The combination of valproic acid and lithium delays hematopoietic stem/progenitor cell differentiation

Hematopoietic stem/progenitor cells (HSPCs) have a great clinical potential, but many of their applications are limited by insufficient numbers. Therefore, attempts have been made to define culture conditions, which would amplify stem cells in an uncommitted state and maintain their potential in vivo.^1-3  However, so far there has been limited success in preventing HSPC commitment and differentiation in vitro without genetic manipulation (eg, overexpression of HoxB4, Bmi1, β-catenin).^4-6  Alternatively, small molecules could be used as new tools for manipulating cell fate decisions. A number of these compounds have been identified to affect developmental processes, stem cell self-renewal and differentiation, or reprogramming efficiency of somatic cells to a pluripotent state.^3,7-11  Two of such molecules, valproic acid (VPA) and lithium (Li), have been shown to affect normal and malignant hematopoiesis.^12,13 

VPA is a pleiotropic histone deacetylase (HDAC) inhibitor (HDI), which can modulate histone acetylation by preventing its deacetylation.⁸  During commitment and differentiation, the chromatin structure of stem cells undergoes major epigenetic changes, which affect cellular transcriptional programs. Therefore, factors interfering with chromatin remodeling enzymes, such as HDIs, might modulate the behavior of stem and progenitor cells.^14,15  For instance, it has been suggested that VPA can enhance stem cell proliferation, while retaining self-renewal in ex vivo stem cell expansion protocols.^16-18  Furthermore, several groups suggested that HDIs can affect multilineage differentiation of hematopoietic cells.^19-22  The hematopoietic effects of Li have been described for the first time in psychiatric patients, which included disturbed differentiation into several lineages and increased numbers of CD34⁺ cells.^13,23,24  In vivo and in vitro studies with human and murine BM cells confirmed the stimulatory effects of Li on hematopoiesis.^25,26  Furthermore, experimental findings suggest that Li can increase transplantable HSPCs in mice.²⁵  Although the mechanism by which Li affects HSPCs is unclear, multiple molecular targets have been proposed, including glycogen synthase kinase 3β, a negative regulator of Wnt pathway.²⁷  Wnt signaling, as well as epigenetic events, play an important role in control of HSPC fate decisions.^6,14  Taken together, data suggest that VPA and Li might serve as chemical regulators of both stem cell maintenance and differentiation.

We hypothesized that combinations of small molecules might be useful to promote maintenance of HSPCs in culture by inhibiting cell commitment. In this study, the combined effects of VPA and Li on HSPC proliferation, differentiation, and self-renewal were investigated. We document potent synergistic effects of both chemical compounds in a variety of in vitro and in vivo assays and showed strong anti-differentiation activity at the level of uncommitted and committed progenitors. Moreover, we identified genetic networks and putative targets that are transcriptionally affected by both compounds. Our data support the notion that chemical compounds may be of use in ex vivo cell therapy protocols.

Female B6.SJL (CD45.1) mice, originally purchased from The Jackson Laboratory and bred in our local animal facility, and female wild-type B6 (CD45.2) mice (Harlan) were used as a source of donor HSPCs. In the transplantation settings, we used female B6.SJL donors, female wild-type B6 (Harlan) BM as competitors, and female F1:B6xB6.SJL (CD45.1/2) as recipients. All animal experiments were approved by the local animal ethical committee of the University of Groningen.

Mice were anesthetized and killed by cervical dislocation. Unfractionated BM cells were obtained by crushing femora, tibia, and pelvic bones. Cell suspensions were filtered through a 100-μm cell strainer (BD Biosciences) to remove debris, and the cells were counted on a Medonic CA620 analyzer (A. Menarini Diagnostics). After erythrocyte lysis cells were stained with lineage cocktail (A700-Mac1, A700-Gr1, A700-Ter119, A700-CD3, A700-B220), Pacific Blue-Sca1, PE-ckit, FITC-CD34, and PECy7-CD16/32 (Biolegend). After staining, the cells were resuspended in propidium iodide (PI) solution (1 μg/mL) and uncommitted (Lin⁻Sca1⁺ckit⁺ [LSK]) and committed (common myeloid progenitor [CMP], L⁻S⁻K⁺CD34⁺CD16/32^mid; granulocyte macrophage progenitor [GMP], L⁻S⁻K⁺CD34⁺CD16/32^high) hematopoietic stem/progenitor subpopulations were sorted by a MoFlo XDP cell sorter (Dako Cytomation).

