Hematopoietic stem cells (HSCs) maintain lifelong hematopoiesis by remaining quiescent in the bone marrow niche. Recapitulation of a quiescent state in culture has not been achieved, as cells rapidly proliferate and differentiate in vitro. After exhaustive analysis of different environmental factor combinations and concentrations as a way to mimic physiological conditions, we were able to maintain engraftable quiescent HSCs for 1 month in culture under very low cytokine concentrations, hypoxia, and very high fatty acid levels. Exogenous fatty acids were required likely due to suppression of intrinsic fatty acid synthesis by hypoxia and low cytokine conditions. By contrast, high cytokine concentrations or normoxia induced HSC proliferation and differentiation. Our novel culture system provides a means to evaluate properties of steady state HSCs and test effects of defined factors
Discussion updated.
Hematopoietic stem cells (HSCs) maintain lifelong hematopoiesis at steady state and after stress by producing multipotent progenitors (MPPs), lineage committed progenitors, and their differentiated progeny within bone marrow niche (
HSC culture
Hypothesizing that conditions favoring quiescence are present in bone marrow but lacking in many culture systems, we aimed to define a minimal set of factors that would mimic the bone marrow microenvironment and keep HSCs quiescent and functional in vitro. We show here that HSCs have a unique nutrient requirement for fatty acids. We also optimized environmental conditions, including low cytokine concentrations and maintenance of hypoxia, to favor maintenance of undifferentiated and quiescent HSCs. These conditions were equally applicable to both murine and human HSCs. Development of this system opens an avenue to study steady state HSC properties and manipulate defined factors
To assess potential changes in gene expression in HSCs after conventional culture, we performed cDNA microarray analysis comparing HSCs immediately after sorting and HSCs cultured for 16 hours with SCF and TPO with or without serum (
(
(
(
(
(
(
(
Most circulating fatty acid is contained in lipoproteins, which are partially hydrolyzed to produce free fatty acids, which are then bound to albumin as a carrier (
(
(
(
(
(
(
Addition of 4% nBSA to mimic serum albumin levels restored HSC proliferation capacity in the presence of inhibitors of fatty acid synthesis to a level similar to 10% serum conditions, as compared to 0.1% nBSA or 4% FA-free BSA (
Cytokine combinations that favor HSC quiescence
(
(
(
(
In vivo HSCs remain generally quiescent with minimal output of differentiated cells. Thus, we assessed HSC maintenance
Phenotypic HSCs defined by surface markers are not necessarily functional. Thus we characterized function of HSCs cultured in 3ng/mL SCF, 0.1ng/mL TPO, 4% BSA and 1% O2 conditions (hereafter, termed “maintenance conditions”, and see Methods for details). We first assessed cell cycle status in maintenance conditions versus “high cytokine” (100ng/mL SCF and 100ng/mL TPO) conditions favoring proliferation under either hypoxia or normoxia. After 16 hours of culture and a 2hr EdU pulse, only 4% of cultured HSCs under maintenance conditions incorporated EdU in contrast to 40% under proliferation conditions (
(
(
(
To evaluate repopulation capacity, we transplanted fresh HSCs or HSCs cultured 14 or 27 days under maintenance or high cytokine conditions plus 4% nBSA and 1% O2 into lethally-irradiated recipient mice (
To assess fatty acid requirements in maintenance conditions, we cultured HSCs for 28 days in medium containing either 4% nBSA, 4% FA-free BSA, or 4% reconstituted BSA and then performed transplantation. Peripheral blood chimerism in bone marrow was comparable between the nBSA and reconstituted BSA groups, while the FA-free BSA group exhibited no engraftment (
Multipotent progenitors (MPPs) differ from HSCs in terms of lower self-renewal capacity and higher cell cycling activity (
As noted, HSCs cultured under maintenance conditions differ from fresh HSCs in terms of lower repopulation capacity following primary transplantation. To understand why, we performed cDNA microarray analysis to capture the transcriptomic profile (
(
(
(
(
(
(
(
(
(
We next assessed heterogeneity of gene expression in either fresh or cultured HSCs at the single cell level by performing single-cell qPCR (sc-qPCR) analysis (
