Prostaglandin E₂ regulates murine hematopoietic stem/progenitor cells directly via EP4 receptor and indirectly through mesenchymal progenitor cells

Hematopoietic stem cells (HSCs) maintain a balance between self-renewal and differentiation to replenish all blood cell types. HSC regulation requires both intrinsic and extrinsic signals. Cytokines, chemokines, adhesion molecules, proteolytic enzymes, and neurotransmitters are produced by environmental niches functioning together to regulate HSC function and fate.^1-3 

Prostaglandin E₂ (PGE₂) is synthesized from arachidonic acid by cyclooxygenases (COX) and prostaglandin E synthases.⁴  PGE₂ modulates various pathologic and physiologic activities, such as cancer, fever, inflammation, atherosclerosis, blood pressure, stroke, and reproduction, as recently reviewed by Legler et al.⁵  PGE₂ also modulates hematopoiesis.^6-14  PGE₂ is involved in the production of definitive HSCs in the aorta-gonad-mesonephros region and in HSC engraftment.¹⁵  PGE₂ enhances HSC homing, survival, and proliferation.¹⁶  More recently, a stable derivative of PGE₂, 16,16-dimethyl PGE₂ (dmPGE₂), was shown to significantly enhance the engraftment of human cord blood cells in a xenotransplantation model.¹⁷ 

Although PGE₂ could be used in a clinical setting to improve hematopoietic transplantation therapy, the expression profiles of its 4 receptors (EP1, EP2, EP3, and EP4) vary within different organs and tissues.¹⁸  EP signaling pathways differ according to EP subtypes. Binding of PGE₂ to EP2 and EP4 induces production of 3′-5′-cyclic adenosine monophosphate (cAMP) leading to protein kinase A (PKA)–dependent glycogen synthase kinase-3β (GSK-3β) phosphorylation and β-catenin activation. EP2 couples strongly with Gs protein and increases cAMP levels more effectively than EP4.^19,20  EP4 activates β-catenin signaling through phosphoinositide 3-kinase (PI3K) as well as cAMP/PKA.^19,20  In contrast to EP2 and EP4, an activation of EP1 regulates Ca²⁺ channel gating via an unidentified G protein.²¹  The activation of EP3 inhibits adenylate cyclase via Gi protein and decreases cAMP levels.¹⁸  A recent study demonstrated that crosstalk between the PGE₂ and Wnt signaling at the level of β-catenin stabilization in hematopoietic stem/progenitor cells (HSPCs) contributes to the regulation of HSPC development and regeneration.²²  This finding suggests that the increase in cAMP concentrations and GSK-3β phosphorylation via EP2 or EP4 signaling are key events during PGE₂-mediated HSC regulation. However, it is not completely clear whether EP2 or EP4 is primarily responsible for the regulation of adult HSCs.

PGE₂ also affects the bone marrow (BM) microenvironment.^23,24  When mice were continuously administered PGE₂, they exhibited less bone trabeculae and more short-term HSCs and multipotent progenitors.²⁴  Although it was suggested that these effects are the result of enlargement of the physical space in the BM, PGE₂ may also stimulate a nonhematopoietic cell population to indirectly regulate HSPCs.

Here, we analyzed the ability of PGE₂ signaling to directly regulate HSPCs, and we also studied the indirect regulation of HSPC activity by PGE₂ via stimulation of a nonhematopoietic cell population. We found that PGE₂ signaling positively regulates HSCs both directly and via activation of a nonhematopoietic cell population, and that EP4 is a major receptor for the PGE₂-mediated regulation of HSPCs.

C57BL/6 mice were purchased from Japan SLC. C57BL/6 mice congenic for the Ly5 locus (B6-Ly5.1) were purchased from Sankyo-Laboratory Service. Because EP4 knockout (KO) mice do not survive on a C57BL/6 background,^25,26  they were maintained on a mixed background of 129/Ola × C57BL/6. Animals were cared for in accordance with the guidelines of the Keio University School of Medicine for animal and recombinant DNA experiments. All animal experiments were approved by the ethical review board of the Keio University School of Medicine.

Antibodies used for fluorescence-activated cell sorter (FACS) analyses and immunoblotting are listed in the supplemental data.

