The selective prolyl hydroxylase inhibitor IOX5 stabilizes HIF-1α and compromises development and progression of acute myeloid leukemia

Abstract

Acute myeloid leukemia (AML) is a largely incurable disease, for which new treatments are urgently needed. While leukemogenesis occurs in the hypoxic bone marrow, the therapeutic tractability of the hypoxia-inducible factor (HIF) system remains undefined. Given that inactivation of HIF-1α/HIF-2α promotes AML, a possible clinical strategy is to target the HIF-prolyl hydroxylases (PHDs), which promote HIF-1α/HIF-2α degradation. Here, we reveal that genetic inactivation of Phd1/Phd2 hinders AML initiation and progression, without impacting normal hematopoiesis. We investigated clinically used PHD inhibitors and a new selective PHD inhibitor (IOX5), to stabilize HIF-α in AML cells. PHD inhibition compromises AML in a HIF-1α-dependent manner to disable pro-leukemogenic pathways, re-program metabolism and induce apoptosis, in part via upregulation of BNIP3. Notably, concurrent inhibition of BCL-2 by venetoclax potentiates the anti-leukemic effect of PHD inhibition. Thus, PHD inhibition, with consequent HIF-1α stabilization, is a promising nontoxic strategy for AML, including in combination with venetoclax.

Fig. 1. Loss of Phd2 impairs leukemogenesis in murine AML models.

a, PHD2 (EGLN1) and PHD1 (EGLN2) expression in human CD34⁺ cells, BM mononuclear cells (MNCs) and samples from patients with AML (for both EGLN1 and EGLN2 expression; CD34⁺, n = 16 patients; BM MNCs, n = 18 patients; AML, n = 671 patients). AU, abitrary units. b, PHD2 (EGLN1) and PHD1 (EGLN2) expression in control (CTL), cytologically normal with intermediate prognosis (CNI), cytologically abnormal not otherwise specified (CAO) and different cytogenetic subgroups of human AML BM samples. (CTL, n = 198 patients; CNI, n = 1,043 patients; CAO, n = 47 patients; Complex, n = 130 patients; del(5q), n = 7 patients; del(7q), n = 15 patients; inv(16), n = 40 patients; t(15;17), n = 54 patients; t(6;9), n = 4 patients; t(8;21), n = 62 patients; t(9;11), n = 19 patients; t(v;11), patients; Trisomy8, n = 16 patients). c, Phd2^fl/fl (CTL) and Phd2^fl/fl;Vav-iCre (Phd2^cKO) FL c-Kit⁺ cells were co-transduced with Meis1 and Hoxa9 retroviruses, serially re-plated in CFC assays and transplanted into lethally irradiated recipient mice. d, CFC counts of CTL and Phd2^cKO cells after each re-plating (n = 3 mice per genotype). e, Percentage of leukemic cells in the PB of recipient mice (n = 20 CTL mice; n = 21 Phd2^cKO mice) and survival curve of mice transplanted with CTL and Phd2^cKO leukemic cells (n = 12 mice per genotype), respectively. trx, transplantation. f, rtTA (CTL) and shPhd2/rtTA (shPhd2) leukemic cells were prepared as c. CTL and shPhd2 leukemic cells were serially re-plated in CFC assays. g, CFC counts of CTL and shPhd2 leukemic cells after each re-plating ± DOX (+DOX, n = 4 mice per genotype; −DOX, n = 8 mice per genotype). h, Percentage of annexin V⁺ToPro⁺ cells after three rounds of re-plating +DOX (n = 6 mice per genotype). i, LSK cells from iMLL-AF9;Control and iMLL-AF9;shPhd2 mice were sorted for in vitro and in vivo assays. j, CFC counts of iMLL-AF9;Control and iMLL-AF9;shPhd2 cells + DOX (n = 8 mice per genotype). k, Percentage of leukemic cells in PB of recipient mice 6 weeks following transplantation (n = 5 iMLL-AF9;Control mice; n = 6 iMLL-AF9;shPhd2 mice). l, Percentage of leukemic GFP⁺ cells in PB of recipient mice 6 weeks following transplantation (n = 6 mice per genotype). m, Survival curve of mice transplanted with iMLL-AF9;Control and iMLL-AF9;shPhd2 LSK cells (n = 6 iMLL-AF9;Control mice; n = 7 iMLL-AF9;shPhd2 mice). n, LDA in secondary recipients transplanted with indicated doses of CD45.2⁺ BM cells from primary recipients. o, Plot showing Poisson statistical analysis. Circles represent the percentages of negative mice for each cell dose and triangles represent any data values with zero negative responses. Solid lines indicate the best-fit linear model, and dotted lines represent 95% confidence intervals (CIs). LSC frequencies were calculated using the ELDA software. The exact n number per group and analyses from ELDA software are provided in the Source Data. p, CTL and shPhd2 leukemic cells were serially re-plated and transplanted into recipient mice without DOX. At 8 weeks post-transplant, recipient mice were continuously treated with DOX. q, Percentage of leukemic cells in PB of recipient mice (n = 12 mice per genotype). r, Survival curve of mice transplanted with CTL and shPhd2 leukemic cells (n = 12 mice per genotype). Data represent mean ± s.e.m. Comparisons with no P value are not significant (NS). P values were calculated using a two-tailed Mann–Whitney U-test and a paired or unpaired Student’s t-test, unless stated otherwise. Kaplan–Meier survival curve statistics were determined using the log-rank (Mantel–Cox) test. Source data