We next modified the HSC culture medium as a means to establish gene expression patterns more closely resembling those seen in fresh HSCs. Further addition of fatty acids or fatty acids plus cholesterol to 4% nBSA medium dose-dependently downregulated Scd1 and Pmvk expression to levels seen in fresh HSCs, while the megakaryocyte genes or Evi1 or Flt3 remained unchanged (
As noted, cytokine concentration is a major quiescence/differentiation fate-determinant for HSCs
(
(
(
(
(
(
Next, we examined the nBSA requirement for HSC survival at lower cytokine concentrations (3-9ng/mL SCF) under hypoxia or normoxia. In 0.1% nBSA medium, HSCs did not survive at any SCF concentration, and the effect was exacerbated under 1% O2 conditions (
We next assessed our maintenance conditions in cultures of human HSCs. First, we created a cytokine response map as we had in murine HSCs (
(
(
(
(
(
(
(
(
Next we examined fatty acid or cholesterol requirements using human HSCs. Addition of fatty acid alone to cultures containing nBSA dose-dependently decreased HSC number, while further addition cholesterol rescued HSC numbers (
Manipulation of HSCs
Recently, liver-derived TPO was shown to extrinsically regulate HSCs outside bone marrow (
Lower cytokine concentrations are critical to maintain HSCs in an immature state but may insufficient for de novo biosynthesis of essential factors such as fatty acids (as shown here), possibly due to suppressed expression of biosynthetic enzymes. These results are consistent with accumulating finding showing that AKT activation perturbs HSC maintenance and that quiescent HSCs are metabolically inactive (
Although we achieved 1-month maintenance of functional HSCs
The system defined here has numerous potential benefits for HSC research and engineering. A more efficient culture system will allow (1) better analysis of HSC behavior in
We thank all members of the Takubo laboratory for indispensable support; M. Haraguchi and S. Tamaki for technical support and laboratory management; and E. Lamar for preparation of the manuscript. KT was supported in part by KAKENHI Grants from MEXT/JSPS (26115005, 18H02845, 18K19570, 26115001, 15K21751), a grant of the National Center for Global Health and Medicine (26-001, 29-2007), AMED-CREST (JP18gm0710010), an AMED grant for Realization of Regenerative Medicine (JP18bm0704011) and grants from the Japan Leukemia Research Fund, the Japan Rheumatism Foundation, the Takeda Science Foundation, the Senshin Medical Research Foundation, and the Japanese Society for Hematology. HK was supported in part by a KAKENHI Grant (17K16200), a grant of the National Center for Global Health and Medicine (29-1015) and a grant from the Uehara Memorial Foundation. HS was supported in part by AMED-CREST (JP18gm0910011 and JP18gm0710002). MS was the leader of JST ERATO Suematsu Gas Biology until March 2015, which provided infrastructure for multi-photon laser confocal microscopy, essential to accomplish the aim of this current study.
H.K., T.M., A.O., F.H., S.W., and T.H-Y. performed the study and analyzed data; G.N., D.H., H.S., F.A., Y.K., M.S., and T.S. provided scientific advice and materials. H.K. and K.T. wrote the manuscript. K.T. conceived the project and supervised the research.
The authors declare no competing interest.
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Keiyo Takubo (
C57BL/6J mice (8-14 week-old, purchased from SLC Japan or CLEA Japan) were used in all experiments, unless otherwise stated. C57BL/6-Ly5.1 congenic mice purchased from CLEA Japan were used for competitive repopulation assays. Ubc-GFP reporter mice (
Human CD34+ bone marrow cells were purchased from Lonza and stored in liquid nitrogen until use. Frozen cells were thawed in vials in a 37°C water bath and transferred to a 15mL tube. Medium (SF-O3 or αMEM or DMEM/F12) containing 10 % FCS plus 2U/mL DNaseI (Sigma Aldrich) was slowly added to the suspension while swirling gently to fill the tube. The suspension was centrifuged at 200 x g for 15 minutes at room temperature. Supernatants were aspirated and the wash step repeated. An antibody cocktail (50μL of PBS + 2%FCS plus 10μL of anti-CD34-FITC, 2μL of anti-CD38-PerCP-Cy5.5, 5μL of anti-CD90-PE-Cy7, and 10μL of anti-CD45RA-PE) was added to the suspension and kept on ice for 30 minutes. Cells were washed once with PBS+2% FCS and resuspended in PBS+2%FCS with 0.1% PI. Cells were sorted into SF-O3 or αMEM containing 4% BSA + 55μM 2-ME by FACSAria II.