PGE₂ was purchased from Sigma-Aldrich (St. Louis, MO), and dmPGE₂ was purchased from Enzo Life Sciences (Farmingdale, NY). The EP1-selective agonist (ONO-DI-004: 17S,17,20-dimethyl-2,5-ethano-6-oxo-PGE₁), the EP2-selective agonist (16s-9-deoxy-9β-chloro-15-deoxy-16-hydroxy-17,17-propano-19,20-didehydro-PGE₂), the EP3-selective agonist (ONO-AE-248: 11,15-O-dimethyl-PGE₂), and the EP4-selective agonist (16-(3-methoxymethyl)phenyl-ω-tetranor-3,7-dithia prostaglandin E₁) were provided by Ono Pharmaceutical (Osaka, Japan).

The preparation of BM cells and immunostaining for flow cytometry have been described.²⁷  The isolation of endosteal cell populations on the bone surface has been described.²⁸  Stained cells were analyzed and sorted using a FACSvantage DiVa flow cytometer (BD Biosciences, San Jose, CA) and FACSAria (BD Biosciences).

Methods for conventional quantitative reverse-transcription polymerase chain reaction (qRT-PCR, Digital PCR, and Digital Array; Fluidigm), and qPCR array (Dynamic Array, Fluidigm) analyses are described in the supplemental data.

To evaluate the colony-forming abilities of Lin⁻Sca1⁺cKit⁺ (LSK) cells from EP2 KO or EP4 KO mice, LSK cells were sorted from EP2 KO, EP4 KO, or wild-type (WT) mice and cultured in MethoCult GF M3434 medium (StemCell Technologies, Vancouver, BC). The number of colony-forming units in culture (CFU-Cs) containing more than 50 cells was scored after 7 days of culture. To evaluate the number of high proliferative potential colony-forming cells (HPP-CFCs), the number of colonies more than 2 mm in diameter was counted after 2 weeks of culture. To analyze the effects of EP agonists on colony-forming ability, LSK cells were treated with control (ethanol), PGE₂ (0.1, 1, 10, or 100 μM), an EP2 agonist (0.1, 1, 10, or 100 μM), or EP4 agonist (0.1, 1, 10, or 100 μM) in 2% fetal calf serum phosphate-buffered saline (FCS-PBS) for 2 hours on ice. After incubation, treated cells were cultured in MethoCult GF M3434 medium (StemCell Technologies). To determine whether PEG₂ or EP agonists have an effect to stimulate BM niche cell compartment to regulate HSPCs, endosteal fractions (ALCAM⁺Sca-1^–, ALCAM^–Sca-1^–, and ALCAM^–Sca-1⁺ cells) were isolated and were cultured on type 1 collagen–coated plates (BD Biocoat Cellware; BD Biosciences) (7000 cells/well) in MSCBM (Lonza, Walkersville, MD). After 4 days of culture, each endosteal cell fraction was treated with dmPGE₂ (10 μM), EP agonists (0.1, 1, and 10 µM), or ethanol for 2 hours at 37°C and then washed with PBS. LSK cells (400 cells) were cocultured on each endosteal population for 3 days in MyeloCult (StemCell Technologies) without cytokines. Cells were then collected for the colony assay.

To evaluate the effects of treatments on the levels of phosphorylated GSK-3β, nonphosphorylated GSK-3β, and phosphorylated β-catenin in HSPCs, BM Lin^–c-Kit⁺ cells were treated with dmPGE₂ (10 μM), the EP4 agonist (10 or 100 μM), or the EP2 agonist (10 or 100 μM) in 2% FCS-PBS for 30 minutes at room temperature. Cells were lysed and subjected to sodium dodecyl sulfate and polyacrylamide gel electrophoresis and immunoblotting using antibodies against phospho-GSK-3β (Ser9) (D85E12, Cell Signaling Technology, Danvers, MA) (1:1000), nonphosphorylated GSK-3β (27C10, Cell Signaling) (1:1000), or phosphorylated β-catenin (Ser675) (D2F1, Cell Signaling) (1:100). To assess downstream signaling, BM Lin^–c-Kit⁺ cells were treated with dmPGE₂ (10 μM), EP4 agonist (100 μM), or EP2 agonist (100 μM) in the presence of PKA (10-μM H89, Sigma-Aldrich) or PI3K (10-μM LY294002, Sigma-Aldrich) inhibitors in 2% FCS-PBS for 30 minutes at room temperature. Densitometric quantification was performed on scanned immunoblot images using Multi Gauge Version 3.1 (FujiFilm, Tokyo, Japan).