Fig. 2. PHD2 is dispensable for functional hematopoiesis.

a, Steady-state analyses of 8–10-week-old CTL and Phd2^cKO mice. b, PB counts include WBC, white blood cell; RBC, red blood cell; HGB, hemoglobin; HCT, hematocrit; PLT, platelet (n = 7 CTL mice; n = 4 Phd2^cKO mice). c, Total BM cellularity (n = 7 CTL mice; n = 4 Phd2^cKO mice). d, Total numbers of HSCs, MPPs, primitive hematopoietic progenitor cells (HPC-1 and HPC-2) (n = 7 CTL mice; n = 4 Phd2^cKO mice). e, The 5 × 106 CD45.2⁺ BM cells from 8–10-week-old CTL and Phd2^cKO mice were transplanted into lethally irradiated syngeneic CD45.1⁺/CD45.2⁺ recipient mice with 5 × 106 competitor CD45.1⁺ BM cells. Animals were analyzed 6 months after transplantation. f, Percentage of CD45.2⁺ cells in overall PB, B cell, T cell and myeloid compartments in recipient mice (at all time points n = 4 CTL mice; at month 1 and 3, n = 5 Phd2^cKO mice; at month 4 and 6, n = 4 Phd2^cKO mice). g, Steady-state analyses of 13–14-week-old CTL and shPhd2 mice treated with DOX for 8 weeks. h, PB counts (n = 5 CTL mice; n = 5 shPhd2 mice). i, Total BM cellularity (n = 11 CTL mice; n = 10 shPhd2 mice). j, Total numbers of HSC, MPP, HPC-1 and HPC-2 populations (n = 11 CTL mice; n = 10 shPhd2 mice). k, 200 HSCs from 8–10-week-old CTL and shPhd2 mice were transplanted into lethally irradiated syngeneic CD45.1⁺/CD45.2⁺ recipient mice together with 5 × 106 competitor CD45.1⁺ BM cells. Recipient mice were treated with DOX 6 weeks post-transplantation. Animals were analyzed 4 months after transplantation. l, Percentage of CD45.2⁺ cells in the overall, B cell, T cell and myeloid PB compartments in recipient mice. For overall engraftment (month 1, n = 7 CTL mice, n = 9 shPhd2 mice; month 2, n = 8 CTL mice, n = 9 shPhd2 mice; month 4, n = 7 CTL mice, n = 10 shPhd2 mice). For myeloid cell engraftment (month 1, n = 8 CTL mice, n = 9 shPhd2 mice; month 2, n = 9 CTL mice, n = 9 shPhd2 mice; month 4, n = 8 CTL mice, n = 10 shPhd2 mice). For B cell engraftment (month 1, n = 7 CTL mice, n = 12 shPhd2 mice; month 2, n = 7 CTL mice, n = 9 shPhd2 mice; month 4, n = 8 CTL mice, n = 10 shPhd2 mice). For T cell engraftment (month 1, n = 9 CTL mice, n = 12 shPhd2 mice; month 2, n = 8 CTL mice, n = 9 shPhd2 mice; month 4, n = 8 CTL mice, n = 10 shPhd2 mice). Data represent mean ± s.e.m. Comparisons with no P value are NS. P values were calculated using a two-tailed Mann–Whitney U-test and paired or unpaired Student’s t-test, unless stated otherwise. Source data

Fig. 3. Loss of Phd1 compromises AML propagation and maintenance.

a, Phd1^fl/fl (CTL) and Phd1^fl/fl;Vav-iCre (Phd1^cKO) FL c-Kit⁺ cells were co-transduced with Meis1 and Hoxa9 retroviruses, serially re-plated in CFC assays and transplanted into lethally irradiated recipient mice. Leukemic cells collected from primary recipient mice were then transplanted in lethally irradiated secondary recipients. b, CFC counts of CTL and Phd1^cKO cells after each re-plating (n = 5 CTL mice; n = 10 Phd1^cKO mice). c, Proliferation analyses with CTL and Phd1^cKO cells (n = 6 CTL biological replicates; n = 3 Phd1^cKO biological replicates). P value calculated at the 72-h time point. d, Percentage of leukemic cells in the PB of recipient mice in primary transplant (n = 15 mice per genotype). e, Survival curve of mice transplanted with CTL and Phd1^cKO leukemic cells in primary transplant (n = 18 mice per genotype). f, Percentage of CD45.2⁺ leukemic cells in BM of primary recipient mice at the end of the experiment (n = 3 mice per genotype). g, Percentage of c-Kit⁺ cells in BM of primary recipient mice at the end of the experiment (n = 3 mice per genotype). h, Proliferation analyses with CTL and Phd1^cKO cells collected from primary recipients (n = 6 biological replicates per genotype). P value was calculated at the 72-h time point. i, Percentage of leukemic cells in the PB of recipient mice in secondary transplant (n = 22 mice per genotype). j, Survival curve of mice transplanted with CTL and Phd1^cKO leukemic cells in secondary transplant (n = 12 CTL mice; n = 13 Phd1^cKO mice). k, Steady-state analyses of 8–10-week-old CTL and Phd1^cKO mice. l, Total BM cellularity (n = 9 CTL mice; n = 11 Phd1^cKO mice). m, Total numbers of HSCs, MPPs and primitive hematopoietic progenitor cells (HPC-1 and HPC-2) (n = 7 CTL mice; n = 6 Phd1^cKO mice). Data represent mean ± s.e.m. Comparisons with no P value are NS. P values were calculated using a two-tailed Mann–Whitney U-test and paired or unpaired Student’s t-test, unless stated otherwise. Kaplan–Meier survival curve statistics were determined using the log-rank (Mantel–Cox) test. Source data