Mouse bone marrow cells were isolated from two femurs and tibiae. Femurs and tibiae were flushed with PBS+2% FCS using a 21-gauge needle (Terumo) and a 10mL syringe (Terumo) to collect the bone marrow plug. The plug was dispersed by refluxing through the needle and the suspension centrifuged 680 x
SF-O3 medium (Sanko Junyaku) is a mixture of RPMI1640, Dulbecco’s MEM, and Ham’s F-12 medium at a 2:1:1 ratio with added cytidine, glutathione, para-amino benzoic acid, and cholesterol (a formula suggested by Shin-ichi Nishikawa, personal communication) supplemented with recombinant-human insulin, recombinant-human transferrin, sodium selenite, ethanolamine, HEPES, and sodium bicarbonate 2.2g/L. We further added 0.1% v/v of BSA by diluting a 10% BSA concentrate (Sanko Junyaku) or adding 0.1% w/v BSA powder (Sigma Aldrich). Additional BSA powder or fatty-acid-free BSA (Sigma Aldrich) was directly dissolved in the medium at indicated concentrations and filtered using 0.22μm filter. 2-ME was added before filtration at final concentration of 55μM. SCF, TPO, and other cytokines or reagents were added to the filtered media.
For other culture media (αMEM or DMEM/F12 medium), we added human recombinant insulin (Nacalai Tesque) before filtration at indicated final concentrations. Sorted murine or human cells were centrifuged at 340 x
For donor cells, HSCs (CD150+ CD41- CD48- CD34- Flt3- LSKs) from C57BL/6-Ly5.2 mice or HSCs (CD150+CD41-CD48-Flt3- LSKs) from Ubc-GFP mice or cultured HSCs, together with 4×105 BMMNCs from untreated C57BL/6-Ly5.1 or Ly5.2 mice, were used. For recipients, C57BL/6-Ly5.1 or Ly5.2 congenic mice were used. Donor cells were transplanted retro-orbitally into recipients that had been lethally-irradiated (9.5Gy using MBR-1520R (Hitachi Power Solutions), 125kV 10mA, 0.5mm Al, 0.2mm Cu filter). At 1, 2, 3, and 4 months after BMT, peripheral blood was collected and the percentage of donor-derived cells and their differentiation status determined by MACSQuant. Forty to eighty μL of peripheral blood was sampled from the retro-orbital plexus using heparinized glass capillary tubes (Drummond Scientific) and was suspended in 1mL of PBS+heparin. For platelet and red blood cell (RBC) analysis, 5μL of blood suspension was mixed with 50μL staining solution (PBS + 2% FCS with Ter-119-PerCP-Cy5.5 (1:100), and anti-CD41-PE(1:100)) for 30 minutes at 4°C, washed once and analyzed by MACSQuant. For white blood cell analysis, the blood suspension was centrifuged at 340 x
100 x Donor-derived (GFP+ or Ly5.2+Ly5.1-) cells (%) / (Donor-derived cells (%) + Competitor- or recipient-derived (Ly5.2-Ly5.1+ or GFP-) cells (%))
Myeloid cells, B cells, T cells, RBCs, and platelets were marked by Gr-1+ or Mac-1+, B220+, CD4+ or CD8+, CD41- Ter119+, and CD41+ Ter119-, respectively. Total cell chimerism represents the frequency of donor-derived Ly5.2+Ly5.1- or GFP+ cells over the frequency of Ly5.2-Ly5.1+ or GFP- cells in mononuclear or unlysed cells. At 4 months after BMT, the frequency of donor-derived cells in bone marrow was determined using one femur and tibia per recipient. After counting bone marrow cells using an automated counter TC10 (Bio-Rad), equal volumes of cell suspensions (20-30% of total volume) from each recipient were pooled and 2 x 106 of cells were resuspended in SF-O3 medium (0.1% BSA). The cell suspension (200μL/recipient) was transplanted by retro-orbital injection into lethally (9.5 Gy)-irradiated Ly5.1+ recipients with a 1mL syringe and 27-gauge needle. Remaining cells were stained to assess bone marrow chimerism. Anti-CD150-BV421, anti-CD48-PE, anti-Flt3-APC, anti-lineage (CD4, CD8a, Gr-1, Mac-1, Ter-119, B220)-PerCP-Cy5.5, anti-c-Kit-APC-Cy7, anti-Sca-1-PE-Cy7, anti-Ly5.1-Alexa-Fluor700, and anti-Ly5.2-FITC were used for surface antigen detection. All antibodies were 1μL per sample.