EP4 short hairpin RNA (shRNA) sequences were as follows: shEP4-1, 5′-CGAGTGTTCATTAACCAGTTA-3′ and shEP4-2, 5′-CCACCTCGCTGAGAACTTT-3′. The sequence was separated by a 9-nucleotide noncomplementary spacer (TTCAAGAGA) from the corresponding reverse complement of the same 21-nucleotide sequence. A scrambled sequence (5′-GTCCCTATCGCCATTTACT-3′) served as a control. Oligonucleotides were cloned into BglII and HindII sites of the pReGS retrovirus vector provided by Dr. Hara (Tokyo Metropolitan Institute of Medical Science). Retrovirus-shEP4 was produced using the PlatE cell line.

LSK cells were precultured for 1 day, transfected with retrovirus-shEP4 on RetroNectin (Takara Bio, Shiga, Japan) using Magnetofection (OZ Biosciences, Marseille, France), and were incubated for an additional 2 days at 37°C in 5% CO₂. Cells were maintained in serum-free medium (SF-O3) containing 1.0% bovine serum albumin, 100 ng/mL of stem cell factor, 100 ng/mL of thrombopoietin, and 10⁻⁵ M of 2-mercaptoethanol.

Donor cells were transplanted into lethally irradiated recipient mice. Peripheral blood (PB) chimerism was analyzed every month. The percentage of donor-derived cells in BM and the ability of donor-derived cells to differentiate into various cell types were analyzed 3 to 4 months after BM transplant (BMT). Details of the BM reconstitution assay are described in the supplemental data.

To induce myelosuppression, 8-week-old C57BL/6 mice were injected with 5-FU (150 mg/kg body weight, intraperitoneal [IP] route). To evaluate the effects of PGE₂ treatment, 5-FU–treated mice were injected intraperitoneally twice daily for the indicated days with either PGE₂ (6 mg/kg body weight, IP) or vehicle (4% ethanol in PBS) from the day after 5-FU injection. In some cases, 5-FU–injected mice were treated with the EP4 selective agonist (50 μg/kg body weight, IP) or vehicle twice per day for 7 days. BM mononuclear cells (MNCs) were harvested 12 hours after the final injection for analysis.

Statistical significance was determined by a 2-tailed Student t-test, the Tukey test, or the Dunnett test.

To determine the expression of EP receptors in fractionated mouse BM cells, we undertook qPCR analysis (Figure 1A). We found that Ep4 was highly expressed in LSKCD34^-Flt3^–, LSKCD34⁺Flt3^–, and LSKCD34⁺Flt3⁺ cells compared with Lin^– or Lin⁺ fractions. Ep2 messenger RNA (mRNA) expression was highest in LSKCD34⁺Flt3⁺. EP3, a negative receptor that decreased cAMP,¹⁸  was expressed in Lin^–, LSKCD34^-Flt3^–, LSKCD34⁺Flt3^–, and LSKCD34⁺Flt3⁺ cells. EP1 was not expressed in hematopoietic cells. Next, we investigated the mRNA copy numbers of Ep2, Ep3, and Ep4 by Digital Array and found that the level of Ep2 transcripts in HSPCs was much lower than that of EP3 and EP4 transcripts (Figure 1B). The estimated copy numbers of Ep2, EP3, and Ep4 were 9.75 ± 2.48, 15.57 ± 2.59, and 26.94 ± 5.58 in LSKCD34⁺Flt3⁺ cells (per 250 cells), 0.79 ± 0.05, 16.07 ± 2.44, and 15.72 ± 1.63 in LSKCD34⁺Flt3^– cells (per 250 cells), 1.55 ± 1.42, 21.06 ± 4.46, and 29.14 ± 3.39 in LSKCD34^–Flt3^– cells (per 250 cells), respectively.

