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

Abstract

Loss of long-term hematopoietic stem cell (LT-HSC) function ex vivo hampers the success of clinical protocols that rely on culture. However, the kinetics and mechanisms through which this occurs remain incompletely characterized. In this study, through time-resolved single-cell RNA sequencing, matched in vivo functional analysis, and the use of a reversible in vitro system of early G1 arrest, we defined the sequence of transcriptional and functional events that occur during the first ex vivo division of human LT-HSCs. We demonstrated that the sharpest loss in LT-HSC repopulation capacity happens early on, between 6 and 24 hours of culture, before LT-HSCs commit to cell cycle progression. During this time window, LT-HSCs adapt to the culture environment, limit the global variability in gene expression, and transiently upregulate gene networks involved in signaling and stress responses. From 24 hours, LT-HSC progression past early G1 contributes to the establishment of differentiation programs in culture. However, contrary to the current assumptions, we demonstrated that the loss of HSC function ex vivo is independent of cell cycle progression. Finally, we showed that targeting LT-HSC adaptation to culture by inhibiting the early activation of JAK/STAT signaling improves HSC long-term repopulating function ex vivo. Collectively, our study demonstrated that controlling early LT-HSC adaptation to ex vivo culture, for example, via JAK inhibition, is critically important to improve HSC gene therapy and expansion protocols.

Fig. 1. Kinetics of cell cycle progression, survival and loss of long-term repopulation capacity of LT-HSCs during ex vivo culture.

(A) Cumulative quiescence exit kinetics of GT_mPB LT-HSCs determined by phospho Rb (Ser 807 – 811) flow cytometry analysis. Curve is least-squares sigmoidal fit with EC₅₀ = 24.7 h; n=3 biological replicates for 0h, 24h, 62h, 72h and n=4 biological replicates for 6h. Standard error ± 5.186; R² = 0.9797. (B) Cumulative quiescence exit kinetics of EXPER_CB LT-HSCs determined by phospho Rb (Ser 807 – 811) flow cytometry analysis. Curve is least-squares sigmoidal fit with EC₅₀ = 24.72h; n=3 biological replicates for 0h, 6h, 24h, 48h and n=4 biological replicates for 72h. Standard error ± 3.944; R² = 0.9844. (A-B) Dashed line indicates EC_50, time of quiescence exit. (C) Cell cycle phase assignment of GT_mPB LT-HSCs determined by pRb/DAPI flow cytometry analysis. Equivalent repeats as in (A) (D) Cell cycle phase assignment of EXPER_CB LT-HSCs determined by pRb/DAPI flow cytometry analysis. Equivalent repeats as in (B). (E) Cumulative first division kinetics (excluding dead cells) of GT_mPB LT-HSCs. Curve is least squares sigmoidal fit. Representative examples shown (biological replicates: n=3). Dashed line indicates EC₅₀. EC₅₀ = 64.83 h; 95% CI = 62.83 h – 66.87 h; R² = 0.9976. (F) Cumulative first division kinetics (excluding dead cells) of EXPER_CB LT-HSCs. Curve is least squares sigmoidal fit. Representative examples shown (biological replicates: n=4). EC₅₀ = 55.44 h; 95% CI = 54.28 h – 56.70h; R² = 0.9995. (G) Time to first division kinetics summary of LT-HSCs cultured in GT_mPB (E) and EXPER_CB (F) systems (biological replicates: n=3 for GT; n=4 for EXPER). Unpaired t-test shown. (H) Percent of LTRC in GT_mPB CD34⁺CD38^- cells as determined by LDA analysis in the transplanted population. %LTRCs at each time-point +/- 95% CI shown. LTRC frequency estimate for GT_mPB: 0h: 1 in 939 (29 mice); 6h: 1 in 1211 (21 mice); 24h: 1 in 3371 (23 mice); 62h: 1 in 2510 (23 mice). ELDA statistical test shown. Data in Table S1. (I) Percent of LTRC in LT-HSCs cultured in EXPER_CB system as determined by LDA analysis in the transplanted population. % LTRCs at each time-point +/- 95% CI shown. LTRC frequency estimate for EXPER_CB: 0h: 1 in 14.8 (31 mice); 6h: 1 in 15.2 (19 mice); 24h: 1 in 80.6 (31 mice); 72h: 1 in 293.7 (40 mice). ELDA statistical test shown. Data in Table S1. (J) Survival of LT-HSCs cultured in GT_mPB systems determined by Annexin-V/7-AAD flow cytometry. n=3 biological replicates at each time-point. Mean +/- SD shown. (K) Survival of LT-HSCs cultured in EXPER_CB system determined by Annexin-V/7-AAD flow cytometry. n=3 biological replicates at each time-point. Paired t-test shown for 6h and 24h comparison. Mean +/- SD shown.