A total of 5000 HSCs (CD34+ CD38- CD90+ CD45RA-) from frozen samples or the progeny of 5000 cultured HSCs (day 0 equivalent) were transplanted retro-orbitally into irradiated (2.5Gy) NOG mice. At 1, 2, and 3 months after BMT, peripheral blood was collected and the percentage of human cells and their differentiation status were determined by MACSQuant. After cell preparation as described above, cells were resuspended in 50μL PBS +2%FCS with 0.3μL anti-mouse Fc-block. Surface antigen staining was performed using the following antibody panel: mouse CD45-PE-Cy7, mouse Ter-119-PE-Cy7, human CD45-BV421, human CD13-PE, human CD33-PE, human CD19-APC, human CD3-APC-Cy7.
Myeloid cells, B cells, or T cells were determined as human CD45+ CD13/CD33+, human CD45+ CD19+, or human CD45+ CD3+ respectively. Total cell chimerism represents the frequency of PI human CD45+ murine CD45- Ter-119- cells over total PI- mononuclear cells.
Fatty acid sodium salt (a combination of palmitate, oleate, linoleate, and stearate) was dissolved in methanol to a concentration of 4-20mg/mL and cholesterol (Tokyo Chemical Industry Co., Ltd.) was separately dissolved in methanol to a concentration of 4mg/mL in glass tubes (Maruemu Corporation). Solutions were mixed in glass tubes, air-dried, and then heated on water-bath at 50°C until methanol had evaporated. Medium containing 4% w/v BSA was directly added to the tube and sonicated until lipids were dissolved. Medium was then filtered using a 0.22μm filter (Millipore). 2-ME or insulin (if needed) was added just before filtration. Other reagents were added after filtration.
We used four types of BSA in culture media: native BSA (nBSA, Sigma Aldrich), fatty-acid-free BSA (FA-free BSA, Sigma Aldrich), FA-free BSA reconstituted with fatty acids (reconstituted BSA), and nBSA reconstituted with fatty acids and cholesterol (enhanced BSA). Fatty acid stock solutions used for reconstituted BSA were either 1:1 sodium palmitate and sodium oleate or 4:3:2:1 sodium palmitate, sodium oleate, sodium linoleate, and sodium stearate. Fatty acids used for enhanced BSA were 4:3:2:1 sodium palmitate, oleate, linoleate, and stearate. In that case, the final concentration of fatty acids was 400μg/mL and of cholesterol was 20μg/mL.
Cultured HSCs were stained using an Annexin V-PE Apoptosis Detection Kit (BD Biosciences), according to the manufacturer’s instructions. Cells were resuspended in 250μL of PBS+2%FCS +0.1%PI, and apoptotic cells (Annexin V+ PI- cells) were detected by MACSQuant.
Cells were cultured 16 hours under indicated conditions, and 10mM EdU was then added to culture medium to a final concentration of 10μM and incubated at 37°C for 2 hours. Cells were permeabilized to detect intracellular EdU using a Click-iT EdU Alexa Fluor 647 Flow Cytometry Assay Kit (Thermo Fisher Scientific) and analyzed by MACSQuant (Miltenyi Biotec).