Freshly isolated uncommitted and committed hematopoietic progenitor subpopulations were differentiated in a liquid or semisolid medium in the presence of FCS (10% or 30% FCS, respectively), murine SCF (mSCF; 100 ng/mL; Amgen), and recombinant mGM-CSF (10 ng/mL or 20 ng/mL, respectively; R&D Systems). Cells in liquid cultures were plated at a density of 2000 cells/mL in a 6-well plate (Costar) and cultured in StemSpan medium (StemCell Technologies). Cells in semisolid medium were plated in various concentrations. The cells were cultured for 7 days at 37°C in a humidified atmosphere and 5% CO₂ in air with or without 1mM VPA (Sigma-Aldrich) or 5mM LiCl (J. T. Baker), or the combination. Both compounds were dissolved in PBS solution (PAA Laboratories GmbH).

After liquid cultures, cells were harvested; total cell number and cell viability were determined by trypan blue and/or PI exclusion. Cultured cells were centrifuged for cytospin preparations (800 rpm; Shandon cytospin III cytocentrifuge; Thermo Fisher Scientific BV,) and stained with May-Grünwald-Giemsa staining. Cytospins were analyzed with light microscopy by counting ≥ 200 cells per experimental condition. Cells were classified as immature (blast-like) or differentiated on the basis of nucleus and cytoplasm morphology. To analyze the immunophenotype of 7-day differentiated LSK cells, cells were stained with a cocktail of Abs as described earlier and analyzed on the LSR-II (BD Bioscience). Data analyses were performed with FlowJo PC Version 7.6.5 software (TreeStar)

The CFU assay was performed to assess the potential of cells to form granulocyte/macrophage colonies (CFU-GM). Briefly, either freshly isolated hematopoietic progenitor subpopulations or cells differentiated for 7 days were plated in a methylcellulose medium, and after 6 or 7 days colonies with granuloid and/or macrophage cells were scored. Colonies derived from single cells of freshly isolated progenitors were replated, and secondary GM colonies were scored after another 7 days of semisolid culture. To assess the number of hematopoietic progenitor cells or more primitive stem cells cobblestone area–forming cell (CAFC) assays were performed in the population of 7-day differentiated cells. Both assays were performed as described previously.²⁸ 

Seven-day cultured cells (CD45.1) alone or mixed with 0.5 × 10⁶ of freshly isolated BM competitor cells (CD45.2) were injected into the retro-orbital sinus of lethally irradiated (9.5 Gy) mice (CD45.1/2). In transplantations without competitors, recovery of peripheral blood counts was determined up to 32 days by automatic cell counting of blood samples (Medonic CA 620 analyzer). Long-term repopulation potential of cells transplanted in the presence of competitors was monitored by calculation of the percentage of CD45.1, CD45.2, and CD45.1/2-derived cells in the blood. Data were acquired with an LSR-II (BD Bioscience) and analyzed with FlowJo PC Version 7.6.5 software (TreeStar)