Osteoblasts in BM reportedly express PGE₂.^29-31  However, because cells in the endosteum are heterogeneous, it is unclear which populations are responsible for PGE₂ production. Therefore, we analyzed the expression of genes encoding enzymes that participate in PGE₂ synthesis, namely, Cox1, Cox2, prostaglandin E synthase 1 (Ptges1), Ptges2, and Ptges3, in BM endosteal cell populations. Proinflammatory stimuli markedly induce expression levels of Ptges1 and Cox2, accompanied by a marked increase in PGE₂ production. Prostaglandin E synthase 1 is also reportedly functionally coupled with COX2 in marked preference to COX1.^32,33  In contrast, Ptges2 is constitutively expressed primarily in the brain, heart, skeletal, kidney, and liver, and Prostaglandin E synthase 2 is functionally coupled with both COX1 and COX2. The endosteal cell fraction (CD45^–CD31^–Ter119^–) was subdivided into 3 fractions based on activated leukocyte cell adhesion molecule (ALCAM) and Sca-1 expression (supplemental Figure 1). We previously reported that mature osteoblasts, immature osteoblasts, and MPCs were enriched in ALCAM⁺Sca-1^–, ALCAM^–Sca-1^–, and ALCAM^–Sca-1⁺ cells, respectively.²⁸ Cox1 was highly expressed in ALCAM^–Sca-1^– cells, whereas Ptges2 was more highly expressed in ALCAM⁺Sca-1^– and ALCAM^–Sca-1⁺ cells than in ALCAM^–Sca-1^– cells. Cox2 was highly expressed in ALCAM^–Sca-1^– and ALCAM^–Sca-1⁺ cells. It is interesting to note that Ptges1 was highly expressed in ALCAM^–Sca-1⁺ MPCs (supplemental Figure 1). These data suggest that all endosteal cell populations contribute to PGE₂ production to varying extents.

Next, we examined EP4 KO mice to analyze the role of EP4 in PGE signaling in hematopoiesis. Although PB counts were normal, analysis of T, B, and myeloid cells in PB revealed that EP4 KO mice exhibited a slightly higher percentage of T cells (47.7 ± 5.0% in WT and 55.1 ± 6.5% in EP4 KO) and myeloid cells (9.3 ± 2.5% in WT and 14.0 ± 5.2% in EP4 KO) and a slightly lower percentage of B cells compared with WT controls (34.8 ± 5.8% in WT and 23.3 ± 5.0% in EP4 KO) (Figure 2A and supplemental Table 1). The number of CFU-Cs and HPP-CFCs did not significantly differ between WT and EP4 KO LSK cells (Figure 2B). Percentages of LSK, LSKCD34^–Flt3^–, and LSKCD41^–CD48^–CD150⁺ cells in EP4 KO mice were comparable to those seen in WT mice (Figure 2C-D).

Because EP4 KO mice on a C57BL/6 background die of patent ductus arteriosus,^25,26  we performed shRNA-mediated knockdown (KD) of EP4 in HSPCs. First, LSK cells were transduced with retrovirus expressing 2 different shRNAs, shEP4-1, or shEP4-2. Both suppressed EP4 expression in that cell type (Figure 2E). We then examined the LTR capacities of LSK cells transduced with scrambled shRNA, shEP4-1, or shEP4-2. The engraftment of donor cells 3 months after BMT was reduced after transduction of shEP4-1 (6.3 ± 4.0%) or shEP4-2 (4.0 ± 1.5%) compared with LSK cells transduced with the scrambled shRNA (9.2 ± 1.6%) at 3 months after BMT (Figure 2F).

We examined the differentiation capacity of donor-derived cells 4 months after BMT. Similar to the BM of EP4 KO mice, mice transplanted with EP4 KD LSK cells had higher percentages of T and myeloid cells and a lower percentage of B cells, compared with mice transplanted with scrambled shRNA-transduced LSK cells (supplemental Figure 2A-B).

We next evaluated EP2 receptor function in HSPC regulation by analyzing EP2 KO mice. These mice had normal PB cell counts and a normal differentiation capacity (supplemental Figure 3A and Table 2). The number of CFU-Cs and HPP-CFCs did not differ significantly between WT and EP2 KO LSK cells (supplemental Figure 3B). WT and EP2 KO mice had similar proportions of LSK, LSKCD34^–Flt3^–, and LSKCD41^–CD48^–CD150⁺ cells (supplemental Figure 3C).

When the LTR capacity of EP2 KO LSK and LSKCD34^– cells was analyzed, no defect in BM reconstitution was detected (supplemental Figure 3D). Short-term treatment with an EP2 agonist did not change the number of CFU-Cs and HPP-CFCs, suggesting that the colony-forming activity of LSK cells was unaffected (supplemental Figure 4). These data suggest that PGE₂/EP2 signaling does not affect the self-renewal activity or proliferation activity of HSPCs.