Fig. 4. Daprodustat and IOX5 are potent PHD inhibitors with distinct modes of action.

a, View from a crystal structure of PHD2 complexed with NOG (PDB: 5LR9). b, Visualization of the binding mode of Dap to PHD2, based on a crystal structure of PHD2 complexed with CCT-6 (PDB: 5OX5). c, Visualization of the potential binding mode of IOX5 to PHD2, based on a crystal structure of PHD2 complexed with compound 1 (PDB: 6ST3). Mn in the crystal structures substitutes for catalytically active Fe(II). d, Western blot showing dose-dependent HIF-1α stabilization in HEK293T cells treated with IOX5 or vehicle CTL. GAPDH was used as a loading CTL. Roxadustat was used to achieve HIF-1α stabilization as a positive CTL for the HIF-1α western blot analysis. Blot representative of four independent experiments. e, Comparison of the broad-spectrum 2OG oxygenase inhibitors NOG and 2,4-pyidine dicarboxylic acid (2,4-PDCA) and the FIH selective inhibitor NOFD^,, with the PHD-selective inhibitors Dap, molidustat, roxadustat and IOX5 against isolated recombinant forms of 2OG-dependent oxygenases. For conditions see Supplementary Information. f, Synthetic route to IOX5. (i) T3P, DIPEA, DMF, room temperature, 16 h, 86%. (ii) Pyrazole, Pd^tBuXPhos G3, Cs₂CO₃, ^tBuOH, N₂, 60 °C, 16 h, 46%. (iii) Lithium chloride (LiCl), DMAc, 120 °C, microwave, 8 h, 30%. Source data

Fig. 5. Pharmacological inhibition of PHD2 stabilizes HIF-1α and substantially impacts AML growth and survival in vitro.

a, Western blot of HIF-1α in MOLM13, OCI-AML3, MV411 and THP-1 cells treated with Dap, IOX5 or vehicle CTL for 48 h. β-Actin was used as a loading control. Blot is representative of three independent experiments. b, Proliferation of iMLL-AF9- and Meis1/Hoxa9-transformed murine cells treated with Dap, IOX5 or vehicle CTL (iMLL-AF9; n = 9 biological replicates per group. Meis1/Hoxa9; n = 12 biological replicates per group). P values calculated at 72-h and 96-h time points, respectively. c, Annexin V⁺DAPI⁺ analyses of iMLL-AF9- and Meis1/Hoxa9-transformed murine cells treated with Dap, IOX5 or vehicle CTL (n numbers as in b). d, Proliferation analyses of FLT3-ITD-, PML-RARα- and AML1-ETO-transformed murine cells treated with Dap, IOX5 or vehicle CTL (FLT3-ITD n = 6 biological replicates per group; PML-RARα n = 6 biological replicates per group; AML1-ETO n = 3 biological replicates per group). P values calculated at the 72-h time point. e, Annexin V⁺DAPI⁺ analyses of FLT3-ITD-, PML-RARα- and AML1-ETO-transformed murine cells treated with Dap, IOX5 or vehicle CTL (FLT3-ITD n = 6 biological replicates per group; PML-RARα, n = 6 biological replicates per group; AML1-ETO, n = 3 biological replicates per group). f, Percentage of annexin V⁺PI⁺ cells (matched to vehicle CTL) of individual patient samples treated with Dap or IOX5 (n = 12 patients). g, Proliferation analyses of THP-1, KASUMI-1, MOLM13, OCI-AML3, MV411 and K562 cells treated with Dap, IOX5 or vehicle CTL (THP-1: CTL, 0 h n = 12 biological replicates, 48 h n = 8 biological replicates, 72 h n = 3 biological replicates; Dap, n = 6 biological replicates per time point; IOX5, 0 h n = 12 biological replicates, 48 h n = 9 biological replicates, 72 h n = 3 biological replicates; KASUMI-1, n = 6 biological replicates per group; MOLM13, n = 6 biological replicates per group; OCI-AML3, n = 6 biological replicates per group; MV411, n = 6 biological replicates per group; K562, n = 6 biological replicates per group). P values calculated at the 72-h time point. h, Annexin V⁺DAPI⁺ analyses of THP-1, KASUMI-1, MOLM13, OCI-AML3, MV411 and K562 cells treated with Dap, IOX5 or vehicle CTL (THP-1: CTL n = 6 biological replicates, Dap n = 5 biological replicates, IOX5 n = 6 biological replicates; KASUMI-1: CTL n = 5 biological replicates, Dap n = 5 biological replicates IOX5, n = 6 biological replicates; MOLM13: CTL n = 5 biological replicates, Dap n = 5 biological replicates, IOX5 n = 6 biological replicates; OCI-AML3: n = 6 biological replicates per group; MV411: CTL n = 6 biological replicates, Dap n = 6 biological replicates, IOX5 n = 5 biological replicates; K562: CTL n = 6 biological replicates, Dap n = 6 biological replicates, IOX5 n = 5 biological replicates). THP-1 = MLL-AF9 translocation; KASUMI-1 = AML-ETO translocation; MOLM13 = MLL-AF9 translocation and FLT3-ITD mutation; OCI-AML3 = DNMT3A, NRAS and NPM1 mutations; MV411 = MLL-AF4 and FLT3-ITD mutation; K562 = BCR-ABL translocation. Data represent mean ± s.e.m. Comparisons with no P value are NS. P values were calculated using a two-tailed paired Student’s t-test, unless stated otherwise. Source data