Global gene expression was assessed in 7-day cultured cells (growth factor [GF] only, Li, VPA, Li + VPA). All samples were analyzed in independent biologic triplicates. Total RNA was isolated with the RNeasy kit (QIAGEN), according to manufacturer's protocol. RNA concentration, quality, and integrity were measured with the Experion Automated Electophoresis System (Bio-Rad). RNA was amplified with the Illumina TotalPrep RNA Amplification Kit (Ambion/Applied Bioscience) and hybridized to Mouse Ref-8_V2 expression platform (Illumina) according to the manufacturer's instruction. Scanning was performed on the iScan System (Illumina). Image analyses and extraction of raw expression data were performed with BeadStudio Version 3.2 software (Illumina) with default settings, no background subtraction, and no normalization. The raw data were thresholded at 1, log2-transformed and quantile-normalized with the use of GeneSpring-GX11.0 (Agilent). From the starting probe list (25 697 probes) probes not expressed in any replicate of the 4 conditions were excluded. The filtered list represented 20 586 probes. Probes differentially expressed between drug-treated and control sample (GFs only) were defined by Welch t test with Benjamini-Hochberg correction and false discovery rate < 0.05. Next, a 1.5-fold differential expression cutoff was applied. Gene expression analyses of 4 distinct primary cell subsets (LSK, L⁻S⁻K⁺, Gr1⁺, Ter119⁺), as described previously,²⁹  were used for comparison of VPA and Li effects with the physiologic hematopoietic differentiation program. Finally, data were subjected to relevance network analyses as described in Voy at al.³⁰  Networks were visualized in Gephi Version 0.8 beta software with the use of the Yifan-Hu and Fruchterman-Reingold force-directed layout algorithm.³¹  All raw data were deposited in the NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) as accession number GSE34088.

The ChIP assay was performed to determine histone acetylation at promoter regions of specific genes with the use of the method described by Frank et al.³²  In brief, 5 × 10⁶ cells were fixed with 1% formaldehyde for 10 minutess at RT. After cell lysis, cross-linked chromatin was fragmented by sonication to obtain ∼ 200- to 600-bp fragments. Precleared chromatin was precipitated with anti–acetyl-Histone4 (Millipore) Ab. The immunoprecipitated chromatin was washed subsequently with washing buffers.³²  For reverse cross-linking, samples were incubated overnight at 65°C in 1% SDS, 0.1M NaHCO-containing buffer, followed by treatment with Rnase A (Roche) and proteinase K (Fermentas Gmbh). DNA was isolated by ethanol/chloroform/isopropanol precipitation. The promoter regions of genes of interest were amplified by quantitative PCR (Bio-Rad) with the use of gene-specific primer pairs for β-actin (fwd, GGCAGTGTCCACAAGGGCGG; rev, TTGAGGAAGAGGATGAAGAGTTTTGGCG), HoxA7 (fwd, GCCAGTCTTCCAGCATGGCCTG; rev, CTCTGCTGCCCAACGCTCTCTG), and HoxB4 (fwd, GGGAGGGGTAGAGAAGGGGAAATAAACC; rev, GCCACCCGGCCTGCGATTC).

In vitro differentiation cultures are a convenient model to study the effects of new compounds on cell proliferation and differentiation. Purified HSPCs, defined by a LSK phenotype, were differentiated into the granulocyte/macrophage lineage by stimulation with SCF and GM-CSF in the presence or absence of VPA with or without Li. After a 7-day liquid culture the effect of both compounds on cytokine-induced HSPC differentiation was studied with respect to cell proliferation and differentiation.

The addition of Li, either alone or in combination with VPA, led to lower numbers of nucleated cells, whereas this effect was not observed on VPA exposure (Figure 1A). Despite decreased proliferation, the combination of VPA + Li resulted in a significantly higher proportion of viable cells as measured by PI exclusion, used as a marker for dead cells (Figure 1B). Because the cultures were initiated with purified LSK cells, after 7 days of differentiation Lineage, Sca-1, and c-kit marker expression was reanalyzed by flow cytometry. Combined drug treatment (VPA + Li) resulted in a significantly higher (8-fold) number of cells with a LSK phenotype compared with control cells grown with GFs only. On VPA exposure a 6-fold increase in number of phenotypically defined LSK cells was observed, whereas Li had only a marginal (1.5-fold) effect (Figure 1C-D). The immature cell phenotype as measured by flow cytometry was in the agreement with morphology of the cells (Figure 1E-F). Quantification of the percentage of cells with blast-like morphology showed that cells treated with VPA + Li contained 7.5-fold more cells with immature morphology. VPA treatment lead to a 4.5-fold increase, whereas the addition of Li resulted in only 2.5-fold more cells with blast-like morphology over control conditions (Figure 1E-F). These data indicate that VPA and Li treatment preserved the HSPC phenotype and morphology after strong differentiation-inducing cultures.