To evaluate the role of PGE₂ signaling through EP4 in HSPCs, we assessed the effect of the EP4 agonist on HSPC colony-forming activity. When LSK cells were briefly treated with dmPGE₂ or the EP4 agonist and cultured in methylcellulose medium, the number of CFU-Cs increased relative to the control (Figure 3A). In addition, dmPGE₂ and EP4 agonist treatments increased the number of HPP-CFCs (Figure 3B). A high dose (100 μM) of dmPGE₂ was less effective for HPP-CFC formation compared with the 0.1, 1, or 10 μM of dmPGE₂. This was not the case for the EP4 agonist.

Next, we examined the effect of the EP4 agonist on the LTR capacity of LSK cells. LSK cells were treated with dmPGE₂ (10 μM) and the EP4 agonist (10 or 100 μM) for 2 hours on ice and were transplanted into recipient mice. In the PB, 10 μM of dmPGE₂- and 100 μM of EP4 agonist–treated cells showed higher chimerism of donor-derived cells than the control cells 4 months after BMT, whereas no statistically significant difference was observed in the frequency of donor-derived cells between controls and 10 μM of EP4 agonist-treated cells (Figure 3C). On the other hand, in comparison with control cells, 10 μM of dmPGE₂ and EP4 agonist (10 or 100 μM)–treated cells showed greater engraftment in BM MNCs (Figure 3D) and in the LSK fraction (Figure 3E) at 4 months after BMT. These results indicate that PGE₂/EP4 signaling positively affects the engraftment of hematopoietic cells.

Next, we evaluated the function of PGE₂/EP4 signaling in the stabilization of β-catenin in HSPC. PGE₂ signaling reportedly interacts with the Wnt/β-catenin pathway.²²  GSK-3β is a component of the β-catenin destruction complex, and its phosphorylation at Ser9 decreases its ability to associate with other complex members. GSK-3β phosphorylation was monitored in Lin^–c-Kit⁺ cells after treatment with EP4 or EP2 agonists, or with dmPGE₂. Immunoblot analysis showed that treatment with the EP4 agonist, but not the EP2 agonist, increased GSK-3β phosphorylation in Lin^–c-Kit⁺ cells in a dose-dependent manner (Figure 4A). To determine which pathway is activated by PGE₂/EP4 signaling, Lin^–c-Kit⁺ cells were briefly treated with dmPGE₂ or the EP4 agonist in the presence of a PKA inhibitor (H89) or a PI3K inhibitor (LY294002). H89 treatment significantly decreased GSK-3β phosphorylation in Lin^–c-Kit⁺ cells (Figure 4B). These data indicate that the cAMP/PKA pathway was activated by PGE₂/EP4 signaling in HSPCs. In addition, treatment with dmPGE₂ or the EP4 agonist increased phosphorylation of β-catenin at Ser675, indicating that β-catenin was stabilized (Figure 4C).³⁴ 

Next, we examined the ability of PGE₂ signaling to regulate HSPCs by modulating the activity of MPCs. qPCR analysis demonstrated that EP1-4 was expressed more highly in ALCAM^–Sca-1⁺ MPCs than in the other fractions (Figure 5A-D). These findings indicate that MPCs in adult BM may be particularly sensitive to PGE₂. Digital Array analysis was performed to determine the mRNA level of each EP receptor in ALCAM^–Sca-1⁺ MPCs (Figure 5E). A high copy number of Ep4 was detected (280.9 ± 26.9 copies/400 cells). The copy number of the Ep1 receptor (113.92 ± 4.2 copies/400 cells) was higher than that of Ep2 (52.03 ± 8.6 copies/400 cells) and Ep3 (41.87 ± 16.18 copies/400 cells), but was lower than that of Ep4. We investigated whether the hematopoietic-supporting activity of each endosteal cell population was enhanced by PGE₂. When unstimulated, cells formed a greater number of CFU-Cs when cocultured with ALCAM^–Sca-1⁺ than when they were cocultured with other fractions or with total endosteal cells (Figure 5F). It is noteworthy that the pretreatment of ALCAM^–Sca-1⁺ MPCs with dmPGE₂ significantly enhanced CFU-C and HPP-CFC formation of LSK cells (Figure 5F), whereas treatment of ALCAM⁺Sca-1^–, ALCAM^–Sca-1^–, or total endosteal cells with PGE₂ had no effect on the colony-forming capacity of cocultured LSK cells. These data suggest that the MPC fraction promotes HSPC proliferation after dmPGE₂ stimulation. Because ALCAM^–Sca-1⁺ MPCs expressed a higher level of EP4 compared with the other EP receptors, we tested whether the EP4 agonist increased the activity of ALCAM^–Sca-1⁺ MPCs to drive CFU-C and HPP-CFC formation of LSK cells. Treatment of ALCAM^–Sca-1⁺ MPCs with dmPGE₂ or the EP4 agonist increased the number of CFU-Cs and HPP-CFC formed (Figure 5G-H). Colony formation was unaffected when LSK cells were cocultured with MPCs that had been pretreated with the other EP agonists (supplemental Figure 5). These data suggest that PGE₂ signaling via EP4 activates a niche cell population to support HSPC proliferation, and in this way, PGE₂ indirectly regulates HSPC activity.