Murine BMMNCs were pooled from 4 mice per experiment. Cells were treated with 8μL (1:50) of Fc block for 10 minutes at 4°C. Since antigenicity of CD150 and Sca-1 is vulnerable to the fixation/permeabilization process, these markers were stained in advance. Cells were stained with 8μL (1:50) of CD150-BV421 antibody and 8μL (1:50) of Sca-1-Alexa700 antibody at 4°C for 30 minutes and then washed once in PBS + 2% FCS and resuspended in DMEM/F12 medium at 4 x107cells/mL. 50μL of cell suspension was dispensed to each well of 96-well round bottom plates followed by addition of 50μL DMEM/F12 containing 2x concentrations of cytokines. Plates were incubated at 37°C, 5%CO2, and 20% O2 conditions for indicated times. Cells were then fixed for 10 minutes at 37°C by adding 100μL of Fixation buffer I (BD Biosciences, PBS with 4% paraformaldehyde) and then chilled immediately by transferring to ice-cold 1mL of PBS/2% FCS. Cells were centrifuged (600 x
Fresh CD150+CD48-CD41-Flt3-CD34- LSK cells (HSCs) from 10 mice were sorted into 500µL SF-O3 containing 4% w/v BSA +2-ME. 33000 HSCs were split into 2 tubes, one (fresh HSCs) subjected to reverse transcription and preamplification using Fluidigm C1 system (Fluidigm) and the other cultured 7 days in 2 wells of a 96-well round-bottom plate (cultured HSCs) followed by reverse transcription and preamplification. SF-O3 containing 4% w/v BSA + 55μM 2-ME with SCF 3ng/mL plus TPO 0.1ng/mL was used for culture medium. Reverse transcription and pre-amplification was performed according to the manufacturer’s instruction using an Ambion Single Cell-to-CT kit (Thermo Fisher Scientifc) and TaqMan PCR probes (Thermo Fisher Scientific). Cells were diluted in Suspension Reagent (Fluidigm) to ∼1000 cells/µL and loaded to C1-Single-Cell Auto Prep IFC for Preamp (5-10μm). After loading cells, IFC was observed microscopically to ensure that single cells were captured in each well using IN Cell Analyzer 6000 (GE Healthcare). Wells with 0 or >1 or shrunken cells were excluded. A total of 80 wells of 96 for fresh HSCs, and 62 of 96 wells for cultured HSCs were regarded as valid. Thermal cycling conditions in reverse transcription, and preamplification were as follows: 25°C for 10min, 42°C for 1hr, and 85°C for 5 min for reverse transcription, and 95°C for 10min, 18 cycles of 95°C for 15s and 60°C for 4min, and kept 4°C until harvest. Amplified cDNA (about 3.5μL) from each sample was diluted with 25μL C1 DNA Dilution Reagent (Fluidigm) and frozen at -30°C until use. 48 samples of cultured or Fresh HSCs were subjected to qPCR. For single-cell qPCR analysis, a Biomark system (Fluidigm) in combination with Fluidigm 96.96 Dynamic Array IFC was used. For assay mix, 2.5µL of TaqMan Gene Expression Assay (Thermo Fisher Scientific) and 2.5µL of Assay Loading Reagent (Fluidigm) were mixed for each gene. For the sample mix, 2.5µL of TaqMan Fast Universal PCR Master Mix (Thermo Fisher Scientific), 0.25µL of GE Sample Loading Reagent (Fluidigm), and 2.25µL of cDNA were mixed. Assay and sample mixes were loaded onto 96.96 Dynamic Array IFC, and PCR analysis was performed on BioMarkHD.
After selection of c-Kit+ cells using magnetic beads, the murine hematopoietic stem and progenitor fraction was labeled as follows. For staining of C57BL/6J mice, lineage marker (CD4, CD8a, Gr-1, Mac-1, Ter-119, B220)-PerCP-Cy5.5, c-Kit-APC-Cy7, Sca-1-PE-Cy7, CD150-PE, CD41-FITC, CD48-FITC, Flt3-APC, CD34-BV421 were used. For staining of Ubc-GFP mice, lineage marker (CD4, CD8a, Gr-1, Mac-1, Ter-119, B220)-PerCP-Cy5.5, c-Kit-APC-Cy7, Sca-1-PE-Cy7, CD150-BV421, CD41-PE, CD48-PE, Flt3-APC were used. All antibodies used were 0.5µL per mouse. Cells were resuspended in 0.5 to 2mL of PBS +2%FCS + 0.1%PI and sorted by FACSAria II into SF-O3 containing 4% w/v BSA or other medium (αMEM or DMEM/F12) with 4% w/v BSA. For murine cell experiments, when cultures were conducted in fatty-acid-free conditions, fatty-acid-free BSA was used for sorting medium. For human cell experiments, fatty-acid free BSA was not used for sorting medium due to low cell survival rate. Murine HSCs were defined as CD150+CD41-CD48-CD34-Flt3-LSK or CD150+CD41-CD48-LSK (
