Counts were log transformed with a pseudocount of 1 added to mitigate the mean-variance relationship using the log1p function, to reduce skewing of data and account for drop-outs.Batch effects were regressed out with the function combat, using the time-point as the covariates.The scanpy combat function is a wrapper function for the combat package (https://github.com/brentp/combat.py).Highly variable genes were selected with the function highly_variable_genes, which implements the works of 11 .The parameters for the highly_variable_genes function were as follows: min_mean=0.05,max_mean=13, min_disp=0.1, max_disp=3.The anndata object was subset with 9,212 highly variable genes.Principal component analysis was performed using the pca function in the Scanpy package with default parameters.Nearest neighbours for the cells were selected using the bbknn package (version 1.3.7) 12.The function bbknn_pca_matrix was used with the PCA matrix calculated by Scanpy and the batch information as inputs, parameters were set as follow: approx=False, metric= 'euclidean'.The function umap was used to find the UMAP representation of the data, the parameter n_components=3.
Differential expression and definition of "ex vivo modulated genes".Genes from the raw count matrix were filtered out if expressed in < 4 cells.Filtered raw counts were used to perform differential expression analysis with the R package DESeq2 (version 1.36.0) 13 .Batch effects were accounted for in the model where applicable.For Integration 1, the union of all differentially expressed genes in each pairwise time-point comparison was curated, generating a list of 10,010 genes, herein referred to as "ex vivo modulated genes" (Table S5).A variance stabilizing transformation was performed in the DESeq2 package from the differential expression analysis.Batch effects were removed on the variance stabilised matrix using the limma package correction (version 3.52.2) 14 .These VST batch corrected values were used for visualisations of gene expression in violin plots and as input for degPatterns.
Gene set enrichment and variation analysis.Gene Set Enrichment Analysis (GSEA) (v4.2) 15 was performed against the c2 curated pathway database using a pre-ranked list by the stat value (value of the Wald test statistic) from the DESeq2 output specifying 10,000 permutations.Gene-Set Variation Analysis (GSVA) was also performed 16 against c2 curated pathways using the normalised, batch corrected count matrix generated by Scanpy (parameters min.siz=10,max.sx=500).GSEA and GSVA were also performed with curated signatures.All gene signatures were created using the top 100 differentially expressed genes contrasting previously reported populations.Signatures were curated from CD34 lo CLEC9A hi (Subset1) and CD34 hi CLEC9A lo (Subset 2) 17 and from human dormant and activated HSC populations 18 .

Identification of patterns of gene expression along the time course.
The degPatterns function from the DEGreport package 19 was used to group genes based on expression pattern, inputting VST limma corrected matrix values for ex vivo modulated genes.The time parameter was set to the time-point and eachStep=TRUE.8,966 genes remained after degpatterns filtering, removing clusters with <15 genes.A distance matrix was generated from all pairwise comparisons between time-points and hierarchical clustering was performed specifying 13 clusters.Upon visual observation 2 of the 13 clusters were manually split, generating 15 clusters which fit the dataset without showing redundancy in expression pattern (clusters were renamed 1-15 accordingly).
To group GSVA biological pathway patterns based on expression trend, the degPatterns function from the DEGreport package 19 was again used, inputting the GSVA score matrix, with the parameter time set to the incubation time of the samples and eachStep=TRUE.4,367 out of 4,596 pathways remained after filtering and 13 clusters again were specified to fit the dataset.Upon visual observation, 6 of 13 clusters of the GSVA score patterns were manually split further, resulting in 19 clusters with no redundancy in expression pattern observed (clusters were renamed 1-19 accordingly).scEntropy measurements.To measure the transcriptomic order of single cells in the dataset, the package scEntropy was used 20 .The scEntropy per cell is defined as the difference between the cell of interest and an intrinsic reference value calculated by the package (parameter option =RCSA).Scanpy was used to pre-process raw counts from both batches separately and cells were normalised by 1x10 4 reads (normalize_total function with parameter target_sum=1e4).The scEntropy value for single cells was calculated.To account for batch effects, the difference in the mean entropy values of batch 1 and batch 2 was subtracted from each entropy value of batch 1.

