Single-cell assessment of transcriptome alterations induced by Scriptaid in early differentiated human haematopoietic progenitors during ex vivo expansion

Priming haematopoietic stem/progenitor cells (HSPCs) in vitro with specific chromatin modifying agents and cytokines under serum-free-conditions significantly enhances engraftable HSC numbers. We extend these studies by culturing human CD133+ HSPCs on nanofibre scaffolds to mimic the niche for 5-days with the HDAC inhibitor Scriptaid and cytokines. Scriptaid increases absolute Lin−CD34+CD38−CD45RA−CD90+CD49f+ HSPC numbers, while concomitantly decreasing the Lin−CD38−CD34+CD45RA−CD90− subset. Hypothesising that Scriptaid plus cytokines expands the CD90+ subset without differentiation and upregulates CD90 on CD90− cells, we sorted, then cultured Lin−CD34+CD38−CD45RA−CD90− cells with Scriptaid and cytokines. Within 2-days and for at least 5-days, most CD90− cells became CD90+. There was no significant difference in the transcriptomic profile, by RNAsequencing, between cytokine-expanded and purified Lin−CD34+CD38−CD45RA−CD49f+CD90+ cells in the presence or absence of Scriptaid, suggesting that Scriptaid maintains stem cell gene expression programs despite expansion in HSC numbers. Supporting this, 50 genes were significantly differentially expressed between CD90+ and CD90− Lin−CD34+CD38−CD45RA−CD49f+ subsets in Scriptaid-cytokine- and cytokine only-expansion conditions. Thus, Scriptaid treatment of CD133+ cells may be a useful approach to expanding the absolute number of CD90+ HSC, without losing their stem cell characteristics, both through direct effects on HSC and potentially also conversion of their immediate CD90− progeny into CD90+ HSC.