To test whether cells exposed to VPA and Li retained stem/progenitor function, in vitro assays for hematopoietic cell activity were performed. Progenitor potential of 7-day cultured cells was determined with semisolid CFU assay, whereby cells were stimulated into granulocyte/macrophage colonies with GM-CSF and SCF. Cells treated with VPA alone or with a combination of VPA + Li gave rise to a higher number of colonies than control cells (45-fold and 81-fold more, respectively). However, treatment with Li did not significantly influence colony-forming capacity (Figure 2A). To test the functional activity of more primitive cells after 7 days of differentiation culture, we performed CAFC assays. In this assay, progenitor activity is measured by early appearing CAFCs (day-7) and stem cell activity by late appearing CAFCs (day 35). Cells exposed to the combination of both compounds (VPA + Li) showed a 30-fold increase in absolute numbers of early- and a 3.5-fold increase in absolute numbers of late-appearing CAFCs (P < .05) compared with control cells (GFs only). Cells treated with VPA alone showed 7-fold and 2-fold more early and late CAFC activity (P < .05), respectively. In contrast, for Li-treated cells no enhancement of early or late CAFC activity was observed (Figure 2B). As shown by these 2 independent assays, the in vitro functional potential of stem and progenitor cells was significantly enhanced by VPA and VPA + Li treatment, but the combination of both compounds most efficiently preserved HSPC function.

To confirm the functional activity observed with in vitro assays, the in vivo reconstitution properties of cells differentiated in the presence of VPA and/or Li were examined. The progeny of 2000 LSK cells, cultured for 7 days with GM-CSF with or without VPA and Li, was transplanted into lethally irradiated mice. At days 10, 14, 21, 28, and 32 after transplantation survival and recovery of peripheral blood counts were analyzed. Mice that received a transplant with cells exposed to VPA + Li had significantly (P < .05) improved survival (12 of 14 mice survived) compared with control mice that received cells cultured with GFs only (6 of 14 mice survived; Figure 3A). In addition, blood cell counts showed that the red blood cell numbers and platelet numbers recovered faster in recipients of VPA + Li-treated cells compared with control mice, confirming more progenitor cell activity in the transplanted cell population (Figure 3B-C). In contrast, mice reconstituted with cells cultured in the presence of a single drug, either VPA or Li, did not show improved survival or blood recovery over mice that received a transplant with cells differentiated with GFs only (supplemental Figure 1, available on the Blood Web site; see the Supplemental Materials link at the top of the online article). These data strongly indicate that the combination of VPA + Li most potently preserved immature, radioprotective hematopoietic progenitor cells.

To evaluate the effects of VPA + Li treatment on long-term repopulating capacity, the cultured progeny of 2000 LSK cells was transplanted into lethally irradiated recipients together with freshly isolated BM competitor cells. Recipient mice transplanted with cells cultured in the presence of VPA + Li displayed higher chimerism levels through the entire time course of the experiment compared with mice that received a transplant with cells cultured in the presence of GFs only. However, the differences between the 2 groups decreased over time (Figure 3D). The percentage of chimerism was significantly higher in the VPA + Li group up to 13 weeks after transplantation, suggesting robust preservation of short-term repopulating stem/progenitor cells. Furthermore, 28 weeks after transplantation only 2 of 5 recipients transplanted with cells cultured with GFs only reached chimerism levels > 1%, whereas > 1% of donor cell contribution was observed for all mice (5 of 5) in the VPA + Li group. Chimerism levels remained stable and distinguishable between the 2 groups until 36 weeks (supplemental Figure 1). In addition, treatment with VPA + Li did not influence the multilineage capacity of the cells because all mature blood types were found in comparable ratios (supplemental Figure 1). Collectively, these data strongly indicate that addition of VPA + Li to differentiation-inducting cultures preserved cells with in vivo repopulation potential.