Fig. 6. PHD inhibition decreases AML engraftment and increases survival in vivo but does not affect normal hematopoiesis.

a, A total of 100,000 THP-1 cells were transplanted into NBSGW recipient mice. At 14 days following transplantation, recipient mice were treated with Dap, IOX5 or vehicle CTL 2× daily via i.p. injection for 21 days. b, Percentage of human CD45⁺CD33⁺ and human CD45⁺CD14⁺ cells in the BM (vehicle CTL-treated mice n = 5; Dap-treated mice n = 5; IOX5-treated mice n = 4). c, Percentage of human CD45⁺CD33⁺ and human CD45⁺CD14⁺ cells in the spleen (vehicle CTL-treated mice n = 4–5; Dap-treated mice n = 5; IOX5-treated mice n = 4). d, A total of 100,000 OCI-AML3 cells were transplanted into NBSGW recipient mice. At 14 days following transplantation, recipient mice were treated with IOX5 or vehicle CTL 2× daily via i.p. injection for 21 days. e, Percentage of human CD45⁺CD33⁺CD14⁺ cells in the liver, BM and PB, respectively (liver: n = 7 vehicle CTL-treated mice, n = 4 IOX5-treated mice; BM: n = 7 vehicle CTL-treated mice, n = 3 IOX5-treated mice; PB: n = 3 vehicle CTL-treated mice, n = 4 IOX5-treated mice). f, A total of 2,000 LSK cells from iMLL-AF9;Control mice were sorted and transplanted into irradiated recipient mice. At 40 days following transplantation, recipient mice were treated with IOX5 or vehicle CTL 2× daily via i.p. injection for 14 days. g, Survival curve of mice transplanted with iMLL-AF9;Control LSK cells treated with IOX5 or a vehicle CTL (n = 7 mice per group). h, LDA in secondary recipients transplanted with indicated doses of CD45.2⁺ BM cells from primary recipients. i, Plot showing Poisson statistical analysis. Circles represent the percentages of negative mice for each cell dose, triangles represent any data values with zero negative responses. Solid lines indicate the best-fit linear model and dotted lines represent 95% CIs. LSC frequencies were calculated using the ELDA software. The exact n number per group and analyses from ELDA software are provided in the Source Data. j, Steady-state analyses of 8–10-week-old C57Bl6 mice treated with IOX5 or vehicle CTL 2× daily via i.p. injection for 14 days. k, PB counts (n = 6 vehicle CTL-treated mice; n = 8 IOX5-treated mice). l, Total BM cellularity (n = 5 vehicle CTL-treated mice; n = 9 IOX5-treated mice). m, Total numbers of HSC, MPP, HPC-1 and HPC-2 populations (n = 5 vehicle CTL-treated mice; n = 7 IOX5-treated mice). Data represent mean ± s.e.m. Comparisons with no P value are NS. P values were calculated using a two-tailed Mann–Whitney U-test and paired or unpaired Student’s t-test, unless stated otherwise. Kaplan–Meier survival curve statistics were determined using the log-rank (Mantel–Cox) test. Source data

Fig. 7. Targeting PHD2 upregulates HIF target genes, including the pro-apoptotic BNIP3.

a, GSEA showing pathways up- and downregulated in IOX5-treated iMLL-AF9 and Hif1/2a^DKO Meis1/Hoxa9 cells (n = 4 RNA samples per group). False discovery rate (FDR) displayed on the graph. b, Proliferation of Hif1/2a^CTL and Hif1/2a^DKO cells treated with IOX5 or vehicle CTL (all n = 12 biological replicates per group apart from: Hif1/2a^CTL treated with vehicle CTL at 72 h n = 11; Hif1/2a^DKO treated with IOX5 at 24 h n = 9, 48 h n = 9, 72 h n = 9, 96 h n = 8). P values calculated at the 96-h time point. c, Venn diagram of overlapping genes between up- and downregulated genes in IOX5-treated iMLL-AF9 and Hif1/2a^DKO Meis1/Hoxa9 cells. Fisher’s exact test statistical analyses shown. Odds ratio 4.04 and 2.85, respectively. d, Scatter-plot showing the inverse correlation between up- and downregulated transcripts as shown in c. Statistical significance was evaluated using Spearman’s correlation test. Bnip3 is annotated. e, Relative levels of BNIP3 mRNA (normalized to ACTB) in MOLM13, OCI-AML3, MV411 and THP-1 cells treated with Dap, IOX5 or vehicle CTL. RNA extracted from individual biological replicates and plated in triplicate. (MOLM13: CTL n = 4 biological replicates, Dap n = 4 biological replicates, IOX5 n = 3 biological replicates; OCI-AML3: n = 3 biological replicates; MV411: n = 3 biological replicates; THP-1: n = 3 biological replicates) f, Western blot of BNIP3 in THP-1 cells treated with IOX5 or vehicle CTL. β-actin used as a loading CTL. Asterisk indicates a nonspecific band. Blot is representative of two independent experiments. g, MOLM13 cells were transduced with lentiviruses expressing scrambled short hairpin RNA (shCTL) and a shRNA targeting BNIP3 (shBNIP3). Annexin V⁺DAPI⁺ analyses of shCTL and shBNIP3 MOLM13 cells treated with IOX5 or vehicle CTL (n = 3 biological replicates per group). Data represent mean ± s.e.m. Comparisons with no P value are NS. P values were calculated using a two-tailed paired or unpaired Student’s t-test, unless stated otherwise. Source data