HSC (SCID repopulating cells or SRC) 12 . This has been shown to be dependent on the specific HDACi used. Various researchers have demonstrated that HDACis, such as Valproic acid (VPA), Scriptaid (Scr), Trichostatin (TSA), Suberoylanilide hydroxamic acid (SAHA or Vorinostat), CAY10433, CAY10398 and CAY10603 allow greater expansion of UCB CD34+, CD34+CD90+ HSPC and/or early in vitro clonogenic cobblestone area forming cells (CAFC) or long term culture-initiating cells (LTC-IC) in ex vivo short term (up to 9 days) cultures in the presence of cytokines than with cytokines alone [12][13][14][15][16][17][18][19] . Of these, three class I/II HDCAis, VPA, Scriptaid and CAY10433 are reported to generate, albeit to differing degrees, higher absolute numbers of UCB CD34+ and CD34+CD90+ HSPCs when added individually to serum-free cultures with stem cell factor (SCF), Flt-3 ligand (FL), thrombopoietin (TPO) and interleukin-3 (IL-3) for 7 days 12 . Interestingly, both VPA 12,18 or Scriptaid (as presented here) addition to cytokine-driven cultures significantly increases the absolute numbers of HSPCs expressing Lin−CD34+CD38−CD45RA−CD90+CD49f+ biomarkers, which define the main phenotype of uncultured HSCs. In surrogate transplant models, greater frequencies of human CD45+ cell engraftment into the bone marrow of transplanted primary NSG immunodeficient mice (e.g. 100% vs 20% of mice with 2,500 culture initiating cell equivalents infused) and greater degrees of human CD45+ cell chimaerism (on average 2.4 fold higher) at weeks 12-14 post transplant were also observed when human UCB HSPC expanded in VPA with cytokines for 7 days were compared to those expanded with cytokines alone 12,18 . We have also carried out preliminary in vivo repopulation experiments of UCB CD133+ HSPCs expanded in Scriptaid and SCF, TPO and FL cytokines versus these cytokines alone for 5 days on nanofibre scaffolds (the cultures being supplemented with these factors at, and 2 days after, the beginning of the cultures). At week 16 post transplant, we observed a greater frequency of engraftment with the Scriptaid plus cytokine cultured cells as opposed to cytokine alone cultured cells (e.g. 100% vs 40% engrafting respectively into 3 and 5 NSG mice with infusion of 2,500 culture initiating CD133+ cell equivalents) and greater degrees of human CD45+ cell chimaerism (on average 3.6 fold higher; Watt SM et al. unpublished data). Higher in vivo primary NOD/SCID engraftment of human CD34+ cells was also observed with the sequential addition of 5-azacytidine followed by TSA in the presence of cytokines (SCF, TPO, FL) than with cytokines alone 13,14,16 . Given that human HSCs (Lin−CD34+CD38-CD45RA−CD90+CD49f+ long-term-(LT)-SRCs), if their stemness is maintained, are expected to increase 3-5 fold in 5-7-day cultures (estimated median doubling-time 36-48 hours), that LT-SRC display delayed G 0 exit (1st division ~66-75 h), that short-term-SRC proliferate more rapidly, and that HSC develop in micro-environments providing additional regulatory cues 20-22 , we and others have hypothesised that chromatin-modifying agents not only expand the CD90+HSC subset without differentiation and by symmetrical division 19 , but also convert more mature CD90− HSPCs back to CD90+HSPCs.
To test this hypothesis, we cultured overnight cytokine primed human umbilical cord blood (UCB) CD133+ HSPCs on nanofibre scaffolds in serum-free media containing SCF, FL and TPO 23,24 plus either the HDACi Scriptaid or vehicle control and examined Lin−CD34+CD38−CD45RA−CD90+CD49f+ HSPC yield. Here, we show that CD90 was upregulated on CD90− HSPCs after Scriptaid-treatment and 'stemness' genes were maintained in the purified CD90+ subset. Transcriptomic signatures using RNAseq and single cell q-RT-PCR of the sorted Lin−CD34+CD38−CD45RA−CD90+CD49f+ HSPC fraction following Scriptaid-treatment thus support the view that this chromatin-modifying agent can maintain more primitive HSPCs without compromising their phenotypic and transcriptomic stem cell characteristics, both by direct effects on HSC and possibly also converting CD90− more mature HSPCs to less mature CD90+ cells.
Combining cytokines (including SCF and TPO) with Class I/II HDACis (such as Scriptaid or VPA 12 ) for HSC expansion may, at first sight, appear to be counterintuitive, given that SCF and TPO have been shown to stabilise HIF-1α which has been reported to positively influence mammalian HSC self-renewal 31-34 , while Class I/II HDACis are known to repress HIF-1α function 31,35 . However, controversy surrounds the importance of HIF-1α in haematopoiesis, with Vukovic et al. 36,37 . more recently demonstrating that HIF-1α is not essential for 'young' HSC self-renewal, long term in vivo haematological reconstitution nor sustaining haematopoiesis post-injury, at least in mice. In their studies, Hoffman's group 12,18 also primed human UCB CD34+ cells with cytokines (SCF, TPO, FL, IL-3) for 16 hours, but then compared the effects of adding VPA with or without added cytokines for 4-7 days ex vivo. By combining these cytokines with VPA, engraftable HSC numbers were significantly enhanced. Since HSC stemness is characterised by condensed, immature mitochondria, low metabolic states and high glycolytic activity 31,38,39 , they then examined if these characteristics are maintained ex vivo in HSC in the presence of VPA 18 . They showed that addition of VPA to their cytokine based cultures allowed the HSCs, a least in the short term, to retain or acquire a mitochondrial profile similar to uncultured HSCs and to show enhanced glycolytic potential 18 . They further reported that the upregulation of p53 and activation of the p53-MnSOD axis were key to limiting reactive oxygen species (ROS) levels, thereby promoting HSC self-renewal and engraftable HSC www.nature.com/scientificreports www.nature.com/scientificreports/ expansion 18 . It remains unclear if cytokines such as SCF, which reduce ROS levels in HSCs 38,39 , contribute to these HSC effects in a manner not involving HIF-1α. Given these results, it seems likely that Scriptaid would promote immature HSPC expansion in a manner similar to VPA.