Next, we investigated the expression of HSC niche-related genes in ALCAM^–Sca-1⁺ MPCs after treatment with dmPGE₂ or the EP4 agonist (Figure 6). qPCR array analysis revealed that the expression levels of Angpt1, Angptl2, Bmp6, Fgf2, Igf1, Igf2, Dll1, Jag1, Alcam, Fn1, Has1, and Mmp2 were up-regulated by both dmPGE₂ and the EP4 agonist. The expression levels of Angptl4, Bmp2, Fgf1, Flt3l, Il6, Kitl, Lif, Tgfb1, Vegfa, Wnt5a, Tnc, and Spp1 were up-regulated by dmPGE₂ but not by the EP4 agonist. In contrast, levels of Cxcl12, Nov, and Has2 were up-regulated only by the EP4 agonist. These patterns of gene expression suggest the possibilities that PGE₂ accelerated HSPC proliferation by the production of mitogenic ligands. In addition, the upregulation of cell adhesion molecule and extracellular matrix–related genes might induce the cell adhesion of HSPCs.

PGE₂/EP4 signaling induced stabilization of β-catenin in HSPCs (Figure 4), enhanced the LTR capacity of HSPCs (Figure 3), and enhanced the ability of MPCs to support the colony formation of HSPCs (Figure 5). Furthermore, dmPGE₂ treatment reportedly increases the number of white blood cells in PB and BM after 5-FU injection,¹⁵  whereas Cox2 KO mice show delayed BM recovery after 5-FU–induced myeloablation.³⁵  Thus, we investigated the function of PGE₂/EP4 signaling in the recovery of HSPCs after myelosuppression. To analyze the effect of PGE₂ on HSPC recovery, the proportion of LSK cells in 5-FU–injected mice treated with PGE₂ was examined 2, 4, 6, and 8 days after 5-FU injection. At 2 and 4 days after 5-FU treatment, the proportion of LSK cells in PGE₂-treated mice was comparable to that of control mice (Figure 7A-B). In contrast, PGE₂-treated mice demonstrated a higher frequency of LSK cells than that of control mice 6 and 8 days after 5-FU injection (Figure 7A-B). The number of LSK cells increased in mice treated with PGE₂ on day 8 after 5-FU treatment compared with controls (Figure 7C). Gene expression in LSK cells isolated from PGE₂-treated mice or control mice was analyzed 2, 4, 6, and 8 days after 5-FU injection. In the later phase (days 4-8), LSK cells derived from PGE₂-treated mice exhibited higher expression of Cyclin-dependent kinase (Cdk) 4, Cdk6, Ctnnb1 (β-catenin), and Myc than that of the control cells (supplemental Figure 6). In addition, LSK cells isolated from PGE₂-treated mice expressed higher levels of the HSC markers, Slamf1,³⁶ Cxcr4,³⁷ Tek (Tie2),²⁷  and Evi1³⁸  (supplemental Figure 6). These data indicate that PGE₂ treatment accelerates proliferation and/or protects HSCs from 5-FU-induced apoptosis in vivo, which enhances the recovery of LSK cells after myelosuppression in PGE₂-treated mice.