Since our results indicate that VPA and Li enhanced short-term repopulating progenitor cells, the effects of the compound treatment on distinct progenitor cell subsets were characterized. To address this question we isolated by flow cytometry uncommitted (LSK) and committed myeloid (CMP and GMP) cell subsets and performed methylcellulose assays for each sorted cell population (Figure 4A). Cells were stimulated to differentiate with SCF and GM-CSF in the presence of VPA and/or Li. VPA and VPA + Li decreased colony-forming activity of LSK cells, showed no effect on clonogenic activity of CMP cells, but displayed strong potentiating effect on colony-forming ability of the GMP cell population (Figure 4A). GMPs had a limited colony-forming ability (∼ 40 colonies per 1000 cells), which increased in the presence of VPA and VPA + Li by almost 8-fold. Li, as a single agent, showed no effect on colony-forming ability of the tested progenitor cell populations.

To investigate whether the observed effects on clonogenic activity were caused by either proliferation or cell death (or both), single-cell proliferation assay was performed. Single cells of each distinct progenitor population (LSK, CMP, GMP) were sorted into 96-well plates in liquid differentiation medium containing SCF and GM-CSF and were cultured in the presence or absence of VPA and/or Li for 7 days. Cell death was not significantly affected by drug treatment in any of the tested cell populations (Welch t test, P > .05; Figure 4B). Cell proliferation was quantified by scoring the size of the clone grown from a single cell. Clone size was classified as small, medium, or large. Less than 40% of GMPs at the single-cell level were able to grow into medium or large colonies, whereas on addition of VPA and VPA + Li > 80% of colonies grow into larger (medium + large) colonies. The enhancing effect of VPA and VPA + Li on cell proliferation was also apparent at the level of CMPs, however, to a lesser extent. Surprisingly, VPA and VPA + Li decreased proliferative capacity of the LSK cells. At the single-cell level Li alone showed no effects on the proliferation of distinct hematopoietic cell populations (Figure 4B). These data indicate that VPA displayed a major effect on cell proliferation, which was potentiated in the presence of Li.

Next, the effect of VPA treatment on plating efficiency of distinct cell progenitor populations was tested with the use of single-cell CFU assay. Single cells of each of the 3 progenitor populations were sorted into 96-well plate in semisolid medium with SCF and GM-CSF and cultured with or without VPA and/or Li. VPA and VPA + Li decreased the plating efficiency of LSK cells, did not influence CMP plating efficiency, but significantly increased the plating efficiency of GMP at the single-cell level (in accordance with Figure 4A). Strikingly, whereas ∼ 10% of purified GMPs were able to produce colonies in the presence of GFs only, the addition of VPA or VPA + Li increased clonogenic activity of GMPs to 60% (Figure 4C). Next, the self-renewal potential of compound-exposed cells was determined by replating colonies derived from a single cell. From each group, ∼ 30 individual colonies were harvested and replated to new methylcellulose cultures in the presence or absence of VPA and/or Li. In the presence of GF only ∼ 50% of individual colonies formed by single LSK cells were able to generate secondary colonies, whereas on exposure to VPA and VPA + Li 100% of the replated colonies derived from single LSK cells gave rise to secondary colonies. Furthermore, VPA and VPA + Li not only enhanced LSK replating efficiency of individual colonies but also increased the total number of secondary colonies derived from a single cell. Strikingly, although CMP and GMP cells did not show any replating activity by themselves, the addition of VPA and VPA + Li induced replating activity of these progenitor populations, resulting in ∼ 60% and ∼ 35% of secondary colonies, respectively. The addition of Li to primary or secondary single-cell cultures did not affect cell (re-)plating efficiencies (Figure 4D).

Summarizing, the effect of VPA or the combination of VPA + Li on colony-forming ability varied, depending on the differentiation status of the cell. VPA and VPA + Li inhibited proliferation and colony-forming ability of LSK cells but increased its self-renewal potential measured by replating. Although proliferation and clonogenic activity of CMPs were not significantly influenced, VPA and VPA + Li induced replating capacity of this population. GMPs were specifically sensitive to VPA or the combination treatment, because the compounds not only strongly increased clonogenic activity, plating efficiency, and proliferation of this population but also induced their self-renewal capacity.