Conclusion
In conclusion, the rarity of human haematopoietic stem cells and the very limited ability to study these cells in vivo in the human stem cell niche are significant hurdles in understanding the regulatory networks and cues that control the balance between their maintenance, survival, quiescence, self-renewal and differentiation, and may limit their clinical use. Our studies support the single-cell view that specific HDACi, when combined with cytokines, maintain the transcriptome of more immature HSPCs. This has the potential to impact on outcomes of ex vivo gene editing for hematological diseases, where short-term ex vivo cultures require HSC maintenance or expansion.

Methods
Cell isolation. Human umbilical cord blood (UCB) was collected from the John Radcliffe Hospital, Oxford, UK or provided via the NHS Cord Blood Bank, London, and used with informed, written pre-consent and ethical approval from the South Central Oxford C and Berkshire Ethical Committees and approval of the NHSBT R&D committee and all methods were performed in accordance with the relevant guidelines and regulations. Mononuclear cells (MNCs; density < 1.077 g/ml) were isolated by density gradient centrifugation no more than 24 hours after UCB collection. Human CD133+ haematopoietic stem and progenitor cells (HSPC) were enriched by MACS using the CD133 direct microbead kits (Miltenyi Biotec GmbH) and cryopreserved until use 23,24,40,41 . Purity of the cells isolated was routinely assessed by flow cytometry and only donors with >90% CD133+ cell purity were used for expansion experiments.
Expansion culture. Cells were thawed and cultured overnight (20-24 hs) in 24 or 96 well round bottom plates at 200 cells/μl in serum free Stem Span ACF media (Stem Cell Technologies) supplemented with 3 cytokines (C3) at 37 °C, 5% CO 2 , 95% humidity. The following cytokines were used for expansion: Stem Cell Factor (SCF), FLT-3Ligand (FL) (both at 100 ng/ml) and Thrombopoietin (TPO; 20 ng/ml) (all from R&D Systems). On day 0, cells were harvested, counted using Countbright absolute counting beads (Molecular Probes) by flow cytometry [40][41][42] and plated on 3D nanofiber scaffolds (NANEX plates from Compass Biomedicals) at an optimised cell density of 2,500 cells/ml in Stem Span ACF supplemented with 3 cytokines as described above 23,24,41 . Cells were either treated with the HDAC inhibitor Scriptaid (Sigma) at 1 µM in DMSO or an equivalent amount of vehicle (0.1% DMSO, Sigma) and harvested for downstream assays on days 2 or 5.  26,32 . A representative gating strategy is shown in Supplementary  Fig. S5. In some cases, sorting of these HSC and progenitor subpopulations was carried out on a FACS Aria II (BD Biosciences) 40 .
Apoptosis assay. All expanded cells were harvested after 5 days of culture and stained with Annexin V FITC (BD Biosciences) according to the manufacturer's instructions. DAPI was added at 100 ng/ml directly before acquisition on a LSRII flow cytometer (BD Biosciences). Data were analysed with FlowJo software (TreeStar Inc.). Cells were gated on FSC-A against SSC-A and then analysed for apoptosis (AnnexinV+/DAPI−) and cell death (AnnexinV+/DAPI+).

LTC-IC and CFU assays.
Long-term culture-initiating cells (LTC-ICs) assay were following the protocol described previously 25 . The murine stromal cell lines (M2-10B4 and SL/SL mixed at 1:1) and irradiated with 8000 cGy were plated in 96-well collagen-coated microtiter plates (5000 cells/well of each cell line) and cultured in long-term culture medium (MyeloCult H5100; Stem Cell Technologies) supplemented with hydrocortisone 21-hemi-succinate (10 −6 M). Limiting dilution analysis (LDA) for quantitating LTC-ICs present in the Lin− CD34+CD38−CD45RA−CD90+CD49f+ cell subset 5 days after expansion in C 3 -cytokines plus Scriptaid or the vehicle control. Cells were plated at 50 to 150 cells per well (10-100 for unexpanded condition) by flow sorting 43 . Co-cultures were maintained at 37 °C in high humidity and with 50% medium exchange every week.