Fig. 2. Dynamics of gene expression over the first division of LT-HSC ex vivo at single cell resolution.

(A) UMAP of 429 single EXPER_CB LT-HSCs over a time-course of 0h, 6h, 24h and 72h (n=2 independent experiments). UMAP generated using Seurat 4 pipeline following cell cycle regression. (B) 2D pseudotime rank plot of EXPER_CB LT-HSCs over time-course generated following cell cycle regression. (C) Transcriptional allocation of cell cycle status at time-points of EXPER_CB LT-HSC culture; n= 429 single cells. (D) Number of differentially expressed genes (FDR<0.05) at each time-point with respect to 0h. Upregulated genes (green) and downregulated genes (brown). Full list of genes available in Table S4. (E) Broad patterns of gene expression identified over time-course (8,966 genes classified after filtering by DEG Report algorithm). Numbers indicate the percentage of genes showing the specific patterns of gene expression displayed to the right of bar. (F-J) GSVA score of c2 curated pathways showing specific expression patterns: (F) continuous up; (G) continuous down; (H) transient up; (I) transient down; (J) up later than 6h. GSVA score calculated per single cell with line at median and upper and lower whiskers indicate 25th and 75th percentile of expression. (K) GSVA scores of indicated published gene signatures, representative of specific HSPC subsets. Median and interquartile range shown. * indicates p<0.001. (L) scEntropy value at each time-point (calculated for both batches combined; Wilcoxon rank sum test shown; 0h vs 6h p = 0.835). Median and interquartile range shown. * indicates p<0.001. (M) Number of maximally variable genes at each time-point (MVG; see Supplemental Methods; 2792 genes total). (N) Selected biological pathways significantly enriched from MVG (-log₁₀(adjusted p-value) <0.05). Full list of pathways available in Table S7.

Fig. 3. Transcriptional effects of preventing progression past early G₁ during ex vivo culture of LT-HSCs.