To identify molecular events by which VPA and/or Li are able to inhibit differentiation, we performed expression arrays on 7-day differentiated LSK cells in the presence or absence of VPA with or without Li. RNA samples were harvested at the time when in vitro and in vivo assays were initiated. Differential expression analyses (for details, see “Gene expression analysis”) showed distinct effects of Li and VPA on gene expression. Although no differentially expressed genes could be identified on Li treatment compared with GFs only (Figure 5A), VPA significantly affected 60 genes, of which 10 were up-regulated and 50 down-regulated. Strikingly, an obvious more than additive effect was observed when Li was added to VPA in our differentiation cultures. The combined treatment of VPA + Li resulted in differential expression of 360 genes (110 up-regulated and 250 down-regulated), showing a strong synergistic effect (Figure 5A). The lists of all genes differentially expressed on the compounds treatment is presented in supplemental Table 1.

To assess whether VPA + Li treatment inhibited differentiation by preserving a primitive cell transcriptome, the changes in expression of VPA + Li targets (Figure 5A and supplemental Table 1) were compared with normal hematopoietic differentiation. For this purpose, previously obtained transcriptome data of 4 developmentally distinct hematopoietic cell stages, stem cells (Lin⁻Sca1⁺ckit⁺), progenitors (Lin⁻Sca1⁻ckit⁺), myeloid (Gr1⁺), and erythroid (Ter119⁺) cells, were used.²⁹  Analysis found that target genes that were up-regulated by VPA + Li were more abundantly expressed in primary stem/ progenitor cells compared with differentiated myeloid and erythroid cells, whereas VPA + Li down-regulated targets showed the opposite pattern, being higher expressed in differentiated than in uncommitted cells. (Figure 5B)

Since VPA + Li treatment altered expression of 360 genes, an important question is whether expression changes of these targets are all independent or co-regulated. Cellular differentiation is most probably orchestrated by networks of co-regulated genes, rather than by changes in expression of individual genes. To address this question, the method described in detail by Voy et al ³⁰  was used. First, gene correlation studies (using Pearson coefficient) were performed with the same transcriptome data of the 4 developmentally distinct hematopoietic cell stages as used for determining cell type expression abundance of VPA + Li affected targets. The strongest correlations, most probably reflecting biologic relevance, were defined by less than −0.8 or > 0.85 thresholds and visualized with Gephi Version 0.8 beta software.³¹  Such a gene relevance network, reflecting similarity measures of gene expressions in 4 cell populations, is presented in Figure 5C. VPA + Li targets were color-coded according to the expression abundance. Interestingly, it appeared that VPA + Li targets are co-regulated during normal hematopoietic differentiation, creating high connectivity networks. In addition, genes clustered together according to a specific cell type, creating a network consisting of 3 apparent clusters. One combined cluster contained genes most abundantly expressed in either stem (green) or progenitor (yellow) cells, and 2 separate clusters contained genes predominantly expressed in myeloid (blue) or erythroid (purple) cells (Figure 5C). Network analysis found also differential gene induction or repression by the treatment. As expected, genes within a cluster, usually expressed in the same cell type, showed positive correlations (red lines) with each other, suggesting that expression levels of these genes are changed in parallel by VPA + Li. However, strong negative correlations (blue lines) were apparent between genes belonging to separate clusters. Almost all genes within the stem/progenitor cluster displayed inverse correlation with genes within myeloid and erythroid clusters, indicating interesting gene relations between uncommitted and differentiated cells. Collectively, our analyses show that VPA + Li affected the differentiation program of the cell by perturbing the coordinated expression of multiple genes in co-regulated fashion, preserving stem/progenitor genes, and down-regulating differentiation-associated genes.