Fig. 8. PHD inhibition combined with venetoclax ablates AML growth in vitro and in vivo.

a,b, Proliferation and annexin V⁺PI⁺ analyses, respectively, of MOLM13, OCI-AML3, MV411 and THP-1 cells treated with Dap, IOX5, venetoclax (Ven), Dap + Ven, IOX5 + Ven or vehicle CTL (n = 4 biological replicates per group). Statistical significance between Ven only and Dap + Ven or IOX5 + Ven is represented on the graph. The statistical significance between CTL and all other experimental conditions is described below. MOLM13 P values; CTL versus IOX5 P = 0.0397, CTL versus Dap + Ven P = 0.0074, CTL versus IOX5 + Ven P = 0.0104. OCI-AML3 P values; CTL versus Dap P = 0.0325, CTL versus IOX5 P = 0.0047, CTL versus Dap + Ven P = 0.0143. MV411 P values; CTL versus Dap P = 0.0446, CTL versus IOX5 P = 0.0178, CTL versus Dap + Ven P = 0.0183, CTL versus IOX5 + Ven P = 0.0178. THP-1 P values; CTL versus IOX5 P = 0.0397, CTL versus Dap + Ven P = 0.0074, CTL versus IOX5 + Ven P = 0.0104. P values calculated at the 72-h time point (a). MOLM13 P values; CTL versus Dap P = 0.0049, CTL versus IOX5 P = 0.0001, CTL versus Ven P = 0.0033, CTL versus Dap + Ven P = 0.0005, CTL versus IOX5 + Ven P < 0.0001. OCI-AML3 P values; CTL versus Dap P = 0.0327, CTL versus IOX5 P = 0.0065, CTL versus Ven P = 0.0004, CTL versus Dap + Ven P = 0.0228, CTL versus IOX5 + Ven P = 0.0040. MV411 P values; CTL versus Dap P = 0.0028, CTL versus IOX5 P < 0.0001, CTL versus Ven P = 0.0029, CTL versus Dap + Ven P = < 0.0001, CTL versus IOX5 + Ven P = < 0.0001. THP-1 P values; CTL versus Dap P = 0.0236, CTL versus IOX5 P = 0.0484, CTL versus Ven P = 0.0008, CTL versus Dap + Ven P < 0.0001, CTL versus IOX5 + Ven P < 0.0001 (b). c, Percentage of annexin V⁺PI⁺ cells of individual patient samples treated with Dap, IOX5, Ven, Dap + Ven, IOX5 + Ven or vehicle CTL (n = 7 patients). Statistical significance represented as in a,b. P values; CTL versus Dap P = 0.0086, CTL versus IOX5 P = 0.0028, CTL versus Dap + Ven P = 0.0011, CTL versus IOX5 + Ven P = 0.0111. d, 100,000 MV411 cells were transplanted into NBSGW recipient mice. At 14 days following transplantation, recipient mice were treated with IOX5, Ven or vehicle CTL. Dosing regimen consisted of 2× daily via i.p. injection (IOX5 or vehicle) and/or 1× daily via o.g. (Ven or vehicle). After 14 days treatment, half of the cohort were analyzed for human AML cell engraftment, while the other half were observed for survival analyses. e, Percentage of human CD45⁺CD33⁺ and CD45⁺CD14⁺ cells, respectively, in BM (n = 5 mice per group). f, Percentage of human CD45⁺CD33⁺ and CD45⁺CD14⁺ cells, respectively, in the spleen (n = 5 mice per group). g, Percentage of human CD45⁺CD33⁺ and CD45⁺CD14⁺ cells, respectively, in the liver (n = 5 mice per group). h, Survival curve of mice treated with IOX5, Ven, IOX5 + Ven or vehicle CTL (CTL n = 7 mice; Ven, IOX5 and IOX5 + Ven n = 6 mice each). Statistical significance between Ven only and IOX5 + Ven is represented on the graph. Statistical significance between CTL and all other experimental conditions are described below. P values; CTL versus Ven P = 0.0403, CTL versus IOX5 P = 0.0059, CTL versus IOX5 + Ven P = 0.0027. i,j, Proliferation and annexin V⁺DAPI⁺ analyses, respectively, of THP-1 cells treated with Ven, IOX5, FIH inhibitor (FIHi; DM-NOFD), IOX5 + FIHi, IOX5 + Ven, FIHi + Ven, IOX5 + FIHi + Ven or vehicle CTL. Statistical significance between Ven only, IOX5 + Ven, FIHi + Ven or IOX5 + FIHi + Ven is represented on the graph. Statistical significance between CTL and all other experimental conditions are described below. CTL, Ven, FIHi, IOX5 + Ven, FIHi + Ven, IOX5 + FIHi + Ven n = 6; IOX5 n = 6 biological replicates per group at the 0, 24, 48 and 72-h time points, n = 5 biological replicates per group at the 96-h time point, IOX5 + FIHi n = 6 biological replicates per group at the 0, 24, 48 and 72-h time points, n = 5 biological replicates per group at the 96-h time point. P values; CTL versus IOX5 P = 0.0015, CTL versus IOX5 + FIHi P = 0.0054, CTL versus IOX5 + Ven P = 0.0004, CTL versus FIHi + Ven P = 0.0011, CTL versus IOX5 + FIHi + Ven P < 0.0001. IOX5 versus IOX5 + Ven P = 0.0091, IOX5 versus IOX5 + FIHi + Ven P = 0.0015. FIHi versus FIHi + Ven P = 0.0022, FIHi versus IOX5 + FIHi + Ven P < 0.0001. P values calculated at the 96-h time point (i). n = 6 biological replicates per group. P values; CTL versus Ven P = 0.0003, CTL versus IOX5 P < 0.0001, CTL versus IOX5 + FIHi P < 0.0001, CTL versus IOX5 + Ven P < 0.0001, CTL versus FIHi + Ven P = 0.0002, CTL versus IOX5 + FIHi + Ven P < 0.0001 (j). IOX5 versus IOX5 + Ven P < 0.0001, IOX5 versus IOX5 + FIHi + Ven P < 0.0001. FIHi versus FIHi + Ven P = 0.0003, FIHi versus IOX5 + FIHi + Ven P < 0.0001. Data represent mean ± s.e.m. Comparisons with no P value are NS. P values were calculated using a two-tailed Mann–Whitney U-test, paired or unpaired Student’s t-test, unless stated otherwise. Kaplan–Meier survival curve statistics were determined using the log-rank (Mantel–Cox) test. Source data