Next, we examined the effect of in vivo treatment with an EP4 agonist on HSPC recovery after myelosuppression in vivo. 5-FU-injected mice were treated with the EP4 agonist, and the proportion of LSK cells 8 days after 5-FU injection was determined (Figure 7D-F). The EP4 agonist promoted recovery of HSPCs after 5-FU–induced myelosuppression in vivo. A higher percentage and frequency of LSK cells was observed in EP4 agonist–treated mice than in control mice (Figure 7E-F), whereas the percentage of LSK cells in EP4 agonist–treated mice was lower than that in PGE₂-treated mice (Figure 7E). qPCR analysis of LSK cells from EP4 agonist–treated mice or control mice was performed 8 days after 5-FU injection and revealed that treatment with the EP4 agonist up-regulated Cdk4 and Ctnnb1 in a manner similar to PGE₂ treatment, whereas the expression levels of Myc and Cdk6 remained unchanged (supplemental Figure 7).

Finally, we examined the expression of PGE₂ synthesis–related enzyme genes in the BM endosteal cell populations 3, 6, 8, and 12 days after 5-FU injection. FACS analyses indicated that the percentage of cells in the ALCAM^–Sca-1⁺ MPC fraction increased after 3 and 6 days, and decreased 8 and 12 days after 5-FU administration (supplemental Figure 8A). Expression of Ptges1, Ptges2, Cox1, and Cox2 increased during BM recovery in ALCAM^–Sca-1⁺ MPCs (supplemental Figure 8B). These data indicate that PGE₂ production is up-regulated in ALCAM^-Sca-1⁺ MPCs under myelosuppressive conditions.

PGE₂ signaling interacts with the Wnt/β-catenin pathway and contributes to the generation and regenerative capacity of HSPCs.²²  Here, we demonstrated that the EP4 receptor plays a crucial role in the direct regulation of HSPCs by PGE₂ signaling. We also found that PGE₂/EP4 signaling supports hematopoiesis indirectly by stimulating a specialized population of cells in MPCs.

EP2 and EP4 receptors stimulate Gs protein, which leads to cAMP production and activation of the cAMP/PKA pathway,¹⁸  and modulation of Wnt/β-catenin signaling.²²  Digital PCR analysis clearly demonstrated that HSPCs had a higher copy number of EP4 than that of EP2 (Figure 1), and EP2 KO mice did not have defects in the number of HSPCs. EP2-deficient HSPCs showed normal colony formation and LTR capacities (supplemental Figure 3). In addition, we found that EP2 agonist treatment did not affect the colony formation of LSK cells (supplemental Figure 4). In contrast, EP4 KD reduced BM reconstitution (Figure 2), and both EP4 KO mice and recipient mice transplanted with EP4 KD LSK cells had increased percentages of T and myeloid cells and decreased percentages of B cells (Figure 2 and supplemental Figure 2). These findings suggest that EP4, but not EP2, affects the PGE₂-mediated direct regulation of adult HSCs. Furthermore, the EP4 agonist increased the colony-forming capacity and LTR capacity of LSK cells (Figure 3). Further investigations using, for example, the serial BMT assay will be necessary to elucidate whether PGE₂/EP4 signaling enhances the self-renewal activity of LT-HSCs.

We found that EP4 agonist induced phosphorylation of GSK-3β (Ser9) and β-catenin (S675). A PKA inhibitor significantly decreased PGE₂- and EP4 agonist–induced phosphorylation of GSK-3β, whereas a PI3K inhibitor did not (Figure 4). These data suggest that PKA is the main downstream effector of PGE₂/EP4 signaling that is responsible for activating β-catenin in HSPCs.

Wnt/β-catenin pathway is required for the development of leukemia stem cells in acute myeloid leukemia.³⁹  In colon cancer, PGE₂/EP2 signaling interacts with the Wnt/β-catenin pathway through PI3K/Akt to promote cancer cell growth.⁴⁰  Therefore, it would be interesting to determine whether PGE₂/EP4 signaling functions to activate Wnt/β-catenin signaling during the development of acute myeloid leukemia.

We identified that PGE₂ also indirectly regulated HSPCs by modulating ALCAM^–Sca-1⁺ MPCs. LSK cells cocultured with dmPGE₂-treated ALCAM^–Sca-1⁺ MPCs formed more CFU-Cs and HPP-CFCs than LSK cells cultured with control ALCAM^–Sca-1⁺ MPCs (Figure 5). Gene expression patterns in MPCs after PGE₂ treatment suggest that PGE₂ accelerated HSPC proliferation by the production of mitogenic ligands and induced the cell adhesion of HSPCs by the production of cell adhesion molecule and extracellular matrixes in MPCs. These data indicate that PGE₂ can induce the proliferation of HSPCs by activating MPCs. The EP4 agonist also activated ALCAM^–Sca-1⁺ MPCs, whereas the other EP agonists had no effect. These data suggest that the activation of ALCAM^–Sca-1⁺ MPCs by dmPGE₂ requires EP4. However, the effects of the EP4 agonist on HPP-CFC formation and gene expression were less potent than those of dmPGE₂. Therefore, the activation of EP4 and another receptor(s) in a specialized cell population, including MPCs, may be required to completely mimic the effects of dmPGE₂.