To try to identify primary candidates probably involved in cell fate decisions, the VPA + Li target list was filtered for transcription factors (TFs) and other DNA and RNA interacting genes. Table 1 provides a full list of transcriptional regulators that were affected by the treatment. A reduced network retained the 3-cluster structure with strong negative correlations between stem/progenitor and differentiated-cell associated TFs (Figure 5D). Among VPA + Li up-regulated targets were several known stem/progenitor TFs, such as Gata2, Gfi-1, and HoxA7.^33-35  Other TFs, with still unknown functions in HSCs, could also be assigned to the stem cell cluster, including Rbbp7 and Phf5. Moreover, TFs involved in epigenetic events, such as Hmgn3 and Prmt5, were up-regulated by the VPA + Li treatment.³⁶  In contrast, most of the down-regulated targets were correlated with cellular differentiation, including known TFs for myeloid (Erg1/2, Junb),^37,38  erythroid/megakaryocyte (Dab2),³⁹  and lymphoid (Irf5, Nfkbia)^40,41  differentiation. TFs up-regulated by the treatment, showing a negative correlation with differentiation-associated TFs, seem to be especially interesting as targets for differentiation-preventing agents.

Because VPA is an epigenetic modifier, known to affect gene expression by preventing chromatin deacetylation, we tested whether the observed changes in gene expression were caused by altered histone acetylation levels. To this end, ChIP with the use of Abs against acetylated histone H4 were performed. As proof of principle, HoxA7, up-regulated by both VPA and VPA + Li, was selected for ChIP assay. Interestingly, several reports have suggested that another homeobox gene, HoxB4, is a target of VPA in human cells.^16-18  In our study, expression of HoxB4 was not affected by the compound treatment. The ChIP experiments showed that VPA caused hyperacetylation of histone H4 at regulatory sites of the HoxA7 promotor, but not of the HoxB4 promotor. This correlates with the 2.3- and 8-fold higher HoxA7 expression measured by microarray and quantitative RT-PCR (not shown), respectively, and lack of change in HoxB4 expression. As expected, Li did not affect the epigenetic status of promotor regions of tested genes (Figure 5F). This indicates that VPA affected expression of HoxA7 by increasing histone acetylation on regulatory regions of the gene.

Taken together, genome-wide gene expression profiling of VPA- and Li-treated cells showed the synergistic effects of these 2 compounds at the molecular level. Gene relevance networks indicated that these effects required simultaneous changes in multiple genes to prevent differentiation, which was achieved by enhancing genes involved in cell primitiveness and repressing genes associated with cellular differentiation.

To date, attempts to define culture conditions that would favor stem cell self-renewal over differentiation have resulted in limited success. We hypothesized that combinations of cytokines and small molecules might represent an integrative approach to prevent differentiation in culture. In this report, we tested the combination of a well-known HDI, VPA, with the simple cation, Li, on HSPC maintenance under strong differentiation pressure and showed that the combination of these 2 compounds displayed strong anti-differentiation effects on hematopoietic differentiation at the biologic and the molecular levels. Although Li as a single agent displayed limited effects on HSPC differentiation, VPA and the combination of VPA + Li were able to preserve in vitro functionality of LSK and induced self-renewal of committed progenitors (CMPs and GMPs). Strikingly, after a 7-day differentiation culture only the combination of these 2 agents provided radioprotection to lethally irradiated recipients, shortened the time required for platelet and red blood cell recovery, and enhanced in vivo repopulation potential. These observations are consistent with previous reports, suggesting that VPA or Li, can affect fate decision of HSPCs,^{16-18,20-22,25,26}  but to our knowledge this is the first report to show that VPA + Li act synergistically to delay HSPC differentiation.

In accordance with the biologic data, the most compelling evidence that VPA + Li act synergistically to inhibit LSK differentiation was provided by genome-wide gene expression analysis. Strikingly, VPA + Li affected expression of 360 genes, whereas VPA alone affected only 60 genes, and Li did not affect any genes. Network analysis provided evidence that most genes affected by VPA + Li were co-regulated during normal hematopoietic differentiation. In addition, VPA + Li preserved expression of stem cell–related genes, which coincided with repression of genes involved in differentiation. This indicates that the anti-differentiation effects of VPA + Li are caused by coordinated changes in multiple co-regulated genes, rather than in individual genes. Our findings were further illustrated by the fact that TFs (probably being the primary targets in the network) favoring self-renewal (eg, Gata2, Gfi-1, HoxA7)^33-35  were increased, and differentiation-associated TFs were decreased on VPA + Li treatment. It has been suggested that competition and balance between TFs, such as Gata2, PU.1, Gfi-1, Erg1/2, may regulate switches between self-renewal and differentiation and triggers the determination of cell fate.^3,37,42-44  The inhibition of the differentiation program was shown for multipotent uncommitted and also for committed cells, which are at various stages of differentiation. Strikingly, on addition of VPA and VPA + Li, individual CMP and GMP progenitor cells re-acquired replating capacity in methylcellulose assays, suggesting possible cellular de-differentiation. However, whether the same mechanism is responsible for the effects of Li + VPA on committed progenitor differentiation versus LSK differentiation remains to be determined.