Extended Data Fig. 1. Loss of Phd2 reduces murine AML cell growth and LSC frequency in vivo.

(a) PHD2 (EGLN1) expression in AML patient samples with adverse, intermediate and favorable prognosis, respectively. Data from Beat AML1.0 and Beat AML 2.0 datasets, respectively. (Beat AML1.0, Adverse n = 162 patients, Intermediate n = 150 patients, Favorable n = 117 patients; Beat AML 2.0, Adverse n = 175 patients, Intermediate n = 100 patients, Favorable n = 158 patients). P values calculated by Ordinary One-way ANOVA. (b) Proliferation analyses with Control and Phd2^cKO cells (n = 6 biological replicates per genotype). P value calculated at 72 hr time point. (c) Percentage of GFP⁺ cells after 3 rounds of re-plating +DOX (n = 8 biological replicates per genotype). (d) iMLL-AF9 mice were crossed to Contol (rtTA) and shPhd2 (shPhd2/rtTA) mice, generating iMLL-AF9;Control and iMLL-AF9;shPhd2 mice, respectively. LSK cells from iMLL-AF9;Control and iMLL-AF9;shPhd2 mice were cultured in DOX-treated media and GFP expression was measured by flow cytometry. Representative flow cytometry plot (n = 3 mice per genotype). (e) Proliferation of iMLL-AF9;Control and iMLL-AF9;shPhd2 cells in the presence of DOX (iMLL-AF9;Control n = 9 biological replicates, iMLL-AF9;shPhd2 n = 12 biological replicates). (f) LDA in secondary recipients transplanted with cells from primary recipients (n = 6–7 mice per dose). Lower, estimate and upper LSC frequencies were calculated using the ELDA software. Exact n number per group and analyses from ELDA software is provided in the Source Data. Data represent mean ± SEM. Comparisons not marked with a P value = ns. P values were calculated using a two-tailed Mann–Whitney U-test, paired or un⁻paired student t-test, unless stated otherwise. Source data

Extended Data Fig. 2. PHD2 is dispensable for functional hematopoiesis.