High-throughput RNA sequencing and bioinformatics analysis. Expanded cells with the phenotype
of Lin−CD34+CD38−CD45RA−CD49f+ were index sorted (BD FACS Aria II cell sorter) into 4 µl lysis buffer as 100 cells per sample based on the expression of CD90. The mRNA of quadruplicate samples originating from two biological replicates were then used to prepare the cDNA library following the protocol described by Picelli and colleagues 44 . Amplified cDNA libraries were cleaned by Ampure SPRI beads (1:1) and then quality checked using the Agilent high-sensitive analysis chip. About 1 ng of each library was processed for library preparation using the Nextera XT DNA sample preparation kit (Illumina) following the manufacturer's instructions. After cleaning with Ampure beads, samples were pooled to a final concentration of 10 nM. The pooled library was sequenced in multiple lanes using the Illumina HiSeq4000 platform (pair-end 75-bp reads) at the Wellcome Trust Centre for Human Genetics, University of Oxford, UK. Approximately thirty (30) million reads per sample were obtained. Raw files of the same sample from different lanes were merged into a single file using Samtools (1.3). PCR duplications were checked and removed by the MarkDuplicates function on Picardtools (2.3.0) and Samtools. Processed files were then aligned using STAR (2.4.2a) against the human genome (Homo_sapiens/ Ensembl/GRCh37) and the quality was also checked by fastQC. Genes were annotated using BioMart (2.32.1) and non-adjusted read counts for each gene were assessed statistically for global differential expression between the specified populations using the RUVseq (1.10.0) and DESeq2 (1.16.1) packages on R (3.4.3). Genes that are significant (absolute fold change >1.5 or <0.5) at a 5% false discovery rate (calculated using a Benajmini-Hochberg adjusted p-value) are considered differentially expressed between populations as described 45 . Gene Ontology Biological Process analysis was performed on R by package topGO (3.6).
TaqMan gene expression analysis. Multiplex quantitative real time PCR (q-RT-PCR) analysis was performed with the BioMark 96.96 Dynamic Array platform (Fluidigm) and TaqMan Gene Expression Assays (Applied Biosystems) as previously described 46 . Inventoried or Made-to order TaqMan assays (Supplementary  Table S1, Applied Biosystems) were pooled to a final concentration of 2x for each assay. Individual cells were sorted directly into RT-PreAmp Master Mix (2.5 µl Reaction Mix (Invitrogen); 1.25 µl 2x assay pool; 0.6 µl RT/Taq enzyme (Superscript III kit, Invitrogen). Cell lysis and sequence specific reverse transcription were performed at 50 °C for 15 min. The reverse transcriptase was inactivated by heating to 95 °C for 2 min. Subsequently, in the same tube, cDNA was pre-amplified by denaturing at 95 °C for 15 s, and annealing and amplification at 60 °C for 4 min for 22 cycles. PCR products were diluted 5-fold prior to analysis with Universal PCR Master Mix and inventoried TaqMan gene expression assays (ABI) in 96.96 Dynamic Arrays on the BioMark Fluidigm System. Ct values were calculated from the system's software (BioMark Real-time PCR Analysis; Fluidigm). Data were analysed using the ΔΔCt method and results were normalized to GAPDH expression. Only cells expressing GAPDH were included in the analysis. Principal component analysis was performed using the fluidigmSC package (3.6.2) in R (3.4.3) using LoD as 24 on raw Ct value.

Data Availability
The datasets generated during and/or analysed during the current study are available from the first author on reasonable request.