The balance between histone acetylation and deacetylation is thought to play an important role in maintaining chromatin structure consistent with the stem cell state. Epigenetic regulation is believed to be important for cell fate decision, and epigenetic transcriptional repression was suggested to be required for silencing of stem cell genes and subsequent cell differentiation.^14,15,44  VPA has been shown to inhibit HDACs and therefore could interfere with the epigenetic status of the cell.⁸  We confirmed increased histone acetylation at the HoxA7 promotor region, correlating with the increased expression of HoxA7. In addition, other confirmed VPA + Li targets (Gata2) have been suggested to bind to HDAC or to be involved in epigenetic modifications (Hmgn3, Prmt5).^36,45  Moreover, the chromatin conformation maintained by VPA might allow a spectrum of Li targets to be better accessible. Thus, common targets of VPA and Li (eg, Wnt pathway)⁴⁶  in combination with more accessible chromatin conformation may result in the observed synergistic effect of these 2 compounds. Therefore, modulation of epigenetic regulation of gene expression, as shown here with VPA, can lead to coordinated preservation or up-regulation of genes involved in stem cell maintenance and simultaneous down-regulation of differentiation-associated genes. The resemblance with expression patterns during natural differentiation suggests a physiologic mechanism driving expression of multiple genes determining cell fate. Self-renewal and differentiation seem to be strongly co-dependent, thus ex vivo stem cell culturing systems should target both processes to maintain uncommitted cells.

Methods to inhibit stem cell differentiation may be of benefit in ex vivo cell therapy protocols, which favor in vitro HSPC self-renewal rather than commitment and differentiation. So far, culturing systems with “classic” hematopoietic GFs resulted in stem cell loss because of profound differentiation.^1,2  Recently, factors other than classic hematopoietic GFs (eg, RA, VPA, FGFs, SR1, pleiotropin)^{9,10,18,47,48}  showed a promising potential in ex vivo expansion protocols. Our study underlines that combinatorial screening of Li and VPA and other new compounds may uncover possible additive and/or synergistic effects to modulate HSPC self-renewal and differentiation and thereby may lead to novel approaches in culturing systems of ex vivo stem cells.

The authors thank Pieter van der Vlies and Bahram Sanjabi for technical assistance; Henk Moes, Geert Mesander, and Roelof Jan van der Lei for assistance on cell sorting; Adrian Bracken for advice in setting up the ChIP assay in our laboratory; and Brad Dykstra for scientific advice.

This work was supported by the European Community's Seventh Framework Program (FP7/2007-2013; grant 222989; StemExpand), the Landsteiner Foundation for Blood Research (LSBR-0702), Dutch Platform for Tissue Engineering (DPTE; STW-GGT6727), and the Netherlands Organization of Scientific Research (VICI grant ZonMW 918.76.601, G.d.H.) and (VENI grant, G.H.).

Contribution: M.A.W., G.d.H., and R.v.O. designed research; M.A.W., V.v.d.B., S.O., A.A., and M.R. performed research; M.A.W, L.B., V.v.d.B., and G.H. analyzed and interpreted data; and M.A.W. wrote the manuscript with contribution from L.B., G.d.H., and R.v.O.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Ronald van Os, Department of Cell Biology, Section Stem Cell Biology, University Medical Center Groningen, University of Groningen, Groningen 9700 AD, The Netherlands; e-mail: r.p.van.os@umcg.nl.