Male mice (24–31g, 11–13 weeks old, non-fasting) were anesthetized via intraperitoneal injection of urethane (800mg/kg) and α-chloralose (80mg/kg), tracheotomized, and intubated with a handmade Y-shaped tube for mechanical ventilation. Animals were mechanically ventilated with a small-animal ventilator (MiniVent type 845, Harvard Apparatus) with 21% O2 at a tidal volume of 8μL/g and a respiratory rate of 120 breaths/min. The left femoral artery and vein were cannulated to monitor mean arterial pressure (MAP) and intravenous chemical administration, respectively. An arterial catheter connected to a pressure transducer (MP 150, BioPac Systems) was placed in the left femoral artery to continuously monitor MAP and heart rate. Rectal temperature was maintained at 37.0 ± 0.5 °C throughout the experiment using a heating pad (ATC-2000, World Precision Instruments). The head of each mouse was fixed in a stereotactic frame (SG-4 N, Narishige Scientific Instrument Lab) in the sphinx position. The skull bone was exposed by a midline skin incision. Images were acquired using a two-photon laser microscope (FV1000MPE, Olympus) attached to a mode-locked titanium-sapphire laser system (Chameleon Vision II, Coherent) that could achieve a 140-fs pulse width and an 80-MHz repetition rate. To visualize the bone marrow microvasculature, 500 kDa TRITC-dextran (0.2g/kg body weight, Merck) was injected. FITC-albumin (40mg/kg body weight, Merck) was administered intravenously to assess vascular leakage of albumin. FITC-albumin intensity was measured using Fluoview software (version FV10-ASW, Olympus).
Forty to eighty μL of peripheral blood was sampled from the retro-orbital plexus
The serum was subjected to an ELISA using a Quantikine ELISA Kit (R&D Systems)
Fatty acid measurement was performed as described (
Most (170µL) of the medium in wells of a 96-well plate was aspirated and samples were stained with 10μL of antibody cocktail for 30 minutes at 4°C. For murine experiments, antibodies used were anti-lineage markers (CD4, CD8a, Gr-1, Mac-1, B220, Ter-119)-PerCP-Cy5.5, anti-c-Kit-APC-Cy7, anti-Sca-1-PE-Cy7, anti-CD150-PE, anti-CD48-FITC, anti-CD41-APC. All antibodies used were 0.1μL/well. For human experiments, anti-CD34-FITC (0.5µL/well), anti-CD38-PerCP-Cy5.5 (0.1µL/well), anti-CD90-PE-Cy7 (0.25µL/well), and anti-CD45RA-PE (0.5µL/well) were used per well. After incubation, 100µL of PBS + 2%FCS was added to wells, and the plates were centrifuged 5 minutes at 4°C at 400
Approximately 7000-15000 fresh or cultured cells per condition were subjected to RNA extraction using an RNA-easy mini kit (QIAGEN). cDNA was synthesized using SuperScript VILO (Thermo Scientific Technology) in a final volume of 20µL, according to the manufacturer’s instruction.
qPCR was performed using SYBR Premix ExTaq™ IIa (TaKaRa Bio) according to manufacturer’s instruction. A mixture of 45μL SYBR Premix, 0.36μL each of forward and reverse primers, 1.8µL of Rox II dye, 41.48µL of distilled water, and 1µL of cDNA solution was established and 20µL was dispensed to each of 4 wells of an assay plate. PCR analysis was performed using an ABI 7500 Fast Real-Time PCR System (Applied Biosystems) under the following conditions: 95°C for 10s followed by 40 cycles of 95°C for 5s and 60°C for 34s. Expression levels were determined as 2^(Ct value – mean Ct value of GAPDH) and were normalized to control samples, unless otherwise stated.
CD150+CD41-CD48-CD34-Flt3-LSK cells of pooled bone marrow from 10 (for 7-day cultures) or 20 (for 16-hour cultures) mice were sorted into SF-O3 medium and then and either lysed for the fresh sample or cultured. Cultured cells were centrifuged at 340 x
Data are presented as means ± SD, unless otherwise stated. For multiple comparisons, statistical significance was determined by Tukey’s multiple comparison test using the Tukey HSD function or one-way or two-way ANOVA using the anova and the aov function of R software. An unpaired two-tailed Student’s t test was used for experiments with two groups. In some transplantation assays with high variation, the Wilcoxon rank sum test was calculated using the wilcox.eact function of R software. False discovery rate (FDR) was calculated using the qvalue function of the qvalue package from Bioconductor (
Scanning data and normalization were performed by DNA Chip Research Inc. The hybridized gene chip was scanned using DNA MicroArray Scanner (Agilent), and scanned images were analyzed with Feature Extraction Ver.9.5.3 (Agilent). Normalization was performed using GeneSpring software (Agilent). Raw data were first transformed to log2 scale with expression levels <1.0 set to 1.0. Gene expression levels across samples were normalized using the 75 percentile shift algorithm. Visualization of scatter plots of normalized expression with gene name annotation was performed using the ggplot2 package of R software. Genes with a log2 fold difference in expression of >1.5 were shown in dark blue, and genes of interest (i.e. fatty-acid-related genes) were in red. Genes with a “Not detected” flag were removed from the scatter plot.