Bayesian modelling of gene expression and over-dispersions to measure expression variability.
The BASiCS package (version 2.2.4) 21 was used to perform Bayesian modelling of the counts.The amount of ERCC molecules present together with each cell was calculated from the concentration of the ERCC mix added and the information for the ERCC mix acquired online ([https://assets.thermofisher.com/TFS-Assets/LSG/manuals/cms_095046.txt]).Raw counts were filtered to average reads per million > 20, resulting in 8,464 genes.BASiCS objects were created with the filtered counts, ERCC information and batch information.The objects were created independently per time-point.For the 24 h time sample, one cell was not included for the creation of BASiCS object due to a low ERCC quantity present.Bayesian inference of the parameters of gene count distributions were performed using the function BASiCS_MCMC.The parameters for the BASiCS_MCMC function which calculate residual over dispersion (e) were as follow: N=10000, Thin=10, Burn=1000, WithSpikes=TRUE and Regression=TRUE.To identify residual over-dispersed genes for all time-points, the function BASiCS_TestDE was used that compares the value of residual over-dispersion and performs statistical testing between 2 samples.Genes that are differentially over-dispersed based on the residual over-dispersion value were curated for each time-point.To calculate maximally variable genes (MVG) at each time-point, genes that were differentially variable in multiple pairwise comparison were assigned to the time-point with the highest residual over-dispersion value.

Correlations of median expression.
For EXPER_CB conditions, the median expression value for ex vivo modulated genes plus genes differentially expressed in PD treated comparisons (genes changed in 0h vs 72h PD, 72h PD vs 72h UNTR, 0 vs 24h PD, 24h PD vs 24h UNTR) (n=10,903 genes total) was plotted for each comparison and the Pearson's correlation coefficient calculated (95% CI; mean value shown).For GT conditions, the median expression value for the sum of genes changed in 0h vs 62h UNTR, 0h vs 62h PD and 62h PD vs 62h UNTR, was used (5,469 genes) and each cell at each time-point was compared.All gene lists can be found in Table S5.

Supplemental note 1
An increasing body of work is identifying transcriptional and functional heterogeneity within the human LT-HSC pool 17,18,[22][23][24] , indicating the potential presence of multiple subpopulations at the 0 h time-point.0 h LT-HSCs were therefore further split into 2 clusters by k means clustering (Fig. S2D).Differential gene expression was performed with DESeq2, followed by GSEA (Table S3) and GSVA analyses.One cluster expressed significantly higher levels of genes involved in hypoxia regulation (Fig. S2E) than the other and displayed enrichment for the most dormant 18 and multipotent 17 LT-HSCs gene signatures (Fig. S2F).This cluster was denoted "0 h-early" and was subsequently used as pseudotime origin (Fig. S2G).In the other cluster, named "0 h-late", we found higher levels of oxidative phosphorylation genes compared to "0 h-early" LT-HSCs (Fig. S2H; p=0.0580) and significant enrichment of gene signatures related to a myelo-lymphoid restricted self-renewing HSC subset 17 (Fig .S2F).In summary, our scRNA-seq strategy supports recent literature identifying transcriptional heterogeneity within the quiescent human LT-HSC fraction 17,[22][23][24][25] .(E) GSVA scores per cell of "KRIEG_HYPOXIA_VIA_KDM3A" gene-set in the indicated conditions.Unpaired t-test.
(I) 2D Pseudotime rank plot of LT-HSCs over time-course generated with no cell cycle regression applied.

Fig. S7. Representative gates for LT-HSC isolation, colony generation from single LT-HSCs and long-term in vivo grafts generated from transplanted HSCs (A)
Representative gating strategy for isolation of LT-HSC.The example is here is from a 0h mPB CD34 + enriched donor sample.LT-HSC: Zombie -CD19 -CD34 + CD38 -CD90 + CD49f + (top 30% of CD90 and CD49f expression).The same strategy was used for CB LT-HSCs.This study used the antibodies in the table in the panels described below and referred to in the relevant methods sections:  QC thresholds used for filtering "good quality" single cells in each of the independent scRNAseq experiments.NA: not applied.

Fig. S1 .
Fig. S1.Cell cycle status of LT-HSC during GT and EXPER culture

Fig. S3 .
Fig. S3.Dynamic patterns of LT-HSC gene expression observed during ex vivo culture (A-B) Output from DEG Report clustering of (A) gene expression patterns or (B) GSVA pathway patterns.(A) 15 gene expression patterns identified from 10,010 genes classified total; 8,966 genes after filtering.(B) 19 GSVA biological pathway patterns identified from 4,596 total GSVA pathways classified; 4,367 pathways after filtering.TableS5.

Fig. S4 .
Fig. S4.Validation of reversible early G1 arrest in LT-HSCs by PD treatment and the transcriptional effects cell cycle arrest (A) Representative example of pRb/DAPI flow cytometry plot of alive UNTR (left) or PD treated (right) mPB LT-HSCs cultured for 62h in GT conditions.