(A) Cumulative first division kinetics (excluding dead cells) of UNTR/PD treated LT-HSCs cultured in GT_mPB (grey and dark red) or EXPER_CB (black and blue) system. Curve is least squares sigmoidal fit. Representative example shown (n=3 UNTR/PD matched biological replicates for GT and n=2 UNTR/PD treated matched biological replicates for EXPER). Dashed line indicates EC₅₀, time to first division. Untreated (UNTR) and Palbociclib treated 200nM (PD). (B) Divided single cells as a proportion of total alive cells at 96h (EXPER_CB: n=5 biological repeats, n=2 UNTR/PD matched biological repeats; GT_mPB: n=3 matched biological repeats). Paired t-test shown. (C) Representative example of flow cytometry plot for phospho Rb (Ser 807 – 811) Alexa 647 and DAPI staining on UNTR (left) or PD treated (right) LT-HSCs cultured in CB_EXPER medium for 72h. (D) Quantification of pRb⁺ (as % of viable cells) in UNTR/PD treated LT-HSCs cultured for 62h in GT_mPB (n=3 UNTR/PD treated matched biological repeats; no LV transduction) or 72h in EXPER_CB (n=3 UNTR/PD treated matched biological repeats) systems. Paired t-test shown. (E) UMAP visualisation of scRNA-seq from 954 LT-HSCs from the indicated culture conditions (EXPER_CB: 536 single cells: GT mPB: 418 single cells). Cell cycle regression applied. (F) 2D Pseudotime density rank plot of single cells shown in (E). Cell cycle regression applied. (G) Pearson’s correlation coefficient estimate comparing the median expression value of 10,903 genes at time-point/condition comparisons in the EXPER dataset (union of all differentially expressed genes between any 2 UNTR time-points and between PD treated and UNTR conditions; available in Table S5). (H) Pearson’s correlation coefficient estimate comparing the median expression value of 5,469 genes at time-point/condition comparisons in the GT dataset. (union of all differentially expressed genes between any 2 UNTR time-points and between PD treated and UNTR conditions; available in Table S5). (I) Selected Reactome pathways (FDR<0.05) enriched between PD treated and UNTR LT-HSCs cultured for matched durations of 24h in EXPER_CB (purple), 72h in EXPER_CB (pink) and 62h in GT_mPB (black) systems. Full DeSeq2 results and Reactome pathway enrichment available in Table S8.

Fig. 4. Preventing progression past early G₁ during ex vivo culture of LT-HSCs dampens the establishment of differentiation programmes but does not affect loss of long-term repopulation capacity.

(A) GSVA scores of indicated lineage gene expression signatures from at indicated time-points of LT-HSC culture. GSVA score generated per cell and line at median. EXPER_CB : n= 536 single cells; GT_mPB: n=418 single cells. (B) Cell diameters of single CB LT-HSCs cultured in EXPER conditions (n=2 experiments representing n=469 total single cells). Unpaired t-test. * indicates p<0.001. (C) Tetramethylrhodamine (TMRM) staining of bulk CB LT-HSCs cultured in EXPER conditions (n=5 biological repeats for 0h; n = 4 matched biological repeats for UNTR/PD treated 24h; n=3 matched biological repeats for UNTR/PD treated 48 h). Unpaired t-test shown. (D) Workflow of in vivo transplantation of LT-HSCs cultured in EXPER_CB system (24h and 72h) and CD34⁺/CD38^- cells cultured in the GT_mPB system (62h). UNTR/PD treated cells transplanted in matched cell dose experiments. Created with BioRender (license agreement: KE26QKHW50). (E) Graft size (% of human CD45⁺⁺ and GlyA⁺) at 18-week post transplantation of UNTR/PD treated CB LT-HSCs cultured for 72h in EXPER system (n=5 biological experiments; graph representative of engrafted mice only, n=42 PD mice, n = 38 UNTR mice). Two-way ANOVA with Sidak’s multiple comparisons performed (50 cells UNTR vs 50 cells PD p=0.9552; 300 cells UNTR vs 300 cells PD p=0.4084; 700 cells UNTR vs 700 cells PD p=0.971). (F) Graft size (% of human CD45⁺⁺ and GlyA⁺) at 18-week post transplantation of mPB CD34⁺CD38^- cells after GT protocol culture for 62h including LV (n=3 biological repeats; graph representative of engrafted mice only, n=25 mice UNTR, n=26 mice PD). Two-way ANOVA with Sidak’s multiple comparisons performed (all cell doses UNTR vs PD p>0.9). (G) % LTRC in CB LT-HSCs cultured in EXPER system in presence or absence of PD, determined at 24h (n=31 mice UNTR, n=30 mice PD) and 72h (n= 42 mice PD, n=40 mice UNTR). Numerical estimates for LTRC frequency available in Table S2. ELDA statistics (24h UNTR vs 24h PD p=0.405; 72h UNTR vs PD p=0.426). (H) LDA of secondary transplantation experiment from EXPER_CB UNTR/PD 72 h primary mice cohort. Secondary animals were transplanted with sorted CB CD45⁺⁺ from primary recipients (n= 20 mice total; 10 UNTR, 10 PD; n=1 experiment; Table S10). ELDA statistical test performed (p=0.190). (I) LDA of secondary transplantation experiment from GT_mPB UNTR/PD 62 h primary mice cohort. Secondary animals were transplanted with whole mouse BM isolated from primary recipients (n=21 mice total; UNTR = 11 mice; PD = 10 mice; n=1 experiment; Table S10). ELDA statistical test performed (p= 0.860).