(a-c) Analyses of 8⁻ to 10-week-old Control and Phd2^cKO mice. (a) Total numbers of B cells, Granulocytes and Monocytes in the BM (Control n = 7 mice, Phd2^cKO n = 3 mice). (b) Total numbers of LK and LSK cells in the BM (Control n = 7 mice, Phd2^cKO n = 4 mice). (c) Total numbers of Pre-GM, GMP, Pre-MegE, Pre-CFU-E and CFU-E, EryA, EryB, EryC and ProE progenitors (Pre-GM, GMP, Pre-MegE, Pre-CFU-E and CFU-E Control n = 7, Phd2^cKO n = 3; EryA, EryB, EryC and ProE Control n = 8 mice, Phd2^cKO n = 3 mice). (d-f) Analyses of 13- to 14-week-old Control and shPhd2 mice treated with DOX for 8 weeks. (d) Percentage of GFP⁺ cells in total BM, Lin^-, LK, LSK, HSC, MPP, HPC-1 and HPC-2 compartments (n = 5 mice per genotype). (e) Total numbers of B cells, Granulocytes and Monocytes cells in the BM (n = 5 mice per genotype). (f) Total numbers of LK and LSK cells in the BM (n = 5 mice per genotype). (g-j) Transplantation of HSCs from 8- to 10-week-old Control and shPhd2 mice. Recipient mice were treated with DOX 6 weeks post-transplantation. (g) Percentage of GFP⁺ cells in CD45.2⁺ in overall PB compartment. (Month 1, n = 7 Control mice; n = 9 shPhd2 mice. Month 2 n = 10 Control mice; n = 9 shPhd2 mice. Month 4 n = 9 Control mice; n = 10 shPhd2 mice). (h) Relative levels of Phd2 mRNA (normalized to β-actin) in total BM and LSK CD45.2⁺ sorted cells 16 weeks post-transplantation (Data analyzed from RNA extracted from n = 2 mice per genotype, plated in triplicate). (i) Percentage of CD45.2⁺ cells in the B cell, Granulocytes and Monocytes in the BM compartments of recipient mice (n = 4 mice per genotype). (j) The percentage of CD45.2⁺ cells within the LK, LSK, HSC, MPP, HPC-1 and HPC-2 compartments of recipient mice 16 weeks after transplantation (n = 4 mice per genotype). Data represent mean ± SEM. Comparisons not marked with a P value = ns. P values were calculated using a two-tailed Mann–Whitney U-test, paired or unpaired student t-test, unless stated otherwise. Source data

Extended Data Fig. 3. Modeling of PHD inhibitors.

(a) View from a crystal structure of PHD2 complexed with Molidustat (PDB: 6ZBO). (b) Visualization of Dap complexed with PHD2. (i) View of PHD2 complexed with CCT-6 (PDB: 5ox5) and structure of CCT-6. (ii) An overlay of binding mode of Dap (modeled) with CCT-6. (iii) Prediction of Dap binding mode with PHD2 (PDB: 5ox5) and structure of Dap. (c) Visualization of IOX5 complexed with PHD2. (i) View of PHD2 complexed with compound 1 (PDB: 6ST3) and the chemical structure of compound 1. (ii) An overlay of binding mode of IOX5 (modeled) with that of compound 1. (iii) Prediction of IOX5 binding mode with PHD2 (PDB: 6ST3) and structure of IOX5. Note Mn in the crystal structures substitutes for catalytically active Fe(II).

Extended Data Fig. 4. Inhibition curves of PHD inhibitors.

(a) PHD2 inhibition by IOX5 as assayed by SPE-MS. No enzyme and DMSO controls were used for normalization of the HIF-1α CODD peptide hydroxylation. Errors are standard deviations of the mean (n = 2 independent assays). Eleven concentrations assayed in duplicate were used to obtain dose-response curve for IOX5. Values in the graph are means with standard deviation (n = 2 independent assays). IC₅₀ values were calculated using GraphPad Prism. (b) in vitro FIH selectivity profiling of IOX5, as assayed by SPE-MS. Eleven concentrations assayed in duplicate were used to obtain a dose-response curve for IOX5. Values present in the graph are means with standard deviation (n = 2 independent assays). IC₅₀ values were calculated using GraphPad Prism. (c) in vitro JMJD6 selectivity profiling of N-oxalylglycine (NOG), 2,4-pyridine dicarboxylic acid (2,4-PDCA) and the FIH selective inhibitor N-oxalyl-D-phenylalanine (NOFD), Daprodustat, Molidustat, Roxadustat and IOX5 assayed by SPE-MS. Eleven concentrations assayed in duplicate were used to obtain dose-response curves. IC₅₀ values were calculated using GraphPad Prism (n = 2 independent assays). (d) in vitro KDM4A selectivity profiling of IOX5, as assayed by LCMS. Three concentrations were assayed in duplicate. (e) in vitro KDM5B selectivity profiling of IOX5, as assayed by LCMS. Three concentrations were assayed in duplicate. (f) in vitro KDM6B selectivity profiling of IOX5, as assayed by LCMS. Three concentrations were assayed in duplicate. (g) OGFOD1 selectivity profiling of N-oxalylglycine (NOG), 2,4-pyridine dicarboxylic acid (2,4-PDCA) and the FIH selective inhibitor N-oxalyl-D-phenylalanine (NOFD), Daprodustat, Molidustat, Roxadustat and IOX5 as assayed by SPE-MS (the protocol is described in the Supplemental Information section). Eleven concentrations assayed in duplicate were used to obtain dose-response curves for each compound. IC₅₀ values were calculated using GraphPad Prism (n = 2 independent assays). Source data

Extended Data Fig. 5. Therapeutic validation of PHD inhibitors in AML.