Macrophages and monocytes have strong PGE₂ biosynthetic capacity.^7,41,42  In this study, we found that ALCAM^–Sca-1⁺ MPCs expressed high levels of Ptges1 and Cox2 (supplemental Figure 1), which participate in PGE₂ synthesis under proinflammatory conditions. The expression of Ptges1, Pteges2, Cox1, and Cox2 was up-regulated in ALCAM^–Sca-1⁺ MPCs after 5-FU treatment (supplemental Figure 8). Therefore, we anticipate that MPCs facilitate the recovery of HSPCs under myelosuppressive conditions by producing PGE₂. Autocrine PGE₂ activity also activates MPCs and supports HSPC recovery.

PGE₂ treatment accelerates the recovery of hematopoietic cells after 5-FU–induced myelosuppression.¹⁵  On the other hand, the inhibition of either Cox1 or Cox2 decreases the recovery of PB and BM white blood cells.^9,15  Also, Cox2 KO mice show delayed recovery of BM after 5-FU–induced myelosuppression.³⁵  However, the population dynamics of hematopoietic cells during BM recovery have not been understood. Here, we demonstrated that treatment with PGE₂ or an EP4 agonist enhanced the recovery of LSK cells after 5-FU treatment. Gene expression analysis revealed that LSK cells from PGE₂-treated mice expressed higher levels of cell cycle–related genes and HSC markers than the controls (supplemental Figure 6). These data suggest that PGE₂ signaling promotes BM recovery by accelerating HSPC proliferation. The increased expression of HSC markers after 5-FU treatment suggests that PGE₂ may protect immature stem cells.

Although in vivo treatment of mice with the EP4 agonist improved the recovery of HSPCs after 5-FU–induced myelosuppression, its effect was weaker than that of PGE₂ (Figure 7). Therefore, PGE₂ may promote HSPC recovery after myelosuppression by 2 distinct mechanisms: direct activation of HSPCs by binding to EP4; and indirect activation of a particular cell population, including MPCs, by binding to EP4 and other unknown receptor(s).

HSCs are used to treat hematologic malignancies, immunodeficiencies, and BM failure. Because BM contains a low level of HSCs, the ability to culture HSCs ex vivo would be advantageous. Therefore, understanding how HSCs are regulated and the functions of each specialized cell population may facilitate the production of HSCs for transplantation therapy. PGE₂ is already used in various clinical settings and is an attractive potential therapy for patients with depleted HSCs. This study demonstrated that PGE₂/EP4 signaling increases the LTR capacity of HSPCs. The modulation of PGE₂/EP4 signaling may facilitate hematopoiesis in vitro. It would also be interesting to investigate the combinational activation of PGE₂/EP4 signaling and other cytokine signaling for the efficient induction of the expansion of HSCs. PGE₂ plays a role in the regulation of the mesenchymal cell functions in various tissues. Indeed, it has recently been reported that PGE₂/EP4 pathway has an antifibrosis effect in kidney interstitial cells.⁴³  In this study, we found that PGE₂ indirectly induced HSPC proliferation through the activation of MPCs. A more comprehensive understanding of these indirect effects of PGE₂ may facilitate the expansion of HSPCs for use in regenerative medicine.

The authors thank Ono Pharmaceutical for providing the EP selective agonists.

This study was supported by the Funding Program for Next Generation World-Leading Researchers (NEXT Program) and a Grant-in-Aid from the Japan Society for the Promotion of Science.

Contributions: Y.M.I. and F.A. designed and performed experiments, and analyzed and interpreted data; K.H., H.T., and K.T. participated in performing experiments, data analysis, and discussions; T.F. and S.N. provided the EP2 and EP4 KO mice; Y.M.I., F.A., and T.S. wrote the paper.

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

Correspondence: Fumio Arai and Toshio Suda; e-mail: farai@a3.keio.jp (F.A.) and sudato@z3.keio.jp (T.S.)