Figure S6 :
Figure S6: Characterising the effects of RUX treatment on mPB HSCs in vitro and in vivo

( B )
Representative gating strategy for determination of colony size and lineage output from single cell MEM assay.Representative gating of a colony arising from a single sorted mPB LT-HSC after 21 days culture in MEM media.A true colony was determined if >30 cells are observed in each of the CD45 + and GlyA + gates.Colonies were assigned to specific lineages if >30 cells were found in the following gates: myeloid: CD45 + CD11b + ; monocyte: CD45 + CD14 + ; granulocyte: CD45 + CD15 + ; lymphoid (NK only): CD45 + CD11b -CD56 + .Undifferentiated colonies were determined as having >30 cells in (CD45 + & GlyA + ) gates but <30 cells in all the other lineages.Representative gating strategy for analysis of mouse BM after primary transplantation (18-20 weeks).The example here is from a mouse injected with mPB CD34 + CD38 -cells cultured in GT conditions (62h) and transduced with a LV containing GFP. Erythroid (Ery) cells were identified as CD45 -GlyA + .Myeloid (My) cells were identified as CD45 ++ CD33 + .Lymphoid (Lym) cells were identified as CD45 ++ CD19 + .The same gating strategy was used for CB experiments and all secondary transplantation experiments.

Table S2 : Engraftment data for primary Limiting Dilution Assays reported in this study comparing culture duration in untreated conditions.
Number of mice injected and number of mice engrafted for each of the doses tested in primary limiting dilution experiments.Mice were considered engrafted if human cells (hCD45 ++ & GlyA + ) were ³30 cells and represent ³0.01% of Singlets, and if ³20 cells were present in any lineage determination gate.Engraftment measurements for all primary mice are reported in TableS1.EXPER_CB mice were considered engrafted if engrafted thresholds were met in either the injected femur or other bone marrow harvested from hind legs.LTRC: Long Term Repopulating Cell.

Table S9 : Engraftment data for primary Limiting Dilution Assays reported in this study comparing UNTR/PD treated conditions Number
of mice injected and number of mice engrafted for each of the doses and conditions tested in primary limiting dilution experiments.Mice were considered engrafted if human cells (hCD45 ++ & GlyA + ) were ³30 cells and represent ³0.01% of Singlets, and if ³20 cells were present in any lineage determination gate.Engraftment measurements for all primary mice are reported in

Table S1 .
EXPER_CB mice were considered engrafted if engrafted thresholds were met in either the injected femur or other bone marrow harvested from hind legs.UNTR: Untreated; PD: Palbociclib; LTRC: Long Term Repopulating Cell.

Table S10 : Engraftment data for secondary Limiting Dilution Assays reported in this study comparing UNTR/PD treated conditions
Number of mice injected and number of mice engrafted for each of the doses and conditions tested in secondary limiting dilution experiments.GT mice were transplanted with counted, unsorted BM isolated from primary recipients.EXPER mice were transplanted with CD45 ++ cells isolated from primary recipients by flow cytometry.Mice were considered engrafted if human cells (hCD45 ++ & GlyA + ) were ³30 cells and represent ³0.01% of Singlets, and if ³20 cells were present in any lineage determination gate.EXPER_CB mice were considered engrafted if engrafted thresholds were met in either the injected femur or other bone marrow harvested from hind legs.UNTR: Untreated; PD: Palbociclib; LTRC: Long Term Repopulating Cell.

Table S13 : Engraftment data for primary Limiting Dilution Assays reported in this study comparing DMSO/RUX treated conditions.
Number of mice injected and number of mice engrafted for each of the doses and conditions tested in primary limiting dilution experiments.Mice were considered engrafted if human cells (hCD45 ++ & GlyA + ) were ³30 cells and represent ³0.01% of Singlets, and if ³20 cells were present in any lineage determination gate.Engraftment measurements for all primary mice are reported in

Table S14 : Engraftment data for secondary Limiting Dilution Assays reported in this study comparing DMSO/RUX treated conditions.
Number of mice injected and number of mice engrafted for each of the doses and conditions tested in secondary limiting dilution experiments.Mice were considered engrafted if human cells (hCD45 ++ & GlyA + ) were ³30 cells and represent ³0.01% of Singlets, and if ³20 cells were present in any lineage determination gate.RUX: Ruxolitinib; LTRC: Long Term Repopulating Cell.