Figure 5. RUX treatment improves serial replating ability and self-renewal capacity of cultured HSCs.

(A) GSVA score of KEGG JAK/STAT signaling pathway gene-set (same as Fig.2H). GSVA score calculated per single cell with line at median and upper and lower whiskers indicate 25th and 75th percentile of expression. (B-C) Serial replating of human mPB HSC/MPPs (CD34⁺CD38^-CD45RA^-) cultured for 72h in the EXPER system (B) or for 62h in GT (C). Graphs (upper panels) show fold change in colony number compared to DMSO from primary (2 week), secondary (4 week) and tertiary (6 week) plating. n=3 independent mPB samples in (B) and n=4 independent mPB samples in (C), individual donors indicated by shapes. Mean and SD shown. Tables (lower panels) report statistics from generalised-linear-mixed-effect-model (glmer) analysis performed with raw colony counts. Tukey corrected p-values for pairwise comparisons to DMSO where p<0.05 are shown (Table S12 for all comparisons and raw data). (D) Representative image of wells from tertiary replating experiment of mPB HSC/MPPs cultured in 72h EXPER conditions (upper panel) or 62h GT conditions (lower panel) treated with either DMSO (left) or 10nM RUX (right). Circles indicate manually scored colonies. Images brightened by 17%. (E-G) Serial replating of human mPB HSC/MPPs cultured for 62h in GT. Graphs (upper panels) show fold change in colony number compared to DMSO from tertiary (6 week) plating in the conditions indicated: (E-G) RUX (10 nM), (E) 0h: fresh HSC/MPPs, (F) UM171 (35nM), low TPO (20ng/ml), (G) CASi: pan-caspase inhibitor Z-VAD(OH)-FMK (100nM); (E) and (G) n=3 independent mPB samples, (F) n=5 independent mPB samples. Individual donors indicated by shapes and matching across (E-G). Mean and SD shown. Tables (lower panels) report statistics from glmer analysis fitting raw colony counts. Tukey corrected p-values for pairwise comparisons of EM means where p<0.05 are shown (Table S12 for all comparisons and raw data). (H) Workflow of in vivo transplantation of mPB CD34⁺CD38^- cells cultured in the GT system for 62h with LV transduction with RUX (10nM) or DMSO. Secondary transplantations were performed from whole BM of engrafted mice. Created with BioRender (license agreement: UR26QKHL5H). (I) Graft size (% of human CD45⁺⁺ and GlyA⁺) at 18-weeks post transplantation in the BM of mice transplanted with mPB CD34⁺CD38^- cells cultured for 62h in GT protocol with RUX (10nM) or DMSO. n=6 biological repeats; graph representative of n=68 engrafted mice (n=34 DMSO, n=34 RUX). Two-way ANOVA with Sidak’s multiple comparisons performed (30,000 cells DMSO vs 30,000 cells RUX p=0.0059, all other doses DMSO vs RUX p>0.9). (J) Log-fraction plot of limiting dilution model fitted to data in Table S14. Indicates %LTRC estimates from secondary transplantation experiments. Whole BM of engrafted mice from primary transplants was transplanted in NSG-SGM3 mice and analyzed 8 weeks post-transplantation (n=1 experiment; n=35 mice). Slope is log-active fraction. Dotted line is 95% CI. Zero negative response indicated by triangle.