(a) Western blot of HIF-2α in MOLM13, OCI-AML3, MV411 and THP-1 cells treated with IOX5 or vehicle control for 48hrs. α-Histone 3 (H3) used as a loading control. Blot is representative of 3 independent experiments. (b) Proliferation and Annexin-V⁺DAPI⁺ analyses of murine NPM1/Flt3-ITD cells treated with Daprodustat (Dap), IOX5 or vehicle control (n = 5 biological replicates per group). (c) Proliferation analyses of THP-1, MOLM13, OCI-AML3, MV411 cells treated with Molidustat (Mol) or vehicle control (THP-1 n = 5 biological replicates per group; MOLM13, OCI-AML3 and MV411 n = 6 biological replicates per group). P value calculated at 72 hr time point. (d) Annexin-V⁺DAPI⁺ analyses of THP-1, MOLM13, OCI-AML3, MV411 cells treated with Molidustat or vehicle control. (THP-1 Control n = 10 biological replicates, Mol n = 5 biological replicates; MOLM13 Control n = 6 biological replicates, Mol n = 5 biological replicates, OCI-AML3 n = 6 biological replicates per group, MV411 n = 6 biological replicates per group). (e) Proliferation analyses of THP-1 cells treated with Roxadustat or vehicle control (Control n = 3 biological replicates, Roxadustat n = 2 biological replicates). (f) Western blot of HIF-1α in THP-1 cells treated with IOX5 for described time points. For vehicle control and positive control, cells were treated with DMSO and Dap, respectively, for either 48 or 96 hrs. β-actin used as a loading control. Blot representative of 2 independent experiments. Data represent mean ± SEM. Comparisons not marked with a P value = ns. P values were calculated using a two-tailed Mann–Whitney U-test, paired or unpaired student t-test, unless stated otherwise. Source data

Extended Data Fig. 6. Anti-leukemic potential of PHD inhibition in vivo.

(a) Weight of animals pre- and post-treatment with Dap, IOX5 or vehicle control 2x daily via i.p. injection for 21 days. (Day 14, Control n = 4 mice, Dap n = 6 mice, IOX5 n = 6 mice; Day 35, Control n = 5 mice, Dap n = 5 mice, IOX5 n = 3 mice). (b) Percentage of leukemic CD45.2⁺ cells and myeloid cells in PB of recipient mice pre-treatment with IOX5 or vehicle control (Control n = 7 mice, IOX5 n = 9 mice). (c) LDA analyses in secondary recipients transplanted with cells from primary recipients (n = 9–10 mice per dose). Lower, estimate and upper LSC frequencies were calculated using the ELDA software. (d-k) Analyses of 8- to 10-week-old C57Bl6 mice treated with IOX5 or vehicle control 2x daily via i.p. injection for 14 days. (d) Total spleen cellularity (Vehicle control-treated mice n = 5, IOX5-treated mice n = 9). (e) Total numbers of B cells, Granulocytes and Monocytes in the BM (Vehicle control-treated mice n = 6, IOX5-treated mice n = 9). (f) Total numbers of B cells, Granulocytes, Monocytes and T cells in the spleen (Vehicle control-treated mice n = 5, IOX5-treated mice n = 9). (g) Fold change of RBC, HCT and HGB in PB post-/pre-treatment (Vehicle control-treated mice n = 6, IOX5-treated mice n = 8). (h) Total numbers of erythrocyte progenitors in the BM (Vehicle control-treated mice n = 6, IOX5-treated n = 8). (i) Total numbers of erythrocyte progenitors in the spleen (Vehicle control-treated mice n = 5, IOX5-treated mice n = 7). (j) Total numbers of LSK and LK cells in the BM. (Vehicle control-treated mice n = 5 mice, IOX5-treated mice n = 7). (k) Total numbers of Pre-GM, GMP, Pre-MegE, Pre-CFU-E and CFU-E progenitors in the BM (n = 6 mice per group). Data represent mean ± SEM. Comparisons not marked with a P value = ns. P values were calculated using a two-tailed Mann–Whitney U-test, paired or unpaired student t-test, unless stated otherwise. Source data

Extended Data Fig. 7. Dysregulated transcripts following PHD inhibition.

(a) Scatter-plot of transcripts in control and IOX5-treated iMLL-AF9 cells (n = 4 RNA samples per group). Transcripts significantly up- (red) and downregulated (blue) in IOX5-treated iMLL-AF9 cells are highlighted. (b) Volcano plot of differentially expressed genes in IOX5-treated iMLL-AF9 cells. Genes positively- and negatively- regulated by HIF-1α are marked in blue and red, respectively. (c) Expression of Hspa8, Nup98, Kpnb1 and Rbm15 in IOX5-treated iMLL-AF9 cells. Violin plots showing distribution of log2 expression values (transcripts per million). (n = 4 RNA samples per group). (d) Gene expression of Kdm5b, Bnip3, and Fam162a in IOX5-treated iMLL-AF9 cells. Violin plots show the distribution of log2 expression values (transcripts per million). (n = 4 RNA samples per group). (e) Relative levels of BNIP3 mRNA (normalized to β-ACTIN) in MOLM13 cells transduced with lentiviruses expressing scrambled short hairpin RNA (shCTL) and a shRNA targeting BNIP3 (shBNIP3) (n = 1 biological replicate per group, plated in triplicate). (f) Proliferation analyses of shCTL- and shBNIP3-transduced MOLM13 cells treated with IOX5 or vehicle control (n = 3 biological replicates per group). Data represent mean ± SEM. Comparisons not marked with a P value = ns. P values were calculated using a two-tailed Mann–Whitney U-test, paired or unpaired student t-test, unless stated otherwise. Source data

Extended Data Fig. 8. FIH expression across healthy and AML samples.

(a) FIH (HIFAN) gene expression, respectively, across human AML bone marrow samples, healthy CD34⁺ BM cells and BM MNCs. (CD34⁺ n = 16 patients; BM MNC n = 19 patients; AML n = 671 patients). Source data