Gene ontology (GO) analysis was performed using GeneSpring software (Agilent) for differentially expressed genes (DEG) between 2 samples. DEG was determined as genes with a fold difference of expression >2.0 between 2 samples.
Normalized expression data were assessed using GSEA v2.0.13 software (Broad Institute). Gene sets were obtained from the Molecular Signatures Database v4.0 distributed at the GSEA website (
Ingenuity pathway analysis (QIAGEN) was performed using microarray data from fresh HSCs in serum-free conditions (16 hours). The top 200 differentially-expressed genes were analyzed, and significantly enriched canonical pathways or upstream regulators in serum-free conditions were listed.
Doubling time in
Upper estimate was:
7 / log2 ((Mean of total cell count -SD) / (Input cell number))
Lower estimate was:
7 / log2 ((Mean of total cell count +SD) / (Input cell number))
When (Mean-SD)/Input <1, the result was denoted as NA.
Data were analyzed using Biomark qPCR analysis software (Fluidigm). Melting curves showing low quality (Quality threshold <0.65) and a Peak Ratio Threshold of <0.8 were interpreted as not detected. The baseline was corrected by linear correction. Data were exported to a .csv file and downstream analysis was performed using R software. Cells lacking expression of Gapdh, Hprt1 or Actb were excluded, leaving 46 of 48 fresh cultures and 42 of 48 cultured HSCs. Genes not expressed in any cell (such as Cdkn2a, Cdkn2b, Csf1r, and Il7ra) were also excluded from analysis, leaving 92 of 96 genes. For each sample, Ct values were normalized to expression levels of housekeeping genes by subtracting average Ct values of Actb and Gapdh (normalized Ct values were designated ΔCt values). ΔCt values for not detected genes were set to the maximum ΔCt value of each gene + 3.5. For principal component analysis, the prcomp function implemented in R software was used on ΔCt values. Housekeeping genes (Actb, Gapdh, and Hprt1) were removed prior to analysis. Data scaling was not applied. The first 2 principal components were used to generate the plot. Hierarchical clustering with Euclidean distance and complete linkage clustering was performed on the correlation coefficient matrix of selected genes calculated by the cor function with Pearson’s correlation coefficient. Genes with a p-value <0.05 (calculated from ΔCt values using the t.test function) between fresh and cultured HSCs were selected for the cluster analysis. Hierarchical clustering was performed using the hclust package and the heatmap.2 function from the gplots package. All figures, including violin plots of ΔCt values, PCA plots, and hierarchical clustering, were visualized using the ggplot2 package. For visualization of expression levels in violin plots, the maximum ΔCt of each gene + 3.5 – ΔCt was shown, setting expression levels of not detected genes to zero. The false-discovery rate of differences between expression levels in fresh versus cultured cells was calculated using the qvalue function in the qvalue package after creating a list of p-values for each gene using the t.test function.
Cell number or fold-change of phosphoproteins under various cytokine conditions was visualized followed by smoothing using R software, so that the global landscape of HSC behavior in the cytokine concentration space is readily recognized. Local regression was performed to fit a polynominal surface to data points using the loess function with parameters set as, degree = 2, and span = 0.25. X-, and Y-axes were expanded from 7 x 7 matrix to a 61 x 61 matrix, and data prediction was performed using the predict function. When data were log scale (as in total cell number or megakaryocyte number), the raw data value was increased by 1 to bring the data point above 0. 2D contour plots were generated using the ggplot2 package, and 3D plots were generated using the rgl package.
cDNA microarray data generated here are accessible in the GEO database under the accession number GSE117515 and GSE117516. All software packages and methods used in this study have been properly detailed and referenced under “QUANTIFICATION AND STATISTICAL ANALYSIS”.
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(
(